Copyright is owned by the Author of the thesis.  Permission is given for 
a copy to be downloaded by an individual for the purpose of research and 
private study only.  The thesis may not be reproduced elsewhere without 
the permission of the Author. 
 
  
Measuring performance in farming: A comparative analysis 
of dairy production systems in New Zealand and Chile 
 
 
 
A thesis presented in partial fulfilment of the requirements for the degree of  
Master  of AgriScience  
at Massey University, Turitea, Palmerston North,  
New Zealand  
 
 
 
 
 
Licy Maren Beux Garcia 
 
2013  
  
 i 
Abstract 
The purpose of this work was to identify, examine, and compare the key performance indicators and 
drivers of success of pasture- based dairy systems in New Zealand and Chile . Key similarities and 
differences between dairy farming systems in these countries were identified by analysing data 
provided by DairyBase and,  its Chilean counterpart, TodoagroBase. Comparable observations were 
nested using country- specific classification systems based on existing knowledge , followed by the 
estimation of efficiency scores for each individual observation within these classes using Data 
Envelopment Anal ysis (DEA). Efficiency scores were then attached to the original datasets and used 
as the response variable in several country- specific Regression Partitioning Trees . This procedure 
identified the most relevant benchmarks in each country and showed that there are various pathways 
to high efficiency. Knowledge gains  provided by this research are expected to  influence farming 
practices and management, research and extension, and to encourage future cooperation between the 
two countries. 
Dairy farmers in New Zealand and Chile benefit from low - cost production advantages because of 
their favourable environment for pasture- based dairying, efficiently and profitably producing milk at 
a lower cost than the world’s average. However, a large variability in farming systems within the 
countries was identified, as were different benchmarks . In New Zealand, herd productivity and 
labour played key roles  in defining efficiency, while in Chile, herd productivity and supplements fed 
per litre of milk produced were key indicators  explaining efficiency. In New Zealand, operating cost 
per kg of milk solids, r eturn on Assets (ROA), operating profit margin  (OPM ) , operating profit per 
hectare, and asset turnover (ATR) were also major indicators. In Chile, gross farm revenue per cow, 
cost of production per litre of milk produced, wages per litre, operating profit per cow and ATR  
were also highlighted. The absence of indicators such as ROA in Chile was noticeable. 
Reasons for d ifferent key performance indicators occurring in each country stem from history to 
geography, and have resulted in differences in values and goals. New Zealand farmers are 
profitability and cost- focused, looking alternatively to both OPM and the capital invested. C hilean 
farmers are revenue- focused and respond strongly to milk:feed price ratio and to the efficiency in the 
use of supplement. In both countries, the systems are evolving in similar ways, gradual ly increasing 
intensification levels and specialisation. In both countries, consistently high performing farms are 
efficient at producing both milk and revenue, and are more likely to have higher herd productivity 
and labour efficiency than poorer performers. In New Zealand, consistently efficient farms also had 
significantly better asset use as reflected by their ROA and ATR. In Chile better performers used 
significantly less supplement per litre of milk produced.  
Keywords: pasture- based, farming system, ef ficiency, benchmarks, New Zealand, Chile  
 
ii 
 
 
 
 iii 
Acknowledgements 
Many  thanks go to the New Zealand Ministry of Foreign Affairs and Trade and to Sylvia 
Hooker and her beautiful team: Olive, James, Leuiana, and Sue from the NZAID programme. A 
special thank you, too, to Natalia Benquet from the International Support Office; Alfredo and I 
will always be grateful to and remember you all.  
Thank you to all the direct and indirect contributors to this study. After being introduced to 
Nicola Shadbolt in 2010 by Daniel Conforte , I got to know Nicola as an advisor, a professor, 
and as a supervisor. For a number of reasons, Nicola taught me what kiwi humility really 
means. Ever since 2010, it has been an honour; thank you. Special thanks to Elizabeth Dooley 
for being my supervisor, for looking after me, and for teaching and guiding me during the highs 
and lows of this process. Thanks also to Chris Dake and Jonathan Godfrey for sharing a bit of 
your brilliant minds and ideas with me. From the department, thanks Lee - Anne, Kate, Linda , 
and Denise Stewart for always being willing to help. From Uruguay, Ariel Asuaga has been a 
friend and a great source of inspiration and confidence. 
At Massey, during my first year in New Zealand, I had the pleasure of assisting to  the classes of 
Marta Camps, Dave Gray, Michael Hedley, David Horne, Matthew Irwin, Peter Kemp, Cory 
Matthews, Alan Palmer, Ranvir Singh, and Mike Tuohy. Thank you also to Lisa Haarhoff and 
Ian Furkert. In particular, I am grateful to Tania, Julia Rayner, Jacqui Burne, and Lois 
W ilkinson from the Centre for Teaching and Learning , for your time and patience; you have 
been a  great help all the way!   
Thanks to the Modelling and Breeding Club for sharing your knowledge and experiences; to 
some of you also for being our friends. To my office mate, Federico: it has been good fun 
despite the ‘disturbed environment’. Thank you to the Kay (one special thought goes to our 
favourite chestnut, Squash, and grey, Saigo), the Lane, the Claridge, and the Guscott families 
for helping us to understand several aspects of New Zealand farming systems, combining both 
fun and work.  
To our ‘gringo’ friends: the Wier (incre dible Harry and  Chloe, Laura, Tessa, Peter, and the 
lovely Ashley), the Charteris, and the Dobson (including all your beautiful horses) families. 
Thanks Aotearoa. Haere ra. Finally, we offer a special acknowledgement to our friends from the 
Latin American community at Massey University  and Palmerston North.  
Finally, thanks go to our many friends in Uruguay, including Quique Iturralde and Carlos 
Bautes, who ar e gone forever. To our families; you are always in our hearts and thoughts. Thank 
you for coming and sharing this awesome place and also for patiently waiting for us to come 
back home, including ‘abuela Marficia’,… somewhere.                                               Licy Maren  
iv 
 
 
 
 v 
TABLE OF CONTENTS  
ABSTRACT .................................................................................................................................. i 
ACKNOWLEDGEMENTS ....................................................................................................... iii 
LIST OF TABLES ..................................................................................................................... ix 
LIST OF FIGURES ................................................................................................................... xi 
1. INTRODUCTION. . ................................................................................................................ 1 
1.1 The Country Context  ............................................................................................................... 3 
1.2 Research Motivation  ............................................................................................................... 4 
1.3 Statement of Purpose  ............................................................................................................... 4 
1.4 Research Questions and Specific Objectives  .......................................................................... 4 
1.5 Outline of the Study  ................................................................................................................ 5 
2. LITERATURE REVIEW. . .................................................................................................... 7 
2.1 Pasture –Based Dairy Systems  ................................................................................................. 7 
2.1.1 Milk output from pasture.  ................................................................................................. 8 
2.1.1.1 Seasonality. ............................................................................................................. 10 
2.1.1.2 Feeding strategies. .................................................................................................. 11 
2.1.1.3 Stocking rate. ........................................................................................................... 12 
2.1.1.4 Calving systems. ...................................................................................................... 13 
2.1.1.5 Genotype. ................................................................................................................ 14 
2.1.1.6 Economics. .............................................................................................................. 15 
2.1.1.7 Diversification or specialisation strategy? ............................................................. 16 
2.1.2 The challenges. ............................................................................................................... 16 
2.2 Farming Systems Approach  .................................................................................................. 19 
2.2.1 Values and goals. ............................................................................................................ 19 
2.2.1.1 Lifecycle stage and age of the business. .................................................................. 20 
2.2.1.2 Business operator’s age. ......................................................................................... 21 
2.2.2 Structure.  ........................................................................................................................ 21  
2.2.2.1 Land ownership. ...................................................................................................... 22 
2.2.3 Family businesses.  .......................................................................................................... 23 
2.2.4 Management in farm businesses.  ................................................................................... 24 
2.2.4.1 The manager’s entrepreneurial orientation. ........................................................... 27 
2.3 Metrics Used for Success Appraisal in Pasture–Based Systems ........................................... 28 
2.3.1 Physical Key Performance Indicator s (KPIs).  ................................................................ 28 
vi 
2.3.2 Financial KPIs. ............................................................................................................... 30 
2.3.2.1 Profitability. ............................................................................................................ 30 
2.3.2.2 Solvency, liquidity, and wealth. ............................................................................... 34 
2.3.2.3 Resilience measures. ............................................................................................... 36 
2.4 Macro- environment ............................................................................................................... 38 
2.4.1 Economic factors.  ........................................................................................................... 38 
2.4.2 Political and legal factors.  .............................................................................................. 42 
2.4.3 Ecological and climate factors.  ...................................................................................... 43 
2.4.4 Socio- cultural and consumer factors. ............................................................................. 46 
2.4.5 Technological factors. .................................................................................................... 48 
2.4.5.1 New Zealand dairy farming systems. ...................................................................... 49 
2.4.5.1.1 The pasture resource and feeding systems. ...................................................... 52 
2.4.5.1.2 New Zealand dairy calving systems. ................................................................ 53 
2.4.5.1.3 New Zealand breeds. ........................................................................................ 54 
2.4.5.2 Chilean dairy farming systems. ............................................................................... 55 
2.4.5.2.1 The pasture resource and feeding systems. ...................................................... 57 
2.4.5.2.2 Chilean dairy calving systems and breeds. ...................................................... 57 
2.4.5.3 Milk payment systems. ............................................................................................. 58 
2.4.5.4 Recent empirical research. ...................................................................................... 60 
2.5 Benchmarking Farming Systems  .......................................................................................... 63 
2.5.1 Why benchmark? ............................................................................................................ 64 
2.5.2 Be nchmarking and benchmarks. .................................................................................... 64 
2.5.2.1 Benchmarking forms and types. .............................................................................. 65 
2.5.2.2 Benchmarking as a farm management tool. ............................................................ 66 
2.5.2.3 Data envelopment analysis (DEA): a method for benchmarking. ........................... 67 
2.5.2.4 Regression or Recursive partitioning Tree ............................................................. 71 
2.5.3 Generic competitive strategies.  ...................................................................................... 72 
2.5.3.1  Differentiation. ....................................................................................................... 73 
2.5.3.2  Focus. ..................................................................................................................... 73 
2.5.3.3 Overall cost leadership. .......................................................................................... 73 
2.5.3.4 Competitiveness within groups. ............................................................................... 75 
2.5.3.5 Competitiveness across groups. .............................................................................. 75 
2.5.4 The International Farm Comparison Network. .............................................................. 76 
3. METHOD. .............................................................................................................................. 79 
3.1 Research Strat egy .................................................................................................................. 79 
3.2 Materials and Data Analysis Methods  .................................................................................. 79 
3.2.1 DairyBase and TodoagroBase overview.  ....................................................................... 80 
3.2.1.1 Data quality and limitations. ................................................................................... 80 
3.2.2 Review of methods.  ........................................................................................................ 82 
3.2.2.1 DEA models. ............................................................................................................ 82 
3.2.2.2 Recursive partitioning tree. ..................................................................................... 83 
3.2.2.3 Metafrontier envelope. ............................................................................................ 85 
3.2.3 Procedure: the data analysis.  .......................................................................................... 85 
 vii 
3.2.3.1 Sampling. ................................................................................................................. 86 
3.2.3.2 Variable correction and creation. ........................................................................... 88 
3.2.3.3 The NZ classification system. .................................................................................. 89 
3.2.3.4 The Chilean classification system. .......................................................................... 93 
3.2.3.5 Class-specific DEA. ................................................................................................. 94 
3.2.3.5 Recursive Partitioning Trees fitting and KPI selection. .......................................... 96 
3.2.3.6 Benchmarks sets. ..................................................................................................... 96 
3.2.3.7 Metafrontier analysis. ............................................................................................. 97 
4. INTERNATIONAL FARM COMPARISON NETWORK RESULTS.  .......................... 99 
4.1 Typical NZ and Chilean Dairy Farms Comparison  ............................................................... 99 
4.2 Discussion and concluding comments on the IFCN comparison  ........................................ 103 
5. NEW ZEALAND RESULTS.  ............................................................................................ 109 
5.1 Data Envelopment Analyses  ............................................................................................... 109 
5.2 Recursive Partitioning Trees results  .................................................................................... 115 
5.2.1 Milk Solids Results  ...................................................................................................... 115 
5.2.2 Gross Farm Revenue Results.  ...................................................................................... 120 
5.2.3 The fifth NZ tree  .......................................................................................................... 125 
5.3 NZ Benchmarks Set  ............................................................................................................ 125 
6. CHILEAN RESULTS. ........................................................................................................ 133 
6.1 Data Envelopment Analyses (DEA) .................................................................................... 133 
6.2 Regresion Partitioning Tree results  ..................................................................................... 135 
6.2.1 Milk  results.  ................................................................................................................. 135 
6.2.2 GFR results.  .................................................................................................................. 140 
6.2.3 The fifth Chilean tree  ................................................................................................... 145 
6.3 Chilean Benchmarks Set  ..................................................................................................... 146 
7. METAFRONTIER ANALYSES. ...................................................................................... 153 
7.1 New Zealand Metafrontier Results  ..................................................................................... 154 
7.2 Chilean Metafrontier  ........................................................................................................... 157 
7.3 Metafrontier across countrie s .............................................................................................. 161 
8. GENERAL DISCUSSION.  ................................................................................................ 163 
8.1 The Non- financial Benchmarks  .......................................................................................... 165 
viii 
8.2 The Financial Benchmarks  .................................................................................................. 170 
8.3 The Benchmark Sets Applied to Different Performance Groups  ........................................ 176 
8.4 The Metafrontier Analyses  .................................................................................................. 179 
8.5 The Metafrontier Across Countries  ..................................................................................... 181 
8.6 Review of Method  ............................................................................................................... 182 
9. CONCLUSIONS. ................................................................................................................ 187 
10. REFERENCES.  ................................................................................................................ 189 
 
APPENDICES. .......................................................................................................................  215 
Appendix One: DairyBase variables and ratios  ........................................................................ 216 
Appendix Two: TodoagroBase Variables and Ratios  ............................................................... 224 
Appendix Three: R Codes  ......................................................................................................... 228 
Appendix Four: The NZ Classification System, an Iterative Process  ....................................... 229 
Appendix Five: IFCN Results  ................................................................................................... 239 
Appendix Six: 5 th Chilean Tree Including Assets as Input Variable  ......................................... 240 
Appendix Seven: 5 th NZ Tree Using ROA as Response Variable  ............................................ 241 
Appendix Eight: Frequency of Consistently Efficient Farms  ................................................... 242 
 
 
 ix 
 List of Tables 
 
Table 1 Average Figures for Normal Raw Milk at 20oC ................................................................. 9 
Table 2 Basic Comparative Figures.............................................................................................. 39 
Table 3  Economic Indicators of Selected Countries .................................................................... 40 
Table 4 Dairy Industry Facts ........................................................................................................ 46 
Table 5 Advantages and Disadvantages of Respective Dairy Industries ..................................... 47 
Table 6 Operating Structure of Dairy Farms in Both Islands and at the National Level .............. 51 
Table 7 Frontier Methods Taxonomy .......................................................................................... 68 
Table 8 Comparison of Representative Dairy Systems ................................................................ 77 
Table 9 Sustainability Areas and Indicators for Dairy Farms Worldwide .................................... 78 
Table 10 Original NZ and Chilean Samples .................................................................................. 87 
Table 11 Number of DMUs Included in the Analysis per Year and Country ................................ 88 
Table 12 Estimated Seasonal Prices for Output (MS)  ................................................................. 89 
Table 13 Averaged Indicators Showing the Relevance of Grouping into Farming System ......... 90 
Table 14 NZ Classification System ............................................................................................... 92 
Table 15 Chilean Classification system ........................................................................................ 94 
Table 16 Inputs Included in the Output DEA Models by Country ................................................ 95 
Table 17 Weighted Averages for DEA Results by Year .............................................................. 109 
Table 18 DEA Results per Class Within Season 2006/07 ........................................................... 111 
Table 19 DEA Results per Class Within Season 2007/08 ........................................................... 112 
Table 20 DEA Results per Class Within Season 2008/09 ........................................................... 112 
Table 21 DEA Results per Class Within Season 2009/10 ........................................................... 113 
Table 22 DEA Results per Class Within Season 2010/11  .......................................................... 114 
Table 23 Key Performance Indicators as Revealed by MS Trees ............................................... 120 
Table 24 Key Performance Indicators as Revealed by GFR Trees .............................................. 124 
Table 25 NZ Benchmarks Set Comprising Physical and Financial KPIs ...................................... 126 
Table 26 Descriptive Statistics on Selected Physical KPIs .......................................................... 127 
Table 27 Comparison of Means for the Output Produced per Cow Indicator ........................... 128 
Table 28 Comparison of Means for the Output produced per Labour Unit Indicator ............... 128 
Table 29 Descriptive Statistics on Selected Financial KPIs......................................................... 129 
Table 30 Comparison of Means for Dairy ROA .......................................................................... 130 
Table 31 Comparison of Means for Operating Profit Margin ................................................... 130 
Table 32 Comparison of Means for Operating Profit per Hectare ............................................ 131 
Table 33 Comparison of Means for Operating Expenses per Kg of Output ............................... 131 
Table 34 Comparison of Means for Assets Turnover Ratio in Different Classes ........................ 131 
Table 35 Weighted Averages for DEA Results by Year .............................................................. 133 
Table 36 DEA Results per Class for the Period 2007-11 ............................................................ 134 
Table 37  Key Performance Indicators (KPIs) as Revealed by MS Trees .................................... 140 
Table 38 Key Performance Indicators as Revealed by GFR Trees .............................................. 144 
Table 39 Chilean Benchmarks Set Comprising Physical and Financial KPIs .............................. 146 
Table 40 Descriptive Statistics in Physical KPIs.......................................................................... 147 
Table 41 Comparison of Means for Litres per Cow Indicator (L/cow) ....................................... 148 
x 
Table 42 Comparison of Means for Supplement Used Per Litre of Milk produced (g/L) ........... 148 
Table 43 Descriptive Statistics of Financial Indicators .............................................................. 149 
Table 44Comparison of Means for the Wages ratio (CLP$/L) ................................................... 150 
Table 45 Comparison of Means for the Production Cost Indicator (CLP$/L) ............................. 151 
Table 46 Comparison of Means for the Operating Profit per Cow Indicator (CL$/Cow) ........... 151 
Table 47 Descriptive Statistics and Inputs/Output used for Metafrontier Efficiency in New 
Zealand ...................................................................................................................................... 154 
Table 48 Comparison of Means for the Metafrontier Scores .................................................... 155 
Table 49 Comparison of Means for the CPI -adjusted Gross Farm Revenue ............................. 156 
Table 50  Comparison of Means for the CPI-Adjusted Indicator Operating Expenses .............. 156 
Table 51 Descriptive Statistics and Inputs/Output used for Metafrontier Efficiency in Chile ... 157 
Table 52 Comparison of Means for the Metafrontier Scores .................................................... 158 
Table 53 Comparison of Means for the CPI-Adjusted Gross Farm Revenue .............................. 159 
Table 54  Comparison of Means for the CPI-Adjusted Operating Expenses .............................. 159 
Table 55 Comparison of Means for the Physical Input Cows .................................................... 160 
Table 56 Comparison of Means for the Physical Input Supplements per Cow .......................... 160 
Table 57 Comparison of Means for the Metafrontier across Countries .................................... 162 
Table 58 Production, Labour and Feed Benchmarks ................................................................. 165 
Table 59 Financial Benchmarks and Similarities Across Countries ............................................ 171 
 
List of Tables in Appendices 
 
Table A-1 List of Variables DairyBase Analyses ......................................................................... 216 
Table A-2 List of variables TodoagroBase Analyses .................................................................. 224 
Table A-3 Descriptive Regional Analysis of Selected KPIs.......................................................... 230  
Table A-4 Descriptive Seasonal Analysis of Selected KPIs ......................................................... 232 
Table A-5 Descriptive Farming System analysis of  selected  KPIs............................................. 233 
Table A-6 IFCN Approach New Zealand Versus Chile ................................................................ 239  
Table A-7 Frequency of Consistently Efficient Farms Across Years, Region, Farming 
System, Irrigation and Size ........................................................................................................ 242 
  
 xi 
List of Figures 
 
Figure 1. NZ$ versus US$ exchange rate. .................................................................................... 40 
Figure 2. Chilean $ versus US$ exchange rate. ........................................................................... 41 
Figure 3. New Zealand trend in the average herd size and number of herds............................. 42 
Figure 4. New Zealand and Chile dairy regions and latitudes. .................................................... 44 
Figure 5. Evolution in MS production per cow and per effective area since 1992/93. .............. 49 
Figure 6. CRS and VRS DEA envelopes. ....................................................................................... 71 
Figure 7. Averaged farm operating expenses within NZ classes. ................................................ 93 
Figure 8. Economic performance of the typical NZ farm relative to the world’s average. ......... 99 
Figure 9. Economic performance of the typical Chilean farm relative to the world’s average. . 99 
Figure 10. NZ strong and weak profile. ..................................................................................... 101 
Figure 11. Chile strong and weak profile. ................................................................................. 101 
Figure 12. Typical Chilean dairy farm compared to the typical NZ dairy farm. ........................ 103 
Figure 13. MS DEA partitioning tree. ........................................................................................ 116 
Figure 14. Modified MS DEA partitioning tree. ......................................................................... 118 
Figure 15. GFR DEA partitioning tree. ....................................................................................... 121 
Figure 16. Modified GFR DEA partitioning tree......................................................................... 123 
Figure 17. MS DEA partitioning tree. ........................................................................................ 136 
Figure 18. Modified MS DEA partitioning tree. ......................................................................... 138 
Figure 19. GFR DEA partitioning tree. ....................................................................................... 141 
Figure 20. Modified GFR DEA partitioning tree......................................................................... 143 
Figure 21. Boxplot of metafrontier efficiency across countries. ............................................... 161 
 
List of Figures in Appendices 
 
Figure A-1. Herd size by region and intensification level .......................................................... 234 
Figure A-2. Stocking rate by region and intensification level.................................................... 235 
Figure A-3. FTEs per hectare by region and intensification level .............................................. 235 
Figure A-4. Milk solids per hectare by region and intensification level. ................................... 236 
Figure A-5. Milk solis per cow by region and intensification level. ........................................... 236 
Figure A-6. Operating expenses per hectare by region and intensification level. .................... 237 
Figure A-7. Operating profit per hectare by region and intensification level. .......................... 237 
Figure A-8. Operating profit margin (%) by region and intensification level. ........................... 238 
 
  
xii 
 
Chapter I: Introduction 
1 
1. Introduction. 
 
A farming system might simplistically be defined as the particular combination of resources 
operating over time, under similar socioeconomic conditions and producing a certain set of 
outputs (Cochet &  Devienne, 2006; National Research C ouncil, 2010) . Most importantly, 
farming systems are complex entities, exposed and sensitive to  all types of ex ternal fluctuations 
and particularly prone to potential rapid change in system drivers. R esearch in this field has 
often been identified as high- priority, and , as such, it continues to be encouraged. The N ational 
R esearch Council [NRC]  (2010) stated :   
Agricultural research and development programs should aggressively fund and pursue 
integrated research and extension on farming systems that focus on interactions among 
productivity, environmental, economic and social sustainability outcomes. Research 
should explore the properties of farming systems and how they could make the systems 
robust and resilient over time (p. 22) . 
Support for these concepts can be found worldwide and in almost every industry, coming from 
major entities  that promote extension and research such as Dairy  NZ  (2010)  and Consorcio 
Lechero  (2010) . Irrespective of  the industry, the a vailability of relevant information is 
increasingly needed, since the farm business  ability to succeed and sometimes to survive rests 
on making  informed decisions ( Dillon, Hennessy, Shalloo, Thorne, & Horan, 2008). Some of 
this information may be provided by empirical research, and particularly by the long term 
assessment of different farming systems (NRC,  2010). According to Thorne and Fingleton  
( 2006 ) , there is an increasing requirement f or transparency and information- sharing around the 
world in response to increasing trade liberalisation and globalisation which have led to farmers 
seek ing more resilient and profitable farming systems. International efficiency analyses and 
benchmarking activities have the added value of encouraging increased sampling of firms from 
a range of geographies, policies and market contexts , allowing comparisons of performance and 
the development of methods for improved systems (McGuckian, 1996). This suggests that 
international comparisons of business performance are actually useful excercises.  
D airy farming is important for a number of reasons including the nutritional value of milk , the 
significance of its global production, and the trade that the sector underpins. M ilk  is perhaps the 
most important single agricultural commodity in the world:  40 per cent of the world ’s  
population consumes milk daily (Blasko, 2011) , contributing significantly to human health and 
food security (Swinnen & Squicciarini, 2012). The milk indus try is large and has a strong 
growth forecast; according to Blasko ( 2011) ,  711 million tonnes of milk  were produced in 2010, 
Chapter I: Introduction 
2 
and future production may rise to over 794 million tonnes in 2017. This increase in supply 
would be supported by a demand that is projected  to escalate in both developing and developed 
countries (Popp, 2010). As a consequence, the  dairy sector has a promising outlook , even 
though additional supply is expected to drive prices down from current levels beyond 2012 
( Ministry of Agriculture and F orestry [MAF], 2011; Organisation for Economic Co - operation 
and Development [OECD], 2011).  
Incre asingly, milk producers operate  in a global economy, interacting through diverse linkages 
boosted by free trade policies. This setting provides both opportunities and threats, and forces 
successful farmers to continually assess their external environment and internal resources to 
meet their long term goals ( Kay, Edwards ,  & Duffy, 2012; Olson, 2010 ). Focusing on the 
positive side of this dichotomy, globalisation represents an opportunity for those  farming 
systems, countries and regions to specialise where  they have a comparative advantage. F or 
producers of commodities in general and dairy farmers in particular, c ost advantage provides a 
competitive strength, and low- cost leadership is the generic strategy that best fits their external 
environment ( Langemeier, 2010; Olson, 2010; Shadbolt, 2011). As pasture is the cheapest 
source of ruminant feed, pastoral farming systems are well- suited to that strategy of cost 
leadership. In particular , dairy farming systems show a strong inverse relationship between 
costs and graz ing because feed is the highest single cost component ( Dillon, et al., 2008 ;  
Lapple , Hennessy & O’Donovan , 2012; van Vuuren, 2006). Therefore, the  recognition of the 
advantage of low- cost production is important and dairy producers that benefit from a 
favourable climate for pasture production need to exploit this  as a comparative advantage 
(Lapple,  et al., 2012).  
Furthermore, pasture - based farming systems are less controversial than supplement intensive 
farming systems in terms of human food security, animal welfare , and environment. 
Consequently, pasture is used worldwide as the main source of feed for a range of animal 
production systems including dairying. However, given the biological nature of agriculture g reat 
variation can still be found among pasture- based systems. A particular farming system can be 
distinguished from the rest, in terms of environment and te chnology, by looking at the resources 
available and system configuration. Griliches (1987) defined technology as the state of 
knowledge concerning ways of converting resources or inputs into product or output. The 
output achieved by using that knowledge, r elative to the maximum amount of output physically 
achievable by mastering a given technology and inputs, is defined as efficiency (Diewert & 
Lawrence, 1999). Exogenous conditions that are likely to affect efficiency are soil, climate, and 
other physical parameters (Fraser & Cordina, 1999) which vary across the terrain and affect the 
technology adopted (Sumner & Wolf, 2002). Therefore, a ssuming the existence of different 
Chapter I: Introduction 
3 
technologies is customary in inter- country comparisons in the same way that the spatial 
dimension is clearly significant in terms of efficiency (Alston, 2002).  
1.1 The Country Context 
Many  rural economies have developed with agricultural production as their main activity, 
including those in N ew Z ealand (NZ)  and Chile. The urban and rural populations are distributed 
similarly in both countries ( 86% NZ, 89%  Chile), but the labour force occupied in agriculture in 
Chile is almost double that of NZ (13% to  7%) (C entral I ntelligence Agency [CIA] , 2011). In 
turn, the contribution of agriculture to Gross Domestic Product (GDP) , including downstream 
activities, is around 15% in both countries. The countries also have a number of other 
commonalities relevant to the macro- environment in which they operate. NZ exports are 
delivered to more than 150 countries  as a consequence of a consolidated modern open market 
system; d eveloping countries are the main destination while k ey markets are China, the U nited 
S tates (US) , Japan , and the European Union ( EU )  (Blasko, 2011). Similarly, Chile has shown a 
longstanding commitment to trade liberalisation, having 59 bilateral or regional trade 
agreements with various regions and countries including the EU, the US, M ERCOSUR 1 , China, 
India, South Korea, Mexico, Singapore , and New Zealand (Agosin & Bravo Orteg a, 2009).  
N Z ’s land area is about one third that of Chile and its population less than one quarter of Chile’s 
population (CIA, 2011). In addition, the adjusted self - sufficiency in milk supply data shows 
that, while NZ’s contribution to dairy world trade is large, the C hilean contribution is negligible 
(CIA, 2011) . The level of development in the dairy industry is a major difference between the 
two countries. NZ accounts for just over 2% of the total world milk production, but makes up to 
one- third of the world’s cross - border trade in dairy products (Blasko, 2011). N ationally and 
internationally, the NZ dairy industry is seen as a pastoral giant, where cohesiveness and co -
operation abound and the transparency of milk pay- outs operates as a powerful driver for 
collective action (MAF, 2011). Virtually all NZ milk is processed into dairy products and 
exported. The NZ Insti tute of Economic Research (2011) , estimates that the dairy industry 
contributes NZ$ 10.4 billion of export ea rnings or 2.8% of the national GDP  including 
downstream activities such as marketing and transport  to the economy. In contrast, dairy 
production in Chile has traditionally supplied the domestic market demand but  several factors 
including social, political, and economic instability have adversely affected the development of 
its dairy industry (Dobson, 2003). In  1977, Chile produced 1,003 million litres of milk that were 
100% internally marketed (Zegers & von Baer, 1978) , and nowadays dairy still provides a 
                                                          
1 MERCOSUR stands for Mercado Comun del Sur, meaning The Southern Common Market. It is an association 
between Argentina, Brazil, Paraguay, Uruguay and Venezuela, as full members. Bolivia, Chile, Ecuador and Peru 
have the status of associated countries, while Mexico is also included as an observer. The objectives of MERCOSUR 
are to allow for the free movement of goods, services and factors of production between the countries through the 
elimination of customs duties and non-tariff restrictions; to establish a common external tariff and to adopt a 
common trade policy in relation to third States or groups of States (MERCOSUR, n.d.) 
Chapter I: Introduction 
4 
minor input into the Chilean government revenue ( Banco Central de Chile [ BCCL ] , 2012) . 
However, d airying is seen as a sector full of growth opportunities, with Chile having recently 
become a net exporter of milk products  (Barton, Gwynne & Warwick, 2007).  
1.2 Research Motivation 
This research is motivated by the need for further knowledge about pasture - based dairy farming 
systems. A major premise is that a rigorous analysis of the past may contribute to knowledge 
gains, which could directly influence future farming practices and management, policy making, 
research, and marketing programmes. A second premise is that NZ and Chilean pastoral dairy 
farming systems may mutually benefit from being benchmarking partners . Therefore, an 
exploration of  these systems was proposed here based on key metrics testing performance.  This 
summary set of performance indicators allowed the different farming systems to be compared. 
Despite the us e of only quantitative measures to analyse the farming systems, this thesis aims 
for a holistic view by incorporating rich contextual information. This study intends to use  a 
systems approach, including external environment  considerations and diagnoses of the techno-
economic performance, to search for the reasoning behind common practices leading to the 
production of milk.  
1.3 Statement of Purpose  
The purpose of this study is to identify,  examine and compare key performance indicators and 
drivers of efficiency of pasture- based dairy systems within New Zealand and Chile . This will be 
achieved using standardised metrics and methodology to look at empirical data generated by 
two national databases over five consecutive seasons.  
1.4 Research Questions and Specific Objectives  
The research questions in this study were :  
1. Which are the most relevant benchmarks to assess and compare dairy farming systems in 
Chile and NZ?  
2. Are these benchmarks the same in both countries?  
3. If the benchmarks were not the same, why this would be ?  
4. How can ‘performance’ best be measured in NZ and Chile?  
5. What is the relative efficiency of the normal NZ and Chilean pasture - based dairy farming 
system? 
 
  
Chapter I: Introduction 
5 
The specific objectives were :  
1. To gain a deeper understanding of pastoral based dairy farming systems in NZ and Chile, in 
terms of their internal resources, key performance indicators, and drivers of success.  
2. To identify differences and similarities between the country- specific farming systems.  
Some key  assumptions were:  
1. A farming system may be described with reasonable accuracy, after understanding physical 
and socio- economic factors as part of the whole agrarian system context,  and through 
reviewing accounting records.  
2. The use of large databases would allow general conclusions to be drawn from statistical 
inference. 
3. Benchmarks can be used to identi fy performance gaps, while benchmarking practices can 
identify how to close these gaps ( for example, imitation of superior practices or correction 
of low- grade practices) (Camp, 1989)  
4. The assessment of the external environment and the on- farm competitive advantages, are 
important for reaching meaningful conclusions. The environment and strategic management 
both influence a farming system’s structure and shape competitive advantages (Porter, 1985, 
in Shadbolt, 2011), which if effectively exploited, will  allow a farm to outperform 
competitors and exhibit greater performance over time (Pertusa- Ortega, Molina- Azorin & 
Claver- Cortes, 2010).  
1.5 Outline of the Study 
The justification and purpose for the study ha ve been set out in this chapter. In Chapter Two, the  
relevant literature on pasture- based dairy systems, farming systems approach es, performance 
measurement, the macro - environment, and benchmarking as a farm management tool , is 
discussed. The method used to assess performance and answer the research questio ns is 
presented in Chapter Three. The results and summary statistics for the major performance 
indicators are presented in Chapter Four, Five and Six. The results of the metafrontier analyses 
and the cross- country comparison are described in Chapter Seven . In Chapter Eight  the main 
findings are discussed and compared with the literature. Finally, in Chapter Nine  the 
conclusions from this study are drawn and areas of improvement and further research are 
suggested. 
  
Chapter I: Introduction 
6 
 
Chapter II: Literature Review 
7 
2. Literature Review. 
 
In t his chapter the generalities of pasture- based dairying are described and the farming systems 
approach is conceptualised; these ideas build up the research framework  for introducing a 
summary of commonly used metrics for appraisal in pasture- based systems, including physical, 
financial, and efficiency indicators. This is followed by an overview of the NZ and C hilean 
macro- environments and dairy industries which sets the scene and enables a richer discussion 
about specific key features  that, in one  way or another,  have shaped dissimilar dairy farming 
systems. Finally, this literature review addresses the competitiveness of pastoral syst ems and the 
applications of benchmarking on farm management and farm ing systems research. 
2.1 Pasture-Based Dairy Systems 
Pasture- based dairy systems have always been dominant in Oceania and most parts of South 
America. Recently,  these systems have gained interest in other parts of the world where dairies 
have traditionally been based on confinement systems. Pastoral dairies typically feature outdoor 
herds all year round, have a match between milk production and pasture  growth, and g enerally 
have lower feed, culling and total production costs than other systems (Harris & Kolver, 2001) . 
Consistently, in Oceania and South American countries , this type of dairy farming has below 
world’s average costs and a share of feed around 60%  of total costs, while in the world this 
share ranges between 45- 74%  (IFCN, 2011).  
The criteria for systems to be described as ‘pasture - based’  can vary, partly due to pasture 
production differing by regions due to differences in plant growth which depends upon climate 
and soil. According to Taylor and Foltz (2006) , pasture- based dairy systems are those which use 
pasture as the primary feed source during the grazin g period. Other authors are more specific 
and require  animals to directly obtain a minimum of 40% of their forage needs during the 
summer months from pasture (Hanson, Cunningham, Ford, Muller, & Parsons, 1998). 
Alternatively, a ‘grazing operation’  may be one where at least 25% of the annual forage 
requirement is obtained via pasture and the animals graze for at least four  months of the year 
(Dartt ,  Lloyd, Radke, Black, & Kaneene, 1999). Pasture- based farming systems are physically 
feasible only if there is a suitable climate throughout the year, and they are optimally established  
in the temperate regions (Holmes, 2003).  Temperate pastoral systems, like those developed in 
Oceania and South America,  typically show a higher commitment to grazing, wh ich usually 
makes up more than 7 0% of the diet  ( IFCN, 2011).  
In addition, pasture is also a flexible d escription that encompasses low to highly productive 
swards and poor to high quality botanical mixtures or monocultures (Waghorn, Burke , & 
Chapter II: Literature Review 
8 
Kolver, 2007).  Typical temperate species are usually C3 2 , with best performances at an 
optimum temperature of 20- 25 °C, outside which photosynthesis and, hence, production  
decreases considerably below 10 and above 30°C . Conner, Hamilton, Sheehy, Stuth, and 
Kreuter  (1998) defined the  temperate zone as regions where the annual mean sits around 15°C  
and where winter temperatures that may fall below 5°C . Characteristically, the average rainfall 
across the temperate zone fluctuates from 500 to 1,000 mm (although large deviation can be 
expected),  evapotranspiration may vary greatly between seasons, and water deficit in the 
summer may be more or less important as it interacts with soil properties. Four well - defined 
seasons with potentially large seasonal amplitude, soil  water retention ability and rainfall 
patterns through the seasons are the other vital features defining a temperate environment 
( Carter, Murphy, & Cheal, 2003) .  
Overall , t emperate grasslands include a variety of environments resulting from factors such as 
season, latitude, altitude, aspect , and distance from the sea. In spite of this highly variable 
environment, t he normal conditions are very favourable for livestock production. Consequently, 
temperate grasslands around the globe have developed relatively low input systems shaped by 
several different management strategies that cope with this natural variability. These strategies, 
in turn, have logically allowed for a range of farming systems to take  place differing, for 
example , in how much pasture they grow, supplement they use , and the amount of desirable and 
undesirable outputs they produce. 
2.1.1 Milk output from pasture. 
This section reviews the definition of milk and milk components , as well as some aspects 
related to milk quality and measurement. Milk is the normal mammary secretion of milking 
mammals, including all commercially available types from cows, goats, sheep, and mares. R aw 
milk  is that which has not been in any way processed, reconstituted, or recombined (Draaiyer, 
Dugdill, Bennett , & Mo unsey, 2009). The composition and characteristics of milk largely vary 
as a function of species, breed , and other factors. Table 1 focuses only on raw bovine milk and 
presents average figures for parameters of interest. 
 
 
 
 
                                                          
2 According to the online Dictionary of Botany, C3 is a type of plant that produces phosphorglyceric acid 
as the first step in photosynthesis, which contains three carbon atoms. Most C3 plants exhibit 
photorespiration when temperature is greater than 25oC and, therefore, are relatively inefficient 
photosynthetically compared to C4 plants, particularly at higher temperatures.  
Chapter II: Literature Review 
9 
Table 1  
Average Figures for Normal Raw Milk at 20oC 
 
Adapted from: Draaiyer et al., 2009.  
Milk  is both a suspension and a solution, fundamentally composed of water and solids that are 
mainly fat, protein , and lactose. As a result, milk  quantity can be measured in volume or weight 
and, on average, one litre of milk produced by a Friesian type cow weighs 1.03 kg (Porter, 
1991). Most payment systems in the world are based on volume as fluid milk  is required for 
home consumption. However, as even these systems inclu de milk solids  ( MS )  in some way, it  is 
also appropriate to measure the mass of milk solids,  or the weight of milk ( Draaiyer et al., 
2009) . MS content can be measured by using estimation readings  from the lactometer, or by 
drying the milk and weighing the solids , and  by using rapid testing methods (Siddique  & 
Gulfraz, 2009). While fat  has traditionally been the first compositional quality parameter to be 
included in a milk payment system, non- fat solids (SNF 3 ) ha s recently become an important 
parameter.  
Mi lk  composition and hygiene are major  quality factor s, and, as such, milk  payment systems 
often offer bonuses or penalties based on quality grading systems.  The somatic cell count 
(SCC) measures the number of somatic cells (SC) present in a sample of milk,  typically 
comprising three quarters white blood cells and one quarter  epithelial cells from the secretory 
                                                          
3 SNF are the portion of MS minus the fat component; SNF includes proteins, lactose and minerals. 
Parameter Value range (%)
Water 85.5 - 89.5
Fat 3.2 - 5.5
Solids Non-Fat 8.2 - 10.0
Total Solids or Milksolids 10.5 - 14.5
Protein 2.6  - 3.6
Lactose 4.6 - 5.0
Acidity 6.6 - 6.7
pH 0.14 - 0.18
Specific gravity 1.032
Somatic Cell Count (000) 100 - 300
Chapter II: Literature Review 
10 
tissue of the udder (Draaiyer et al., 2009) . A high concentration of SC  per millilitre of milk 
indicates an abnormal condition in the udder and can make the milk susceptible to rejection  ( Le 
Marechal, Thiery, Vautor , &  Le Loir, 2011).  
Milk output from pasture shows distinct variability both in volume and quality, affecting 
production costs and returns. A lower degree of control over intake and diet quality  can be 
expected  in pasture- based feeding systems, a long with poorer individual cow performance 
(Gazzarin, Frey, Petermann , & Hoeltschi, 2011). Milk produced from pasture  also shows higher 
variability in fat and protein contents over time; a lower proportion of saturated fatty acids 
(Wyss, Mauer, Frey, Reinhard, Bernet, & Hofstetter , 2011) and lower total somatic cell counts 
than milk from  indoor systems (Pedernera, Garcia, Horagadoga, Barchia, & Fulkerson, 2008; 
Wyss et al., 2011). All of these vary for the different farming systems but they particularly vary 
over time responding to seasonality, a common attribute of all pasture - based dairy systems. 
2.1.1.1 Seasonality.  
Su pply fluctuation is a common feature among food commodities, although it is not always part 
of a seasonal pattern. Milk production from pasture is seasonal in nature. Seasonality can be the 
result of natural, economic, social , or institutional causes, and is defined as the systematic, but 
not necessarily regular, movement or set of movements within a year or less that happen in a 
temporal series (Marin, Cavalheiro, & Anschau, 2011). In Chile, seasonality can be expressed 
as the ratio, litres of milk produced in Spring - Summer  relative to litres produced in Autumn-
Winter (Lerdon, Baez, & Azocar, 2008). S eason affects not only milk volume , but also milk 
quality;  composition, fat, protein ,  and lactose content;  and the herd’s reproductive performance . 
From an industry standpoint, seasonality i nfluences the milk processing ability , the final product 
class, and even the monthly milk prices ( Holmes et al., 2003).  Consequently, there is an overall 
negative effect on the standardisation ability of the industry that begins with the heterogeneity 
of milk supply. In addition, s easonality is also a challenge with respect to scheduling and 
utilisation of plant infrastructure because the handling of the milk during the peak season  
requires  an adequate processing capacity  while plants are often idle during autumn and winter 
months ( Holmes, 2003) . This excess capacity adds substantial processing costs to the system, 
which are paid by all, dairy producers  and processors.  
Boehlje and Schiek (1998)  suggest that complex  matters like th ese need to be studied using a 
complete systems approach, placing emphasis on the entire value chain . S everal strategies have 
been designed to cope with the issue of seasonality, directed to improve the non- seasonal yield 
of milk . In NZ, these strategies are generally based on a proportion of the herd calving in the 
autumn and bonus payment schemes for winter milk production  (Holmes et al ., 2003 ).  In Chile, 
bonus payments for winter milk can be accompanied by penalties for spring milk (Lerdon et al., 
Chapter II: Literature Review 
11 
2008). However, pursuing a more level production pattern is costly to farmers,  and although 
fresh winter milk usually gets rewarded,  marketing arrangements sometimes provide little 
incentive for such activities. In conclusion, seasonality of milk production is the most common 
pattern for pasture- based dairying and one of the biggest challenges for the milk industry as a 
whole.  
2.1.1.2 Feeding strategies.  
Maximising the grass utilis ation and animal performance from grazed pasture has been cited as 
one of three key areas of innovation important for the sustainability of dairy farming (Dillon et 
al., 2008). The most typical and appealing characteristic about pastoral systems is that milk can 
be produced from inexpensi ve pasture at lower costs; as demonstrated by the data for nine 
countries, a  higher proportion of pasture in the diet is negatively correlated with cost of 
production (Dillon et al., 2005). This finding has been supported by  broader sample studies 
( IFCN , 2010;  2011; 2012 ).  In addition, supplements integrated with the pasture supply add 
flexibility to the systems and may produce extra milk and extra profit . As a consequence, t he 
optimum balance between pasture and supplements is often system- specific and has proven to 
be difficult to find. 
There are different feeding strategies aimed at several objectives:  exploit ing the natural 
production peak af ter calving;  increasing the milk fat or protein content;  extend ing the lactation;  
or improving the body condition score over the dry period (Dairy  NZ, 2 011) . The diet 
influences not only volume of production and every aspect of milk quality and  composition, but 
also reproductive performance which, in turn, affects the whole system’s physical and economi c 
efficiencies. Feeding strategies concerned with both productive  and reproductive performance 
have to deal with the negative association between milk yield and fertility  ( de Vries & Risco, 
2005; Washburn, Silvia, Brown, McDaniel , & McAllister, 2002). This  is, in part, explained by 
the antagonist correlation between milk yield and fertility , and in part, explicated by  the energy 
partition and balance of the lactating cow (Hansen, 2000) .  
The theory and practice of grazing manageme nt and feeding supplementary feed sometimes 
conflict and add complexity to these matters. F eeding strategies for pasture- based dairy systems 
have been developed looking at both grazing manage ment and supplementary feed research; 
however, grazing management research h as proven to be difficult, resulting in slower progress 
in strategy development. R esearch in this area has partially removed subjectivity from grazing 
management, propagating the use of certain rules, such as target pasture mass at key times and 
optimum seasonal grazing management  (MacDonald & Penno, 1998) . Although these rules 
meet the dual objective of feeding the herd w hile maintaining healthy swards and quality , the 
persistent dualism has determined that grazing management stays sub - exploited and 
Chapter II: Literature Review 
12 
underdeveloped. M eanwhile, the use of a wide range of supplements continues to gain 
popularity across the most diverse dairy regions in the world. Milk production responses of 
grazing cows offered supplements reported in the literature vary between 30 and 150 g of milk 
solids per kg of dry matter intake, however, positive quadratic responses to increasing amounts 
of supplement have been observed for yield of milk (Auldist, Marett, Greenwood, Hannah, 
Jacobs, Wales, 2013).  
2.1.1.3 Stocking rate.  
Stocking rate, also known as stocking density, is  the number of animals supported per hectare or 
other unit of area. Although the weaknesses of the simplest ratio, ‘cows/ha’ , are widely 
recognised, it has persisted because of its simplicity rather than its accuracy. The stocking rate 
(SR) a farm may  support is limited by the availability of pasture and the intensity of the use of 
supplements (Penno, McGrath, Macdonald, Coulter, & Lancaster, 1999) . Potential for pasture 
production depends mostly on the quality of the land. In pastoral systems, t he reliance on 
pasture production means that SR  affects not only physical and economic performance, but also 
a farm’s risk profile  (Penno, 1999 ) . Usually , most types of risk associated with  farming are 
positively correlated with increasing SR . Recent work by Anderson and Ridler (2010) analysed 
the optimisation of resource allocation in dairy production systems in NZ  and found that there is 
less financial risk  when a farm is slightly under- stocked than when it is highly stocked.  
S tocking rate is , therefore, an effective management tool since it can be manipulated to provide 
a wide variety of management options. SR adjustments throughout the seasons are common 
place for most farmers, and often the potential of the land is challenged by the available 
supplement cropped or purchased. In general terms, the milk production response for an 
incremental SR  change is negative on a per cow basis, while a strong positive relationship exists 
to a point between SR  and milk production per ha (McCarthy, Delaby, Pierce, Journot, & 
Horan, 2011; Penno, 1999). Also, a farm may be able to carry  more cows (smaller)  per hectare, 
in which case a financial disadvantage per head may be partly or completely offset by higher 
stock numbers and performance per hectare.  
Physical resource constraints, such as soil type, have real and significant negative impacts on 
the probability of increasing SR  because of their effects on pasture production ( Penno, 2000) . 
MacDonald, Penno, Lancaster , and Roche (2008 ) , after conducting a three- year- study on a 
seasonal calving system, concluded that there is a tendency for most production variables to 
behave quadratically with increasing SR . They showed the need for a more robust measure than 
cows per hectare and introduced the concept of a comparative SR.  Using this concept, the 
carrying capacity of the farm is defined by the live weight of the cows and the potential of the 
land to produce pasture. According to Macdonald, Penno, Nicholas, Lile, Coulter, and Lancaster 
Chapter II: Literature Review 
13 
( 2001) , for NZ farming systems, the economically optimum comparative SR appear s to lie 
between 80 and 90 kg of live weight  (LW)  per tonne of dry matter (DM) produced.  
2.1.1.4 Calving systems.  
A calving system is defined by time and pattern of calving, two key element s in the construction 
of any pastoral dairy farming system. The dairy cow can conceive, calve and lactate 
successfully at any time of year as long as enough energy can be provided when it is required  
(Garcia & Holmes, 1999) . Therefore, the calving system  can be used as a management tool. 
This offers complete flexibility in the selection of a calving system. How ever, such flexibility is 
partially offset by the seasonality of pasture production that implies greater restrictions in 
respect to the times at which economical feed can be provided in the required quantities. Hence, 
there is a considerable productive challenge in meeting feed supply with the increased demands 
of pregnancy and lactation considering the herd need to get in calf again soon after calving. The 
combination of management strategies, production, and fertility governs the pressure of animal 
demand on the grazing area, shape s the lactation curve, and regulates the animal performance, 
all of which affects farm’s  profit (Steinwidder et al., 2011).  
A compact calving pattern at the appropriate time is the major reproductive target for the 
seasonal dairy farm ( Holmes et al., 2003 ) . Cost- focused dairy pastoral systems in NZ  tend to 
implement a location- adjusted strategy which  requires a labour efficient, concentrated spring 
calving that matches the animal demand with availability of forage resources (Adams, Clark, 
Klopfenstein , & Volesky, 1996). This strategy has been quoted as one of three key areas for the 
sustainable development of dairy farming (Dillon et al., 2008). According to Garcia  and Holmes 
(1999) , early spring  calving ensures sufficient feed in early lactation and more days in milk  than 
a later calving pattern.  
Nonetheless , a variety of results suggest  that for maximum profitability , no calving time is 
optimal in all environments. Alternative systems are autumn calving, a combination of both 
spring and autumn, and all year round calving . All three systems can be found in Chile, but the 
combined spring- autumn tends to predominate in pasture- based dairies (Lerdon, Baez & 
Azocar, 2008). In general, one or other of t hese may be particularly suitable if pasture growth is 
slower in summer than in winter; if affordable supplements can be fed at any time of the year; 
and if there is a price incentive for winter- milk (Hodgson & Chesnutt, 1999). Production results 
show that autumn calving systems can have very good annual milk yields, mainly sustained on 
longer milking periods , and higher yields in late lactation. A review of comparisons between 
autumn and spring calving systems showed that autumn- calved cows require more su pplements 
during winter and usually have lower daily milk yields at peak lactation than spring - calved 
cows (Garcia & Holmes, 1999). According to Garcia, Cayzer, Holmes, and  MacDonald, ( 1998) ,  
Chapter II: Literature Review 
14 
economic results are diverse and are mainly a function of the premium received for milk 
produced in winter and the prices paid for supplements. 
2.1.1.5 Genotype.  
The use of profitable high quality genetics to increase the herd’s performance  from grazed 
pasture has been mentioned as one of the three crucial areas of improvement important for the 
sustainability of milk production (Dillon et al., 2008). There are so many interactions that the 
choice of a suitable cow genotype for a particular farming system remains a large and 
continually evolving area of research. The value of certain breeds depends on the system’s 
configuration and resources. However, the breed itself does not completely explain genetic 
merit as it does when strain is also taken into account.   
The NZ  breeding programme managed by the Livestock Improvement Corporation (LIC)  has 
put emphasis on milk solids ( MS )  production and,  therefore,  has produced a strain of Holstein-
Friesian  (HF)  cow which yields higher MS  and increased protein:fat ratio  than overseas’ HF . 
McCarthy et al. (20 07) compared the economic efficiency of three divergent strains of HF cows ;  
high- production North American (HP), high- durability North Ameri can (HD) , and New Zealand 
(NZ) across a variety of Irish p asture- based production systems. The economic performance of 
each strain and feed system was affected by different production scenarios, milk quota s,  
predicted future prices, costs, and potential land availability. The results showed that in a fixed 
milk quota scenario, the NZ strain ret urned the highest profitability while the HD strain proved 
to be the next most profitable. The  HP animals were least profitable in all systems because the 
productivity gains achieved were outweighed by associated increases in reproductive costs. 
Typically,  in a different environment to what cows were bred for,  genetic selection solely for 
increased milk production may result in reduced farm profitability.  
South American dairy production systems have been highly  influenced by genotypes from 
North America  whose strains are typically capable of very high milk yields  with low MS  
produced by mobilising large amounts of body reserves. On pasture alone , this metabolic effect 
reduces lactation length and leads to reproductive issues (Harris & Kolver, 2001; Harri s & 
Winkelman, 2000).  Thomson, Turner, Lopez - Villalobos and Glassey (2005) evaluated the effect 
of genotype on milk composition, milk value, and dairy farm profitability by means of a trial 
including a low and a high merit strain from NZ , and another high merit strain from America. 
Their results clearly showed that, within the same breed and under the NZ payment system , the 
American strain resulted in high milk  production but lower farm profitability than the NZ  strain. 
Overall, t he potential for improving the performance of different dairy systems through breed 
substitution and by identifying the more appropriate strains has been frequently  examined  
(Piccand ,  Cutullic, Schori, Keckeis, Gazzarin, Wanner, & Thomet, 2011) . Crossbreeding, 
Chapter II: Literature Review 
15 
breeding and multi- factorial selection are increasingly valid options to include in any farming 
system’s design. The  results presented above reinforce the economic value of genetic 
improvement based on a selection index pertinent to the production environment.  
2.1.1.6 Economics.  
Empirical data  has demonstrated that pasture- based dairy systems consistently have lower costs 
than other farming alternatives (IFCN, 2011) . This organisation also indicated a mean world 
cost of milk production of  US$40/ 100 kg milk in 2010 4  in a study which included all major 
producers and farming systems in the world. L ong term experiment s by White, Benson, 
Washburn, and Green (2002)  on pasture- based systems showed what may be considered a 
regular pattern: lower m ilk production per cow, inferior feed and culling costs and lower labour, 
plant and equipment  costs. This has been supported by other authors who claim that pastoral 
systems exhibit consistent ly higher net returns per cow than confinement systems ( Rust ,  
Sheaffer, Eidman, M oon,  &  Mathison, 1995). However, there is often controversy when 
undertaking these  sorts of comparisons because the attributes to be measured and the 
appropriate metrics to be used are often system- specific. For instance , return on feed cost is less 
relevant if analysing a pasture- based system due to the inherent difficulty in estimating dry 
matter cost from long lived pasture ( Barnard & Boehlje , 1998) . In contrast, some financial 
indicators, including solvency and liquidity  measures,  have been regarded as mandatory 
( Barnard & Boehlje , 1998 ;  Hanson et al. , 1998) . This is consistent with the inherent cash flow 
difficulty in all land- based industries (Oltmans, 2007).  
Profitability, which is possibly the most important determinant of long term business 
performance, has several indicators which cannot be considered in isolation and which need to 
be contextualised adequately.  There are several financial indicators, and depending on the 
indicator and the context where it is used, a n empirical comparison might exhibit conflicting 
results. For instance , in the US  where a vast range of dairy farming systems coexist , a pasture-
based system may exhibit a higher gross margin than a confinement system ( Parker, Muller &  
Buckmaster , 1992;  Tozer, Bargo, & Muller, 2004 ) . Yet , the confinement system can be found to 
have the highest profit measured as net income per cow when the milk:feed price  ratio is 
favourable (Tozer, Bargo, & Muller, 2003) . In NZ, t he same pattern was found by Jensen, 
Clark, and Macdonald (2005)  who analysed different levels of intensification within pasture-
based systems. They found that increased input systems sustained greater SR  and were able to 
duplicate the output, while the indicator return on assets r evealed that profitability was governed 
by the relationship between milk payout: feed price. When the payout was low at NZ$ 3.50/ kg  
MS , the low input system was competitive irrespective of land price ($18,000 or $37,000/ha) or 
                                                          
4 Considering all cash and non-cash costs and including indoor and outdoor systems from 49 countries. 
Chapter II: Literature Review 
16 
supplement cost (18c or 24c/kg DM). Conver sely, at a higher payout of NZ$4.5/kg MS , the 
higher input systems showed higher profitability.  
2.1.1.7 Diversification or specialisation strategy? 
M ilk production usually represents more than 85% of the total dairy farm’s income, which 
originates from a multi- product system resulting from vertical diversification, which may 
include feed or stock sales ( IFCN, 2011) . Assuming that the rationale for portfolio 
diversification is to decrease non- systematic risk, this m ultiplicity of outputs can be seen as an 
advantage due to effective diversification ( Rumelt, 1982). However, according to Ethier (1982) ,  
it can also be perceived as a disadvantage due to a lack of specialis ation. In some cases, certain 
outputs are actually by- products, such as bobby calves 5  and cull cows, and not deliberate 
diversification. According to Rumelt (1982) , t rue diversification occurs when a firm produces or 
sells a product which has zero or close to zero cross price - elasticity with each of the firm’s other 
products. Different diversifi cation strategies that exist in farming can be classified according to 
variations in their related or unrelated activities. The latter, also called primary diversification , is 
perhaps the most common practice for farmers all over the world, and usually involves those 
activities that share common skills or resources.  
However, why would pasture- based dairy farmers choose specialisation instead of 
diversification? Many factors influence this decision, and the perceived risk and attitudes 
towards risk play a m ajor role (Chatter jee & Wernerfelt, 1991). Risk averse  farmers are more 
likely to increase portfolio diversification. Initial farm size and current level s of diversification 
are also likely to influence f uture diversification decisions:  large farm size may  be associated 
with greater diversification of unrelated farming activities. Chatterjee and Wernerfelt (1991) 
also demonstrated that the resource profile can partially explain the type of diversification 
strategy the firms engage. Therefore, some resources and core skills can be taken as competitive 
advantages (Porter, 1985) and also as key factors explaining diversification. A resource- based 
approach would consider three classes of resources: physical, financial , and intangibles. 
Typically, financial resources are more flexible than the other two. In contrast, physical 
resources usually have fixed capacity , and excess physical capacity will lead to related 
diversification. Similarly, the presence of intangible assets also le ads to related diversification,  
and conversely, the availability of internal funds, equity capital or unused debt capacity , will 
favour unrelated diversification (Chatterjee & Wernerfelt, 1991).  
2.1.2 The challenges.  
Pastoral dairy farmers, regardless of their setting in the world, face several diverse challenges. 
These may be classified into at least three areas: productive, environmental , and consumers or 
                                                          
5 In NZ, these are calves up to five days old. 
Chapter II: Literature Review 
17 
general. Those challenges from the productive area are often addressed and researched. 
Challenges concerning to the environment and consumers areas are possibly even more complex 
than those related to production, and are probably among the most debated in modern 
agriculture. Both are closely connected because environm ental issues are likely to influence 
preferences, public opinion , and consumer trends. The present section briefly addresses some of 
these challenges.   
Among the production challenges, ensuring an adequate feed supply to high - yielding dairy cows 
in a highly variable environment would be the greatest challenge (Ferris, 2007). Feasible  
grazing systems require the use of high quality, high yielding pastures aimed at increasing  either 
farm productivity, profitability , or both (Waghorn et al., 2007). In the future, land availability 
may limit the expansion of pasture - based dairy systems so that the maximisation of productivity 
per hectare could become critical (Shalloo, Dillon, O’Loughlin, Rath, & Wallace, 2004, cited in  
Patton, Shalloo, Pierce & Horan, 2012). A more general , and for some perhaps greater 
challenge, is how to get the most out of the existing competitive adv antages by using a number 
of options spatially across the farm and temporally across the seasons (Lapple, et al., 2012). 
Plenty of alternatives have been designed to buffer the systems and cope with variability,  
mainly by spreading and increasing feed supply (V alentine & Kemp, 2007). Despite these 
alternatives existing , and variability being intrinsic to most pasture- based farming systems, not 
all farmers adjust their tactics and strategies to match heterogeneity and uncertainty.  
Sustainable  profitability is a challenge that combines the productive and environmental aspects. 
As Beale (2006) claimed in the Financial Times 6 , contemporary businesses want not just large 
dividends, but also want them to be ethically driven, environmentally concerned and  
ecologically sustainable. Pasture- based dairy farming is required to be profitable on a 
sustainable basis; this means establishing high, lasting , and economic productivity and 
improving the quality of life without damaging the environment (Derpsch & Moriya, 1998; 
McGuckian, 1996). McGuckian (1996) also emphasised the enhancement of the natural 
resource base over time as a condition for sustainability. Therefore,  farm sustainability is only 
achieved when all objectives, obligations  and requirements associated with the farm system are 
fulfilled in a reconciliatory way ( van Eijk, 1998). O n- going financial and environmental 
problems related to agriculture prove that the reconciliation of all factors linked to farm 
sustainability is if not the greatest challenge,  at least a difficult proposition.  
Many common practices in farming have the potential to  negatively impact on the atmosphere, 
water or land qualities. Among environmental challenges, agriculturally produced greenhouse 
                                                          
6 The Financial Times is one of the world's leading business news and information organisations, 
publishing in several countries and languages worldwide. 
Chapter II: Literature Review 
18 
gases stand out as a current concern (Surinder, Bolan, Bhandral, Hedley & Luo, 2004). Methane 
( CH 4 ) is a greenhouse gas over 2 0 times more effective in trapping heat in the atmosphere than 
carbon dioxide (CO 2 ), and is emitted from a variety of natural and human- influenced sources. 
Howev er, 18%  of total greenhouse emissions are produced by fermentative processes in the 
digestive system of ruminants (FAO, 2006). The quantification and management of these 
undesirable outputs is under study, as well as other interrelated bodies of research dedicated to 
understanding and generating mechanisms to cope with global warming and its consequences 
( Fischer , Shah , & van Velthuizen, 2002 ). In the meantime, environmental regulations threaten 
the image and competitiveness of pastoral farming systems.  
E nvironmental impacts affecting land and water are more ‘conventional’  and longstanding, but 
not less challenging issues. Relatively recently , regulatory agencies, environmental groups, and 
concerned citizens have focused their attention on livestock operations potentially harmful to 
the environment. This is significant because, in any industry,  the public perception influences 
present and future consumption and trends, sways t he people willing to take part in the business , 
and contributes to the growing development of environmental standards and regulations (Hicks 
& Dietmar, 2007). For example, agriculture  through historic over- application of phosphorous 
and nitrogen from commercial fertilizers and organic wastes, has contributed to many off- site 
impacts through surface runoff and subsurface transport. Consequently, w astewater and 
fertiliser- related pollution problems have regularly been reported as resulting from farm 
operations (Nahuelhual, Engler, Carrillo, Moreira , & Castro 2009). Dairy farms  are particularly 
critical because of the high organic matter and nutrient levels contained in dairy effluent, which 
can have detrimental effects on drinking water, water supplies, fish resources, and recreation  if 
not adequately  managed (Rahelizatovo & Gillespie, 2004).  
Finally, there are the challenges around the adoption or adaptation of technology to a changing 
world, markets and regulations. Griliches (1987) defined technology as the state of knowledge 
concerning ways of converting inputs or resources into outputs or product. L and- based 
industries, soils, climate , and landscape, in particular, are likely to affect efficiency and  impose 
restrictions on the selection and type of technology used. Therefore, technology is not only 
limited but also shaped by the environment in which it is applied (Alston, 2002). According to 
Sumner and Wold (2002) , land differences influence the amount and type of resources, the 
opportunity cost of land, and the level of scale economies. Pasture- based systems developed on 
superior land classes have the potential to produce more pasture and milk, outperforming other 
poorer- resourced systems. More importantly, well - managed and resource- rich pastoral farms 
could generate fewer adverse environmental impacts and, thus, would be  socially more 
acceptable (Benson, 2008). C onversely, less well- managed farms or farms making use of more 
marginal, fragile land  types could be easily damaged under pressure. Increased pressure on 
Chapter II: Literature Review 
19 
existing resources may only be tolerable if it is both profitable and sustainable in the long term. 
Consequently, if significant increases in milk production are to be made from higher 
productivity grasslands, this may require efficiency increases, technology adjustments, or even 
paradigm changes.  
2.2 Farming Systems Approach 
F arming systems are multi- factorial, complex entities that may not be able to be properly 
assessed unless their drivers are fully understood (Kristensen & Jakobsen, 2011).  The systems 
based on pasture may exhibit additional  intricacy, resulting from  biotic and abiotic components, 
economic and genetic factors, and environmental and management functions. M any 
relationships explaining variation in dairy performance can be assessed thr ough financial data 
analysis,  but others may only be explained by contextual factors and managerial be haviours. 
According to Garcia and Holmes (1999) , due to the large number of relationships and 
interactions present in pasture- based systems,  meaningful research should only be approached 
by means of systems research.  
The first premise of systems research is holism. Wilson a nd Morren (1990) proposed that the  
first step towards system thinking and practice is to develop the ability to explore experience s 
holistically, assuming that no matter how the individual parts are studied,  the emergent 
properties within the whole cannot be understood, nor can the original complex situation be 
improved, unless it is studied in its entirety. A  farming systems approach may contribute 
significantly not only to the data analysis, but also to the  understanding of actual farming 
practices (Tripp , Anandajayasekeram, Byerleee, & Harrington,  1990). The use of hard, 
empirical data enables unbiased studies, while the systems approach may be  needed to reach 
significant conclusions. These conclusions may in turn contribute to the continuous 
development and acquisition of relevant technology , since the processes related to technology 
mastering or advancement need to be grounded in a full knowledge of the specific farming 
system into which it is to be implemented ( van Eijk, 1998).  According to Ronan and Clearly 
(2000) , a systems approach and appropriate participatory activities that include active farmers 
are central for any rural industry to be able to react proactively and appropriately to most 
challenges. This may be the case within the dairy sector, where these type s of initiatives have 
been increasingly promoted at both national and international levels.  
2.2.1 Values and goals.  
Performance is a relative term and how to gauge it depends on the businesses’  values and goals, 
along with other important factors such as its lifecycle stage and age. The complex nature of the 
interactions within a farming system and the significant influence that the farmer, the family and  
Chapter II: Literature Review 
20 
the values and goals involved have on the operation of the farming system must be recognised 
(Tripp, 1991). Gasson (1973) defined values as  
… a more permanent property of the individual, less liable to change with time and 
circumstances. A value is a conception of the desirable referring to any aspect of a 
situation, object or event that has a preferential implication of being good or bad, right 
or wrong. Values are felt to be justified by reason, moral or aesthetic judgements. …are 
ends in themselves, pursued for their own sake. They  serve as standards influencing the 
selection from among available modes, means and ends o f action. ( p. 522 )  
So , why  values and goals are so important in practice? Olsson (1988) affirmed that farmers with 
well- established values follow a structured decision- making process, which provides them with 
the greatest opportunity to succeed. F our major categories related to dominant values that are 
likely to be associated with farming  were created by Gasson (1973) . If i nstrumental values 
predominate, then farming is viewed as a means of obtaining income and security. Similarly, if 
prevailing values are social in nature, farming is undertaken for the sake of interpersonal 
relationships. However, th e dominance of expressive values means that farming is used as a 
means of self- expression or personal fulfilment . Finally, if intrinsic values outweigh other 
values, then farming is valued as an activity in its own right. Later , Gasson and Errington (1993)  
stated that typical goals in farming were maximising profits, controlling a larger farm, reducing 
levels of debt, having a well- managed tidy farm and improving lifestyle. Nowadays, th ese goals 
might still be a sound representation of typical values for most farmers around the world. The 
vision and mission statements encouraged across many industries, including agriculture, are 
related to both values and goals,  and act as long term expressions  of the basic purposes and 
values of the business (Steiner, 1997) .   
2.2.1.1 Lifecycle stage and age of the business. 
Lifecycle describes the development of the business over  time. To assess a lifecycle’ stage, two 
aspects must be considered:  the business’ age and the business  operator’s age . According to 
Gale (1994) , a factor that contributes to considerable variability in the farm sector is the 
interaction between the farm and the farmer life cycles, which may be related but are not 
necessarily in synchrony. Boehlje (1973; 1992) identified three stages that can apply to both the 
farm and the farmer: establishment, growth and survival, and disinvestment. E ach stage can last 
for several years, may have particular characteristics including objectives and goals, a nd would 
require specific skills or management.  
In most industries and particularly in farming, the  age of the business is an important factor 
when understanding its goals and objectives, and for interpreting its results. Theoretical and 
empirical works have  shown that a business’ growth’  rate usually declines with age (Evans, 
Chapter II: Literature Review 
21 
1987; Jovanovic, 1982). Similarly, there is a well- established pattern which shows that failure 
rates also fall as businesses age (MacDonald, 2007) . Moreover, w ell- established businesses 
have greater survival prospects than younger businesses (MacDonald, Korb, & Hoppe, 2007) . 
The importance of a business’ age  holds, irrespective of the farm size , whereby large farms are 
more likely to succeed , but the effects of a business’  age are generally even larger than the 
effects of farm size and can be similar to the implicit of business’ operator’s age. Apparently, 
k now- how and business’ network s seem to provide an important advantage to well- established 
firms, regardless of their size. 
2.2.1.2 Business operator’s age. 
The operator’s  age has a profound impact on the firm’s performance . Despite the existence of 
mixed results, a general consensus is that farms operated by middle- aged managers are more 
likely to survive  and grow ( Gale, 1994) . Gale also partially explained this , suggesting  that while 
older farmers generally lack the motivation for growth, new and young farmers encounter 
financial constraints that limit the size of the farm they operate. The farm’s lifecycle hypothesis 
suggests that farms entering the industry should have younger operators than farms exiting  the 
industry, which has been supported by Gale’s results (1994). Experienc e continues to matter at a 
diminishing rate as farm managers’  age. In addition, according to Perkin (1992) , older farmers 
are more risk averse than younger ones, possibly due to previous negative experience and / or 
because they are close to retirement. Also, young farmers may need to take more risk to get 
higher returns and get ahead Perhaps 45 - year old farmers and older also attach more importance 
to social goals and values (Gasson, 1973).  
2.2.2 Structure. 
In the context of this research, the  business’ s structure refers to the configuration of a firm’ s 
internal resources, which are  associated with production, ownership, and management factors. 
The business’ s structure is regarded as being pertinent because it may uncover those values and 
goals that are likely to be prevalent in a particular farm. According to Spulber (2004) , whatever  
the structure, it  has to enhance implementation of the firm’s strategy; consequently , changes in 
the goals or strategic shifts have to be reflected in the business structure, sometimes favouring  
some business units to the detriment of others. In New Zealand farming systems, ‘structure’  
describes the farm’ s ownership. According to Dairy  NZ (2012) business in NZ are as follows:  
Owner operator: The operator owns both cows and land or leases them from an external party ; 
typically these are family businesses. This business structure may have a manager. 
Owner with 50-50 sharemilker: The operator ownw the land but not the herd and contracts a 
sharemilker or equivalent . A variation is when the operator owns the herd but not the land and 
receives a variable milk income that genera lly fluctuates between 40 – 60%.  
Chapter II: Literature Review 
22 
Owner with contract milker: The owner of the land and the cows and contracts either 
contracts a milker or  a variable order sharemilker  to milk the cows who receive a percentage of 
income. 
Operator variable order sharemilker: The operator is a variable order sharemilker  and gets 
approximately 22% of milk income . The owner owns the land and cows. 
Other or Diverse: The operator can be in any other ownership type, typically with a 
combination of more than one role or activity. It could also relate to companies and equity 
partnerships that are either corporate or family- centred with a number of equity partners  that 
own shares. Any of these may have a manager; otherwise one of the equity partners may operate 
the farm. 
So me of these categories are NZ specific as these business  structures are relatively unique to 
this country. Ot her structures,  such as ‘ owner- operator’  and ‘ diverse’,  are common to most 
countries. Whatever the case, it is generally accepted that ownership , and particularly land-
ownership, causes noise when analysing the economics of essentially land- based businesses. 
Therefore, a proper separation from production activities is needed prior to any analysis 
(Gardner & Hargreaves, 1987; Oltmans, 2007). Land te nancy will be discussed further because 
of the importance of land,  both as an input for production and as an asset for investment.  
2.2.2.1 Land ownership. 
For most  agricultural activities, land is the single largest asset accounting for more than 40% of 
total assets in Oceania and most countries in South America, including Chile. In part because of 
this, there is an inherent cash flow difficul ty in all land- based industries. Correspondingly, debt  
on real estate tends to dominate the liability side of the balance sheet in farming (Oltmans, 
2007). In theory, a  business should generate a return commensurate with the investment and the 
risk involved, a s well as being able to meet all outgoings. In other words, the market price must 
maintain a strong relationship with the estimated price by capitalizing expected farm income at 
current prices (Decimavillaa, San Juanb, & Sperlichc, 2008) . According to Schönhaut  ( 1999 in 
ODEPA, 2009)  land quality is just one factor affecting the dynamic price of the land; other 
factors are demand, location, potential land use s and versatility, size of the property,  
improvements incorporated to the property, public infrastructure in the area,  among others. As a 
consequence, the relationship between income and value has not been traditionally that clear 
with farm land. Some of the factors just mentioned can be summarized  in what has been called 
the lifestyle market effect. In NZ for example, the signific antly higher market value of land in 
the Waikato region is influenced by the demand for dairy and for other purposes (IFCN, 2012). 
Land not only has ‘productive’ value but also a ‘place value’ that  creates problems when 
Chapter II: Literature Review 
23 
analysing the pure farm business (Ol tmans, 2007). Sometimes, the place value may exceed the 
productive value, and different tools have been put in place to overcome this undesirable effect.  
Leasing is a process by which a firm can obtain the use of land or other fixed assets , for which it 
must pay a series of contractual, periodic, tax deductible payments called rent. The lessor is the 
owner of the asset, and the lessee is the one who benefits from the use of the assets under the 
lease contract. According to Gardner and Hargre aves (1987) , the reasons for lease land from the 
lessee’s standpoint are as follows:  
• It is a step to farm ownership;  
• It may improve capital utilis ation through scale increase;  
• It can provide flexibility, minimis e risk , and increase profits;  and 
• It can allow  for return on limited capital. 
According to Gardner and Hargreaves (1987) , farmers have traditionally preferred to own rather 
than lease farm land. However, the long term option in the buying or leasing decision depends 
upon the actual relationship between interest rates, rents and land values, and upon the 
expectations a farmer has about these variables in the future. The perceived uncertainties about 
each of these play a role, as does  the risk involved in the expected variation in the return on 
investment (Gardner & Hargreaves, 1987).  
Wealth generation , defined as the increase in equity is achieved in NZ farming  through a 
number of land purchases and sales during the life cycle of a farm business (Gardner & 
Shadbolt, 2005). Therefore, it is evident that the related concepts of wealth, asset s and equity 
have significant importance in NZ farming systems. In contrast, these concepts are expected to 
be less important in Chile where no reports on wealth creation related to farming have been 
found. In Chile , land often has a share of total assets greater than the 40% estimated by IFCN 
(2011). S urvey information provided by Chilean dairy farmers showed that land was the main 
component of total assets (70% ) , followed by capital in cattle (16% ) , machinery (7%) and 
construction (7%) (Silva, 1997). F arming in Chile is characterised by far more stable farm 
ownership patterns with fewer opportunities to purchase land, sinc e land is held by farming 
families for many generations.  
2.2.3 Family businesses. 
The family business model is the most commonly in use in New Zealand. A  farm business may 
be classified as a family business if ownership is combined with managerial control;  the 
household is identified with the work place ;  and/ or the family provides a significant part of the 
labour requirements . According to Gasson and Errington (1993) , this  definition has changed 
over time because the labour criterion has become less relevant,  as the principle switched from 
Chapter II: Literature Review 
24 
labour to capital. A family business is a partnership of people who are, in a social sense, already 
partners ( Robbins  & Wallace, 1992 ) . If this is so, t hen, as noted by G ersick, Davis, McCollom ,  
Hampton, and Lansbe rg (1997) , a three dimensional model may be a sound conceptual 
representation of the family business. This multi- dimensional model comprises three 
independent subsystems:  the family, the business itself,  and the ownership on its own as an 
investment. As the subsystems coexist and overlap, they also generate subsequent combinations,  
totalling seven zones, including a multiple combination of the three subsystems. These special 
configurations, exemplify the complexity of this distinctive type of business. Such  complexity 
needs to be acknowledged in order to understand or predict the drivers and the management 
process.   
F amily farm businesses exhibit clear advantages, such as flexibility  in responding to the 
environment or a greater confidence in the family working team ( Gasson & Errington, 1993) . 
The degree of elasticity in task allocation may be a positive factor comparable to that related to 
the cash flow; reliance on family labour may signify that wages do not have to be paid rigidly. 
This also features positively from a liquidity point of view , especially during unfavourable 
periods provoked, for example, by depressed markets . Moreover, as the firm’s goals and values, 
the business nature, and the operator and employees all build a business’ s level of performance, 
a higher level of commitment is expected from family members than for strangers  which, in 
turn, may at least partially compensate any other disadvantage  ( Gersick et al., 1997 ) .  
However, although common identity, values ,  and language provide strength to the family 
business, at the same time they can work against the executive behaviour of the employed 
family. Furthermore, the family business also exhibit s weaknesses, such as conflict of interests 
and lower rationality, than other types of businesses because the different demands are 
linked ,and family and business respon sibilities do not naturally mix (Robbins & Wallace, 1992) , 
but tend to overlap and often to conflict. In addition, t he family business is typically less 
rational than other firm’s structures that  often behave according to more objective, unemotional 
management. Another drawback  of family employment is that in some cases it can result in a 
smaller range of skills being available than if human resources were supplied from the market. 
If the range of skills were broader than needed , sometimes the opportunity cost for family 
labour could also be neglected (Gasson & Errington, 1993).  
2.2.4 Management in farm businesses. 
The importance of management in farm businesses is such that sometimes a change can result in 
a lift in profitability greater than that expected from the adoption of new technology. From a 
management perspective, if  the existing technology  is used to its full potential through 
optimising technical efficiency, the savings achieved are translated into a farm’s  higher chance 
Chapter II: Literature Review 
25 
of surviving and prospering (Bravo - Ureta & Rieger, 1991).  According to Bennett (1982) ,  
management research used to common in early farm economic research but it has gradually lost 
importance as research moved towards more quantitative theory and methodology. However, 
the human factor is a determinant of performance given certain technology or productive 
environment. Agriculture, being based on three major res ources, land, labour,  and capital, 
without the contribution of management farming would be random (Nuthall, 2006) .  
According to Spulberg (2004) , the management strategy is defined as a broad plan of action to 
achieve the business goals, and includes the following five main components:  
1. B usiness goals;  
2. E xternal environment ;   
3. The competitive advantages;   
4. The competitive strategy;  and  
5. Business structure   
In turn, m anagement as a process, rather than as a strategy ,  is defined in a slightly different way. 
It is typically represented by effective knowledge, planning , and control. The management 
process has been shown to be able to be used to develop more successful farming systems than 
other approaches (van Eijk, 1998) . According to Barry , Ellinger, Hopkin ,  and Baker (2000) ,  the 
strategic management process can be organized into the following systematic set of six 
interconnected steps:   
1. D eveloping the firm’s mission ;  
2. Form ulating its objectives ;   
3. Assessing the firm and its environment;   
4. B uilding and implementing a strategy;    
5. E valuating its performance;  and 
6. I mplementing any corrective action.  
Impediments to the management process sometimes arise from the management style itself, or 
from limitations in observation skills and the ability to order, analyse , and process complex 
information (Barry et al., 2000) . Decision - making is an input component in the complex 
management process. The farmer’s decisions are influenced by many interconnected factors: 
values, financial and non- financial goals, knowledge ,  and previous experience are well - known 
factors providing strong guidance in successful decision- making. Willock  et al. ( 1999) pointed 
out that attitudes towards stress, risk, technology , and nature, and also personality and other 
psychological characteristics, play an important role. S tress levels and control, innovativeness 
Chapter II: Literature Review 
26 
and intelligence have proven to be significant in determining a farmer’s success in decision 
making ( Willock  et al., 1999).   
Individual characteristics related to the manager’s  style and skills in the context of farm 
management have been the subject of different studies and types of taxonomy. In NZ, multiple 
management styles, age s and gender differences were identified by Fairweather and Keating 
(1990) through an empirical study of farmers. D issimilarities were found not only in values and 
objectives , but also in preferences, likes, and dislikes.  Three management styles were identified 
resulting from this variability. The Dedicated Producer is associated with young men who 
thrive on hard farm work, with achievement , and a high quality of production. Their means are 
full time work and care in planning and financial management. In contrast, the Flexible 
Strategist tends to be older and female. This style thrives on being conscious about nature, 
responding to the environment, reducing work load,  and pursuing off- farm ventures. Its means 
are higher level understanding, effective marketing , and asset  diversification. Thirdly, and 
lastly, Lifestyler’s are oldest and comprise both males and females who allocate even higher 
value on nature, and thrive in working with family and sustaining or increasing lifestyle by 
several means.  
Regardless of the world setting, variation in farm  profitability is a consequence of interacting 
outputs, inputs and prices , but especially a result of business skills applied to available resources  
( Kay & Edwards, 1999; Wilson, 2011). The external environmental plays a major role as a 
determinant of success or failure in pastoral farming, but  the variability of results within a 
particular environment has suggested that the role of farmers as general managers and their 
managerial abilities need to be observed ( de Lauwere, 2005) . According to Olson (2010) , the  
functions farmers accomplish are many, diverse ,  and related to major areas such as production, 
marketing, finance and human resource management. Farmers need to be effective , and unlike 
managers in other sectors, versatile. Covey (1989) highlighted seven habits of highly effective 
people in general, that could also be applied to managers and operators  in the farming sector:  
1. To be proactive;  
2. To be goal minded (principle ce ntred) ;  
3. To prioritise (put first things first) ;  
4. To be positive;  
5. To seek underst anding first and then to be understood;  
6. To synergise (seek interdependence and put value in team work) ; and 
7. Sharpen the saw (cultivate observation skills) . 
The farm management literature states that a successful farm manager requir es a broad set of 
skills as reported by Boehlje, Dobblins , and Miller (2001) : strategic, procur ement and selling, 
Chapter II: Literature Review 
27 
relationship, leadership , and risk management skills. Nuthall (2010) also identified that 
observation,  keeping of records ,  and the use of a scoring system of facts were core skills in farm 
management. According to Barry et al. (2000) ,  superior managers show a methodical approach 
to the strategic management process, strong analytical skills, decisiveness, and the capacity to 
admit the consequences and to review decisi ons as conditions warrant. From the skills  listed 
above, some  may be referred also to as contributors to a particular characteristic known as 
entrepreneurship.  
2.2.4.1 The manager’s entrepreneurial orientation.  
Entrepreneurial orientation refers to behaviours, processes, or practices that lead to a sustainable 
competitive advantage, especially when this involves proactiveness, innovativeness, and 
financial risk taking.  In economic theory, an entrepreneur is defined as a the person who makes 
judgement  decisions about the management of scarce resources (Gasson &  Errington, 1993). 
According to van Praag (1999) , entrepreneurs are held responsible for economic development 
by introducing and implementing innovative ideas, including product, process esing, market and 
organisational innovations. Taken  to the extreme, an entrepreneur may be that person who has a 
single objective of profit maximisation and always select s the appropriate means to attain this 
goal. As this is not prevalent in farming, the pure definition of entrepreneur is ignored while 
entrepreneurial orientation is discribed as another management style.  
Entrepreneu rial orientation influences performance and is an important factor enabling firms to 
create, modify , and apply their resources in more efficient ways. However, b oth the resource 
configuration and the environment may modify the impact of entrepreneurship. Recent research 
indicates that, despite many others characteristics being evaluated,  only proactiveness has 
shown a significant positive contribution to business performance (Kraus, Rigtering , & Hughes, 
2012). Although numerous forces drive the entrepreneurial orientation in agriculture, 
information and knowledge are important contributors . As in other industries characterised by 
negotiated or personal linkages, those individuals with unique and accurate information and 
knowledge have greater power and control. This would imply having greater capabilities to 
transfer risk to other  less capable individuals ( Boehlje &  Schiek , 1998) . According to de 
Lauwere (2005), performance in farming depends on many internal and external conditions, but 
the role of the agricultural entrepreneur is extremely critical  in doing this. A study of Dutch 
farmers showed that the positive entrepreneurial characteristics were achievement- orientation, 
market - orientation, perseverance, creativity, leadership, initiative, self - criticism, and inspiration, 
which were highly correlated, as were the negative entrepreneurial characteristics,  ‘ love of ease’  
and ‘ passivity’ . In addition, the farmers who scored higher in the positive characteristics 
exhibited better performance measured as annual income and expectations  achieved ( de 
Lauwere, 2005) . 
Chapter II: Literature Review 
28 
2.3 Metrics Used for Success Appraisal in Pasture-Based Systems 
Ensuring and enhancing farm performance requires effective management , and measurement of 
both financial and physical resources (Dairy NZ , 2011). The selection of the best measure of 
performance in farming can be controversial because the way success is gauged depends on the 
particular business and its goals. Parmenter (2010) distinguished between a key result indicator 
(KRI) and a key perf ormance indicator (KPI): a KRI tells what has been done and can be either 
a financial or non- financial measure, while a KPI tells what can be done to increase 
performance dramatically and is always a non- financial measure. However, i n this research KPI  
are quantitative measures that can be financial or non - financial, and serve as testimony of the 
past, reflect in some essential way the purpose of a firm, and can be used to encourage 
improvement (Hansen, Stokstad, Hegrenes, Sehested, & Larsen, 2005).  
Despite being useful, KPIs have some intrinsic limitations (Bogetoft & Otto, 2011). A limitation 
of these indicators is that KPIs  make assumptions about the relationship between inputs and 
outputs. For instance, when businesses of different sizes are compa red, constant returns to scale 
(CRS) are assumed. A second limitation of the KPI approach is that it usually involves partial 
evaluations. Often, a substitution effect between inputs can be expected and this is why par tial 
benchmarks can lead to  misleading comparisons. To avoid these, the use of certain 
combinations of benchmarks may be required. A third limitation is known as the Fox’s paradox, 
which suggests that even if one firm has higher values in all its partial benchmarks, it might 
have poorer overall results than other firms. This paradox highlights the importance of 
achieving the right combination of outputs so that to do well, a firm must make adequate use of 
those sub processes that have relatively higher productivity than others, and not just per form 
each process well. In this section, an approach to success appraisal in pasture - based dairy 
farming is proposed based on the review of relevant literature. This includes physical and 
financial key performance indicators part icularly suited for pasture- based farming systems.  
2.3.1 Physical Key Performance Indicators (KPIs).  
Physical performance indicators, especially ratios, often assist financial indicators in providing a 
more holistic view of the business by showing how certain resources have been allocated 
(Rawlings, 1999). Physical indicators are valuable for tracking performance over time and help 
to identify strengths and areas for improvement. Measurements are often given in metric units 
and many have been defined by Fingleton (1995) for Europea n farming systems. They may be 
set as targets to monitor and control in the short term, and also used retrospectively to make 
comparisons between seasons, farms, or competitors.  
Production measures and indicators defining the system such as grazing area, pasture supply, 
use of supplements, or stocking rate, must be included in a basic set of physical KPIs as follow s:  
Chapter II: Literature Review 
29 
1. Dairy area (may or may not exclude dry stock): sometimes acres; usually hectares (ha);  
2. Herd size: peak milked cows as in New Zealand or total cows as in Chile;  
3. Labour input : total full time equivalent (FTE) (corresponding to 24,000 hours of work 
time); these can be paid FTE s, non- paid FTE s, or ratios such as cows/FTE or kgMS/FTE ;  
4. Production: total litres, kg MS, E nergy Corrected M ilk (ECM) (kg );  
5. Production ratios: any of the previous  measures on a per cow, area, or labour unit  basis; 
such as L itres per cow (L/cow) ; Litres per hectare (L/ha)  (Bywater, 2010); and  
6. Stocking rate: cows/ha, Animal unit/ha, Livestock unit/ha, k g Live weight/ton grass grown.  
 
Many of these indicators, such as area, herd size, or labour units, are self - explanatory  inputs, 
and all are common, significant and well known  as physical indicators. Outputs have more 
variation across countries, but commonly, production is expressed as litres (L) or kilogram (kg) 
of milk. Dairy systems analysed using a raw milk volume  do not account for possible variation 
in milk constituents. In such cases, the resulting conclusions may be biased in favour of those 
producers with lower levels of milk solids. To overcome this, two standard measures of 
production have been proposed: kg MS (which ignores volume of milk) , and energy corrected 
milk (ECM)  which normalises different milk qualities to match a MS content of  4.0% fat  and 
3.3% protein  (IFCN, 2011, 2012).  
Production may be better expressed as annual litres of milk per cow or dairy area if the payment 
system rewards volume, for example, L/ha/year and L/cow/day. In Chile, production of milk (L) 
per year and per unit of area has been indicated as the most reliable indicator to measure dairy 
farm efficiency (Bywater, 2010; Department of Agricultural Economics at the Catholic 
University of Chile, 1991, cited in Silva, 1997). Bywater (2010) also recommended the use o f 
litres per cow in Chile. Under European conditions  where cows are housed, the ratio per cow 
has been indicated as significantly more important than the ratio per hectare (Hansen et al., 
2005). A Brazilian study of a pastoral dairy system also found the i ndicator per cow to be more 
important in explain ing variability in return on capital (Neto, Campos, de Olivera, & Gomes, 
2012). Macdonald et al. (2008) went further , and proposed including cow size by expressing 
production relative to size or body weight ( for example, kgMS/kg LW).  
Estimates of how much pasture is available or eaten in terms of dry matter per unit of area (kg 
DM/ha), how much feed energy each hectare produces (Mega Joules of Metabolizable  
Energy/ha), and how many supplements the cows eat (k g DM/cow/year) are also valuable 
indicators. In pasture - based systems, these indicators, although valuable, are based on 
information that is usually unavailable or unreliable (Smit, Taweel, Tas, Tamminga, & 
Elgersma, 2005). Cases where indicators are acces sible often depend on backwards’ 
Chapter II: Literature Review 
30 
calculations, so their reliability needs to be addressed (Bywater, 2010; Mattiauda, Tamminga, 
Gibb, Soca, Bentancur, & Chilibroste, 2013).  
2.3.2 Financial KPIs. 
The observation and study of causes underlying differences in performance began in the early 
1900s and marked the foundation of farm management and farm business analysis (Kay & 
Edward, 2012). A financial KPI is often a ratio of an output to an input and, although most 
industries have very specific KPIs, some financi al indicators are used generically. These include 
Return on Assets (ROA), Gross margin, and Debt ratios (Bogetoft & Otto, 2011). The Farm 
Financial Standards Council (FFSC) (1997) developed and published recommendations for the 
preparation of farm financial performance measures which many organisations have adopted as 
a guide. Later a task force created by the American Agricultural Economics Association 
developed guidelines for the estimation of farm commodity costs and returns. In addition, 
groups in New Zealand and other countries have developed their own sets of KPIs to be applied 
within their particular farming systems.  
 
Nowadays, differences in methodology, terminology, and definitions remain across countries, 
especially when farming systems vary. Furthermore, recent research has demonstrated that no 
single dominant measure describes a farming system’s performance over a reasonably long 
period. Hadrich and Olson (2011)  suggested that a single indicator, such as ROA, may not 
capture enough variability, and multiple indicators should be used jointly  to determine a farm’s 
performance.  
2.3.2.1 Profitability.  
In this section, a basic set of profitability indicators is presented and explained. Profitability 
may be the closest to an efficiency measure of all the financial indicators because it captures 
both inputs and outputs. Several measures of profitability have been recommended by 
specialised entities around the world. The FFSC (1997) recommended examining profitability 
based on four major indicat ors: the Net farm income (NFI ) as an absolute measure; and three 
relative measures: Operating Profit Margin (OPM), ROA, and ROE. In New Zealand, Operating 
Profit (OP) per hectare, Farm Working Expenses (FWE)  per kgMS, ROA, and ROE are 
important KPIs for da iry farming systems (DairyBase, 2009).  
 
Net Farm Income (NF I) is one aspect of profitability , and encompasses Gross F arm I ncome 
(GFI)  -  (Total farm expenses + Depreciation). Depreciation is an imputed cost based on market 
values and a standard depreciation rate depending on the asset. NFI is a n useful measure for 
comparing profitability year to year within a business, but is not suited for comparative analysis 
of firms of different size. NFI represents the returns to  unpaid labour, management, and the 
Chapter II: Literature Review 
31 
owner’s equity (Olson, 2010). It is the amount by which revenue exceeds expenses plus any 
ordinary gain or loss on the sale of capital assets (Kay & Edwards, 2012).  
 
O ther farm financial ratios can be derived from NFI , such as the r elationship between operating 
expenses and gross farm income, also called the Revenue ratio. Kohl and Wilson (1997) 
indicated that a revenue ratio greater than 80% denotes a weak financial position; a ratio 
between 80 and 65% would be considered stable; and less than 65% indicates a strong position. 
These threshold values increase by five to ten per cent when assets are mostly leased rather than 
owned. From NFI, operating profit (OP) can be derived. OP can be summarised  as:   
 
OP = NFI – Livestock - Labour - Man agement- Feed Adjustments + Debt Interest + Other Farm 
Income.  OP is the return after the livestock inventory change adjustment , non- paid labour, and 
management adjustments; supplementary feed inventory change; and depreciation, all of which 
are included in the calculation of operating expenses (DairyBase, 2009).  
 
Although OP is a measure of business performance, it is not an indicator of operating efficiency 
(Shadbolt & Gardner, 2005). In New Zealand, owned run- off adjustment is also included in the 
OP calculation and the us e of relative measures expressing  OP  per kilogram of milk solids, cow, 
or per hectare is encouraged. Operating profit margin (OP M) can be calculated from OP relative 
to gross farm revenue (GFR) ; it is the ratio OP/GFR , and thus, is an expression of the 
input/output ratio. When leasing costs are not deducted, OPM is an accurate indicator of the 
business’s operating efficiency in producing profit from its revenue (Shadbolt, 2011). By 
definition, it is also an indicator of how well  costs are controlled. Kohl and Wilson ( 1997 ) 
indicated the following reference values for OPM positions: strong, >25% ; stable, 10- 25% ; and 
weak, <10%.  
 
Anderson and Ridler (2010) studied resource allocation optimisation in dairy farming systems 
to determine the importance of assessing expenses as well as production and income to optimise 
profit through resource allocation. Farm cost measures and detailed expenditure contain 
valuable information whose value increases with intensification. In Norway, for exa mple, the 
most important KPIs are both fixed and variable feed costs, and the difference between milk 
income and variable feed costs (Hansen et al., 2005). Relevant KPIs are Operating Expenses 
(OpEx), including both cash and non- cash expenditure, and Farm Working Expenses (FWE), 
which only includes cash expenditure (Shadbolt & Gardner, 2005 ). The distinction between 
cash costs and economic costs is particularly important when comparative analysis is 
undertaken and both need to be used when appropriate. According to Dillon et al. (2008) and 
Lobos , Soto, Zenteno, and Prizant  (2001) , total economic costs are considered to be the leading 
Chapter II: Literature Review 
32 
indicator of the sector’s ability to compete, adapt, and expand. Total economic costs include all 
resource input costs, includ ing family labour, equity capital, and owned resources, such as land. 
Based on this, operating expenses are a priori preferable over FWE. Both , OpEx and FWE can 
be expressed in absolute value ($), as a ratio using unit of output such as FWE/kgMS, or as a 
ratio using unit of input such as O pEx/ha. Thorne and Fingleton (2006) , based on Fingleton 
(1995) , proposed working under the format of expenses relative to kilograms of milk solids 
rather than area to examine the competitiveness of Irish milk production.  
 
In Chile, production costs and direct cost s of production are both important indicators measured 
in CLP$ per L  of total milk produced (Lerdon et al., 2008). ‘Direct costs’ involve only costs 
directly related to the management of the herd, the milking facility o r dairy shed, feed, taxes, 
rent and labour costs, including wages and social security. ‘Production cost’ includes  several 
items as follows:  
• wages,  
• electricity,  
• fuels and lubricants,  
• contributions and rent,  
• maintenance of machinery, buildings, milking facility,  fences, roads , etc.,  
• fertilizers, chemicals, seeds and forage conservation,   
• concentrates and off- farm grazing,  
• animal health and milk control,  
• insemination,  
• replacement and purchase of cows,  
• depreciation,  
• administration,  
• general expenditure, including overheads, compensation manager, freight, and livestock 
sales fees.  
 
Relative profitability indicators are widely used across many industries and types of businesses. 
ROA and ROE are the recommended indicators of the earning capacity of the asset or equity 
base (Doehring, 2001). ROA represents the average interest rate being earned on all investments 
in the business. If assets are valued at market value, ROA can be looked at as the opportunity 
cost of farming versus alternate investments (Olson, 2010). Its calculation is as follows:  
ROA = (Net Farm Income + Debt Interest – Management/Unpaid labour reward + Other Farm 
Income) / Total Opening Farm Assets  
Chapter II: Literature Review 
33 
ROA is the term used by the FFSC, but it can also be referred to as Return on Investment (ROI) 
or O perating Return on Dairy Assets and hence, it is the return on assets employed at the 
beginning of the year to generate the income. It is the dairy operating profi t plus owned run- off 
adjustment  less rent, divided by opening assets (DairyBase, 2009). In contrast, Kay and 
Edwards (2012) recommended the use of average assets, calculated as the average of the 
beginning and ending total assets at market valuation. In Chile, return on total assets, including 
land, is calculated by using a verage assets and market valuation (Lerdon et al., 2008).  
Because the ROA shows profitability per dollar of assets, it can be used to compare businesses 
of different sizes and even belonging to different industries (Boehlje, 1994). As ROA is affected 
by opportunity costs (labour and/or management), care is needed in calculating it. According to 
Kohl  and Wilson (1997) , there are different critical values when benchmarking ROA, depending 
on the ownership of the assets. A strong position is indicated by ROA greater than 5% where 
assets are mostly owned, but higher than 12% if assets are mostly leased. Similarly, a weak 
position is indicated by 1 and 3%, respectively.  
ROE  is a measure of the return achieved by the owner’s equity; in other words, it is the int erest 
rate earned on the farm’s net worth (Olson, 2001). ROE is calculated similarly to ROA and is 
defined as the dairy OP plus owned run - off adjustment, plus net off - farm income, less rent and 
interest, as a percentage of opening equity. A crucial relatio nship exists between ROA and 
ROE, which tells whether or not acquired assets are generating a return greater than the cost of 
borrowing (Doehring, 2001). If ROA is greater than the cost of debt, then ROE is greater than 
ROA; conversely, if ROA is less than  the cost of debt, then ROE is less than ROA. ROE is very 
sensitive to the cyclical nature of agricultural production and, with both ROE and ROA, the 
absolute difference between the two values depends on the debt to equity ratio and the standing 
interest rate. Long term, the ROE should be higher than the ROA because it is assumed that the 
manager is using debt leverage advantageously. Therefore, there is a trade- off between a high 
return on equity and high risk as the two are positively correlated.  
A downside to ROA and ROE is their reliance on asset valuations. A sound representation of 
asset valuation, such as government valuations must be used consistently to ensure a that 
meaningful interpretation is derived from these measures. Other indicators that may  be largely 
affected by asset valuation are measures of financial efficiency, such as Asset Turnover Ratio 
(ATR). This is calculated as the relationship GFR/Average total farm assets. According to Kohl 
and Wilson (1997) , there are no critical values for ATR defining strong, stable, or weak 
positions since these depend on the type of operation and whether assets are owned or leased.  
Chapter II: Literature Review 
34 
2.3.2.2 Solvency, liquidity, and wealth. 
Solvency is a financial performance indicator of ownership risk (Shadbolt &  Gardner , 2005) . A 
business is solvent if assets exceed liabilities and insolv ent if vice versa. Ratios in the area of 
solvency are designed to measure the long- term solvency of a farm and are concerned with the 
levels of equity and debt. According to Kay and Edward s (2012) , solvency measures the 
liabilities of an operation relative to the amount of the owner’s equity invested in the business. It 
also provides an indication of the ability to pay off all the business’s financial obligations. The 
most commonly used measure of financial risk is the l everage ratio, from which other indicators  
can be derived. Therefore, both  solvency ratios, Debt:Assets and Equity:Assets, are strongly 
related. The Debt:Asset ratio can be defined as:   
Debt:Asset  Ratio = Total Farm Liabilities / Total Farm Assets.  
Trading in an insolvency situation is illegal since this is a high- risk situation for any lender. 
According to Barry et al. (2000) , the desirable value for the Debt:Assets ratio should be lower 
than 0.5, and the optimum leverage ratio may vary but increases with farm size and leased land. 
The critical value for this ratio relates to the firm’s ability to meet all the expenses and service 
the debt with an acceptable margin of safety. If a farm business is efficient and has a high return 
on assets, then it would be desirable to have a certain level of debt in order to ‘lever up’ return 
on equity. Kohl and Wilson (1997) stated that there is nothing inherently wrong with high levels 
of debt, provided the opera tion is able to service that debt. Therefore, a Debt:Assets ratio of zero 
is not necessarily desirable, but can represent a lower financial risk. If debt payments get first 
claim on profits and equity is a residual return, Debt:Assets will need a relativel y higher return 
to compensate for higher risk. 
According to Shadbolt and Gardner (2005), while profitability could be considered an opinion,  
cash is a fact and therefore, liquidity is  very important. They also claimed that lack of liquidity 
is the most common reason for farm business failure. Chatterjee and Wernerfelt (1991) stated 
that all the standard measures of liquidity can be used as proxy for availability of unused 
internal capital. Four liquidity indicators are commonly used in farming: Working capi tal, 
Current ratio, Operating cash, and Disposable income. The available Working capital and the 
Current ratio are measures of liquidity that show the ability of a business to meet financial 
commitments. Both are relatively easy to understand and are straightforward calculations:   
Working Capital = Total Current Farm Assets -  Total Current Farm Liabilities;  
Current Ratio = Total Current Farm Assets/Total Current Farm Liabilities.  
A positive value for worki ng capital is desirable, but too high a  value may indicate that too 
many lazy assets are being held (Kohl &Wilson, 1997). The Current ratio shows the number of 
Chapter II: Literature Review 
35 
times the current assets cover the cost of the current liabilities; hence, a ratio below one could 
indicate a potential developing cash flow problem. A very high value may not be desirable 
either because it may indicate that too many assets are tied up in conservative investments that 
have lower rates of return. Both the absolute measure of Working capital , and the ratio, are 
relevant because they show whether or not an operation can cash in its current assets without 
compromising its operational status. These indicators express the business’s health; if fixed 
assets have to be sold to cover current liabilities then the operation is affected in undesirable 
ways. Bankruptcy occurs when a business is unable to cover its commitments without 
liquidating fixed assets (Barry et al., 1995). These measures are usually snapshots at the end of 
the year, which do not tell the story of seasonal fluctuations , and therefore, their usefulness  for 
farm seasonal dairy businesses is limited. 
In farm businesses where indebtedness is high , Debt servicing capacity has proven to be another 
useful liquidity ind icator (Shadbolt & Gardner, 2005 ). There are various methods for ca lculating 
Debt servicing capacity , being t he following the most comprehensive:  
 
Debt servicing capacity = (Interest + Principal paid + Rent) /Gross farm revenue  
 
Financial attributes may affect operational efficiency, both on a daily and on a long term basis. 
Nasr, Barry, and Ellinger (1998) explained the different theories about the relationships between 
financial structure and technical efficiency. They noted that the Agency Cost hypothesis implies 
a negative relationship between debt and efficiency. S imilarly, the Free Cash Flow theory 
implies that excess cash flow results in more relaxed management , and thus lower efficiency. 
The Credit Evaluation theory postulated that more efficient farms are expected to require 
investment, and hence, are likely to have higher debt levels. Chavas and Aliber (1993) earlier 
suggested the Embodied Capital hypothesis, which implies that indebted farmers are more 
focused and tend to have higher efficiency levels. Bierlen and Featherstone (1998) investigated 
the effects of financial constraints in farm investment and concluded that a trade- off between 
financial stability and efficiency may exist. Supporting this, Handley et al. (2001) found a 
negative relationship between indebtedness and technical efficiency. Lopez, Featherstone, 
Langemeier, and Grunewald (2006) found that mean technical efficiency was lower when farms 
faced financial constraints, whereas allocative efficiency was found to be greater  for financially 
constrained farms. They additionally noted that farm size influences these relationships. Hallam 
and Machado (1996) reported that larger farms tend to be more efficient. Weersink, Turvey, and 
Godah (1990) found that efficiency was positively associated with size but at a decreasing rate.  
 
Chapter II: Literature Review 
36 
Operating farm cash  surplus or deficit, and discretionary  income are also regarded as relevant 
liquidity indicators in New Zealand. Farm c ash operating surplus is calculated as a farm’s total 
cash income from sales, minus the sum of purchases and cash expenses (FWE) generated 
through the year. Makeham and Malcom (1993) suggested that cash surplus/deficit is the best 
liquidity measure for a season al business. An alternative to discretionary income is known as 
Discretionary Cash (Martin, 2009). In New Zealand, Discretionary c ash is considered to be an 
important KPI, representing the cash available to meet capital purchases, debt repayments, 
drawings, and extraordinary expenses. Discretionary cash t can be defined thus:  
 
Discretionary cash  = Cash operating surplus – (Rent + Interest + Tax) + (Net non- dairy cash 
income + Income equalisation + Net off - farm income)  
 
This calculation excludes interest, rent, and taxes, which, in contrast, are included in the 
definition of discretionary Income. Hence, farm cash surplus can be calculated by subtracting 
net debt paid, drawings, and net capital investments from the disposable income on adding in  
existing off - farm income. This is the indicator of cash available to meet debt repayment, 
personal or family requirements, and to invest after having paid imputable taxes, leasing, and 
debt obligations (Shadbolt & Gardner, 2009).  
 
The relative importance of wealth and wealth creation in farming is different in NZ and Chile . 
However, t he standard measure of absolute wealth, equity , is international. Wealth creation may 
be represented by a positive change in equity , which is commonly taken as a measure in New 
Zealand (Shadbolt & Gardner, 2009). Growth in equity can be calculated as the change in 
owner’s equity over  the year as a percentage of opening equity. A positive change can be the 
result of an appreciation in value or the retention of increased earnings. Conversely, a negative 
change in wealth may be the result of depreciation of the value of the assets or negative returns. 
2.3.2.3 Resilience measures. 
Resilience may be defined as a system’s capability to absorb shocks, avoid crossing a threshold, 
and reorganise after an adverse effect (Resilience Alliance, 2009). The concept emerged to 
conceptualise human- environment systems in sustainability science, and was  originally used in 
relation to the robustness of populations and ecosystems (Folke, 2006). Earlier, Walker ,  Hollin, 
Carpenter, and Kinzig (2004) defined resilience as the on- going capacity of a system to cope 
with whatever the future may bring without experiencing  undesirable changes. Resilience 
applied to farming systems in the area of farm management is a relatively novel concept that has 
been defined as the capability to adapt to alterations in the environment and to take advantage of 
opportunities created by them, while maintaining productive capacity despite potential 
Chapter II: Literature Review 
37 
variability in financial, market, and production related factors (Shadbolt, Rutsito, & Gray, 
2011). Maintaining f arm viability in the long run depends on the ability of the system to adjust 
to change in a way that still generates a profit without compromising land integrity and 
environmental quality. Overall, resilience is a desirable attribute of farming systems and relates 
to what ecology and biology have defined as robustness, plasticity, and rusticity (Sauvant & 
Perez, 2010). Aven (2011) pointed out that a key difference between robustness and resilience is 
that the first relates to a fixed threat, whereas resilie nce applies to any type of potential event 
and associated uncertainties in a given environment.  
Rodriguez, deVoil, Power, Cox, Crimp, and Meinke (2011) claimed that flexible  farming 
systems are more resilient than more rigid systems. Research suggests that flexibility is an 
important attribute for farming systems, just as phenotypic plasticity is a key element in the 
functioning of organisms (DeWitt & Schneider, 2004). Rodriguez et al. (2011) explained that 
where farm management is highly contingent on, for instance, environmental conditions, farm 
managers often vary crops and inputs based on the availability of limited and variable resources 
(land, water, finances, labour, machinery) and signals from the operating environment (rainfall, 
markets), with the objective of achieving a number of often competing objectives such as 
reduced risk or increased profits. According to this, t he ability to respond quickly to changes 
from the external environment is critical  in maintaining profit margins and extract ing profits 
from variability. Dairy  NZ (2009) has noted that:  
The need to compete against low cost producers in a volatile international market means 
that dairy farmers must remain focused on resilient farming systems with relatively low 
fixed costs (p. 6).  
This means that businesses with a higher proportion of variable costs may be more responsive to 
changing market conditions. Conversely, firms with higher proportions of fixed, structural costs 
may not be able to adjust to change. Information management and quick recognition of 
volatility, followed by appropriate adjustments and corrections, are vital, and resilience studies 
may help farmers to identify strong and weak areas  (Rodriguez et al ., 2011). However, t here are 
two key issues around resilience studies on farming systems. First, the observation of resilience 
in the farm management discipline is a difficult task since it requires empirical data for a 
minimum period which depends upon the variability of the environment; it cannot rely on 
snapshots as resilience implies behaviour, which needs  to be observed over time. To overcome 
this, alternative methodologies, including the use of modelling, have been proposed. Second, 
indicators or surrogate variables able to offer an appropriate empirical measure of resilience are 
needed (Bennett, Cumming, & Peterson, 2005). Walker  et al. (2004) postulated that latitude, 
resistance, and precariousness are three major attributes of a system’s resilience. Latitude may 
Chapter II: Literature Review 
38 
be defined as the amount of stretch that a system may support without losing the ability to 
recover its original form; in other words, its flexibility. Resistance may be defined as the ease 
required to create a change in the system. Precariousness is the closeness of the system to the 
point that leads to permanent change. Shadbolt et al., (2011) proposed that these attributes can 
be estimated through surrogates commonly found among key performance indicators in regular 
use in farming: liquidity can be assimilated to latitude; efficiency to resistance; and solve ncy 
can express the degree of precariousness (Shadbolt et al., 2011).  Resilient farms are expected to 
be efficient overall, have increased liquidity, vary according to market conditions, and have a 
low Debt:Assets ratio under unfavourable market conditions.  Along with this, Scholz, Blumer, 
and Brand (2012) pointed out that resilience is linked to risk and vulnerability; it incorporates 
the capability of a system to cope with the adverse effects or risks that a system has been 
exposed to.   
2.4 Macro-environment  
Farm businesses are an integral part of the food and fibre industries and sit within the wider 
national and international macro- environment (Nell & Napier, 2005). The IFCN (2011) noted 
that the viability of dairy farms internationally rest on their ability to cope with the risk to which 
they are exposed. There are several types of risks,  from productive to political risk, from 
technical to financial risk , and from personal to market risk. Robinson and Barry (1987) defined 
a risk as that situation wh ose outcome is not known with certainty  and which alters the decision-
maker ’s  well- being. The EPEST approach has been recommended for the macro - environment 
assessment in farming because farming systems are naturally exposed to m any risks . This 
approach, al ternatively and systematically, looks at economic factors (E), the political and legal 
environment (P), ecological and climate factors (E), social - cultural and consumer aspects (S), 
and technological issues (T). Farming systems often depend on the external  environment and, 
therefore, need to be highly responsive to it. To enhance this responsive ability, it is advisable to 
understand the external environment dynamics and to monitor emerging or potential strategic 
opportunities and threats. This section presents the country background using  the EPEST 
approach, with an emphasis on the New Zealand and Chilean dairy farming systems and 
industries, and also focusing  on some relevant empirical research in both countries.  
2.4.1 Economic factors. 
Different social r ealities exist in New Zealand and Chile, beside peculiar economies, currencies, 
and farming systems. S ocial realities have been shaped by particular cultures, histories, and 
contemporary characteristics. As a consequence, the New Zealand and Chilean dairy industries 
differ in their development and importance to their economies (Challies, & Murray, 2006), 
despite having many similarities, such as shared geographical characteristics, climatic features, 
Chapter II: Literature Review 
39 
production patterns, and ideological convergence around free trade. Table 2 provides a 
comparative description of the two countries.  
Table 2  
Basic Comparative Figures 
 
Source: CIA (2012) ; *Robobank (2012)  
New Zealand has a land area just over one third that of Chile, with a population density of 16 
people per square kilometre compared to 23 in Chile, and a n agricultural labour force occupied 
approximately half that of the Chilean one. Directly, agriculture contributed to 4.6% of the New 
Zealand Gross Domestic Product (GDP) (Statistics NZ, 2012) and 5.1% of Chile’s (INDEX 
Mundi, 2012). Table 2 shows that the dairy sector directly and indirectly throughout dairy-
related activities contributed to 16% and 15% of the  New Zealand and Chilean economies. 
However, farming, and particularly dairy, has always been of far greater importance to New 
Zealand than to Chile (Painter, 2007 ). Table 3 shows the contrasting macroeconomic indicators 
of the target countries. 
 
 
 
 
 
 
 
Country New Zealand Chile
Land area (1000 km2) 268 744
Population (million) 4 17
Population density (person/km2) 16 23
Labour force occupied in Agriculture 7% 13.2%
*Agriculture contribution to GDP (%) 16 15
*Self-sufficiency in milk supply (%) >1000 115
Chapter II: Literature Review 
40 
Table 3  
Economic Indicators of Selected Countries 
 
Source: CIA (2011)  
Table 3 presents figures in US$ because there is no exchange rate between New Zealand and 
Chile and some type of standardisation is needed to compare the two countries. GDP per head 
and average wages have typically been much higher in New Zealand, but the inflation rate 
fluctuates in both countries under inflation targeting regimes. Both countries are important 
commodity exporters, having a similar degree of trade openness in their economies. The 
promotion of free trade has been a longstanding policy that enjoys wide  support in both 
countries, based on the belief that every country has a comparative advantage which must be 
used to develop a trade pattern for its own benefit and that of its trading partners (OECD, 2011).  
Figure 1 presents the NZ$ to US$ exchange rate, which shows a clear increasing tren d for the 
same period (1999 to 2012).    
 
Figure 1. NZ$ versus US$ exchange rate. 
Adapted from Robobank (2012).  
The NZ$ has appreciated against the US$ since 1998, reaching its historical maximum value in 
2011. Sharples (1990) warned that, especially when making international comparisons, the 
Country New Zealand Chile
Per capita GDP (US$) 27,900 16,100
Average wage rate (US$/hr) 11 3
Inflation rate (%) 4.5 3.3
Chapter II: Literature Review 
41 
exchange rate used may affect the conclusions. The author clarified that v olatile exchange rates 
may show very high or low competitiveness relative to other countries, and therefore advises the 
use of averages over a certain length of time when comparing measures of competitiveness. 
Harrison and Kennedy (1997) noted that exchang e rates may influence competitiveness 
measurements in the way that devaluation of the domestic currency results in a decrease in the 
price of domestically- produced goods sold on the world market , and thus, an increase in 
domestic firms’ profits and market shares. This is also stressed by Krugman (1994) , who points 
out that a nation’s competitiveness may be falsely created by devaluing the domestic currency ,  
thus increasing the value of exports, whereas the nation’s standard of living may actually 
decline if the purchasing power is not favourable. Brinkman (1987) explained that when the 
demand for the currency of a competitive country is high, this actually strengthens its exchange 
rate. Chile has traditionally shown higher level of international trade openness than New 
Zealand, with the greatest absolute difference being greater than 20% in 2008. Thereafter, the 
continually declined, reaching very similar s table levels by 2010 (OECD, 2011 ). Interestingly, 
almost all credits and deposits are denominated in local currency: the NZ dollar (NZ$) and the 
Chilean peso (CLP$ ).  Figure 2 shows the Chilean exchange rate to US$ from 1999 to 2012.  
 
Figure 2. Chilean $ versus US$ exchange rate. 
Adapted from: Central Bank of Chile, (2012).  
The exchange rate between the CLP$ and the US$ has had an erratic trend, sharply devaluing 
from 1998 to 2003, appreciating thereafter until 2008, when again the exchange rate abruptly 
reached CL P$650. This level of variability and volatility makes converting l ocal currency into 
US$ for analytical purposes problematic. Bureau, Butault , and Hoque , (1992) pointed out that 
comparing production costs between countries is also made difficult by the variation in 
exchange rates across time periods.  
Chapter II: Literature Review 
42 
2.4.2 Political and legal factors. 
Political and legal factors in New Zealand and Chile are similar. The deregulation of the New 
Zealand economy in the 1980s generated the conditions for restructuring the agricultural sector 
and for changing farming practices towards higher efficiency. This is recognised as an important 
turning point that affected the relative profitability of agricultural industries, forced successful 
farmers to become internationally more competitive and pushed less successful farmers out of 
the business (Frangley & Engelb recht, 1998). Illustrating this, Figure 3 shows the evolution for 
the last 30 seasons in number of dairy herds and their average size.  
 
Figure 3. New Zealand trend in the average herd size and number of herds. 
Source: NZ dairy st atistics 2010/ 11, LIC & Dairy  NZ (2011)  
In New Zealand over the last 30 years, the number of herds has decreased from 16,000 to less 
than 12,000; conversely, the average herd size has increased from just over 120  cows to more 
than 380 cows. Amalgamations f or bigger scale were common across the country, resulting in 
these trends (Journeaux, 2002). Jaforullah and Devlin (1996) pointed out that the average herd 
size also increased because land holdings and herds owned by corporate farms increased, 
especially in the South Island. This process  of amalgamation was not in isolation from other 
phenomena of similar relevance associated to the pattern of land use and the manufacturing 
industry. Conversions from sheep and beef farms to dairy units following the changes in relative 
profitability of these agricultural activities provoked a clear change in land use (Jaforullah & 
Devlin, 1996; Johnston & Frengley, 1994). Johnston and Frengley (1994) indicated that 
following deregulation, sheep and beef farming was displaced by dairying or forestry. The 
manufacturing industry also suffered dramatic changes after a series of mergers and acquisitions 
took place. The number of manufacturing cooperatives decreased from 30 in 1980, to 16 in 
1996, and four at beginning of the 21 st Century (Holmes, 2003). According to the Fonterra 
Chapter II: Literature Review 
43 
Cooperative Group’s Annual Report 2002, economies of scale, efficiency of transport, and the 
retention or gain associated with market share were major reasons for the amalgamations.   
Similar changes were occurring in Chile and worldwide . Zwanenberg  (2001) stated that in the 
1990s a worldwide dairy industry consolidation took place and highlighted  the key drivers for 
this as being milk demand growing faster than supply , and increasingly powerful customers 
having more requirements. A similar dairy expansion accompanied by conversions happened in 
Chile, but the turning point occurred later, in 1994. According to Escobar, Mladinic, Sanhueza, 
and Diaz (2006) , the  beginning of the change for the dairy sector occurred when it was 
announced that Chile would join the MERCOSUR. Before that, the Chilean dairy industry 
showed a sustained rate of slow growth, a result of a stable macroeconomic policy and 
appropriate sectorial policies, including spec ific tariffs on imports, implementation of standards, 
sanitary rules, and technology transference programmes. However, since 1995, the dairy sector 
has grown significantly, particularly in the Central–Southern territory as a consequence of the 
modifications to the strategy after joining the MERCOSUR.  
2.4. 3 Ecological and climate factors. 
New Zealand and Chile share a number  of ecological and climatic factors, including latitudes, 
mountainous landscapes, soils originating from volcanic ash, and similar lit hology, climate, and 
isolation (Challies & Murray, 2006). Geographic location and associated features mean that 
both countries benefit from a complete set of natural barriers. While New Zealand is a group of 
islands in Oceania in the South Pacific Ocean, C hile is located in Southwest South America, 
with the Pacific Ocean coastline stretching more than 4,300 kilometres long to the west; the 
Andes7  to the east; the Atacama which is th e world’s most arid desert  to the North; and the 
Antarctic to the south. Due  to the geographic isolation, strict border controls, and similar 
farming practices, both countries are free from a range of diseases that are endemic in the rest of 
the world.  
The variability in the New Zealand climate is due to its turbulent mid - latitude location, which is 
influenced by many features from the Tropics to the sub- Antarctic (National Institute for Water 
and Atmospheric Research [NIWA], 2010). Much of the variability  in the New Zealand climate 
originates from the interactions of global air circulation and orography being the west to east 
gradient in rainfall the most notable effect. Evenly distributed, frequent damaging events can 
include floods, droughts, wind storms, and heat waves (NIWA, 2010).  
                                                          
7 The Andes is the longest continental mountain range in the world, being a continuous range of 
highlands along the western coast of South America. This range is about 7,000 km long, about 200 km to 
700 km wide, with an average height of about 4,000 metres. 
Chapter II: Literature Review 
44 
Chile extends between 22 and 55 degrees of  latitude South, with strong climatic influences form 
the sea and the Andes. A large area has a mountain climate with perpetual snow and glaciers. 
However, a large percentage of the population lives where agriculture thrives in the lowlands of 
Central and South ern Chile. This part of Chile tends to be wet all year round, with changeable 
weather (Escobar et al., 2006). Annual precipitation can be as high as 5,000 mm (200") , much 
of which falls as snow farther south and on the higher mountains. On the coast, winters are 
rarely extremely  cold, but summers are cool and cloudy. Figure 4 shows the countries, their 
latitudes, with  their key dairy farming regions highlighted in darker shading.  
 
Figure 4. New Zealand and Chile dairy regions and latitudes.  
Adapted from: Challies & Murray (2006)  
In New Zealand, grassland occupies approximately 45% of the country’s area and can be grown 
and intensively grazed all year round (McCall & Clark, 1999). However, NZ grasslands still 
Chapter II: Literature Review 
45 
show large variation in land characteristics, winter temperatures, and summer soil moisture 
(Hodgson, 1999; Holmes, 2003). In the majority of the New Zealand’s non - mountainous 
territory the climate is temperate. Mean daily temperatures range from 8°C in July t o 17°C in 
January, but summer temperatures can reach over 30°C in many places (NIWA, 2011). The 
average rainfall ranges widely from less than 400 mm in the Central South Island (Otago) to 
around 1,000mm in most regions (Gaul & Hughes, 1996). For the majori ty of the North Island 
and northern South Island, the driest season is summer (December –February), but for much of 
the South Island, winter (June –August) is the driest season.  
Over the period 2006 to 2011, N IWA reported that there were several extreme events. The  
2006/ 07 was a good climatic year, but t he 2007/ 08 summer was the driest in a century in parts 
of the North Island, particularly in the Waikato region and the eastern South Island (Dairy NZ, 
2009). Pasture growth, pasture covers, milk production, and stock body condition were severely 
affected by the drought. Dairy  NZ (2010) indicated that the 2008/ 09 season was fairly typical 
climatically in most regions, but dry weather was  again a factor in some major dairying regions, 
such as the Waikato and Bay of Plenty, and also in the Auckland region, Northland, central 
North Island, and the north and east of the South Island. In contrast, it was a wet year in 
Gisborne, Manawatu, and parts of Westland. A Dairy  NZ  economic survey (2011) , reported that 
2009/ 10 was a difficult climatic year for most regions, mainly due to a late summer/autumn 
drought affecting most of the upper North Island and some parts of the South Island. These 
widespread autumn droughts resulted in early drying- off of dairy herds in many regions, 
although the effects of the drought were less severe in the South Island as many Canterbury 
farms irrigated their pastures and parts of Otago –Southland had 42% more rainfall than average 
in early 2010.  
The Chilean southern plains and their temperate grasslands extend south from 38º Latitude  S up 
to approximately 44º S . Mean monthly temperatures range between 5 ºC in the winter months 
and 12- 15 ºC in summer  (Vera, 1996) . Rainfall increases in the north to south  direction from 
about 1,400 mm to over 2,000 mm, with 70% of the rain concentrated from early autumn to 
early spring (Balocchi & Lopez, 2001). Summer rainfall represents between 10 to 17% of this 
total in the north and south, respectively. This wheather pattern creates a brief water deficit 
during peak summer in the northern part of the region, which does not exist at the southern end. 
Over the period 2007 to 2011, according to the Dirección Meteorologica of Chile (2007, 2008, 
2009, 2010 and 2011) , rainfall patterns we re varied. In 2007, key dairy ing regions such as 
Valdivia  and Osorno were drier than normal (about  1,200 mm of rain in Valdivia  and just less 
than 900 mm  in Osorno) . In 2008, rainfall was 1,995 mm and 1,028 mm respectively, while in 
2009, Valdivia had 1,950 mm and Osorno a maximum of 1,345 mm. In 2010, Valdivia had an 
Chapter II: Literature Review 
46 
annual rainfall closer to average of 1,491 mm and Osorno also totalled 1,041 mm, similar to 
those observed in 2011: 1,618 mm and 995 mm respectively. 
2.4.4 Socio-cultural and consumer factors. 
The New Zealand and Chilean dairy industries have been through similar economic transitions. 
However, the development of the industry, the government paradigm shifts, a nd the trade and 
production strategies have been very different. New Zealand dairying has al ways been 
principally an export - oriented activity, while the Chilean dairy sector emerged to supply the 
domestic market. The basic dichotomy that brought about the two contrasting dairy industries in 
New Zealand and Chile remains in force today and is reflected in the information  summary 
presented in Table 4.  
Table 4  
Dairy Industry Facts 
 
Source: I FCN (2011)  
 
Chile has many more, considerably smaller dairy farms than New Zealand. The number of 
milking cows in Chile is approximately one tenth of those in New Zealand. The Chilean 
industry is made up of mostly heterogeneous and small- scale farms, close to market, 
representing around 80% of produce rs and 40% of the dairy herd. In New Zealand, given the 
external market focus, location close to domestic markets is not imperative, farm structure is 
relatively similar, and herds are markedly larger. New Zealand produces not only more milk 
Country
Profile 2009 2010 2009 2010
Dairy farms 11691 15903
Cows (000) 4253 4397 466 423
Average herd size 376 27
Protein (%) 3.7 3.36
Fat (%) 4.91 3.74
Milk solids (%) 8.61 7.1
Milk yield (t/cow) 4.36 4.36 4.91 5.72
Land (USD/ha) 22315 11000
New Zealand Chile
Chapter II: Literature Review 
47 
than Chile but a different type of milk that yields higher milk solids (MS) and is more suited to 
industrial processes. Recently, milk solids content has become part of the Chilean development 
strategy, which states that the country needs to produce milk higher in MS, w ith a view to more 
efficient industrial processing into milk powder ( Consorcio Lechero, 2012). The goal is to 
increase MS from the current value of 7.1% to 7.6% at a rate of 0.05% per year. Although it is 
clear that industry guidance is needed, how to ac hieve the goals remains unclear. S everal 
variables, such as season and time of calving, breed, and diet, are being evaluated since these 
are major factors which can affect milk composition.  
 
According to Winkler (1999) , both countries have  some disadvantages and many benefits in 
dairying as shown in Table 5. 
Table 5  
Advantages and Disadvantages of Respective Dairy Industries   
 
 
In Chile, the advantages are clearly lower wages and capital costs due to cheaper land than New 
Zealand. A major advantage in New Zealand is that virtually all proc essing is dominated by a 
farmer- owned cooperative (Painter, 2007; Willis, 2004) , Fonterra, which became the largest 
dairy exporter  in the world after the New Zealand Dairy Board, the New Zealand Dairy Group, 
and Kiwi Cooperatives merged to create a single entity  ( Fonterra Cooperative Group, 2002) . 
According to the same source, this consolidation has resulted in greater efficiency in operating 
scale, transport, and labour management, while securing m ilk supply, access to capital, market 
share and power, and ensuring New Zealanders and not foreigners receive the potential benefits. 
This means that there is great capacity and limited industry competition within NZ , with 
Fonterra having a 89% share of the market, down from 98% in 2010 (Fonterra, 2012).  
 
Chile has two major companies , Nestlé and Soprole , which account for approximately half of 
the total milk processing, and many more, smaller co mpeting processors (ODEPA, 2012). There 
Country Advantages Disadvantages
New Zealand Labour productivity Cost of capital
   Industry vertically integrated Public concerns about dairy farming
      Dairy specialization
Chile Cost of labour   Industry structure
Cost of capital   Politic-Economic environment
Chapter II: Literature Review 
48 
is strong competition between companies that engage in vertical and horizontal integration to 
control the oligopolistic market (Amtmann & Blanco, 2003). According to Consorcio Lechero 
(2012 ) , production and seasonality have increased steadily. The industry structure has also 
changed, with the number of companies decreasing from 25 to 14, and processing plants from 
38 to 25. In the period 1977- 2011, the fluid milk production increased by 2.8 times, mainly in 
the form of sterilised milk; milk powder production has increased 2.4 t imes, and whey powder 
has also had a significant increase.  
 
In 2011, the total processing capacity of the industry for fluid milk reached 12.5 million litres 
(L) per day , and milk powder, fluid milk, a nd cheese, reached 6, 4.7 and 1.95 million L a day, 
respectively. The regional distribution shows that the processing capacity in the South (X, XIV 
regions) is 78% of full capacity, at 98% of milk powder production, 88% of cheese production 
capacity, and 35% for processing fluid milk. The Central zone is at 65% capacity for fluid milk 
processing. To maintain current growth conditions and seasonality, the processing  capacity 
would be adequate  until 2020 for powdered milk and fluid milk, but is at limited capacity for 
cheese making, which could be saturated by 2016. However, as plant capacity may be sensitive 
to increases in volume and seasonality, this potential scenario may occur  up to three years 
earlier ( Consorcio Lechero , 2012).  
 
New Zealand and Chile are partners and competitors in mor e than one market and have signed 
numerous and varied treaties in the search for promotion of longer- term synergies and 
cooperation (Barton et al., 2007). The two dairy industries have been in direct competition since 
2006 when the Trans Pacific Strategic Partnership came into force, whereby international tariffs 
are expected to be fully suppressed by 2017. Trade liberalisation between the countries is 
expected to promote efficiency, cooperation, and mutual technological advancement across a 
broad range of sectors. These strategic cooperation and competition have been seen as a benefit,  
but also as an issue, by some Chilean dairy farmers (Challies, & Murray, 2006). Other 
interactions also exist, particularly in the form of New Zealand investment in Chile (Es cobar et 
al., 2006). Neither New Zealand nor Chile is of major significance to the other in terms of trade. 
Until 2005, bilateral trade between the countries had been minor (Challies, & Murray, 2006). As 
cooperation and competition occur at the country level and extend to the individual farms within 
and between countries, an approach to generic competitive strategies becomes relevant and is 
addressed later in this chapter.  
2.4.5 Technological factors. 
Technology is defined as the systematic knowledge and set of features that provide context to 
actual farming practices, while technique refers to skills and performance in a more mechanical 
Chapter II: Literature Review 
49 
context (Villano, Boshrabadi, & Fleming, 2010). The estimation of technical efficiency 
determines the scope for improving farm performance through better farming practices, but is 
subject to existing technological constraints. Any comparison across countries acknowledges 
that the underlying technology is different since it is shaped by such issues as ecological and 
climatic factors, industry, logistics, policies, and human resources.  
2.4.5.1 New Zealand dairy farming systems. 
This section addresses New Zealand’s export - oriented dairy industry and the key technological 
aspects of New Zealand dairy farming systems. As discussed before, t he dairy industry has 
expanded rapidly over the last few decades, with the national herd doubling between 1980 and 
2009. Traditionally, da iry farming was restricted to summer- safe, flat to rolling land with mean 
annual rainfall over 2,000 mm  a year. However, the expansion has  moved into regions that 
receive 600 - 1,000 mm rainfall , and which are highly reliant on access to irrigation in summer 
months or are subject to high drought risk . New Zealand dairy farming systems are based  on a 
seasonal operation, a compact spring calving pattern of eight to 12 weeks, and a completely dry 
milking herd for at least eight weeks a year (Holmes, 2003). The d istance and associated costs 
of transport between New Zealand and its main export markets, combined with an environment 
suitable for pastoral production, has resulted in a focus on efficient  dairy production systems 
(Moot , Mills,  Lucas &  Scott , 2010). The principles of this system are pasture feeding, a 12-
month calving interval, and owner –operated low labour units (LIC & DairyNZ, 2011). 
However, increasing use of imported feed has been observed, which has allowed for higher 
stocking rates, leading to an escalation in the production per hectare and in the use of the fixed 
assets. Figure 5 presents the milk solids production ratios, per cow , and per effective hectare, 
since 1992/ 93.  
 
Figure 5. Evolution in MS production per cow and per effective area since 1992/93. 
Source: LIC & Dairy  N Z ,( 2010)  
Chapter II: Literature Review 
50 
The increasing trend, the seasonal variation in production in response to climate and rainfall, 
and the increased gap between production per cow and per hectare since 2003/ 04 are worth 
noting. Using ‘herd’ as the sample unit, the average annual increase in MS production since 
2001/ 02 has been 4,721 kilograms. According to LIC & Dairy  NZ (2011) , this annual growth 
rate of 4.5% from 2001/ 02 to 2010/11 has several contributing factors :  
 
• Mo re hectares dedicated to dairy- annual growth in milking area of 3.8 ha (+3.1% per year);  
• Larger herds –  annual growth of 12.8 cows per herd (+3.9% per year);  
• Higher stocking rate – annual growth of 0.02 cows per ha (+0.8% per year);  
• More MS per cow – annual gr owth of 1.7 kg (+0.5% per year); and  
• More MS per hectare – annual growth of 11.8 kg (+1.4% per year).  
 
The annual growth in cows per farm has been slightly faster than the growth in milking hectares 
per farm, and therefore stocking rate has increased sligh tly over the last 10 years. This increase 
in stocking rate, coupled with the small growth in MS per cow, has resulted in a greater annual 
increase in MS per m ilking hectare. These continuing  increases in production,  and the New 
Zealand approach to pastoral ism, have generated interest from farmers all around the world. 
Since the beginning, the New Zealand dairy industry has had a strong export focus, easily 
supplying the limited domestic market demand, which today sits around 5% of total NZ 
production (Dairy NZ, 2009). Most dairy production is sold at international market prices, 
resulting in milk prices that closely track the prices of commodities on the world market. 
Exports are directed to a variety of countries in approximately 200 differ ent product forms. New 
Zealand is the largest butter exporter in the world ( Indexmundi , 2012) and also an important 
exporter of whole and skim milk powders, contributing  44%, 38% , and 27% respectively to 
world trade (FAO, 2010).  
 
New Zealand’s primary producers have a strong cost leadership focus shaped by low milk 
prices at the farm gate (Holmes, 2003). Low - cost milk production is supported by the 
availability of pasture all year round, direct grazing, and  a the match between maximum herd  
feed demand and peak pasture production in spring. As a consequence, New Zealand dairy 
farmers have become world cost leaders and diversified along the value chain into the 
processing and marketing of dairy products through a farmer - owned industrial sector (Painter, 
2007). D evelopments in New Zealand’s dairy farming have included: a focus on productivity 
and labour saving technologies; the use of nitrogen fertiliser and supplements; on- going 
adaptation to new technologies; and larger farm sizes  (Emerson & Rowarth, 2009) . Some other 
factors have also been branded as critical to the New Zealand dairy sector: international market 
access; effective national and international political support; effective adaptability of industry 
Chapter II: Literature Review 
51 
structure; maintaining a disease– free status; and the farmers themselves (Conforte, Garnevska, 
Kilgour, Locke, & Scrimgeour, 2008).  
 
The New Zealand dairy farmers’ engagement, not only in their managerial role,  but also in the 
farm operation, makes them invaluable human capital for the country. The fa rmers’ 
commitment is well complemented by other characteristics such as people on farm being 
energetic, technically skilful, and optimistic (Holmes, 2003). It is debatable whether or not 
human capability is a limiting factor in New Zealand  already but it is undebatable that labour is 
increasingly becoming an issue (LIC, 2011). Herringbone milking structures are common, and 
rotary sheds are incrasingly being adopted to increase labour productivity. These innovations 
have been regarded as cost efficient, comp etitive, simple, and profitable (Dobson, 1996) , and 
assumed to be the consequence of particularly well - honed management skills resulting from 
world market exposure (Painter, 2007). Table 6 presents a summary of the regional and national 
statistics showing the operating structure of dairy farms in NZ. 
 
Table 6  
Operating Structure of Dairy Farms in Both Islands and at the National Level 
 
Source: Adapted from LIC & Dairy NZ Statistics 2010/ 11 ( 2011 )  
The majority of NZ dairy farms have traditionally been family owned and operated (Sankaran & 
Luxton, 2003). According to the LIC and Dairy NZ statistics (2011), in 2010/11, 65% of dairy 
farms functioned under an owner- operator structure. Another 35% of all  herds were operated by 
sharemilkers or under other structures. Interestingly, the same source reports that, on average, 
sharemilkers on less than 20% agreements have the larger farms, the highest production per 
herd and per effective hectare.  
Business structure
Owner Operator 5,830 76% 1,847 24% 7,677 65%
Share milker 50/50 1,815 81% 434 19% 2,249 19.2%
Others 1,302 72% 507 28% 1,809 15%
Total 8,949 2,788 11,735
Total 76% 24% 100%
NZSouth IslandNorth Island
Chapter II: Literature Review 
52 
A noticeable characteristic is that New Zealand farmers typically challenge the status quo: they 
like seeking growth  in profit and efficiency by using different strategies and by being innovative 
(Williams, 2006). One of these  strategies is the separation of actual milk production from 
breeding, production of replacement heifers, and feed production (Robertson, 2010). This 
specialisation phenomenon occurs through contracting arrangements: a dairy operator may own 
the milking facilities but contract out par t of the feed production and replacement heifer 
growing phases. Some advantages of this separation are human and capital resource 
specialisation, economies of scale, and increased overall efficiency (Purdy, Langemeier, & 
Featherstone, 1997). A further dime nsion of this separation can be observed in the ownership 
and operation of the resources, which is also relevant for sharemilking agreements. Although 
the specialisation in resource use decreases flexibility, the participation in only one stage may 
increase the options for negotiating with partners. According to Boehlje and Schiek (1998) , the 
specialisation strategy allows farmers to leverage volume by investing funds in only part of the 
needed total fixed assets, while maintaining a certain degree of contr ol of the other phases 
through contract specification.  
2.4.5.1.1 The pasture resource and feeding systems.  
The grassland resources, which occupy approximately 45% of the country’s area and can be 
grown and intensively grazed all year round, play a major role in New Zealand’s farming 
success (McCa ll & Clark, 1999). However, these grasslands  show large variation in pasture 
production mainly determined by land characteristics, winter temperatures, and summer soil 
moisture (Hodgson, 1999; Holmes, 2003). The profile of pasture production for most regions 
typically peaks in spring and, to a lesser extent, in early autumn (Robinson, 1999). Furthermore, 
pasture growth patterns are location- specific. For instance, Gaul and Hughes (1996) noted that 
pasture growth in Canterbury and North Otago more closely matched this  seasonal pattern of 
production than other regions of New Zealand. This production pattern and the widespread use 
of irrigation have made dairy farming viable in the Canterbury region.  
 
New Zealand past ures have traditionally been sown in perennial ryegrass ( Lolium perenne) and 
white clover ( Trifolium repens) (Robertson, 2010) that allow an annual productivity ranging 
from nine to 18 tonnes of dry matter per hectare (tonDM/ha) (Valentine &  Kemp, 2007). 
Nationally, and at a commercial level, mean annual pasture has been reported as 12 tonnes  
DM/ha/yr, while top quality production may range from 10 - 16 ton DM/ha/yr (Holmes, 2003). 
New Zealand feeding systems have predominantly been reliant on direct grazing of pasture , or 
crops, such as oats, brassicas, swedes, chicory, and fodder beet, along with variable 
contributions of hay, maize silage, palm kernel, meals, or other supplements (Harris et al., 1998; 
Penno et al., 1999). According to Holmes (1998) , conventional farms  in the past were self-
Chapter II: Literature Review 
53 
contained for feed, but by 1998, most farms were using some kind of imported feed, including 
grazing off - farm. More recently, DairyNZ 8  classified New Zealand dairy feeding systems 
according to the use of imported feed and/or off farm cow grazing as follows:   
 
• System 1: Self - contained – no imported feed: All stock on the dairy platform and all 
supplement harvested off the milking area.  
• System 2: 4 - 14% of total feed imported or grazing off farm: generally  for dry stock and 
replacements with large variation in quantity  depending on the region. This system 
overlaps in part with system three and must be identified by the objectives and the pattern 
in the use of supplements. 
• System 3 : 10- 20% of total feed imported with  imported feed used to extend lactation in 
autumn or to winter the herd.  
• System 4: 20- 30% of total feed imported and used during both early and late lactation.  
• System 5: More than 30% of total feed imported and  used at all times of the year, incl uding 
lactation and dry period. 
 
2.4.5.1.2 New Zealand dairy calving systems. 
In New Zealand, 90% of dairy farming systems are based  on spring–calv ing herds supported by 
seasonal- supply, pasture - based feeding systems (Garcia & Holm es, 1999; Holmes, 2003). 
Spring - calving herds offer many advantages to management, such as synchrony between pasture 
growth and feed demand, and a simple work schedule, although it requires a higher seasonal, 
concentrated labour input. Other advantages are concerned with animal welfare and animal 
health (Holmes, 1986). The seasonal calving system allows for calf rearing in groups instead of 
the individual rearing of calves. First calving usually takes place at 24 months and the calving 
interval is 364 days  ± 31 days (Macmillan & Moller, 1977). Calving dates differ between  
regions with a clear gradient from a median calving date in the first week of August in the north, 
to a median calving date 30 days later in the south (Holmes, 2003). I n some areas where very 
dry summer conditions are common, split calving has been adopted to smooth feed demand 
(Robinson, 1999). However, research at the farm and commercial levels has shown that a range 
of autumn and spring calving systems (from 100% of c ows in autumn through to 100%  of cows 
in spring) can be operated successfully in New Zealand, depending on local climate and 
resources (Holmes, Brookes, Garrick, Mackenzie, Parkinson & Wilson, 2003).   
                                                          
8 DairyNZ is the dairy industry organisation representing farmers and is funded by a levy on milk solids 
and through government investment. Its purpose is to enhance the profitability, sustainability, and 
competitiveness of New Zealand dairy farming (DairyNZ, 2012). 
Chapter II: Literature Review 
54 
2.4.5.1.3 New Zealand breeds. 
The prevalence and main characteristics of New Zealand dairy breeds have changed over time. 
Nowadays, the New Zealand Holstein –Friesian  (HF)  is predominant, making up 42% of total 
dairy cows, followed by 36% Holste in–Friesian and Jersey crossbred cows, and 13% Jersey 
(Dairy NZ, 2 011). The remaining 9% is made up of other breeds, including 0.8% Ayrshire. 
According to Porter (1991) , the Holstein–Friesian originated in North America (NA) from 
imported Dutch black and white cows through selection as specialist milk producers in 
confinement systems. This resulted in a larger, single purpose dairy cow which can produce a 
very high volume of milk low in MS content. The New Zealand Friesian differs from the NA 
Friesian, being a smaller biotype and producing milk with a higher MS content. T he third breed 
of importance, the Jersey, is a dairy biotype that originated in the British Islands of  the English 
Channel. It has the notable ability to adapt to diverse climates,  phenotypically a good dairy 
conformation, moderate milk yields of very high  MS content  and it is less suited to meat 
production than other breeds (Porter, 1991) . Finally, Ayrshire is a dairy breed native to Scotland 
that typically provides good milk and meat on low - cost forage, has an excellent udder, and can 
adapt to a wide range of climates (Porter, 1991).  
The shifts in breed use and crossbreeding 9  feature strongly in the history of New Zealand 
dairying (Holmes, 2001). By the early 1970s, the dominant breed had changed from Jersey , to 
Friesian s that had been bred under local conditions from animals imported from the USA before 
1925 (Harris & Kolver, 2001). Since the 1960s, northern hemisphere Holstein- Friesian genetics 
were introduced again with an expectation that traits for very high milk production would be 
expressed in New Z ealand conditions. However, this breed  did not yield expected results and 
crossbreeding was later adopted to reduce the impacts of ‘Holsteinisation’ (Harris & Kolver, 
2001) , which included shorter lactations and r eproductive issues ( Lucy, 2001) . Farmers have 
focused on breeding cows more suited to pasture–ba sed systems (Montgomerie, 2004) including  
crossbreeding to tak e advantage of heterosis effects comprising improved fertility (Harris, 
2005). More recently, semen from elite Jersey –Friesian cro ssbred sires (‘KiwiCross’) has been 
commercially available, not just in New Zealand, but also around the world.  
According to LIC and Dairy NZ (2011) , in the 2010/ 11 season, Holstein–Friesian and 
crossbreed cows exhibited the highest MS production (kg), while the HF cows produced more 
protein. Jerseys produced the highest percentages of milk fat and protein and lowest absolute 
milk volume as shown by the average per cow statistics for each region. North Canterbury had a 
larger proportion of Holstein  and HF than the rest of the country, and recorded the highest per 
cow milk volume (4,916 litres), milk fat (221 kg), and protein (186 kg). In turn, Taranaki, with a 
                                                          
9 A crossbreed is defined as having, at most, 13/16 of any one breed (LIC & DairyNZ, 2011) 
Chapter II: Literature Review 
55 
larger proportion of Jersey cows, recorded the highest percentage for milk  fat (5.07% ) and milk 
solids (8.95% ).  
2.4.5.2 Chilean dairy farming systems. 
Dairy farming is among the most important agriculture sectors in  Chile, being particularly 
significant in the Southern regions (Escobar et al., 2006; Lobos et al., 2001). The dairy farming 
systems in Chile vary by region, tending towards smaller, more intensive operations in the 
Northern reg ions and larger, export - oriented and more seasonal operations in the South 
(Challies & Murray, 2006). At the country level, Chilean dairy farming systems can be groupe d 
into three or more classes, ranging broadly from confinement to pasture –based systems. The 
pastoral systems may also house the cows over the winter period, and all three classes rely to 
some extent  on the use of supplements all year round (Smith, Moreira, & Latrille, 2002). 
Evidence suggests that most farm businesses can be classified as System 4 or 5 according to the 
DairyNZ classification system.  
 
The Northern farming systems are often confined or  semi- confined dairies, where animals are 
fed on concentrates, irrigated lucerne, and maize or grass silage (Challies & Murray, 2006; 
Smith  et al., 2002). These  operations are domestic market - oriented with all year round 
production supplying Santiago de Chile, the capital city, and its six million citizens. The 
proximity to the metropolitan region allows relatively small- scale, intensive, and niche 
production to occur, and as such, the price paid to producers is often higher than the national 
average (Chall ies &  Murray, 2006). Domestic market - oriented producers supply the raw milk 
for a range of high value- added products, such as yoghur ts and flavoured milks (Barton, et al., 
2007). In this context, dairy producers are usually classified as small (up to 50 L per day), 
medium (100 to 700 L per day), or large (more than 700 L per day), according to the volume of 
milk delivered (Salas, 1995).  
 
The southern central regions of Bio - Bio, also called VIII, La Araucania (or IX), Los Lagos (X), 
and Los Rios (XIV ) have the potential to establish successful pasture based dairy systems. In 
these regions, land values are lower and more extensive dairying occurs almost natural ly 
(Bravo - Ureta & Fuentes, 2003 and Troncoso & Tobar, 2005, cited in ODEPA, 2009). The 
distance from the major market means that less perishable items, such as cheeses and milk 
powder, are more likely to be produced (Barton, et al., 2007).  
 
The pasture- based dairy farm in Chile is usually family owned, but frequently not owner -
operated, mostly because o f absentee landlords (Anrique & Latrille, 1999). Even owners who 
have completed agricultural studies tend to focus on professional or advisory careers rather than 
Chapter II: Literature Review 
56 
on- farm management (Smith et al., 2002). At the operational level, human capability is often a 
limiting factor because of lack of skills and education, the poor social status of contract milkers, 
and sub- standard working conditions  on farms (Vargas, 2001). Frequently, workers and farmers 
have little, if any, training in currently accepted management techniques. In terms of 
infrastructure, herringbone milking sheds are the most common, and twice a day milking is 
common (Carrillo, Moreira, & González, 2011).  
 
Heterogeneity characterises Chilean dairy farming systems, and this heterogeneity is not ju st 
limited to variation across regions, but also intra- regionally (Carrillo et al.,  2011). In the 
southern regions, low - technology dairies co- exist with technologically sophisticated enterprises 
across all systems. According to Smith et al., (2002) , the gr eatest variability is found among the 
medium- sized farms. As a result of this heterogeneity, the costs of m ilk production are 
extremely variable, depending on the area, the intensification level, the technology used, and the 
management skills of farmers (S alas, 1995). According to Escobar et al. (2006), technological 
changes represent the most important development within the dairy sector in the last few years, 
and only the producers that have successfully incorporated hard and soft technologies have been 
able to remain competitive. In contrast, those farmers who were unable to change exited the 
industry, or converted to sheep and beef farming.  
 
Chilean dairy farms are often not specialised, an d usually allocate land to a variety of 
agricultural activities, such as beef or cropping, which generally occupy more than 50% of the 
total land area (Anrique et al., 2004; Lerdon, Munoz, & Moreira, 2010). In particular, the beef 
and dairy markets are clo sely related, mutually influencing the  decision- making process. For 
most specialised farming systems, any change from dairy to beef is a long - term decision; 
however, for Chilean dairy systems which tend to be not specialised,  it can be either a long-  or a 
short- term decision (Winkler, 1999 ). However, what can  be advantageous for the farmer can 
also be a serious disadvantage for the industry as a whole, mainly because this type of 
diversification policy increases both price and volume instability. This policy may have 
contributed to limiting the necessary specialisation for export production. However, th e current 
situation minimises the risk, with a portfolio of products instead of one product marketed into a 
portfolio of importing countries. However, accordi ng to Escobar et al. (2006) , medium and large 
producers tend to specialise and to incorporate technology to pursue higher standards of milk 
quality (hygiene and milk solids content).  Recently, foreign investment that originated in New 
Zealand has come to dominate Chilean dairy product processing, triggering an export -
orientation, and more seasonal operations have increased in number as a result (Escobar et al., 
2006).  
Chapter II: Literature Review 
57 
2.4.5.2.1 The pasture resource and feeding systems.  
Pasture–based Chilean dairy systems are characterised by grazing herds with a variable 
percentage of the diet being supplements of various classes (Smith et al., 2002). These systems 
utilise ‘natural’ 10  and sown pastures year round, forage or pasture hay and silage, and different 
types of concentrates (Schnettler, Lopez, & Barchiesi, 2002). Very wet winters and short - term 
soil water deficits during peak summer are the major restrictions to pasture growth and grazing 
all year round in most southern- central regions. The pasture depending upon soil characteristics 
and management may comprise spontaneous clovers ( Trifolium sp.), Lotus sp., and fine grasses 
(Balocchi & Lopez, 2001). Typical sown pastures rely on some naturalised species, such as 
Trifolium repens, Bromus catharticus, and Dactylis glomerata , but most frequently include 
Lolium perenne, and sometimes Festuca arundinacea. Short - lived pastures, such as annual  
ryegrass and red clover pastures, are also common. Pastures in central - south Chile may yield 
between 12 and 18 tonnes DM/ha.year ( Teuber, 1996). Well - managed natural pastures can 
provide similar yields and equivalent nutritional quality (Balocchi & Lopez, 2001).  
 
In Chile and other parts of the world, supplements are a way to improve milk yields, bo th per 
cow and per hectare, and at the same time, reduce climatic risk associated with pasture– based 
systems (Edwards & Parker, 1994). Moreover, under grazing conditions, dry matter intake is 
frequently insufficient to meet dairy cows’ high genetic potential and supplements are needed 
(Kolver & Muller, 1998). However, supplements increase the cost of production, and the 
profitability of their use depends on the milk:feed prices ratio and the response in milk 
production per gram or kilogram of  feed fed to the cows. Milk production responses of grazing 
cows offered supplements reported in the literature vary between 30 and 150 g of milk solids 
per kilogram  of dry matter intake, but positive quadratic responses to increasing amounts of 
supplement have been observed for yield of milk, energy - corrected milk (ECM) , and fat and 
protein (Auldist et al., 2013).  
2.4.5.2.2 Chilean dairy calving systems and breeds. 
Da iry calving systems in Chile are diverse, from all year round  calving to concentrated spring, 
autumn or both (Smith et al., 2002; Winkler, 1999). First calving usually takes place at 26  to 32 
months, and the calving interval is 365 to 400 days. Artificial insemination is widespread 
(Ferreira, Mujica, Uribe, Lanuza, Quinteros, & Concha, 2010;  Winkler, 1999 ). An important 
feature of the Chilean farming systems is that there is no veal market, which means that male 
and female calves have to be raised (Winkler, 1999). This affects the selection of the type of 
breed; there is a tradeoff b etween milk and meat production which is consistent with the multi-
product system and the diversification strategy explained earlier.  
                                                          
10 ‘Natural pasture’ means that pasture species are spontaneous or naturalised but not necessarily 
native.   
Chapter II: Literature Review 
58 
Traditionally, Chilean dairy farming systems have used predominantly dual purpose breeds, 
such as British Friesian and German Red, although the se breeds becoming less important due to 
their lower milk production potential. Currently, herds are largely based on Holstein cattle, with 
a smaller proportion of European type dual purpose breeds and crosses. T he main Chilean breed 
is now the American Holstein and North American farm manage ment principles are 
predominant with Litres per cow (L/cow) being the industry standard benchmark for efficiency 
(Ferreira et al ., 2010). However, increasingly  “the health of the herd tied up with the genetic 
improvement is very much linked to the search of higher solid content of the milk” (Escobar et 
al., 2006, p.45). This means that crossbreeding and the use of other breeds is increasingly being 
considered due to the industry objective of increasing MS content before 2020 (Consorcio 
Lechero, 2010).  
2.4.5.3 Milk payment systems. 
This section describes the different milk payment systems in New Zealand and Chile. If strong 
industry guidance is required for dairy  farming systems to succeed, there is no clearer signal 
from the market than that coming from the milk payment  system. According to Draaiyer et al. 
(2009) , understanding  the reasons for different payment systems can be gained through 
consideration of the following factors:  
• Stage of development of milk collection in the country or region;  
• Level of trust in the relationship between producers and the industry;  
• Educational level of farmers;  
• Readiness of farmers  to accept payment systems;  
• Present dairy legislation;  
• The hygienic quality of the milk available;  
• Existing adulteration practices;  
• Obtainable products from the raw milk; and  
• Available testing methods. 
 
Ideally, a payment system must be accepted and understood by the producers and all parties 
need to be convinced that the payment system introduced is the most appropriate. Moreover, 
dairy legislation is something to carefully consider before introducing a payment system, and 
national and local regulations might dictate what a payment system can or cannot do. A 
payment system must address the strengths and weaknesses of the surrounding productive 
environment; for instance, if hygienic conditions at milk collection are low, a payment sy stem 
based on hygienic quality is a priority and, conversely, if these conditions are  high, testing and 
paying for hygienic quality might not be the highest priority. Similarly, if adulteration practices 
are known, then test ing and payment systems should be adjusted accordingly.  
Chapter II: Literature Review 
59 
Accordingly, farmers must consider the final market for the products made from the milk they 
produce and consider this market in their decisions (Boehlje & Schiek, 1998). In the future, 
milk producers may be  increasingly called upon to produce milk of varying composition for 
specific food uses. As milk solids content determines the yield of milk in dairy products, if the 
market for these is greater than the market for  fluid milk, the industry might have a payment 
system based on individual components or total MS. However, in many countries, a difficulty in 
developing such a system is the ability to measure the components in the milk coming off the 
farm (Draaiyer et al., 2009). Therefore, any payment system should be based on raw milk  
testing methods that are available, trusted, and economical.  
New Zealand has a small domestic market compared to its export markets, and therefore, far 
greater demand for dairy products than fluid milk. Consequently, the New Zealand system 
rewards dairy farmers for producing milk solids and penalises water content. Milk containing 
higher MS results in reduced processing costs for a given level of production, and therefore, 
increased profitability within the dairy industry as a whole (Bates, 1998). In addi tion, very high 
hygiene requirements at collection means farmers and processing firms have a strong focus on 
testing and information availability which is vital for meeting quality standards and avoiding 
commercial issues and penalties. Whenever possible t he processing companies send information 
back  to producers (even on a daily basis) , allowing farmers to make changes in management to 
favour the production of one component or another.  
For most New Zealand dairy farmers, milk income is derived from two sources of income as a 
result of a dominant, fully vertical, integrated cooperative with farmer own ing shares in the 
processing company, matching  kilograms of milk solid produc ed in a 1:1 ratio with shares. 
Thus, the typical milk pay - out for the farmer- owned cooperative, Fonterra, has two components: 
firstly, that due to the individual farm output and, secondly, that due to the company’s 
performance in the international market. Nevertheless, it is worth noting that milk price is 
completely separate from dividends. A farmer’s monthly income reflects the estimates of milk 
price for the season, while the share dividend is received in February and November each year 
and reflects the added value component of the product generated by the company performance 
in the previous season (exports). The milk component of the price is paid partially in advance by 
the monthly estimates, with a final payment at the  end of a 15- month period (Fonterra, 2012). 
The minority of the processing industry is made up of an investor- owned company, which 
reward producers just for the amount of MS delivered.  
Overall, two clearly different payment systems in New Zealand and Chile have promoted the 
production of dissimilar types of milk and have contributed shaping the different  farming 
systems. In Chile, the milk is currently paid on volume and composition. Between 1997 and 
Chapter II: Literature Review 
60 
2007, the Chilean Agricultural Census showed an increase in the volume of milk produced and 
a reduction in the size of the  national dairy herd; this is consistent with the way producers are 
rewarded. Farmers have responded by increasing the individual cow productivity to the 
detriment of herd size and MS content (Ferreira et al., 2010) . However, there is no  single raw 
milk payment system in Chile; milk is paid for under seve ral arrangements according to the 
company, and there are also differences between producers within a company. Milk is paid on a 
volume basis (L), using a base price and several other factors to make up the final price. 
Annually, each farmer negotiates the base price per L, and the final price includes bonuses and 
penalties according to:  
• Somatic Cells Count  <  3 million cells/ml;  
• Colony forming units < 90.000 /ml;  
• Volume supplied (L);  
• Winter milk bonus;  
• Protein > 3.0%;  
• Fat > 3.0%; and 
• Presence/absence of a cooling tank (Prolesur, 2011).  
 
The base price relates to the litres of milk  produced, with 3.0% fat and 3.0% of protein. The 
final price is only known after the bonuses and penalt ies have been applied. The 3.0%  milk fat 
and 3.0  % milk protein have a built in bonus, resulting in a fixed amount paid per Kg of fat 
protein above 3.0% . Analyses to determine MS content are undertaken fortnightly, not daily. 
The value of protein is almost four times higher than the value of fat ( Prolesur, 2011) ; this is 
also true for NZ and other countries . Other bonuses relate to the hygienic quality of the farm and 
the milk itself.  
2.4.5.4 Recent empirical research on pasture-based farming systems. 
Empirical research requires direct and indirect experience or observation (Hand, Mannila, & 
Smith, 2001). Empirical evidence has shown that pasture- based dairy farming systems in 
temperate regions have lower production costs per unit of output through lower feed and labour 
expenses  (IFCN, 2010) . These attributes are fully expressed where:  
• pasture production potential is high;  
• variability in seasonal pasture supply and quality is low;  
• milk production itself accounts for a large proportion of total production; and  
• large areas of land are readily available at relatively low cost.  
 
Chapter II: Literature Review 
61 
This last attribute is particularly true in Chile where the cost of pasture land is approximately 
half that for similar quality land in New Zealand (CIA, 2012). O ther relevant costs, such as 
wages and supplementary feed, are much higher in New Zealand than in Chile, triggering 
correspondingly higher labour efficiency (IFCN, 2011; Winkler, 1999). However, both pasture -
based farming systems allow higher labour efficiencies than other more intensively capitalised 
systems. Research on resilience has suggested that this should add greater global sustainability 
in a scenario of increased supply and possibly lower prices. Pasture–based systems are expected 
to not only be more sustainable, but also have increased product quality and improved animal 
welfare (Benson, 2008; Dillon, et al., 2005).  
Acknowledging the value of empirical data, DairyBase 11 was established in 2003 by Dairy  NZ 
and has been addressing issues existing in the New Z ealand dairy industry, such as the lack of 
robust information, terminology inconsistencies, and different approaches to the calculation of 
key performance indicators (Shadbolt, 2009). DairyBase collects and stores large amounts of 
data functional to the levy paying New Zeland dairy farmers  and the industry as a whole. 
Statistics New Zealand is another source of hard data from a more random selection of NZ dairy 
farmers of national coverage. Every year, information from both DairyBase and Statistics New 
Zeal and is used in different studies.  
Dairy  NZ is also a routine user of both the DairyBase and Statistics New Zealand  datasets. A 
recent study has showed that milk production has consistently increased over the last decade , 
but at the same time, the average annual total factor productivity has declined by 0.5% (Dairy  
NZ, 2012). This apparent contradiction is explained by the fact that the  extra production 
averaged 5.6% per year  as a result from a 6.1% annual increase in inputs, such as farm working 
inputs, land, and  cows. Over the same period, labour productivity, measured as additional cows 
per full time equivalent (FTE) per year, has increased at a rate of 3.1%. Therefore, farm working 
inputs other than labour that resulted in decreased total factor productivity need to be identified  
from the inputs.    
Recent research based on the Dairy  NZ farm systems classification using the amount of 
imported feed, investigated the drivers of financial success for each of the different systems. 
The study concluded that the cost leadership strategy was pursued by all the five systems in 
New Zealand through effective cost control over the period of analysis, and that the most 
important measure of profit was ROE (Shadbolt, 2011). Higher input systems also consistently 
produced more milk per hectare, but were less able to cope with deteriorating market conditions 
exhibiting inferior performance under depressed milk price scenarios. The author concluded 
                                                          
11 DairyBase is supported by the New Zealand Institute of Chartered Accountants and the NZ Institute of 
Primary Industry Management. 
Chapter II: Literature Review 
62 
that, regardless of the adopted feeding system, better performers are tho se that have both high 
capital and superior operating efficiencies.    
Rouse, Harrison, and Chen (2010) applied a non- parametric efficiency analysis to a case study 
of New Zealand dairy farms data f or 2004/ 05. They found that farms within the Waikato region 
were more efficient than Taranaki farms, and concluded that the differences were mostly 
explained by herd size. Their results revealed that around 300 cows was optimum  herd size , and 
farms associated with higher soil quality combined with less hilly cont ours were more efficient. 
 
Econometric studies have also played a role in analysing DairyBase information (many actually 
analysing its precursor, ProfitWatch) . According to Laca- Viña (2010) , empirical evidence 
suggests that a variety of farming systems have emerged as a result of dairy farming 
geographical expansion in New Zealand. Encouraged by this evidence, Laca- Viña tested for 
differences in technology between regions of New Zealand  and studied the robustness of the 
different stochastic production functions responding to input/output selection ( Laca- Viña , 
2010) . The author reported that sample farms from Farm Monitoring data located in the South 
Island of New Zealand were, on average, more productive than  those farms in the Waikato –
Taranaki region. Given the differences in technology and scale, it was not clear from his results 
which factor influenced his conclusions the most. The author also claimed that the high levels of 
technical efficiency of farms in both regions suggests that the industry could not expect to 
achieve major productivity gains  through efficiency improvements;  rather, these efficiency 
gains may only be achieved through technical progress or technology gains that increase both 
labour and capital productivity. The robustness of farm technical efficiency estimates to the 
input/output s chosen was addressed by comparing farm level estimates obtained by the different 
models using the Pearson and Spearman correlation coefficients. Evidence suggested t hat 
efficiency estimates were not very sensitive to the choice of input/output set.  
 
Motivated by the proposed Emissions Trading Scheme and its effects on dairy farming, Jiang 
(2011) looked at the regional efficiency performance of dairy farming in New Zea land using the 
stochastic frontier framework for the period 1998/ 99 to 2006/ 07. It was found that, given the 
different climate, soil, and farming history, the North and South Islands do not share the same 
technology. The subsamples in both islands did exhi bit similarities, like decreasing returns to 
scale, but dairy farming technologies are markedly different. For example, labour and electricity 
were relatively more important for dairying in the North Island, while farming systems in the 
South Island relied  more heavily on capital and fertiliser. Interestingly, when evaluated against 
their own regional frontiers, the North Island farming systems  had higher technical efficiencies 
than those in the South Island. However, a direct comparison between North and S outh I slands 
Chapter II: Literature Review 
63 
showed that farming technology was more advanced in the South Island. Jiang concluded that 
efficiency encouragement in the South Island should focus on lifting individual technical 
performance, while a technological change is required in the N orth Island to lift the whole 
region’s efficiency frontier.  
 
Empirical data shows that Chilean dairy farming systems have traditionally shown greater 
variability in efficiency, profitability, and other variables’ results. A study conducted in the 
Central area estimated a range between 3.3 and 13.6% of return on invested capital for dairies 
with varying levels of technology (Aichele, 1979). Similarly, in the Central area, Campos et al. 
(1995) determined that the average return on capital investment for the production of milk 
ranged between 5 and 14.9%. Another study by Silva (1997) surveyed 28 dairy producers in the 
VII Region and reported that return on assets for the larger producers fluctuated between 10 and 
12%, but both variability and profitability obse rved with the medium producers were higher, 
ranging from 10 to 17%  ROA. Diaz and Williamson (1998) looked at profitability, 
infrastructure, and management of pasture- based dairy farms in Southern Central Chile during 
1995 and 1996 , and observed a 3% ROA for producers with medium sized dairies  and 8% for 
large producers. They concluded that the effect of scale on farm profitability was significant. 
They also found that size was positively correlated with management improvements and 
infrastructure.  
According to several studies cited in Lobos et al. (2001) , the most important f actors determining 
the economic performance of a dairy farm are feed, genetics, reproductive results, animal 
health, scale, production per cow, pasture grown per hectare, cost control, a nd investment. In 
Chile, the evidence suggests that a series of management factors directly controlled by the 
farmer determine efficiency (Lerdon, Muñoz, & Moreira, 2010). This study revealed that costs 
of production, stocking rate, and quality of the milk  were the key factors determining the 
productive success of the farm. Moreira and Bravo - Ureta (2003) measured technical efficiency 
and technological change using a highly unbalanced panel of 48 Chilean dairy farms from 
1996/ 97 to 2001/ 02. They found that m ean technological efficiency varied from 69% to 77%. 
Over the same period, but using a different methodology , stochastic metafrontier analysis, the 
same authors evaluated technical efficiency and the technological gap of dairy farms in 
Uruguay, Argentina, and Chile (Moreira & Bravo - Ureta, 2010). They found an even lower 
mean technical efficiency of 66% in Chile, which was the lowest in the region.  
2.5 Benchmarking Farming Systems 
This section addresses the motivation behind comparative analysis as well as relevant 
definitions and applications of benchmarking found in the management and efficiency literature.  
Chapter II: Literature Review 
64 
2.5.1 Why benchmark? 
Recording information and benchmarking are of partic ular importance in agriculture, and both 
have developed to varying degrees, depending on the country and the industry (Whitehead, 
2008). According to recent studies, the smaller the scale of the farms, the greater  the industry 
guidance in the form of technical and economic recommendations is needed (Bernard, Le Gal, 
Triomphe, Hostiou, & Moulin, 2011). T his form of leadership from the industry has included 
benchmarking, monitoring, process control, and other extension activities have been included 
(van Eijk,  1998). Similarly, information, analysis, and evaluation are necessary for the enhanced 
design and management of viable and robust farming systems (NRC, 2011). This implies the 
acknowledgement of others doing well at something, the necessity for shared information, and 
the drive for excellence through continuous learning and improvement. The practice of 
benchmarking has had broad application in vari ous contexts, including problem solving, 
planning, goal and strategy setting, process improvement, and innovation.   
2.5.2 Benchmarking and benchmarks. 
Benchmarking is the exercise or process of searching for best practices that lead to superior 
performance (Camp, 1989). Accordingly, it is  defined as an on- going process that involves 
learning and benefitting from others’ experience , leading to superior performance (Bogan & 
English, 1994). McGuckian (1996) is supportive of t hese definitions and also stated that a firm 
can develop methods to improve its own by looking at the performance of other businesses. 
Anderson and McAdam (2007) pointed out that many definitions are valid for benchmarking 
and argued that they essentially share the same common ideas. According to Jack and Boone 
(2009) , a comprehensive definition of benchmarking should include information sharing, trust, 
collaboration, learning, and adoption of best practices to improve performance. However, to 
undertake these activities, a comprehensive ben chmarking system, including a range of 
benchmarks, is needed.  
Benchmarks are metrics that may be used as operating statistics to identify performance gaps 
and the competitive position of a particular business (Martin, Zwart, Gardner, & Parker, 2005). 
According to Benson (2008) , necessary physical and financial performance benchmarks for 
pasture– based farming systems include ratios per cow, per unit of milk sold, and per unit of land 
area. The stated objective of benchmarking studies is to get insight through comparative 
evaluation; relative performance is assessed for firms which transform the same type of 
resources (inputs) into the same type of products (outputs) (Bogetoft & Otto, 2011). 
Benchmarking c an be used to make longitudinal  panel comparisons, where the performance of 
firms in different time periods is compared. Other stated objectives and uses  of this practice 
may include decision making, coordination, control, and motivation (Bogetoft & Otto, 2011).  
Chapter II: Literature Review 
65 
2.5.2.1 Benchmarking forms and types. 
Martin et  al. (2005) stated that benchmarking activities may be classified according to 
objectives  in or four recognised forms as follows:  
Industry group benchmarking: Comparative analysis where various benchmarks from a business 
operation are contrasted to similar measures in other businesses belonging to the same industry 
group.  
Best practice benchmarking: “A particular systematic approach in which a business evalua tes its 
own operations and procedures through a detailed comparison with those of another business, in 
order to establish best practice and to improve performance” p. 60 (Jack & Boone, 2009) .  
Competitive benchmarking: This is the study of a competitor wit hout their cooperation, for the 
purposes of product or process quality improvement. This form is unlikely to reveal practices 
for improving performance. 
Cooperative benchmarking: This is the study of production functions aimed at improving 
performance. It is done between benchmarking partners. This form has the potential to reveal  
innovative practices.  
Similarly, Martin et al. (2005) suggested three distinctive types of benchmarking:  
Process benchmarking: This type focus on acquiring the know how to achie ve operating 
practices of excellence. In agriculture, this tends to focus on technical issues.  
Performance benchmarking: This  type concentrates at the tactical level on results that are 
consistently being achieved, and seeks to identify the practices that deliver such results.  
Strategic benchmarking: This type concentrates on longer term, strategic decisions that have 
been made in a business, involving the  triggers and the positive outcomes that resulted from the 
change. It is usually long- term focused and rarely industry- specific.  
The most common use of benchmarks for diagnostic purposes and performance analysis in what 
is also called ‘comparative analysis’ (Jack & Boone, 2009) or industry group benchmarking 
(Martin et al., 2005). Watson (1993) identified four key components to benchmarking: the 
planning stage; conducting the research; analysing the data; and adapting, improving, and 
implementing the findings. The process of benchmarking involves a  definition in the scope of 
the study and the selection of a benchmarking partner; determination of the relevant indicators 
or benchmarks; data collection; analysis focusing on strengths and weaknesses; discussion of 
results and implications; and seeking new  benchmarks.  
Chapter II: Literature Review 
66 
2.5.2.2 Benchmarking as a farm management tool. 
Benchmarking has been a longstanding and widespread practice. Discussion groups and monitor 
farms, started in France in the 1940s and in New Zealand in the 1980s , respectively, are 
considered early examples of benchmarking to inform, transfer technol ogy, and motivate 
learning and action (Campbell, 2008). According to the Food Chain Centre (FCC, 2007), 
farmers value benchmarking because for : cost understanding, working practices improvement, 
understanding business drivers, increasing returns, and improving quality of products and 
processes. Benchmarking  has been reported as a common exercise among top performing dairy 
farmers (Wilson, 2011). Verissimo and Woodford (2005) stated that top farmers are generally 
information- rich and use many strategies to acquire and filter relevant information, including 
benchmarking. Benchmarking as a farm management method can help identify the necessary 
changes to improve financial position and is also useful for strategic planning.   
 
However, while benchmarking is widely used, its adoption does not appear to influence 
profitability (Gloy & LaDue, 2003). Benchmark data are powerfu l diagnostic and research tools 
which enable the assessment of the prevailing farm management performance, as well as a 
monitoring the trends and changes in farming (Jack & Boone, 2009). Benchmarking , and the 
search for competitive advantage, can address whether differences in performance are due to 
random events or to managerial ability (Langemeier, 2010). Langemeier (2010) also indicated 
that it is hard to perform well consistently, and concluded that benchmarking over several 
seasons is essential for determining competitive advantages. In addition, according to Benson 
(2008) , judgements about relative performance of one or more particular f arming systems 
should be based on a large sample from different regions and various practices. 
 
Benchmarking can be used to assess  an individual farm business’s performance , and to examine 
the relative competitiveness of different farming systems. However,  Makeham and Malcom 
(1993) criticised benchmarking , pointing out that the emphasis on accounting and recording 
leaves out the critical technical and human aspects of management, and it also has a past, not a 
future orientation (Anderson & McAdam, 2004) . Ba rnard and Nix (1979 )  cited in Fleming, 
Farrel, Villano, and Fleming  ( 2006) noted five key problems with comparative analysis  
benchmarking  identified in the academic literature:  
 
1. It was not a holistic approach;  
2. It failed to incorporate some key economic p rinciples;  
3. It failed to establish causal relations between farming practices and performance;  
4. It had limited scope for action; and  
5. It neglected the effect of risks and uncertainty in farm decision –making.  
Chapter II: Literature Review 
67 
 
According to Fleming et al. (2006) , failure to identify causal relationships between performance 
and factors such as input- output prices or rainfall often led analysts into interpretative errors. 
Therefore, a holistic approach is  needed to adequately contextualise benchmarks. Possibly the 
most damaging criticism of comparative analysis was that it neglected economic principles. 
Malcom (2004) postulated that economics is the core discipline of farm management , and 
without its contribution results from comparative analysis have little prescriptive value. 
Benchmarking businesses to identify differences and similarities in financial and productive 
farm performance may not be enough to comprehensively assess a farming system (Anderson & 
Ridler , 2010). A farming systems approach interpreting the outcomes of the comparative 
analysis may overcome these faults. However, according to Fleming et al. (2006) , 
benchmarking partial performance measures means that benchmarks may only convey partial 
information on the farm’s  performance. A more comprehensive measure is then needed to get 
an accurate picture of the whole farm performance.  
2.5.2.3 Data envelopment analysis (DEA):  a method for benchmarking. 
The literature reviewed has shown that most performance measures are system- specific and that 
there is no consensus on the best set of benchmarks to concisely describe a dairy farming 
system. However, efficiency can fulfil the role of being a simple, comprehensive, and versatile 
indicator that naturally stands out as a benchmark because it is always a relative measure. 
Efficiency measures the ability of a certain business to use the existing technology or set of 
resources in the best possible way. Using  an output- oriented perspective, it is the observed 
output of a particular farm compared to the maximum attainable output that can be produced 
under the current production frontier or technology set (Farrel, 1957). Villano et al. (2010) 
defined efficiency as the ratio of the observed output to the corresponding technological frontier 
or benchmarking peer, conditional on the level of inputs used. Although it is debatable whether 
technical efficiency is the ultimate measure of performance as claimed by Diewert and 
Lawrence (1999) , it is undoubtedly important beca use if a firm does not attain technical 
efficiency, it means that resources are being wasted.  
It is important to distinguish between efficiency and productivity. Jiang (2011) highlighted the 
differences and similarities between both terms, and claimed that these were not interchangeable 
since productivity is an absolute measure. DairyNZ ( 2012) defined productivity as the ratio, 
aggregated inputs to aggregated outputs. However, Bravo - Ureta (1986) warned that the 
aggregation of inputs and outputs is subjective and can affect the conclusions. Despite 
differences in methodology, warnings, and acknowledged limitations, m odern benchmarking 
analyses increasingly use best practice or frontier analysis methods. Table 7 presents a 
taxonomy of frontier  methods adapted from Bogetoft and Otto (2011).  
Chapter II: Literature Review 
68 
Table 7  
Frontier Methods Taxonomy 
 
In the choice of deterministic or stochastic methods, the key question is whether the analysis 
requires flexibility in the mean structure or precis ion in the noise separation. Data Envelopment 
Analysis (DEA) is a deterministic, non - parametric method that responds to flexibility,  without 
the need to know the underlying production functions or to have price information. Gattoufi, 
Oral, and Reisman (2002) reported that top DEA article journals include the European Journal 
of Operations Research, Management Science, Journal of Productivity Analysis, Applied 
Economics, Journal of Econometrics, and Journal of Banking and Finance. In the last decade, 
DEA has increasingly gained popularity among more diverse areas of research because of its 
simplicity, versatility, and relevance.  
Farrell (1957) first envisioned the use  of the production frontier in terms of efficiency, 
observing that there existed an efficient function from which all the observed points deviate 
randomly but in the same direction. Later, Charnes , Cooper and Rhodes (1978) introduced DEA 
as a technique based on mathematical optimisation, which had the ability t o evaluate relative 
efficiency and to set individual firm efficiency targets. More recently, Coelli (1997) pointed out 
that a firm was only efficient if it operated on the frontier, and further, if all associated ‘input 
excesses’ were zero.  A distinct advanta ge of DEA is that it provides explicit, real peer units for 
benchmarking. Rouse et al. (2010) described DEA as a formal method, either for efficiency 
measurement, benchmarking, or both.  
DEA owes its name to the way the frontier envelops observations, whic h are typically 
represented by a performance score that ranges between zero and one and  an expected skewed  
to the right distribution for the efficiency scores with most scores falling above 0.5  (Cooper, 
Seiford, & Tone, 2006). The production frontier has several important properties, such as non -
negativity, non - decreasing values, concavity, and weak essentiality. The first to third conditions 
Deterministic Stochastic
Parametric Corrected Ordinary Least Squares (COLS) Stochastic Frontier Analysis (SFA)
Aigner and Chu (1968), Lovel l  (1993), Greene (1990, 2008) Agner et a l . (1977), Battese and Coel l i  (1992), Coel l i  et a l  (1998a)
Non-parametric Data Envelopment Analysis (DEA) Stochastic Data Envelopment Analysis (SDEA)
Charnes  et a l . (1978), Deprins  et a l  (1984) Land et a l . (1993), Olesen and Petersen (1995), Fethi  et a l . (2001)
Chapter II: Literature Review 
69 
are obvious in their meanings, while the weak essentiality property means that the production of 
positive output is impossible without the use of at least one input (Coelli, Prasada Rao, 
O’Donnell, & Battese, 2005).  
DEA models separate the data variables into input or output types such that input variables are 
causal factors to the outputs. The proper treatment of inputs and outputs in efficiency studies 
has been regarded as being very important; the variables have to be consistent with the 
phenomena they are supposed to capture. According to Coelli et al. (2005) , inputs may be 
classified into five categories: capital ( K); labour (L); energy (E); material inputs (M); and 
services (S). Often E, M and S  categories are aggregated into a single category ‘EMS’ . Capital,  
unlike  the other variables which are consumed in the production process, is a durable input 
which may be measured from the various assets of a particular Decision Making Unit (DMU) or 
farm. In farming, K is typically composed of land, livestock, buildings, and plant. Labour 
constitutes another primary input, normally measured using a single variable which aggregates 
paid and unpaid labour and management. It may be expressed in physical full - time- equivalent 
units, hours of work, or in monetary terms. None of these  measures take into account the 
composition and quality of the labour force such as  skills, age, gender, and education levels 
(Coelli et al., 2005).  
Moreover, the number of efficient DMUs relies on the number of inputs plus outputs (Cinca &  
Molinero, 2004). The greater the dimensionality of the production function the more constraints 
the model has and the less discerning the analysis (Jenkins & Anderson, 2003). The analysis 
ought only to include inputs and outputs that are definitely relevant to the proposed study. 
According to Rouse et al (2010) there are some other rules about their proportion:  
1. The number of farms (firms or decision - making units)  must exceed a minimum of three 
times the number of inputs plus outputs; or  
2. The number of farms must be higher than the product of inputs and outputs. 
 
One important point to note is that inputs need to represent a production function or technology 
set such that the proposed output is achievable (Coelli et al ., 2005). Another rule is that input 
variables should be controllable variables. The transformation of multiple inputs into outputs is 
affected by controllable and non- controllable variables, as well as observable and non-
observable managerial characteristics. Therefore, there is a lack of infor mation about the true 
underlying technology that needs to be acknowledged. This, in turn, leads to an efficiency 
estimate that is always higher than the real value of efficiency. The efficient DMUs (peers) 
resulting from a DEA provide concrete evidence o f performance improvement targets for both 
inputs and outputs. As peer units are usually interpreted as those demonstrating how a firm can 
Chapter II: Literature Review 
70 
improve, it is relevant that changes are actually feasible, which is only possible if inputs are 
fully controllable.  The analysis of the DEA peers is out of the scope of this study; exploring the 
explicit, real peer units would be a field for further research.  
Economic or ‘overall’ efficiency (EE) is one of the major factors explaining differences in firm 
survival and growth (Olson & Vu, 2009).  It has two components: price or ‘Allocative’ 
Efficiency (AE) and technical efficiency (E), such that  
EE = E x AE (Farrel, 1957; Olson & Vu, 2009).  
AE reflects the ability of a business to use inputs and produce outputs in optimal proportions to 
maximise profit, given their respective prices (Farrell, 1957). In turn, E is a measure of how a 
business transforms inputs into physical outputs and includes two components: pure technical 
efficiency (TE) and scale efficiency (SE) (Moreira & Bravo - Ureta, 2010). According to Barnard 
and Boehlje (1998) , economic efficiency is determined to a large extent by pure technical 
efficiency because it reflects managerial ability especially in smaller firms. Allocative, scale, 
and pure technical efficiencies all impact profit, but the first is always related to changes in 
input and/or output prices, neither of which are  controlled by farmers. In contrast, technical 
efficiency refers only to discretionary variables or those variables that can be influenced by 
someone’s discretion, judgement, or preference (Harrison, Rouse & Amstrong, 2012). 
According to Latruffe (2010) , in most circumstances an improvement in technical effi ciency can 
result from a more competent use of the existing technology: that is by producing the same 
output by using less inputs or more output with the same level of inputs. A second way to 
improve efficiency is through economies of scale. Figure 6 shows  a graphical representation of 
both constant ( CRS )  and variable returns to scale ( VRS )  DEA envelopes for one output and one 
input condition.  
Chapter II: Literature Review 
71 
 
Figure 6. CRS and VRS DEA envelopes. 
Adapted from: Coelli (1997)  
Scale refers to a  firm’s size; scale efficient firms have a scale elasticity of one and work under 
Constant Returns to Scale (CRS), while scale inefficient firms could exploit either scale 
economies or diseconomies, which imply Variable Returns to Scale (VRS) , including bo th 
decreasing (DRS) and increasing returns to scale (IR S) (Coelli, 1997).  DRS takes place when a 
proportionate increase in all inputs results in a less than proportional increase in output; the 
latter conversely happens when increased inputs lead to a more than proportionate increase in 
output. ‘ Scale’ interacts with technical efficiency in predictable ways: in larger scale operations, 
managerial input can be spread too widely and this might generate inefficiencies, while 
excessively small scale operations can result in high costs, because the fixed investment is 
spread over relatively low output levels (Barnard & Boehlje, 1998). TE, when measured under 
the assumption of CRS, only represents the true technical efficiency if the firm has an optimal 
scale of operation (Rouse, Harrison, & Chen, 2010). Otherwise, the pure TE component, which 
refers solely to management practices rather than the firm’s operating scale, must be calculated 
under the VRS assumption. SE may also be of interest, and can be calculated as the ratio 
CRS:V RS. Different DEA assumptions define diverse DEA methods that differ in the ex- ante 
assumptions involved. A major DEA assumption is that data is measured without error , and 
thus, differences in efficiency scores are due to inefficient transformation of input into outputs, 
not allowing for noise or error. The free disposability assumption stipulates that unnecessary 
inputs and unwanted outputs can be freely discarded. 
 
Chapter II: Literature Review 
72 
2.5.2.4 Regression or Recursive Partitioning Tree. 
Regression partitioning tree is a data mining  method that recursively partitions data into sets,  
each of which is simply modelled using regression methods ( Wit ten & Frank , 2000). RPT is a 
non- parametric statistical technique used to classify observations into homogeneous groups 
(Godfrey, 2011). DEA e fficiency scores can be used as the response variable in a second stage 
approach, to find the best summary set of benchmarks  that explains a  dairy system’s  
performance. Since often the databases provided many more variables than those required to 
build such a summary set, an effective reduction of the multidimensionality of the datasets is in 
most cases desirable. RPT  can regress discretionary and non- discretionary inputs on its analysis. 
This approach not only generated a ranking of variables, but also revealed the interaction 
between independent variables, which it is not possible to do with other forms of multivariate 
analysis (Curran et al., 1993). RPT approach can identify the relationships between different 
variables of interest as reported by Hansen et al. (2005). Rossi et al. (2005) also pointed out that 
RPT provides a variable ranking that is more stable than stepwise variable selection.  
2.5.3 Generic competitive strategies. 
Competitiveness is defined as the ability to profitably produce and provide value at an equal or 
lower cost than that of other sellers in a specific market (Harrison & Kennedy, 1997). Since a 
competitive advantage is not in itself a guarantee of achievement, in order to be successful, a 
firm needs to define the position and the competitive strategy that it is pursuing. By defining 
these position and strategy, firms get closer to the best way to succeed in a particular industr y. 
Similarly, by executing a generic competitive strategy, the business is able to put its competitive 
advantage into practice. The competitive advantage of a firm is the extra value it creates and this 
is the difference between the value created by the firm and the potential value created by its 
competitors (Spulber, 2004).  
Prosperous farm businesses are increasingly utilising and embracing well- known concepts and 
principles to improve their competitive position, efficiency, and productivity. However, the  
traditional techniques for being efficient , which are reducing cost and increasing margins, are 
not always sufficient to ensure successful agricultural production (Boehlje & Schiek, 1998). 
Three positive ways to achieve the best possible strategy, differentiation, focus, and overall cost 
leadership, are determined by the firm’s competitive advantages (Porter, 1980). The first step  
towards defining the best strategy is to scan the external business environment and assess the  
business's internal reso urces and capabilities. According to Gray and Boehlje (2004) , once the 
activities upon which a business may build a competitive advantage have been identified, the 
strategic position from the three options above can be defined.  
Chapter II: Literature Review 
73 
2.5.3.1  Differentiation. 
The transition of agriculture from a commodity industry to one of differentiated products 
indicates a dramatic paradigm shift in the industry (Boehlje & Schiek, 1998). The main aim of 
the differentiation strategy is to be perceived as unique industry - wide. Differentiation can be 
gained by selecting one or more attributes that clients perceive as important and trying to 
position the firm in such a way that the attributes of its products are considered by the consumer 
to be different from the rest of the industry’s products. Some of the most common attributes that 
consumers perceive as different are the firm’s delivery service and marketing approach (Porter, 
1985). Differentiation takes many form s including design or brand image, special features, 
customer services, and technology. Differentiated firms have  a low level of competition and 
rivalry as consumers have a lower sensitivity to price due to brand loyalty. 
2.5.3.2  Focus. 
Boehlje and Schiek (1998) stated that differentiation in the commodity industry has been 
particularly noteworthy when combined with a focus on the consumer and manufacturing 
approach to production. Focus strategy is based on choosing a market segment or group of 
segments in the industry by product line or geographic market. Once the marke t segment is 
chosen, the firm’s efforts should be focused to serve that segment to the exclusion of others. 
The focus approach can achieve low- cost or differentiation for that market segment , but not for 
the whole market (Porter, 1985).  
2.5.3.3 Overall cost leadership. 
This strategy has a broad scope and is used in agriculture and many other industries and 
industry segments. A business has a cost advantage if cost efficiencies allow it to consistently 
outperform competitors and earn greater economic profits (Porter, 1980). In order to achieve 
and maintain a cost leadership position, an aggressive construction of efficient scale processes 
and on- going cost minimisation through innovation is required (Spulber, 2004). More 
specifically, Barney and Hesterl y (2006) indicated that there are six main sources of cost 
advantages for firms that successfully adopt cost leadership as described below:  
1.  Size differences and economies of scale;  
2.  Size differences and diseconomies of scale;  
3.  Experience differences and learning - curve economies;   
4.  Differential low - cost access to productive inputs;  
5.  Technological advantages independent of scale; and  
6.  Policy choices.  
Chapter II: Literature Review 
74 
Further, the authors explain that the ability of a valuable cost leadership strategy to create a 
sustainable competitive advantage is conditional upon the strategy being difficult and costly to 
imitate. At the farm level, cost structures are explained by fixed and variable costs. According to 
economic theory and empirical research, the cost explanatory variables are U shaped, meaning 
that costs of production decrease to a minimum point where economies of scale are achieved. 
Subsequently, diseconomies of scale occur and costs increase (Smy th, Butler, & Hennessy, 
2008) .  
Agricultural systems’ cost advantages are seen  in those countries or regions where the particular 
climate, soil conditions, or labour supply advantages allow more efficient farming than in other 
regions (Kay, Edwards, & Duffy, 2012). The competitive strategy that best represents pasture–
based dairy farming systems is cost leadership due to typically low feeding costs, since feed is 
the highest single cost component for these farming systems (IFCN, 2011). Low variable costs, 
and particularly low feed costs, have been reported as being important underlying causes of 
efficiency at the farm level which are responsible for achieving a high gross margin (Flaten & 
Giaever, 1998, cited in Hansen et al., 2005).  
The New Zealand dairy industry has historically adopted a strategy of “ broadly- targeted cost 
leadership” (Akoorie & Scott - Kennel, 1999, p.12 ) . This strategy has been used throughout the 
New Zealand dairy industry to compete against other players in the global marketplace. Under 
the cost- leadership strategy farmers decide on the combination of inputs they wish to use and 
they make  these decisions in order to minimise costs. In other words, if the input market is in 
perfect competition and under a cost- leadership strategy, the best technically feasible input -
output combination is that which minimise production costs. These assumptions are considered 
to be appropriate for countries where farmers cannot accurately predict the volume of output, 
take input market prices as given and do not have any control over milk pay- out such as NZ and 
Chile.  
Shapiro (1985 ) discussed the nature of the ‘fit’ between  a firm’s logistics system and its overall 
strategic focus. He considered three possible foci (product innovation, differentiation through 
customer service, and cost- leadership) that were broadly related to Porter’s (1998) typology of 
strategies. For each focus, Shapiro clarified the nature of the task that logistics needed to  
accomplish. Thus, under a strategy of cost leadership, the logistics function should seek to 
minimise costs while keeping service at an acceptable level. Further, the ways suggested t o 
control costs include,  but are not limited to: exploitation of economies of scale; centralisation of 
inventories to the extent possible; lowest cost routing; and automation of materials handling and 
order processing. The need for a strategic fit between a firm’s competitive strategy and its 
supply chain strategy was also espoused by Chopra and Meindl (2001).  
Chapter II: Literature Review 
75 
2.5.3.4 Competitiveness within groups.  
The competitiveness of a particular firm or a group of firms can be defined as the ability to face 
competition and to be successful when doing so. According to Latruffe (2010) , competitiveness 
may be the capability to sell products that meet demand requirement , such as price, quality, and 
quantity. In addition, this capability must result in profits over time and enable the firm to 
prosper. Competitiveness is a relative measure, and a broad concept with many definitions and 
disagreements about how to measure it precisely. However, there is consensus that in order to 
evaluate the competitiveness of a given farm or farming systems, it is necessary to have data 
from peers that allow benchmarking  to occur.  
Data envelopment analysis ( DEA) has proven to be an effective tool for analysing productive 
efficiency, and benchmarking the competitiveness of agricultural units and farming systems 
(Coelli, 1995). A wide range of benchmarking approaches and frameworks are described in the 
literature, which are often  based on non- parametric methods, such as DEA ( Bogetoft & Otto, 
2011). This method allows the estimation of individual farm technical efficiency (TE) relative to 
the maximum feasible production possibilities, given certain technology availability to a 
particular group.  
2.5.3.5 Competitiveness across groups.  
Benchmarking across groups that have  different production opportunities can still be achieved 
using DEA, although it needs to be taken further using a  metafrontier approach. O’Donnell et al. 
(2008) used this extended methodology to compare the TE of firms that could be classified into 
different groups. They used data drawn from FAO Statistics, containing observations from 97 
countries classified into four groups (Africa, America, Asia, and Europe) from 1986 to 1990. 
The data set comprised observations on just one output variable, re presenting the aggregation of 
185 commodities, and five input variables related to agricultural production. This study showed 
that African and European countries produced agricultural outputs under conditions that were 
more restrictive than in other regions of the world with an average metatechnology ratio that 
indicated they could at best produce only 89% of the output that could be produced using the 
unrestricted metatechnology. America could produce 91% at best, and Asia (including Australia 
and New Zealand) could produce 93% of the output that could be produced using the 
unrestricted metatechnology.   
Likewise, Moreira and Bravo - Ureta, (2010) used a single –output, multi– input approach and 
found different TE and metatechnology ratios for dairy farms in Chile, Argentina, and Uruguay. 
They also suggested that technical efficiency can be interpreted as a relative measure of 
managerial ability given certain technologies or productive environments. The findings from 
this study were that technology was indeed different for each country and that the average 
metafrontier efficiencies were 79.6%, 83.8%, and 91.4%, for Chi le, Argentina, and Uruguay 
Chapter II: Literature Review 
76 
respectively. However, metatechnology ratios were much closer with 65.8%, 72.8%, and 73.4%, 
respectively. One attribute of the whole pooled sample was that farms were all operating at a 
sub- optimal size.   
2.5.4 The International Farm Comparison Network. 
The International Farm Comparison Network (IFCN) is an organisation that benchmarks farms 
across the world. In 201 0, 90 countries that account s for 98% of the world’s total milk 
production were benchmarked (IFCN, 2011). The IFCN vision is to be “…the leading, global 
knowledge organization in milk production” (p. 8), while the mission advocates for the creation 
of “…a better understanding of milk production worldwide” (p. 8). The IFCN has been defined 
as a global network of dairy researchers from companies and other stakeholders in  the dairy 
supply chain with a focus on milk production. The IFCN particularly addresses the dairy chain’s 
resources and costs, emissions created, and the political challenges that come from producing 
milk  itself.  
In 2010, the typical farms’ analysis comprised 157 dairy farms from 60 regions and 49 
countries, which accounted for 87% of total milk production. Typical farms under the IFCN 
method represent those that produce the highest share of milk within their  region. Table 8 shows 
a comparison of selected measures for both Typical New Zealand and Chilean dairy farms. Note 
that these figures are from actual farms representing each country’s distinct dairy farming 
system and are not national averaged values. 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
Chapter II: Literature Review 
77 
Table 8  
Comparison of Representative Dairy Systems  
 
ECM kg = Milk kg corrected to 8.5% MS; Stocking rate: 1 Livestock unit = 650 kg LW  
Source: IFCN (2011)  
The IFCN has specific methods for the calculation of major KPIs. Net farm income (NF I) is 
obtained by deducting total expenses from total cash income and then making the necessary 
non- cash adjustments. Milk price is calculated by dividing the milk returns by the total 
kilograms  of ECM sold. The costs calculation for the dairy enterprise consists of the following  
costs: milk production, raising replacement heifers, and home grown feed production cost 
and/or cost of feed purchased for dairy cows and replacements. Total costs are composed of 
expenses included in the profit cash and non- cash costs, including depreciation and opportunity 
costs for factors of production such as family labour, own land, and own capital. Both returns 
and total costs are expressed in US$ and per hundred k ilograms of Energy Corrected Milk 
(ECM).  Cost of production includes farm working expenses, wages, and debt servicing, but 
exclude s management wages and depreciation. The cost of labour (US$/ 100 kg milk) is similar  
Country NZ Chile
Location Waikato Region X
Herd size (cows) 348 421
Stocking rate 3.3 1.0
Labour units 2.3 7.1
Kg MS/cow 394 685
Kg MS/ha 945 753
Total assets (US$/100 kg) 126 48
Total Capital cost (US$/100 kg) 2.4 0.5
Equity 70 100
Milk price (US$/100 kg) 34 35
Cost of Production (US$/100 kg) 34 27
Average wages on farm (US$/hr) 18 6
Cost of labour (US$/100 kg) 6 6
Labour productivity (kg/hr) 290 100
Chapter II: Literature Review 
78 
in both countries despite the number of labour units being more than three times higher in Chile 
than in New Zealand. This is explained by labour productivity which is almost three times 
higher in NZ . These considerable differences in labour units could be afforded in Chile because 
the average wage was US$ 6 per hour wh ile in NZ th is was US$ 18 per hour . 
The entrepreneur’s profit measures the financial sustainability of the typical farm and shows the 
total profitability of the dairy enterprise. It is calculated by subt racting the opportunity costs for 
factors of production as above from the NFI. The return to labour, in turn, indicates how much 
profit an employee or farmer generates per hour of farm work . It is calculated by adding wages 
for hired labour and the opportunity costs of family labour to the entrepreneur’s profit, and 
dividing this partial result by the total hours of labour used for the dairy enterprise. It is driven 
by the wage level of the country, the labour input on the farm, and the economic performance of 
the farm. Average wages on the farm include the gross salary plus any social costs the employer 
has to cover. For hired labour, the cash labour costs effectively incurred are used, and for unpaid 
family labour, the average wage rate which a hired worker would be paid to perform a similar 
task in the same region is u sed.  
The IFCN also studies the sustainability of dairy farming worldwide . Seven areas, where 
economic, social and environmental sustainability  stand out, are considered. Table 9 presents 
the areas of importance, their indicators and their sustainability scale numbers (IFCN, 2011) . 
Table 9  
Sustainability Areas and Indicators for Dairy Farms Worldwide 
 
Considering the indicators in Table 9 and the data available in Table 8 the typical NZ farm is in 
a more risky situation than the Chilean farm. This is mainly due to a lower entrepreneur’s profit 
and higher intensification. Also of interest is noting that among the benchmarks chosen to 
measure sustainability, many of them are commonly used and of ea sy quantification .  
AREA INDICATOR UNIT GOOD BAD
ECONOMIC Entrepreneur's profit US$/100 kg milk ECM > 1 < -1
RISK Operating profit margin % > 15% < 12%
COMPETITIVENESS Return to labour % > 110% < 90%
EMISSIONS Carbon footprint g CO2/100 kg milk ECM 0 - 130 > 140
RESOURCE USE Water footprint L/kg milk ECM 0 - 1800 > 2000
INTENSIFICATION Stocking rate LU/ha < 1 > 1.2
JOB CREATION Employees /100 ton of milk Labour units >1.1 <0.9
Chapter III: Method 
79 
3. Method. 
 
F arming systems and their available technology are shaped by natural resources;  the physical, 
economic, and social environment;  policies;  the characteristics of the labour force;  and the 
existing market for potential outputs. Hence, it is expected that f armers in different regions or 
countries face different production opportunities and make choices from different sets of 
feasible input- output combinations. This chapter presents the research method used in this study 
which was in accordance with these concepts that is framed by a systems approach but makes  
use of predominantly quantitative methods.  
3.1 Research Strategy 
To answer the research questions posed in this research,  a variety of methods were used after 
reviewing the literature which provided the background to better understand the t hree dairy 
databases used in this research. The first part of the study was straightforward and compared 
two dairy farms from NZ and Chile . The second part of the study required g rouping and 
classifying observations from the country- specific dairy bases. Country- specific classifications 
for production systems in NZ and Chile  were developed on an annual basis for NZ  (June - May )  
and Chile (January- December ) . After the existing observations were classified into the most 
appropriate ‘ system’  based on the review of supplementary feed and regional information, a 
comprehensive measure of performance was found for individual observations by applying an 
efficiency analysis separately to each class. Corresponding efficiency scores were then 
associated with the DMU in the  original data set and a regression method was applied to 
determine the relative importance of independent variables in determining efficiency in each 
dataset. Finally, an overall measure of efficiency combining  both datasets was calculated for 
each observation to give the metafrontier for the final comparative analysis. 
3.2 Materials and Data Analysis Methods 
This section introduces the sources of data and the methods applied to the present study. The 
International Farm Comparison Network (IFCN) provided data for  the farm to farm comparison 
from its world- wide dataset for 2011. Data  was analysed using specific features within the TIPI -
CAL 5. 3 spread sheet. The country- specific datasets were “ DairyBase ”  from NZ and 
“ TodoagroBase”  from Chile. R statistical software was used to analyse data from these two 
datasets. Computerised statistical packages such as R are time savers for  both quantitative and 
qualitative research (Patton, 2002). R software was used for the core part of this research, to 
calculate the efficiency scores and to run the regression that identified the key performance 
indicators. The R statistical software system for Windows was used because of its versatility 
Chapter III: Method 
80 
and flexibility in undertaking a range of tasks with large datasets, including both statistica l and 
non- statistical analyses. The R codes 12  used rpart and lpSolve packages.  The data analysis 
combined input oriented D ata E nvelopment Analysis (DEA) , followed by R ecursive or 
Regression  Partitioning Trees (RPTs) , descriptive statistics, and a metafrontier approach, which 
together provided the means for extr acting relevant conclusions from the data available.  
3.2.1 DairyBase and TodoagroBase overview. 
Databases typically record historical information through systematic collection and, if required, 
conversion of quantitative and qualitative observations into numbers or codes. DairyBase and 
TodoagroBase are independent data collections of dairy farm data, annually  produced at the end 
of each season or year. These datasets created by industry- owned and private agricultural 
advisory services in NZ and Chile respectively , were used in this research as secondary data 
sources. Each year, f arm businesses are included on a voluntary basis; therefore , the cooperating 
units are self- selected rather than a random sample. The information, validated at the time of 
entering, has been made available for research purposes. Each database contained  unbalanced 
panel data as the total number of participants varied throughout the seasons, years, farming 
systems, regions, and countries.  
DairyBase included information for five consecutive years:  2006/ 07, 2007/ 08, 2008/ 09, 
2009/ 2010, and 2010/ 11, while TodoagroBase comprised information for five consecutive 
production years:  2007, 2008, 2009, 2010,  and 2011.  Participants represented diverse locations 
across NZ’s North and South Islands,  and Central- South Chile. The datasets only shared a few 
common variables but most of those were expressed in different units ( for example, milk output 
was expressed as kilograms of milk solids  for DairyBase and as litres of milk  for 
TodoagroBase) . F inancial variables and most ratios were often exclusive to one or the other  
dataset. Shared and unshared variables acq uired relevance at the final stage of this study, 
particularly in the cross- country comparison. For most analyses, the datasets were analysed 
separately and the original data were preferred; d ata conversions were only made when 
necessary.  
3.2.1.1 Data quality and limitations. 
Data quality and appropriateness are  just as important as the analytical techniques themselves  
because regardless of how powerful a technique may be, it cannot overcome problems that 
fundamentally reside in the data (Coelli  et al.,  2005) . I naccuracy of production data, a chronic 
problem to be aware of, may or may not be evident to the analyst . A priori , the quality  of the 
available data was considered high because farmers volunteer their data to the databases, and 
therefore, have a vested interest in both the collection and the use of accurate data.  
                                                          
12 The R codes written by Lecturer and PhD in Statistics, Jonathan Godfrey (Statistics and bioinformatics 
group Massey University) can be found in Appendix 3 
Chapter III: Method 
81 
DairyBase supplied New Zealand farm- level data consisting of information on more than 200 
variables and ratios ( Appendix 1) . Each observation consists of a set of records; the database 
originally contained 2,616 sets of financial records and 657 sets of physical observations 
records throughout the period. A balanced panel would have included the same 523 farms each 
year and totalled 2,615 observations. However, the dataset was unbalanced, with only 39 farms  
consistently surveyed every year across the whole period. A similar situation was found among 
Chilean observations, and although the lack of balance  in both databases was acknowledged and 
managed accordingly, the small  proportion of farms present each year was a limitation in this 
research. This posed the question of how  much of the inter- annual variation found was 
attributable to the relative ‘weight’ of different f arms present each year. Non - random, missing 
data on DairyBase was not a concern because the sets of observations were complete for the 
variables of interest.  
TodoagroBase supplied Chilean farm- level data consisting of information on more than 90 
variables and ratios ( refer to Appendix 2 ) . The database originally contained 1082 sets of 
combined financial, economic and physical records  for a range of farming systems, from 
exclusively pasture- based farms to supplement intensive systems feeding up to 4, 734 k ilograms 
of supplement per cow per year. After reviewing the literature, a  cut- off value of 2,700 
kg/cow/year  fed was used to define a pasture- based system;  this resulted in 1,028 pasture- based 
cases being available for further study (95% of the original sample) . Missing data was a  concern 
when it involved important variables for the dairy enterprise. Missing data  was an issue, and  
included observations on depreciation, and remuneration among others. This was important in 
calculating ratios or as part of operating ex penses. The value of assets was calculated backwards 
using the original profitability value but this resulted in another issue which was incoherent 
data. This issue was a concern when involved variables used for the efficiency analysis because 
implied a lower confidence in the data. Case deletion was applied in a few cases. However, the 
most drastic measure was complete variable exclusion with the assets, whose calculated value 
was not used in the main DEA analyses.  
 
In summary, t here were a few concerns which prevented the research from reaching fully 
generalizable  conclusions regarding a respective class or farming system. There were two major 
issues:  unbalance and non- randomness in the two datasets. Due to non- random sampling it was 
uncertain whether the samples were fully representative of the NZ and Chilean  pastoral dairy 
farms populations. The second concern was that presumably, in both NZ and Chile, the samples 
comprised above- average- managed dairy farms. In NZ , available information showed that the 
DairyBase sample had been above average in indicators of interest. According to Jiang (2012) ,  
differences in production have traditionally existed  between DairyBase farms, and the national 
Chapter III: Method 
82 
statistics. While DairyBase samples have traditionally consisted of farms with management and 
performance above the average, there has not been  a big difference in the direction and 
magnitude of the bias between regions. Therefore, since a large sample does not necessarily 
prevent bias (Hawl ey, 2010) efficiency estimates and other outcomes should be interpreted  with 
care as they may not reflect the true situation in the respective farming systems.  
3.2.2 Review of methods. 
Quantitative research permits the diagnosis of historical and current situations, and developing 
trends (Taylor , 2005) . In order to achieve the objectives of this research, this study used a  
combination of predominantly quantitative  methods, both parametric (descriptive statistics) and 
non- parametric, enabling valid, obje ctive descriptions and conclusions. This section describes 
the data envelopment analysis models, the recursive partitioning technique , and the metafrontier 
approach in the order in which they were applied.   
3.2.2.1 DEA models. 
DEA owes its name to the way the  frontier envelops observations, which are typically 
represented by a performance score that ranges between zero and one  (C ooper, S eiford, & Tone, 
2006) . In DEA terminology, all observations are  so- called decision making units or DMUs. The  
envelopment identifies those efficient DMUs ( those with a score=1) and locates them on the 
frontier; in that way they serve as benchmarks to rank all the inefficient DMUs . The term 
‘ frontier’  is used to emphasise the fact that the function gives the maximum output that  is 
technologically feasible. The maximum value of one means that the DMU has achieved  
maximum output with minimum expendab le inputs or resources; in contrast, a minimum of zero  
means nothing is achieved by using certain inputs (which is hard to find in practice) . DEA uses 
mathematical optimisation or linear programming (LP)  to calculate efficiency scores, based on 
one or more outputs and commonly utilises multiple inputs.  
 
DEA can operate on either input or output orientation, but both models estimate the same 
frontier, identifying exactly the same set of DMU’ s as being efficient (Farrel, 1957) . In contrast, 
the efficiency scores associated with inefficient DMU’s may  differ between the two methods 
(Coelli, 1997). The frontier may also be determined assuming a transformation of inputs into 
outputs under constant returns to scale (CRS) proposed by Charnes et al. (1978) , or  variable 
returns to scale (VRS) reported and explained by Banker, Charnes and Cooper (1984)  and later 
by Fare, Grosskopf , and Lovell, (1994). Under the latter assumption, decreasing (DRS)  and 
increasing returns to scale (IRS)  may occur. DRS  takes place when a n increase in inputs results 
in a less than proportional increase in output; conversely, IRS happens when increased inputs 
lead to a greater than proportional increase in output. In this research, DEA models were set up 
to use VRS and an input- oriented approach. The use of VRS is customary in agriculture  and 
Chapter III: Method 
83 
commonly preferred in related studies. The input orientation was used in this study since a 
manager has greater control over inputs than over outputs, which are influenced by external 
factors such as climate and prices.  
 
The selection of Regression Partitioning Tree as the best possible method to use in a second 
stage DEA was the end result of an iterative process that started looking at linear models. After 
reviewing the literature, linear models were chosen as it was necessary to incorporate multiple 
predictors to predict efficiency and the Gauss - Malcom theory behind them is robust and well 
understood (Hazelton, pers. com.). The multi - dimensionality in the dataset required a method 
capable of handling several parameters; therefore, multiple linear regression (MLR) was tested 
to regress efficiency scores attached to the original sets of observations previously cleaned of 
trivial variables. The NZ dataset was used to test the parametric model (MLR) behaviour, which 
was set up to capture main effects and also interactions which were considered unavoidable. 
After a primary cleaning of trivial variables, the number of variables was still large (81) and 
hence MLR failed to achieve the expectations. According to Strobl, Malley, and Tutz (2009), 
standard regression methods commonly fail in these conditions and recursive partitioning trees 
are an alternative approach. 
3.2.2.2  Regression or Recursive partitioning tree.  
Regression partitioning is a technique that produces either classification or regression trees, 
depending on whether the dependent variable is categorical or numeric, respectively  (Strobl et 
al., 2009) . In this study DEA efficiency scores were used as the response variable in a second 
stage approach, to find the best summary set of indicators that explains a dairy system’s 
performance. Since the databases provided many more variables than were required to build 
such a set, and following the principle of parsimony, an effective reduction of the 
multidimensionality of the datasets was pursued. In a second stage RPT, DEA scores from the 
first stage were regressed on discretionary and non- discretionary inputs in a country- specific but 
common- to- all- groups analysis. This approach not only generated a ranking of variables, but 
also revealed the interaction between independent variables, which it w as not possible to do 
with other forms of multivariate analysis (Curran et al., 1993).  
 
RPT generates a  flow- chart structure or tree which has a root node, secondary and tertiary  nodes 
and so on, and finally terminal nodes called leaves. Trees reveal the variables’ ranking of 
importance with the root node representing the top level indicator. Each node provided  
information on the splitting criteria (variable’s  name)  and the threshold value above or below 
which a class is defined. Each leaf presents the cla ss’ s mean efficiency, and the number of 
DMUs included  (n). To reach a leaf node, the pathway starts at the root node. From each non-
Chapter III: Method 
84 
leaf node the procedure splits the data to the left, or to the right , based on the condition and 
value of a certain variable, whose index is stored in the observed node . I f the observed value 
satisfies the condition, the procedure goes to the left, otherwise, to the right.  
 
The partition is created such that observations with similar efficiency values are grouped by a 
set of rules. Once a rule is selected and splits a node into two, the same process is applied to 
each “child” node which is why RPT  is called a recursive procedure (Witten & Frank , 2000).  In 
general, s plitting and stopping follow the principle of impurity reduction, whereby each split 
results in a child node which is more pure than the parent node. In each node , the variable that i s 
most strongly associated with the response variable ( y) is selected for the next split  (Strobl  et al.,  
2009) . Finally, a  response class is predicted in each terminal node of the tree (leaf  or class) . 
Splitting stops when the software detects that no further gain can be made, or some pre - set 
stopping rules are met. In this research a t each node the recursive procedure stopped splitting 
the node further in any of the following cases:  
1. The number of observations in the node was less than the specified value (n = 50);  
2. All the samples in the node belonged to the same class (or, in the case of regression, the 
variation was too small), because it was  not statistically possible to split the node 
further;  or 
3. The best split found did not give any noticeable improvement compared to a random 
choice. 
The pathway to any particular leaf showed the main effects and the interactions between 
variables. The RPT operates as a reduction, classification, and prediction tool for observations 
meeting certain set of rules. A key property of this technique is  that it is possible to compute the 
importance or relative decisive power of each participating variable (Strobl et al., 2009) . A 
variable may be present more than once, and this only reinforces its importance. To avoid the 
analytical or the prediction procedure getting stuck in a  certain node,  the technique also 
provides so- called surrogate splits, in addition to the best ‘ primary’  split (Strobl et al., 2009).  
These surrogate splits can be used for observations with incomplete records and yield nearly the 
same results as the primary splits. If a farm fulfils all the node or partitioning conditions on a 
particular pathway, then it is likely with 95%  of confidence that its efficiency will be mean 
efficiency at leaf +/ -  deviance.  
 
In summary, RPTs were expected t o identify important benchmarks or indicators defining 
efficiency classes and also to present a number of pivotal values to expose the pathway or 
combination of conditions useful to predict technical efficiency. These pivotal values were the 
thresholds indicating a single feature, which may split the dataset such that the variance of 
Chapter III: Method 
85 
subsamples in each partition is less than their variance in the whole dataset. The pathway from 
the root to a leaf node was a sequence of decision - making step s, and the leaf node function was  
evaluated to provide a numerical prediction of the averaged value of efficiency for each leaf.  
3.2.2.3 Metafrontier envelope. 
The measurement of efficiency across groups is only possible if efficiency is measured relative 
to a common metafrontier which is the boundary of a metatechnology or unrestricted 
technology that envelops the restricted group frontiers ( Battese & Prasada Rao, 2002; 
O’Donnell et al. 2008) . This study used a metafrontier to compare dairy farm efficiencies in NZ 
and Chile. By definition, a metafrontier envelops all farms and all group frontiers. In t his 
research, the main reason for using a metafrontier envelope was to provide comparable 
efficiency scores for businesses operating in different environments existing either across 
seasons, regions and countries. The metafrontier also enabled the technology gaps to be 
estimated for firms under different technologies, relative to the potential technology available to 
the industry as a whole (Battese, Prasada Rao & O’Donnell, 2004). A  priori , t he mean technical 
efficiencies relative to the metafrontier are always expected to be smaller than t hose calculated 
from the group frontiers. The difference between the group scores ( E1)  and the metafrontier 
scores ( E2) represented  the restrictive nature of the particular environment. There is a gap 
between a farm and the group frontier, which represents the differences in performance between 
the efficient farms (TE = 1) and the particular DMU , and is called ‘ d’ . Then, the re is another 
gap ‘e’ between this DMU  and the metafrontier. Both indi cators may be used to calculate the 
metatechnology ratio as the relationship d/e ( Battese  & Prasada Rao, 2002; O’Donnell et al. 
2008) . This ratio is of considerable interest because it measures the potential improvement in 
performance resulting from changes in the production environment. 
3.2.3 Procedure: the data analysis. 
The data analysis comprised five key stages including sampling, grouping, efficiency analysis, 
RPT , metafrontier , and benchmarking. The procedure in more detail is presented in 
chronological order as follows:  
1. Analysing IFCN data relevant to the countries of interest, NZ and Chile ;  
2. Developing a classification system for each country ’s  data which yielded 27  NZ and 10 
Chilean classes using a combination of season/ye ar, region and supplements  use;  
3. Running the DEA (VRS) separately for each class which created a DEA score for each 
observation so that the multiple effects of season/year, region and system were  
minimised in the complete data set for that country;  
4. Obtaining DEA scores ( 0- 1)  for each observation within each class and adding these 
scores to the original databases;  
Chapter III: Method 
86 
5. Testing stepwise MLR using DEA scores as response variables  with all non- trivial 
variables contained in the datasets as predictors;   
6. Looking for alternatives to M LR , which was discarded due to the high dimensionality 
of the dataset plus multi- collinearity13  and missing data issues;  
7. Substituting MLR with  RPT in the second stage, using efficiency scores as the response 
and input variables as predictors;   
8. Identifying country- specific sets of major KPIs  for NZ and Chile  based on RPT outputs ;  
9. Describing the samples using descriptive statistics on the identified KPIs;  
10. Standardising the output and input variables to be used on the metafrontier envelop to a 
common currency when necessary; then running the VRS DEA code in R  on the 
combined NZ - Chilean data using these variables.   
11. Summari sing the differences between farming systems, within , and between countries, 
using selected variables with the highest explanatory power , and the metafrontier 
scores.  
3.2.3.1 Sampling. 
S ince the research is based on existing datasets, this section actually r efers to the original and 
final configuration of the samples after cleaning the data. The samples in the databases 
consisted of dairy farms across the countries over five agricultural years, where the sampling 
unit was the dairying enterprise. The original number of observations contained in the NZ and 
Chilean datasets for each season is shown in Table 10.  
 
 
 
 
 
 
                                                          
13 This is a term commonly applied to the statistical phenomenon in which two or more predictor 
variables are highly correlated. In this situation the coefficient estimates may change erratically in 
response to small changes in the model or the data (Adler & Golany, 2001).  
Chapter III: Method 
87 
Table 10  
Original NZ and Chilean Samples 
 
The NZ data w ere allocated to eight NZ regions based on the DairyBase clustering criterion. 
The number of observations for the period 2006- 11 is shown in brackets f or each region from 
the top of the N orth Island to  the bottom of the S outh Island  as follows: Northland ( 268 ) , 
Waikato ( 864 ) , Bay of plenty ( 215 ) , Taranaki ( 292 ) , Lower North Island ( 182 ), West C oast-
Tasman ( 226) , Ma rlborough- Canterbury ( 324 ) ,  and Otago - Southland ( 243 ). Waikato farms 
represented 33% of the original sample , which shows the importance of this region to the NZ 
dairy industry. Similarly, the southern centra l dairy territory in Chile consisted of five agro-
ecological zones: Llano Central Sur ( 404), Ribera de Los Lagos ( 308) , Litoral Sur (82 ) , IX Sur 
and XIV Norte  (2 21 ) , and Los Angeles- Angol ( 66) . The first two regions comprised more than 
66% of the total data , thus demonstrating  their significance to the Chilean dairy industry. 
Observations  were excluded from the raw data  for the following reasons:  
• Business struc ture other than owner operators (NZ data only) ;  
• Extreme outlier data ;  
• Important missing data that could not be inferred or interpolated;  
• Incoherent data  (Chilean data only); and 
• Systems other than pasture- based, that is using above 2700 14  kg supplement per cow per 
year (Chilean data only ) . 
 
O nly data for owner operators were analysed in the NZ data to  attain a homogeneous possible 
sample which was representative of this type of business which is the most prevalent business 
                                                          
14 Equivalent to 45% of the diet based on supplements (year round). 
Season DairyBase TodoagroBase
2006/07 625 259
2007/08 628 245
2008/09 497 237
2009/10 567 205
2010/11 297 135
Period 2614 1081
Chapter III: Method 
88 
structure in NZ. Table 11  shows the number of observations included in the analysis at the end 
of this cleaning process.  
Table 11  
Number of DMUs Included in the Analysis per Year and Country 
 
3.2.3.2 Variable correction and creation. 
For the purpose of this study, it was appropriate to adjust the variables for inflation that 
occurred throughout the period even though inflation rates for the period were negligible in both 
countries. Since TodoagroBase provided Consumer Price Index ( CPI )  corrected values for the 
year 2011,  variable correction was only needed for the NZ dataset and it was carried out using 
the same criterion to be equivalent to the Chilean index 15  and an appropriate CPI coefficient for 
each year published by Statistics New Zealand  ( 2012). The general CPI for the  ‘all groups’  
index was selected. The available price deflators were as follows: 20060/ 07 ( 1.119) ; 2007/ 08 
( 1.067 ) ; 2008/ 09 ( 1.065 ) ; 2009/ 10 ( 1.055); and 2010/ 11 (1).  
The next  step was to check TodoagroBase and to decide which variables in the NZ data should, 
and could,  be created based on the available information. The following set of variables was 
created:  
-  Farm Working  Expenses ( FWE) and ratios (F W E /cow; F W E/ ha) ;  
-  Operating Expenses (OpEx) and ratios ( OpEx/cow; OpEx/ ha) ;  
-  Operating Profit (OpProf it) and ratios (OpProfit/cow; OpProfit/ ha) ;  
                                                          
15 A more correct adjustment would have been to use the dairy Producers Price Index as the adjuster 
which shows that on-farm inflation on dairy farms over the period was generally higher than measured 
by the CPI, and hence using the CPI under-deflated the figures. 
Season DairyBase TodoagroBase
2006/07 619 245
2007/08 614 231
2008/09 496 231
2009/10 564 197
2010/11 295 125
Period 2588 1029
Chapter III: Method 
89 
-  Milk price:Feed  price ratio;  
-  Operating Profit margin (OPM) ;  
-  Cash Operating surplus ;  
-  Discretionary Cash + Tax ;  
-  Total Assets; and  
-  Asset Turnover ratio. 
3.2.3.3 The NZ classification system. 
Battese et al.  (2004)  pointed out that farms operating in different regions or under different 
technologies are not strictly comparable unless a special methodology is used to analyse the 
data. However , in such cases, the observations can be grouped according to similarities and 
differences in technology,  and other conditions among the nominated periods or regions 
(Haghiri, Nolan, & Tran, 2004). For instance, t he seasons 2007 / 08 and 2010/ 11 both had 
exceptionally high  milk prices as shown in Table 12.  
Table 12  
Estimated Seasonal Prices for Output (MS) 
 
 
The physical information contained in DairyBase also suggested that farmers responded to price 
information by feeding more supplements to the herd in higher milk price years , which led to 
farming systems shifting towards greater use of imported feed. Physical input was also used to 
improve the understanding about dairy farming systems in the different regions. The most feed-
Season Milk ($/kgMS) % Average
2006/07 4.15 0.69
2007/08 7.36 1.22
2008/09 5.22 0.86
2009/10 6.16 1.02
2010/11 7.36 1.22
Period 6.05 1.00
Estimated price (NZ$)
Chapter III: Method 
90 
intensive region in NZ, measured as supplement imported per hectare, was found to be 
Canterbury, followed by  the L ower North Island and Waikato  regions. The least intensive 
region was Tasman- West Coast , closely followed by Northland and Taranaki , which are the  
regions with the highest rainfall averages. The Dairy NZ Farming system s (reviewed in Chapter 
II)  in the NZ dataset had a mean of 2.9  and a median of 3, with the most representative regions 
being Bay of Plenty  and Southland. The use of imported feed was positively associated with  
stocking rate, measured as both Cows/ha and LWT/ha. The correlation was  significant at the 1% 
Pearson coefficient, so is  moderately high in both cases ( 0.629 and 0.642 respectively ) . In the 
period under study (June  2006 to May  2011) , farming systems were characterized by quite 
distinctive indicators, as presented in Table 13.  
Table 13  
Averaged Indicators Showing the Relevance of Grouping into Farming System 
 
The clear rainfall gradient shows that the volume of rain is negatively associated with the use of 
imported feed, and hence,  the farming system in place. In contrast, the farming system is 
associated positively with the use of nitrogen (kg/ha)  and feed grown (ton/ha) , as well as with 
production of milk solids per unit of labour ( kgMS/FTE).  These findings supported the need to 
recognise the year and region effects, as well as their interactions with use of imported feed.  
Therefore, the samples were grouped by year and then classified into several sub- samples 
representing location and farming system, ensuring as much homogeneity of exogenous 
conditions as possible within sample group.  
Indicator 1 2 3 4 5
Rainfall (mm) 1764 1766 1301 1346 1187
Nitrogen (kg/ha) 70 116 202 169 169
Hay (%) 20 17 13 11 7
Supplement (ton/ha) 0.23 0.71 1.45 3.19 5.44
Pasture&Crop (ton/ha) 10.69 11.31 12.1 12.79 13.61
MS/FTE 45811 48670 53997 60045 62570
Farming system
Chapter III: Method 
91 
There was an iterative process that guided the classification systems in both countries. In order 
to decide on the best categorization  system for existing differences in technology across NZ,  the 
three approaches to nesting were evaluated: Regional consisting of regions (8) across five years 
and five DairyNZ farming systems;  Seasonal consisting of years across region and systems; and  
Farming systems regardless of year or region. Three intensification levels (IL) were made by re-
coding DairyNZ farming systems so that farming systems 1 and 2 composed LOW, 3 
represented MEDIUM , while 4 and 5 composed HIGH  intensification levels. After the 
evaluation process, observations contained in DairyBase were classified according to year of 
record, region and intensification level. The classification process is partially explained in this 
chapter and complemented by information contained in Appendix 4. Finally, t he N Z 
classification system consisted of 27 classes (Table 14), with half of them being a combination 
of year, region, and farming system, and the other half, a combination of year and region. The 
latter can be observed for those classes where the number of observations did not allow for a 
further split at the intensification level criterion. The NZ classification system is presented in 
Table 14 . 
A priori, concerns regarding all the mentioned classification systems were identified. One 
concern when categorising by year was whether or not the underlying technology changes over 
time in a relatively short period. I t was found that some particularly responsive variables 
showed both intra-  and inter- annual variation. These variables included: input and output prices, 
pasture growth, use of imported feed, and rainfall, which were highly variable between seasons. 
A second concern was that the dataset was unbalanced with only 39 f irms present across all 
seasons, which prevented a fully meaningful comparison between seasons. Similarly, i ssues 
around categorisation by Dairy  NZ farming systems were recognised. Previous empirical work 
had shown little or non- significant differences between systems in efficiency, which were 
explained by greater variation in intra -  than in inter- feeding systems. In addition, other factors  
were prone to affect the farming system, meaning that a variety of technologies could be 
represented by the same feeding system. Also, there were difficulties when categorising by 
region alone because of large intra- region variability.  
 
 
 
 
Chapter III: Method 
92 
Table 14  
NZ Classification System 
 
Season Class (code) Regions Farming Systems
2006-07 67 16 Northland-TasmanWestCoast All
67 78 Canterbury-Otago All
67 2345LOW Bay of Plenty, Waikato, Taranaki, Lower NI 1 and 2
67 2345MEDIUM Bay of Plenty, Waikato, Taranaki, Lower NI 3
67 2345HIGH Bay of Plenty, Waikato, Taranaki, Lower NI 4 and 5
2007-08 78 16 Northland-TasmanWestCoast All
78 7 Canterbury All
78 8 Otago All
78 2345 LOW Bay of Plenty, Waikato, Taranaki, Lower NI 1 and 2
78 2345MEDIUM Bay of Plenty, Waikato, Taranaki, Lower NI 3
78 2345HIGH Bay of Plenty, Waikato, Taranaki, Lower NI 4 and 5
2008-09 89 16 Northland-TasmanWestCoast All
89 78 Canterbury-Otago All
89 2345 LOW Bay of Plenty, Waikato, Taranaki, Lower NI 1 and 2
89 2345 MEDIUM Bay of Plenty, Waikato, Taranaki, Lower NI 3
89 2345 HIGH Bay of Plenty, Waikato, Taranaki, Lower NI 4 and 5
2009-10 910 16 Northland-TasmanWestCoast All
910 7 Canterbury All
910 8 Otago All
910 2345LOW Bay of Plenty, Waikato, Taranaki, Lower NI 1 and 2
910 2345MEDIUM Bay of Plenty, Waikato, Taranaki, Lower NI 3
910 2345HIGH Bay of Plenty, Waikato, Taranaki, Lower NI 4 and 5
2010-11 1011 16 Northland-TasmanWestCoast All
1011 78 Canterbury-Otago All
1011 2345LOW Bay of Plenty, Waikato, Taranaki, Lower NI 1 and 2
1011 2345MEDIUM Bay of Plenty, Waikato, Taranaki, Lower NI 3
1011 2345HIGH Bay of Plenty, Waikato, Taranaki, Lower NI 4 and 5
Chapter III: Method 
93 
The final combined classification approach was useful in overcoming the heterogeneity faults in 
the dataset as shown in Figure 7  for total operating expenses (NZ$ thousands) .   
 
Figure 7: Averaged farm operating expenses within NZ classes. 
As Figure 7 shows there was huge variability across classes. The greatest variability was found 
when comparing across regions, especially when Canterbury ( region 7) and Southland ( region 
8) are taken into account. The year effect was also evident with operating expenses increasing 
steadily from 2006/ 07 to 2008/ 09, then stabilising in 2009/ 10 a nd decreasing slightly in 
2010/ 11. The classification system normalised these and other effects, such as asset value and 
land quality (refer to Appendix 4). However, it  was still complex, comprising  27 combinations 
of year, region, and intensification leve l. By  performing DEA for each of these combinations 
separately, this study recogn ised differences in the environment (year effect) , and farm specific 
characteristics related to region and intensification level. This provides more relevant efficiency 
estimators as farms are scored only with respect to their peers. 
3.2.3.4 The Chilean classification system. 
A similar approach categorizing by year  and farming systems to give a comparable 
classification system was used for the Chilean grouping criteria. Observations contained in 
TodoagroBase were grouped into a classification system based on a combination of year and 
use of supplement per cow. This combined approach was similar to that used for NZ  but was 
restricted by sample size, and thus, could not discriminate by regions. However , it was 
considered appropriate and sufficient for the Chilean dataset. Two intensification levels 
(MEDIUM and HIGH) were drawn  by grouping according to the use of supplements looking at 
the variable ‘Supplement kg/Cow’ . The critical value for MEDIUM was below 2000  
Chapter III: Method 
94 
kg/cow/ year fed which was arbitrarily selected to generate classes of similar size. The final 
classification system is presented in Table 15 .   
Table 15  
Chilean Classification system 
 
 
3.2.3.5 Class-specific input oriented DEA. 
After developing these classification systems, each  country- specific class was analysed to 
calculate efficiencies using input oriented DEA. All input and output variables were read by a 
DEA code in R software . This code allowed for CRS, VRS , or both models to be run and 
efficiency scores to be calculated in any of the mentioned model orientations. However, in this 
study, only VRS was used.  The first step to set up for DEA is to classify  variables as inputs or 
outputs,  such that input variables are causal factors to the outputs. DEA assumes precise 
knowledge regarding which variables are  outputs and inputs. The output selection were 
straightforward: the milk  physical production indicators, exp ressed as kilograms of milk solids 
in NZ or litres of milk in Chile, and the financial indicator gross farm revenue were used 
separately in a single- output approach. These outputs were user- defined and were used so that 
each of them captured different effects.  
 
Year Class (code) Regions Farming Systems
2007 2007 MEDIUM All 3 and 4
2007 HIGH All 5 or more
2008 2008 MEDIUM All 3 and 4
2008 HIGH All 5 or more
2009 2009 MEDIUM All 3 and 4
2009 HIGH All 5 or more
2010 2010 MEDIUM All 3 and 4
2010 HIGH All 5 or more
2011 2011 MEDIUM All 3 and 4
2011 HIGH All 5 or more
Chapter III: Method 
95 
I nput variables were classified into controllable and non- controllable variables. The selection of 
inputs used is very important in efficiency studies and therefore, variables were selected with 
care. First, DEA input variables were selected from the sub - set of controllable variables and,  
second, they were consistent with the phenomena they were supposed to capture, for example  
milk production  in the physical analyses (using either milksolids or litres depending on the 
country) . Several inputs were meaningfu l from a productive standpoint, but some were omitted 
following the principle of parsimony and to avoid multi- collinearity16 . The first principle, also 
called simplicity, states that large numbers of variables in DEA may lead to  over- estimation of 
relative efficiency (Adler &  Golany , 2001) . In this study , the ratio between numbers of DMUs 
and total inputs plus outputs was kept much higher  than 3:1 which, according to Bowlin (1998) 
and Rouse et al. ( 2010) , ensures reliability of the results. Golany and Roll (1989) state  that the 
number of DMUs  should be at least twice the number of inputs and outputs considered and 
Dyson et al. (2001) recommend a total of two times the product of the number of input and 
output variables. These conditions were met in this study and all finally selected inputs were 
standard production factors including capital, labour, operating costs and cows as presented in 
Tables 16 . 
Table 16  
Inputs Included in the Output DEA Models by Country 
 
 *Full time paid and unpaid labour Equivalents: 1=2,400 hours of work a year  
The variables presented in Table 16 were selected because they represent different aspects of the 
farming systems. Capital is a standard candidate variable in farming that typically includes land, 
livestock, buildings and plant . It may be represented by equity  plus liabilities,  as was done for 
NZ , or by  the sum of various assets, as could have been  done for Chile if data was more 
                                                          
16   Inputs that correlate highly with one another were eliminated based on correlation analysis. 
NEW ZEALAND CHILE
Cows (peak number of cows milked) Cows (mass cow; all cows on farm)
Labour (FTE*) Supplement (kg/cow)
Operating Expenses (NZ$) Wages paid (CLP $)
Opening Equity (NZ$) Operating Expenses (CLP $)
Opening Liabilities (NZ$)
Chapter III: Method 
96 
reliable. However, incoherent data made impossible the inclusion of assets as a DEA input in 
the Chilean model. Labour  was another primary input included in both country’s analyses , but 
expressed differently in NZ and Chile. In NZ, it was a single physical variabl e which aggregated 
paid and non- paid labour and management in full time equivalents (FTE). In Chile, it was 
expressed in monetary terms due to the lack of a physical indicator , and only accounted for paid 
labour; therefore, unpaid labour and management wer e not reflected on this measure. Both, the 
labour and the supplement variables did not reflect quality of the input in any way because de 
databases did not contain information on these regards. 
3.2.3.5 Recursive Partitioning Trees fitting and KPI selection. 
Both  the milk  and revenue efficiency scores obtained in the DEA stage  were subsequently used 
in a series of country- specific RP  Trees as the response variable y. The same procedure was 
followed separately with both, the DairyBase and the TodoagroBase datasets. First, a milk  
efficiency y1 and a revenue efficiency y2 were attached to the original datasets. All the variables 
available in the respective datasets were used as predictor variables. The first and second RPT s 
used DEA scores based on the physical output milk (kg of milk solids or litres depending on the 
country) . Tree one regressed DEA estimates on all existing v ariables, including discretionary 
and non- discretionary inputs and ratios, in a common - to- all- groups second stage which involved 
all DM Us. Tree  two was subsequently carried out afte r all milk - related ratios were removed ( for 
example MS/cow, MS/ FTE , MS/ ha; Milk fat/cow, Protei n/cow, Milk L /cow , L/ha). The third 
and fourth trees used DEA scores based on the economic output GFR . Tree three included all 
the dataset’s existing variables, while tree four used all the same variables but was carried out 
after all efficient farms (DEA score = 1) were removed from the dataset. The reason for this was 
to increase the discriminant power in the analysis and uncover potentially new variables that had 
not been revealed by tree number three.  
Two more trees, one per each country are presented in the respective sections and called ‘the 
fifth tree’.  These were used to fullfill different requeriments in both countries. The fifth NZ tree 
included all existing variables in DairyBase and used return on assets as the response variable to 
show what could be done without DEA by using a major variable  instead of efficiency as y.  
The fifth Chilean tree included backward calculated assets to evaluate what could have been 
done and found with better quality information in TodoagroBase .   
3.2.3.6 Benchmarks sets. 
After running the R codes, a series of trees were available for interpretation and they revealed 
those benchmarks used for further analysis in this study. Only the roots, plus the second and 
third level splits in the multiple trees provided enough benchmarks, which were grouped in two 
sets: NZ - specific KPIs and  Chile- specific KPIs. No  common KPIs  were primarily identified. 
Chapter III: Method 
97 
The sets included well known KPIs for each country  and were used for intra- country 
comparisons.  
 
The benchmark  sets were expected to pr ovide answers to the research questions and  were used 
to analyse each database, provided that they were selected as the best possible measures of 
performance for dairy farming systems in each country. Instead of analys ing the pooled 
samples, the datasets w ere split based on the efficiency scores gained in the first stage of DEA. 
All efficient farms (DEA sc ores = 1) overlapping in both milk and revenue efficiency analyses, 
were set apart. The NZ group consisted of 644 DMUs  while doubly efficient DMUs in Chil e 
were 276 . In both countries these figures  were roughly equivalent to one quarter of the analysed 
total observations.  In addition two other groups were of interest. A second class was drawn 
from the efficient group by ordering the database using farm ID  and setting apart the farms 
efficient three or more years. For obvious reasons all those efficient farms that were not present 
in the survey for at least three years were not part of this group. Finally, a third class comprised 
only the inefficient DMUs ( DEA score < 1) , or more precisely all those DMUs that were not 
able to be ‘milk’  and ‘revenue’  efficient simultaneusly.This was just a criterion and it is 
acknowledged here that things could have been done differently, for example, by relaxing the 
cut- off value and considering efficient those virtually efficient observations with score > 0.95.  
 
In NZ , the group of farms efficient three or more years was reduced to 41 , represented by 142 
observations which were only 5.5% of the sample. In Chile, the proportion was larger ,  
increasing to 7.4% with 76 observations and less than 20 farms . In both countries, t hese farms 
were actually consistent in their efficiency score and, therefore, called the group of resilient 
farms. All three groups (i nefficient, efficient, resilient) were analysed in depth and compared 
looking at t he KPIs revealed by the RPTs. The comparisons were made by means of one - way 
ANOVA and Schaffe significance tests.   
3.2.3.7 Metafrontier analysis. 
The metafrontier approach was capable of investigating whether variations in technical 
efficiency responded solely to technical limitations or whether differences in technology 
contributed to some extent to the alleged inefficiency  ( Battese et al., 2004 ) . Metafrontier is the 
boundary of an unrestricted technology that envelops the restricted group frontiers (Battese & 
Rao, 2002); therefore, i nstead of relying on separate frontiers for each class within countries, 
technical efficiency was estimated in respect to a common frontier (Moreira & B ravo- Ureta, 
2010). The relative efficiency scores across regions and countries were estimated by means of a 
common metafrontier. The overarching metafrontier was achieved through linear optimisation 
as for DEA and by  pooling all the data under the same set of output (GFR) and input (cows, 
Chapter III: Method 
98 
supplement kg/cow, wages, and operating expenses other than wages) combinations (Battese  et 
al., 2004). D ata from both databases needed to be modified in order to obtain comparable inputs 
and outputs to be used in the models. For instance, to allow for differences in milk solids 
content, variable standardisation was needed but there was not enough information to use the 
IFCN method for energy corrected milk (ECM)  (IFCN, 2012) , which could still have been used 
by making seve ral assumptions. Finally, GFR was preferred over the physical output milk 
because its transformation to US $ was straightforward and it is also a standard practice when 
comparing across countries. The same procedure, using an annualised exchange rate and CPI 
indexes was repeated for all input an d output variables of interest. 
 
Chapter IV: IFCN Results 
99 
4. International Farm Comparison Network Results. 
 
This section presents information on NZ and Chilean dairy farming  systems extracted from  the 
database developed by the International F arm Comparison N etwork (IFCN) 17 . In 2011,  the 
database comprised information from 171 farms from 61 regions and 51 countries, representing 
90% of the world’s total milk production . Data from t his secondary data source was used to 
compare one typical pasture- based dairy farm in NZ with a typical dairy farm from Chile;  since 
only key findings are highlighted the reader is directed to Appendix 5 for more detailed 
information. This Chapter sets the scene for the general discussion in Chapter VII I.   
4.1 Typical NZ and Chilean Dairy Farms Comparison  
The typical NZ farm  had a herd size of 348 cows ,  a forage area of 120 hectares,  and a stocking 
rate of 3.3 livestock units per hectare ( LU/ha) , whereas the typical Chilean farm had 421 cows ,  
a forage area of 520 ha, and a stocking rate of 1 LU/ha . The typical N Z  farm had a production 
per cow of 4,300 kg of energy corrected milk ( ECM ) and a land productivity of 12,000 kg ECM  
per hectare whereas the typical Chilean farm produced 9,30 0 kg ECM per cow  and 8,000 kg 
ECM/ha . Figure s 8 and 9 show different indicators in comparing the behaviour of these farms 
and their positioning relative to typical farms analysed in 2011 for all countries. The world 
mean is represented by the value at 0%; a value greater than the mean will compare positively 
(to the right), while a value below the world mean will compare negatively (to the left).   
 
Figure 8: Economic performance of the typical NZ farm relative to the world’s average. 
 
Figure 9: Economic performance of the typical Chilean farm relative to the world’s average. 
                                                          
17 Information on this organisation and its particular TIPI-CAL approach was presented in Chapter II. Two 
typical farms from traditional pasture-based dairy regions in both countries were selected from the IFCN 
sample.  
     
    
Chapter IV: IFCN Results 
 
100 
Figure 8 shows that only farm income was positive for the typical NZ far m; in contrast, all 
indicators were positive for the typical Chilean farm ( Figure  9 ) 18 . The farm income reflects the 
total revenue left after all expenses including rent and interest have been paid, and non - cash 
adjustments (depreciations) have been made. The entrepreneur’s profit and the farm income 
relate to milk price, while return to labour relates to the regional level of wages. The 
entrepreneur’s profit was  calculated by subtracting the opportunity costs of farm owned factors 
from the FI. The NZ  farm income was US$161,400,  which is about 30% above the world mean, 
while the Chilean farm income was US$552,3000, which is about  40% above the world’s mean . 
This indicator is driven by milk price and farm productivity . In NZ milk p rice is lower than in 
Chile, but similar in the fact that in both countries it is lower than the world mean. Therefore, 
differences are mainly explained by land productivity and farm size.  
Entrepreneur’s profits were markedly  different, being US$ - 4,5000 in NZ and US$312,600 
thousands in Chile. Also of interest are the returns on investment calculated using the market 
values on all assets, this  was 4% for the typical NZ farm and 9% for the Chilean farm. T he 
differences in return to labour between the typical farms from NZ and  Chile were even more 
noticeable. The return to labour indicates how much profit an employee, or farmer, generates 
per hour of work on the farm and is calculated by adding wages for hired labour and opportunity 
costs for family labour to Entrepreneur’s profit, and dividing this partial result by total hours of 
labour used for the dairy enterprise. It is driven by the wage level of the country, the labour 
input on the farm and the economic performance of farm.  The return to labour was negative for 
the typical NZ farm, while the typical Chilean farm had a very high positive value. Note that the 
return to labour positions the typical farm relative to the world sample and to their own local 
peers, since it uses the national wages to build a ratio . Also of interest is that the average wage 
on the typical Chilean farm was US$ 5.9 per hour and on the typical NZ farm the average wage 
was US$ 17.64 per hour.  
Figure 10 and 11 provide some explanations for these differences. T he NZ and Chilean strong 
and weak profile respectively, compared to the world’s average, again represented by the 0% 
position. The slim vertical lines indicate the bounds where the particular farm is compared to 
the IFCN distribution (weakest 10%, weakest 25%, strongest 10% and strongest 25%). The  
returns profile is presented on top, the profitability indicators are at the bottom, and the costs 
profile is in between. 
                                                          
18 Note that, as explained in Chapter II, all indicators, farm income (FI), entrepreneur’s profit and return 
to labour, are relative to their local prices. 
Chapter IV: IFCN Results 
101 
 
Figure 10: NZ strong and weak profile.  
 
Figure 11: Chile strong and weak profile. 
The return profiles for both countries are quite similar , with  both in the segment of the 25% 
weakest , with minor differences favourable to the NZ farm in the variable ‘ other returns’ , and 
Chapter IV: IFCN Results 
 
102 
also minor differences favourable to the Chilean farm in milk and cattle returns. According to 
IFCN (2012) , a rise in beef prices significantly increased the dairy farm income in most world 
regions. However, the overall result for NZ and Chilean farms were  only slightly affected by 
changes in non- milk returns .  
The differences between the typical farms from NZ and Chile ,  as shown in Figures 8 and 9, in 
return to labour and profitability were clearly explained by two significantly different cost 
profiles. The typical NZ farm had two major weaknesses when compared to the world  sample’s 
mean, land and capital costs,  whereas the typical Chilean farm was weak in land cost only due  
to its large size . U nder the IFCN approach, l and cost consists of regional rent prices for owned 
land, or r ents currently paid by the farmers for rented land per hectare. Capital cost is calculated 
as a percentage of the sum of owned capital, defined as value of assets excluding land, plus 
circulating capital; the real interest rate is assumed to be 3% or 6%  for borrowed funds.  The 
Chilean land cost is higher than the world’s mean but it is far below the NZ  land value. The 
share of land as a proportion of total costs in NZ  is very high compared to the Chilean farm due 
to the relatively higher price of land for dairy. These reflect a land purchase price around 
US$24,000 per hectare in NZ , and around US$12,000 per hectare in Central southern Chile 
(IFCN, 2011). The negative comparison of the typical NZ farm relative to the world’s sample in 
capital cost is driven by higher asset values and a lower equity ratio, which lifts the interest rate 
used from 3 to 6%.  
Both the NZ and Chilean farm s revealed their major strength in relatively low costs of 
production. Since these typical farms represented pasture- based farming systems, they both 
compared positively against the world mean when feed, stock, machinery, energy, building and 
other costs are considered. ‘Labour’, ‘Other’ and ‘ Quota ’  compare positively in both NZ and 
Chile. Quota costs do not exist in either of the two countries . The family farm income compared 
negatively for the NZ farm and positively for the Chilean farm. The entrepreneur’s profit in both 
countries compared more favourably than the farm income, due to lower opportunit y costs in 
these countries compared to the world’s average. Return to labour indicates how competitive the 
farm is relative to the cost of the labour it is using; note that the average wage in NZ is higher 
than the world’s average of wages. Return to labour  features highly positive in both countries, 
but indicates greater labour efficiency in NZ than in Chile  given the wages in each country. 
Fig ure 12  shows a direct comparison between Chilean and NZ  typical farms, where the 
horizontal axis  contains the deviations with respect to the NZ values in percentage terms .   
Chapter IV: IFCN Results 
103 
 
Figure 12: Typical Chilean dairy farm compared to the typical NZ dairy farm. 
The typical Chilean farm compares favourably against the NZ farm in all but ‘Other returns’ ,  
‘Machinery cost’ , ‘Fuel, energy, water’  and ‘Return to labour’ . The typical Chilean farm had 
higher milk and cattle returns due to more milk and beef produced respectively; also had higher 
total returns. L and, capital, animal health, building, labour and other costs all are higher in NZ, 
being land costs the biggest contributor to total costs. Overall, t he bottom line also compares 
very positive for the typical Chilean farm against the NZ farm.  
4.2 Discussion and concluding comments on the IFCN comparison 
Pasture- based systems are a competitive advantage for those countries or regions that have the 
potential to grow grass all year round. These systems have recently re- gained worldwide interest 
due to their generally lower production costs compared to other systems, along with their more 
environmentally friendly profile (Dillon, et al., 2005; Benson, 2008 ). Temperate grasslands 
have a variety of environments as a result of factors such as season, latitude, altitude, as pect and 
distance from the sea. Their normal conditions are sufficiently favourable for livestock 
production to operate relatively low input systems. The variety in natural conditions also allows 
a range of farming systems to be used,  differing, for example, in how much pasture they grow, 
Chapter IV: IFCN Results 
 
104 
supplement they use, or stock they  carry per unit of area. According to expectations, desp ite 
some important similarities between NZ  and Chilean dairy farming systems, the differences 
revealed by the results presented in this chapter were significant.  
In 2011, both the cost of milk production and the profitability of dairying increased on average 
compared to the previous year (IFCN, 2012) . This costs increase was found to be mainly due to 
escalating land, feed and energy prices. The opportunity cost for land also increased in this year 
relative to 2010 since cropping competed with dairy land because of very high prices for cereals 
and oilseeds. A longer term analysis showed “a sharp increase in cost of  milk production in 
most world regions” (IFCN, 2012, p. 13). However, t he relative strength in local currencies in 
many countries reduced the impact of rising energy and fertiliser prices. Profitability improved 
mainly driven by the high world milk price, consequence of a strong unsatisfied demand for 
milk products , which lasted until mid- 2011 (IFCN, 2012) . 
Discussion groups and m onitor farms are considered early examples  of benchmarking. T hese 
activities started in France,  after the Second World War,  and in NZ in the 1980s , in order to 
share information, transfer technology and motivate learning and action among farmers 
(Campbell, 2008). The IFCN provides a modern version of monitor farms at the international 
level. ‘Typical’ farms under the IFCN appr oach represent those that produce the highest share 
of milk within a region , rather than the majority or average of farms in a country. Clearly, the 
NZ farm ,  located within a traditional dairy region (Waikato) , was representative of both its 
region and the country (LIC & Dairy NZ, 2011). T he ‘ typical’  Chilean farm, also belonged to a 
traditional dairy area (X Region, Osorno) , however, the farm was not particularly distinctive of 
its country which is characterised by many heterogeneous, smaller dairy farms that represent 
around 80% of producers , 40% of the dairy herds , and 20% of the total production ( Consorcio 
Lechero, 2010). Although in Chile, the bigger produce rs have the greater share of total 
production ( Escobar et al. ,  2006) , according to all the background information reviewed, the 
typical NZ farm would  still be larger than the normal Chilean dairy farm. In contrast, t he IFCN 
Chilean farm had 21% more cows than a typical NZ farm and 400% more land dedicated to 
dairy. This suggests that the IFCN approach works well representing the average NZ farm, but 
only represents one segment of the Chilean dairy sector.  
Both Chilean and NZ farm s had significant differences in most physical parameters including 
production per cow, stocking rate, and land  and farm productivity. The Chilean farm had greater 
cow and farm productivity than the NZ farm , but lower land productivity due to a significantly 
lower stocking rate (1 to 3.3 LU/ha) . The stocking rate a farm can  support is limited by the 
availability of pasture and the intensity of supplements used, while the potential for pasture 
production depends mostly on the quality of the land  (Penno, 1999 ; Smith, et al., 2009). S everal 
Chapter IV: IFCN Results 
105 
ecological and geographical factors such as latitude, soils and weather patterns  are similar in 
Southern Chile and New Zealand (Challies & Murray, 2006) . H ence, the quality of the land 
would not be expected to be more than three times higher in production of pasture as reflected 
in stocking rate differences . In NZ, the mean annual pasture production at the commercial level 
has been reported at 12  tonnes of dry matter per hectare and year (Holmes, 2003). The quantity 
of grass grown in Chile would in fact be lower than in NZ . This lower pasture production in 
Southern Chile may be due to a series of factors reported by Pinochet (2003) including low 
phosphate and pH values , and poor grazing management.   
However, t he large difference observed in stocking rates is only partially explained by lower 
pasture yields on similar quality land. Another component of this diff erence may be explained 
by the breeds and strains used for dairying in each country and the ‘Livestock unit’ employed 
by IFCN in its database.  Stocking rate describes the number of animals supported per hectare or 
other unit of area and is expressed on a per head ratio, animal unit, or other pre - defined entity 
basis. A smaller NZ cow would allow more ‘ livestock units’  per hectare. Conversely, the bigger 
Chilean Holstein type  will reduce that figure. In other words, the relationship may be 1 to 2.4 19  
(rather than 1 to 3.3) if the stocking rate was measured directly in cows per hectare. Also of 
interest are the implications of the stocking rate  values on the different risk profile s of the farms. 
According to Penno and Kolver  (2000) , and Anderson and Ridler (2010) , since most types of 
risk associated with farming are positively correlated  with increasing stocking rate, the typical 
NZ farm  will have a more risky profile than the Chilean farm.  
In order to complete ly understand the systems, the stocking rate needs to be related not only to 
land productivity but also to supplements fed to the cows. For an increment in stocking rate , the 
milk production response  on a per hectare basis is positive and, conversely, it is negative on a 
per cow basis (McCarthy, Delaby, Pierce, Journot, & Horan, 2011). The typical Chilean farm 
had very high milk yields per cow  ( 9,300 kg ECM) which can only be achieved on pasture by 
using significant amounts of supplements (Gazzarin, Frey, Peter mann, & Hoeltschi, 2011 ;  
N ational Research Council, 2001). Moreover, t he fact that the typical Chilean farm had double 
the milk yields per cow of the NZ farm suggests that  a higher genetic merit for milk production 
is being well capitalised through the use of concentrates or other supplements in Chile, 
favouring individual cow performance rather than land performance. It is also likely  that the 
typical Chilean farm has grown its own supplementary feed on the effective grazing area.   
Comparative benchmarking is particularly valuable for understanding returns, cost, and business  
drivers (FCC, 2007) . The differences between the typical NZ and Chilean farms were clear 
                                                          
19 According to an adult live weight relationship of 400 kg (NZ crossbreed) and 550 kg (American 
Holstein) and the Livestock unit equivalence used by IFCN LU = 650 kg LW 
Chapter IV: IFCN Results 
 
106 
when comparing physical results but not as evident when comparing revenues, costs and results. 
The return profiles shown in Figure 12 showed minor differences in most indicators under 
analysis for total, milk, cattle and other returns . The Chilean farm of 520 ha and 421 cows had 
higher cattle, milk and total returns than the typical NZ farm of 120 ha and 348 cows . This was 
due to higher farm productivity as a result of larger scale in both, effective dairy area and herd  
size , and also higher production per cow. In contrast, ‘other returns ’  were higher for the typical 
NZ farm  due to the dividends generated by shares owned by the farm under the distinctive NZ 
payment system.    
In both countries, dairy farms operated as price takers under conditions sim ilar to perfect 
competition. According to Holmes (2003) , given low milk prices at the farm gate, primary 
producers need a strong cost leadership focus. The cost profiles for the two countries showed 
differences in the order of 7 % in total costs , explained mainly by higher land and capital costs in 
NZ compared to Chile, with NZ having  also slightly higher feed, animal health, building and 
labour costs. L and cost was the biggest contributor to total costs in NZ  because of the higher 
land and rent market values connected to stocking rate and land productivity . Capital costs are 
also related to the value of the assets which are significantly higher in NZ as are  the levels of 
debt. The costs of feed and animal health were confounding factors because the supplementary 
feed used was expected to be lower in NZ than in Chile. A higher proportion of pasture in the 
diet of the cows was expected in NZ, which, according to Dillon et al., 2005, correlates 
negatively with total cost of feed. The NZ  farm had higher feed cost due to high price per unit of 
supplement rather than volume used. Similarly, animal health was expected to be lower in NZ 
than in Chile, where housing increase s the animal health issues and the herd was larger. 
However, it was again more costly for the NZ farm , presumably due to higher veterinary and 
animal health costs. Higher b uilding costs in NZ  may be explained by more infrastructures other 
than housing to maintain in NZ than in Chile. Finally, labour costs were  not surprisingly higher 
in the typical NZ farm , mainly due to significantly higher wages than in Chile. As expected the 
typical NZ farm was less mechanized, used  less fuel, energy and water, and also demonstrated a 
higher labour efficiency as reflected by a return to labour greater than that for Chile. Despite 
these, t he bottom line represented by family farm income and entrepreneur’s profit  compared 
very positively for the Chilean farm, being more than 3% and 6% higher respectively.  
In summary, the IFCN provided valua ble background information for both countries. The farms 
studied shared a series of characteristics and exhibited some differences. One similarity  was that 
both countries produce at low cost although it is not clear if both countries pursued a cost-  
leadership strategy. NZ pastoral systems are well delineate d pasture- based systems and clearly 
pursue cost- leadership. In contrast, the Chilean example showed to be also low - cost but by 
producing a lot of milk rather than using low inputs. Both also were exposed to climatic and a 
Chapter IV: IFCN Results 
107 
price shocks due to natural variability and the absence of subsides and quotes. The differences 
between the countries were mostly in the cost profiles as previously explained, and in the risk 
profiles as a result of very different stocking rate s in similarly uncertain productive 
environments. However,  the comparison of these ‘typical’  farms for only one particular year did 
not allow a good characterisation of pasture- based dairy farming systems in NZ and Chile. This 
reinforces the need for a more comprehensive study, including a broader sample from a  wider 
geographic coverage.  
 
  
Chapter IV: IFCN Results 
 
108 
 
Chapter V: DairyBase Results 
109 
5. New Zealand Results. 
 
This chapter presents the NZ results, consisting of the data envelopment analyses and the 
second stage recursive partitioning analyses.  
5.1 Data Envelopment Analyses 
This section presents the NZ results for the  data envelopment analysis. Table 1 7 presents a 
summary of the NZ results over a five year period from 2006/ 07 to 2010/ 11 showing 
corresponding weighted averages for the outputs milk solids (MS) and gross farm reveue (GFR) 
while Tables 1 8, 19, 20, 21  and 22  show the class results for both MS and GFR  analyses per 
season over the five- year period.  
Table 17  
Weighted Averages for DEA Results by Year  
 
Results per class are not strictly comparable since several DEA have been carried out 
independently, and sample sizes and DMUs observed are not the same over the period.  
Therefore, significance tests would only encourage misleading comparisons, and have not been 
conducted. A score of 1 represents full efficiency;  the range from 0.95 to 1 is referred to as 
virtually efficient; the range from 0.90 to 0.95 is considered very high  efficiency; and from 0.85 
to 0.90 is high. From June 2006 to May  2011 the NZ efficiency proved to be high to very high 
with fluctuations over the period. The milk solids weighted efficiency f or the owner- operators 
Season n MS GFR
2006/07 616 0.885 0.877
2007/08 609 0.884 0.854
2008/09 492 0.892 0.867
2009/10 558 0.884 0.876
2010/11 323 0.92 0.918
Period 2598 0.89 0.88
Weighted Averages
Chapter V: DairyBase Results 
110 
suveyed in DairyBase ranged from 0.884 to 0.920 and the gross farm revenue weighted 
efficiency ranged from 0.854 to 0.918.  
The similarity between MS and GFR efficiencies values was expected since the correlation 
between these variables was very high (0.95) , and GFR  was driven to a large extent by MS and 
milk price. The results suggest that milk  efficiency was also driven by price relationships such 
as the milk:feed ratio, and by the presence/absenc e of weather related shocks such as droughts 
or flooding. Over the period milk and feed prices as well as the climate showed  marked 
variability. International milk prices were exceptionally high in  2007 and the beginning of 2011, 
and a pronounced drop was observed in June 2009 (IFCN, 2012).  
This study looked at the DairyBase physical data to gain context and better interpret the results. 
The DairyBase physical data comprised 657  observations across the period ( 88 observations in 
2006/ 07, 151 in 2007/ 08, 156 in 2008/ 09, 187 in 2009/ 10 and 75 in 2010/ 11 ) . S easons 2006/ 07, 
2007/ 08, and 2009/ 10 had lower rainfall than 2008/ 09 and 2010/ 11, which  led to lower pasture 
growth and quality, and also lower total feed eaten  and productivity in terms of milk solids per 
cow per day. Further consequences of the lower pasture growth were a higher share of 
supplements in total feed eaten, lower annual productivity per cow,  and less days in milk . 
Consequently, the lower physical efficiency found in 2006/ 07, 2007/ 08 and 2009/10 years than 
in 2008/ 09 and 2010/ 11 was expected .   
In 2006/ 07 and 2007/ 08, the weighted mean efficiency converting inputs into milk  was high 
(0.885, 0.884)  and the conversion of inputs into revenue was lower but still high (0.877, 0.854 ).   
In  response to milk price, in 2008/ 09 and despite  a high mean physical efficiency (0.892) , the 
revenue efficiency was again significantly lower (0.867) .  In 2009/ 10, t he mean milk  efficiency 
was high (0.884 ) , while the revenue efficiency was identical to that of the previous year (0.867) .  
Season 2010/ 11 showed a very high overall efficiency;  converting inputs into physical product 
was high (0.920)  as was the revenue efficiency (0.918) , with both being well above the means 
for the five- year period (0.89 and 0.88, respectively) . During the whole period revenue 
efficiency was below milk  efficiency, which is also reflected in the general mean for the period. 
Table 18  shows the DEA outcomes for season 2006/ 07 ( mean efficiencies and standard 
deviations for each class).  
 
 
 
 
Chapter V: DairyBase Results 
111 
Table 18  
DEA Results per Class Within Season 2006/07 
 
 
In 2006/ 07, Rest of North Island - LOW  had the lowest efficiency (0.857)  and the greatest 
variability as denoted by standard deviation 20 . Attention is drawn to the higher efficiency 
classes, Canterbury/Southland (0.902) and Rest of North Island- MED IU M (excluding 
Northland and using medium levels of imported feed; 0.900). The mean efficiencies were very 
close to each other, but Canterbury/Sout hland had a narrower efficiency range and lower 
variability. For the gross farm revenue results, Rest of North Island- LOW (0.849)  had the 
lowest efficiency and was largest variability class. It is  interesting to note that 
Canterbury/Southland’s (0.883) ranking went from being the most efficient class producing MS 
to only third best producing revenue after Rest of North Island - MEDIUM (0.898) and HIGH 
(0.908) . Also of interest is that greater variability is found among the revenue results than 
among the milk  scores.  
Table 19 shows the DEA outcomes  for season 2007/ 08.   
 
 
 
 
 
 
                                                          
20 The standard deviation (SD) of the mean measures the spread of values in a distribution being 
calculated as the square root of the variance. The spread of a distribution is also referred to as 
‘dispersion’ and ‘variability’. All three terms mean the extent to which values in a distribution differ from 
one another. 
                                            Season 2006-07 n M SD M SD
Rest of North Island - Low use of supplements 189 0.857 0.111 0.849 0.116
Northland-Tasman/West coast 119 0.887 0.101 0.878 0.100
Rest of North Island - High use of supplements 67 0.897 0.097 0.908 0.094
Rest of North Island - Medium use of supplements 106 0.900 0.100 0.898 0.096
Canterbury/Southland 135 0.902 0.085 0.883 0.097
MS oriented efficiency GFR oriented efficiency
Chapter V: DairyBase Results 
112 
Table 19 DEA  
Results per Class Within Season 2007/08 
 
The 2007/ 08  season exhibited  a similar pattern to that of 2006/ 07. The milk solids output show s 
that the lower efficiency classes, on average, were Northland - Tasman/West Coast (0.863) and 
Rest of North Island - LOW (0.863)  for a second time,  while higher efficiency classes were again 
Canterbury (0.928) and Southland (0.948 ) . For the revenue  results, a ttention is drawn to the 
consistency among the lower and higher efficiency classes. Rest of North Island - LOW (0.812 )  
and Northland- Tasman/W est Coast (0.817)  had lower efficiencies again and also showed the 
larger variabilities. Canterbury (0.930) and Southland - Otago  (0.923 ) produced revenue very 
efficiently, just under the mean score reached producing milk. G reater variability was once 
again found among the revenue results than among the milk  scores.  
Table 20 shows the DEA outcomes  for season 2008/ 09.   
Table 20  
DEA Results per Class Within Season 2008/09 
 
                                            Season 2007-08 n M SD M SD
Northland-Tasman/West coast 117 0.863 0.111 0.817 0.139
Rest of North Island - Low use of supplements 157 0.864 0.112 0.812 0.121
Rest of North Island - High use of supplements 90 0.871 0.109 0.870 0.098
Rest of North Island - Medium use of supplements 119 0.886 0.107 0.855 0.112
Canterbury 73 0.928 0.081 0.930 0.079
Southland 53 0.948 0.068 0.923 0.090
MS oriented efficiency GFR oriented efficiency
                                            Season 2008-09 n M SD M SD
Rest of North Island - Low use of supplements 109 0.866 0.111 0.888 0.109
Rest of North Island - High use of supplements 63 0.878 0.120 0.868 0.119
Northland-Tasman/West coast 92 0.887 0.112 0.862 0.132
Rest of North Island - Medium use of supplements 123 0.899 0.101 0.869 0.113
Canterbury/Southland 105 0.925 0.083 0.845 0.127
MS oriented efficiency GFR oriented efficiency
Chapter V: DairyBase Results 
113 
In 2008/ 09, the lowest milk  efficiency class was Rest of North Island - LOW (0.863) , while the 
highest mean was exhibited by Canterbury  and Southland- Otago  classes (0.925). F or Rest of 
North Island- HIGH, the variability of these results was much higher than that observed in the 
previous season. For the revenue  results, note the dramatic change in ranking compared to both, 
the milk  results and the previous seasons. Rest of Nort h Island- LOW (0.888)  exhibited the 
highest revenue efficiency. Conversely, Canterbury and Southland- Otago  (0.845) showed the 
lowest mean efficiency and a very high standard deviation.  
Table 21 shows the DEA outcomes  for season 2009/ 10.   
Table 21  
DEA Results per Class Within Season 2009/10 
 
In 2009/ 10, the lowest milk  efficiency class was Northland - Tasman W est Coast (0.847) , 
followed very closely by Rest of North Island - HIGH  (0.850) , with both exhibiting moderate ly 
high milk efficiency. In contrast,  Southland was virtually efficient (0.951) with a low standard 
deviation. Canterbury had the same pattern, having a very high milk  efficiency (0.939). In 
between, Rest of North Island- LOW and MEDIUM exhibited very high efficie ncy and moderate 
variability. The ranking for milk  results almost perfectly matched the ranking for the revenue  
results. However, the lowest refficiency class changed to Rest of North Island - HIGH (0.845) , 
which was at that lowest position for the first time in the series.  
Table 22 shows the DEA outcomes  for season 2010/ 11.   
 
 
 
                                            Season 2009-10 n M SD M SD
Northland-Tasman/West coast 105 0.847 0.126 0.853 0.120
Rest of North Island - High use of supplements 96 0.850 0.126 0.845 0.115
Rest of North Island - Low use of supplements 121 0.878 0.099 0.863 0.103
Rest of North Island - Medium use of supplements 110 0.888 0.103 0.892 0.097
Canterbury 74 0.939 0.076 0.910 0.092
Southland 52 0.951 0.063 0.930 0.083
MS oriented efficiency GFR oriented efficiency
Chapter V: DairyBase Results 
114 
Table 22  
DEA Results per Class Within Season 2010/11  
 
In 2010/ 11, mean efficiencies for both MS and GFR  were markedly higher than for  the previous 
seasons. The MS results show that the lower efficiency class were Rest of North Island - HIGH 
(0.908) and LOW (0.910) , but these values were still in the range of very high milk  efficiency 
range. In contrast, Canterbury and Southland exhibited a very high av erage efficiency (0.947) 
with a low standard deviation. This highest efficiency class also kept i ts place when looking at 
the revenue results (0.937), although the rest of the ranking did change. Interestingly, Rest of 
North Island- HIGH (0.924) jumped from  last to the second best efficiency class and conversely, 
Rest of North Island - LOW went from very high efficiency t o bottom of the table (0.899).  
In summary, variability was evident in both milk and revenue efficiencies  throughout the period 
but the variability was greater among the GFR results than among the MS scores.  The best year 
of the series clearly was 2010/ 11 due to a favourable milk price scenario and the absence of 
weather shocks. The lowest efficiency class for both milk and revenue  for most years was 
Northland- Tasman/ West Coast, while Canterbury and Southland had the higher milk efficiency 
almost every year but failed to deliver the same good results when revenue efficiency was 
analysed.  
The 2007/08 season exhibited a similar pattern to that of 2006/ 07, but the milk price was 
significantly higher and thus, attention is  drawn to the fact that lower and higher milk efficiency 
classes exactly  matched the revenue efficiency class ranking s. Rest of North Island- LOW 
(0.812) and Northland - Tasman/West Coast (0.817) had lower efficiencies again and also 
showed the larger variabilities. Canterbury (0.930) and Sout hland/Otago (0.923) produced 
revenue very efficiently, with just under the mean score reached producing milk. Greater 
variability was once again found among the GFR results than among the MS scores.  In 2008/ 09 
(Table 20) , a year characterised by a high initial milk price followed by a pronounced drop 
                                            Season 2010-11 n M SD M SD
Rest of North Island - High use of supplements 73 0.908 0.095 0.924 0.089
Rest of North Island - Low use of supplements 67 0.910 0.103 0.899 0.106
Rest of North Island - Medium use of supplements 63 0.913 0.089 0.914 0.080
Northland-Tasman/West coast 57 0.923 0.095 0.916 0.097
Canterbury/Southland 63 0.947 0.065 0.937 0.076
MS oriented efficiency GFR oriented efficiency
Chapter V: DairyBase Results 
115 
when the season was well under way, the lower MS efficiency classes generally had the higher 
revenue efficiency averages. In 2008/ 09, clearly only those who effectively  managed the change 
in milk price scenario were able to maintain efficiency producing revenue. In 2009/ 10 the M S 
results positioned Northland- Tasman West Coast (0.847) as the lowest efficiency class, closely 
followed by Rest of North Island- HIGH (0.850). In contrast, Southland was virtually efficient 
(0.951) with a low standard deviation and Canterbury exhibited the second best efficiency 
(0.939) and also low variability . Interestingly, the ranking for MS results almost perfectly 
matched the ranking for the GFR results. In 2010/ 11 mean efficiencies were markedly higher 
than for 2009/ 10. The efficiency at producing milk  had an inverse relationship with the 
efficiency at producing revenue in some classes in the North Island of NZ. In contrast, c lasses in 
the South Island had  a direct association between milk and revenue  efficiency.  
5.2 Recursive Partitioning Trees results 
Both the milk and revenue e fficiency scores obtained in the DEA stage were incorporated into 
the original dataset including all years and regions,  and subsequently used in a series of country -
specific RP trees with the efficiency values used as the response variable y. This section 
presents the milk trees first and later the revenue trees, followed by a summary of key 
performance indicators identified.  
5.2.1 Milk Solids Results 
Figure 13 21  represents a milk solids tree, obtained from the regression of DEA estimates on a 
dataset that included all 176 discretionary and non- discretionary inputs and ratios contained in 
DairyBase (refer to Appendix 1 ). A general concept useful to navigate and understand the trees, 
was that right branches always cluster the subsample with the highest mean efficiency and, on 
the contrary, left branches contains the lowest mean efficiency class. Figure 13 shows 1 3  
terminal nodes within two major branches split by a top level or root variable .  
                                                          
21 See Chapter III: Method for more detail in trees’ structure and how to navigate them. 
Chapter V: DairyBase Results 
116 
 
 
Figure 13: MS DEA partitioning tree. 
Chapter V: DairyBase Results 
117 
Milk solids measured in kilograms per cow (MS/cow) was  the root variable. Based on a 
threshold value of 338.5 kg , MS/cow split the sample into left and right branches. The left  
branch which accounted for 1,150 DMUs or observations that produced less than the pivotal 
value ( 339 kg  MS/cow )  resulted in eight leaves. A virtually efficient class (0.9525) contained  
those 150 DMUs with total labour units lower than 1.305 full time equivalents (FTE). On the 
outer left side, total FT E interacted with a third split on MS/ FTE , which led to two fourth level 
variables, Closing Total Assets and dairy return on assets (dROA) . Following the outer track ,  
the two lower efficiency classes clustered 258 DMUs . The lowest efficiency leaf (0.7075) was 
composed of observations that produced less than the threshold MS/ cow ( 294 kg ), further split 
by operating profit margin (OPM) at a critical value of 22.5% .  
 
The right branch clustered those DMUs with production levels higher than 339 kg MS/cow  into 
four leaves containing 1,438 observations. This branch revealed four more partitioning levels 
using the same variables as in the left branch. However, these variables all acted at different 
levels, revealing varied main effects and interactions between them. Again MS/cow was the 
second order split and in itself was able to define an outer leaf;  this was the largest class (n = 
462) , and was virtually efficient (0.9664) , with production levels greater than 406 kgMS  per 
cow. On the inner side, the second split led to the third level variable, FTE , which revealed the 
most efficient class ( 0.9734; n = 141) ; this had a   combination of moderate yields per cow 
(between 339 and 406 kg MS/ cow), employed less than 1.56 FTE , and comprised just under 
10% of the sub sample.  
MS/cow, FTE, and MS/ FTE were very dominant in Figure 13, and it was considered that this 
could be masking some interesting relationships am ong other less strong variables;  hence,  a 
modification was proposed. Figure 14 shows a second tree generated using the same DEA 
scores based on the physical output MS  with twelve of the  milk - related ratios excluded from 
the dataset ( MS / cow, MS/ FTE, MS/ ha; Milk fat, MFat per cow and ha, Protein , Protein / cow, 
Protein/ha , Total milk , Milk litres  per cow and L/ ha). The analysis regressed DEA  estimates on 
163 variables , yield ing 13 terminal leaves, as shown in Figure 14.  
Chapter V: DairyBase Results 
118 
 
Figure 14: Modified MS DEA partitioning tree. 
Chapter V: DairyBase Results 
119 
The modified tree exposed relationships among variables that had not appeared before. There 
were 1,427 observations with a dROA lower than the threshold value of 4.5% which were split 
to the left ending in nine leaves drawn by means of six levels of partiti on. The second split on 
the left branch FTE  created a virtually efficient (0.9529) , medium sized class (n = 154) , having 
less than 4.5 % of dairy ROA and fewer than 1.3 FTE s. In parallel, following the outer track  the 
sample was further split by operating expenses ratio expressed in  NZ  dollars per kgMS  
(OpEx/kgMS ) at the threshold value of NZ$ 4.935/ kgMS. Net milk sales ( NZ $)  was at the next 
level down, splitting at two different pivotal values. The lowest efficiency class (n = 114, 
0.7247) had an intermediate value for net milk sales, a value of equity equal or greater than 
NZ$1.043 million and paid more than NZ$ 1.99 per kgMS  in interest and rent.  In parallel, the 
left inside third level split also led to net milk sales, then to liabilities  ratio ( NZ $/ kgMS) , and 
finally to gross farm revenue per hectare (GFR/ha), defining four leaves ranging in efficiency 
from 0.9222 to 0.8003.   
 
The root variable dROA also split 1,161 observations for those having a threshold value greater 
than 4.5 % to  the right branch. On the out er side, OpEx was a second level variable at the pivotal 
value of NZ$  4.385/ kgMS clustered obs ervations into the most efficient (0.9599) and largest 
class (n = 582) :  22.5% of the sample fitted into this leaf. On the inner pathw ay, a further split 
originated on the third level variable, Assets Turnover R atio (ATR) , at a threshold value of 
24.5%. Again, on the outer side, a virtually efficient class (0.9548) comprised of 1 49  
observations which were very profitable, as indicated by the dROA,  but less cost efficient,  
however, they  made good use of their assets as indicated by ATR .  
 
Table 23 highlights the order and frequency of appearance of the variables exposed by both 
trees based on milk solids  DEA which were located from the root to the fourth level split in 
Figures 1 3 and 1 4.  
 
 
 
 
 
 
 
 
Chapter V: DairyBase Results 
120 
Table 23  
Key Performance Indicators as Revealed by MS Trees 
 
 
A key property of  the RPT approach is that it identified the relative importance of each variable. 
Not only is the order of entrance of a particular variable important , but also its reappearance at 
different levels and combination with other variables. Table 23  shows the strong variables, such 
as MS/COW, FTE, dROA , and OpEx, which appeared at least twice and sometimes 
‘ participated’  in building up other ratios like MS/FTE. When this is the case, the signific ance of 
the relevant variable was reinforced.   
5.2.2 Gross Farm Revenue Results. 
The following trees (Figures 1 5 and 16 ) used DEA scores based on the economic output GFR  
instead of MS. Figure 15  included all DMUs and all variables used for the first tree MS .  
KPI LEVEL
MS/Cow 1ST
MS/Cow 2ND R
FTE 2ND L
Figure 13 FTE 3RD R
MS/FTE 3RD L
OPM 4TH R
dROA 4TH L
CTASS 4TH L
dROA 1ST
OpEx/kg 2ND R
FTE 2ND L
Figure 14 ATR 3RD R
OpEx/kg 3RD L
OpEx/HA 4TH R
Net milk sales 4TH L
Closing Equity 4TH L
Chapter V: DairyBase Results 
121 
 
 
Figure 15: GFR DEA partitioning tree. 
Chapter V: DairyBase Results 
122 
Figure 15 exposed  13 terminal nodes, six levels of partition,  and overall, was very similar to 
Figure 14. Similarly , t he top level variable was the ratio dROA at the same critical value of 
4.5%, which split the sample into the left branch (n = 1 427) and the right branch (n = 1 161 ). 
Nine leaves were drawn from the left subsample, which had the same structure and complexi ty 
as those shown in tree number two (Figure 15) . The second level variables were again total FTE 
and OpEx/kgMS, although the threshold values were slightly different. In short, Figure 16 
revealed that the highest efficiency class (0.9552) on the right branch, was attainable by having 
higher than 4.5% dairy ROA and operating costs lower than $ 4.295/ kgMS. This was the  
situation for 520 observations  which was approximately 20% of the whole sample. In contrast , 
one tenth of the sample was included in the lowest efficiency class (0.732)  which had ROA 
below 4.5% , total FTE above 1.36 and produc ed less than 315.6 kgMS/cow . This lowest 
efficiency class also had higher closing Equity  than the pivotal value and an operating profit 
margin lower than 22.5% .  
 
Figure 16 shows a tree based only on inefficient farms, that is those non- efficient DMUs left 
after the 570 efficient farms (DEA score = 1) were removed from the dataset  so as to search for 
new interactions. 
 
Chapter V: DairyBase Results 
123 
 
Figure 16: Modified GFR DEA partitioning tree.
Chapter V: DairyBase Results 
124 
This tree yielded 11 terminal leaves, by means of five partitioning levels. Simil ar to tree number 
one (Figure 13 ), the root node split 659 observations to the left and 1131 to the right based on 
the ratio MS/cow. On the right branch and to the outer side , the  highest efficiency class (0.8793)  
was also the largest (n = 513) comprising 27% of the whole sample; it was defined by a n 
operating profit per hectare greater than the critical value of NZ$ 953 and a production pe r cow 
higher than 355.7 kgMS . The second highest efficiency class (0.8737) wa s found on the inner 
left branch. This was a small class (n = 50) having less than 320 kgMS/cow, more  than 23.5% 
OPM and a high labour productivity (> 5228 kgMS/FTE). In contrast, t he lowest efficiency 
class (0.6778) which was  also on the left branch was determined by less than 23.5%  OPM, more 
than 1. 705 FTE and a production per cow lower than 281 kg MS.  
 
Table 24 highlights the order and frequency of appearance of the  variables exposed by Figures 
15 and 16 . 
Table 24  
Key Performance Indicators as Revealed by GFR Trees 
 
 
In summary, the RPT approach  identified those indicators which were important in defining 
efficiency classes in New Zealand ; presented a number of threshold or pivotal values that 
KPI LEVEL
dROA 1ST
OpEx/kg 2ND R
FTE 2ND L
Figure 15 MS/Cow 3RD R
MS/Cow 3RD L
Opening Equity 4TH R
Closing Equity 4TH L
Operating profit/ha 4TH L
MS/Cow 1ST
Operating profit/ha 2ND R
OPM 2ND L
Figure 16 Opening Land&Building Assets 3RD R
MS/Cow 3RD R
FTE 3RD L
MS/FTE 3RD L
MS/Cow 4TH L
Closing Total Assets 4TH L
Chapter V: DairyBase Results 
125 
determine the pathway and the class where a particular DMU is likely to sit ; and ultimately 
exposed the pathway, or combination of indicators, that identified homogenous groups 
accounting for different dairy performance in NZ ( p = 0.05 as pre - defined in RPT code) . These 
results show that there is no one pathway to attain high efficiency, and that many variable 
combinations of good practices can result in high and very high efficiency scores. The variable 
MS/ cow was revealed as a multiple order indicator that came out in different trees and levels, 
sometimes more than once. Similarly FTE was found in all studied trees and also the 
relationship MS/ FTE appeared as a third and fourth level indicator, reinforcing the importance 
of both physical production and labour. dROA appeared repeatedly , from top to fourth order. 
OPM came up four times,  as a second, fourth and sixth level indicator in several trees, having 
the higher level in the last tree which excluded the efficient farms  and looked for a more 
discerning analysis. ATR was revealed  once, as a third level indicator, as did OpEx/ kg.  
5.2.3 The fifth NZ tree 
A fifth NZ tree used ROA as the response variable instead of efficiency scores. This is 
presented in Appendix seven  since it is out of the scope of this work. However, it was 
considered to be an effective way to show relationships between variables without the need for 
estimating efficiency scores using DEA. Such a tree could also be used as a prediction tool ,  
even with a different New Zealand dataset,  for example , for appraisal of past or future seasons.  
5.3 NZ Benchmarks Set 
The NZ be nchmark  sets, identified in the previous analyses, consisted of key performance 
indicators that acted at the first, second and third level  variables, on the trees presented earlier. 
Table 25  presents this set split into two groups, one column showing the physical and other 
column showing the financial indicators drawn from the trees. It is worth noting that the present 
set can be expanded by adding as many levels as required . 
  
Chapter V: DairyBase Results 
126 
Table 25  
NZ Benchmarks Set Comprising Physical and Financial KPIs 
 
This set of benchmarks, which comprises three physical and five financial indicators, was next  
used to describe the data contained in DairyBase. The NZ sample was grouped according to the 
following criteria: inefficient DMUs , with a score < 1  in any of the MS or GFR analyses, made 
up the largest group (n = 1944) , and efficient DMUs, whose score = 1  overlapping in both 
analyses. This second group was further split into two sub sets by ordering observations using 
farm ID : those farms which were efficient just once or twice over the five year period were set 
apart (n = 502)  from those which were consistently efficient three or more times over the same 
period (n = 142) . These resulting groups are thereafter referred to as Inefficient, Efficient , and 
Resilient  respectively.  
Table 26 presents a summary of the descriptive statistics for the physical indicators for each 
group. This is followed by indicator- specific tables (Tables 27 and 28 ) where significance tests 
are presented.  
Physical KPIs Financial KPIs
MS/COW (kg) dROA (ratio)
FTE (units) OPERATING PROFIT/HA ($)
MS/FTE (kg) OPM (ratio)
OpEx/kgMS ($)
ATR (ratio)
Benchmarks set
Chapter V: DairyBase Results 
127 
Table 26  
Descriptive Statistics on Selected Physical KPIs  
 
As shown in Table 26, resilient farms had greater variability in MS/cow and FTE , but on 
average, they produced more output, both per cow and per labour unit expressed in kg of milk 
solids, than the other groups. Differences in total FTE between groups were  minor and not 
significant. Tables 27 and 28 show the means’ comparison for the production ratios  of interest. 
The layout indicates no significant differences in the means when these are shown in the same 
column. The columns represented the number of subsets of means drawn for a confidence value 
of 95% . Significance Schaffe  tests for unequal group sizes,  were conducted to compare the 
groups by using one- way ANOVA.  
 
Resilient 142 391 88 213 650
Efficient 502 376 64 178 564
Inefficient 1944 342 56 118 525
Total 2588 351 62 118 650
Resilient 141 3.04 2.62 .58 12.44
Efficient 502 3.13 2.45 .69 17.56
Inefficient 1944 2.95 1.48 .77 13.24
Total 2587 2.99 1.79 .58 17.56
Resilient 141 59294 19098 16260 102677
Efficient 502 59024 20764 18522 143826
Inefficient 1944 48293 14707 13593 120172
Total 2587 50975 16957 13593 143826
MIN MAX
 
N M SD
MS/cow
Total FTEs 
MS/FTE
Chapter V: DairyBase Results 
128 
Table 27  
Comparison of Means for the Output Produced per Cow Indicator 
 
 
Table 28  
Comparison of Means for the Output produced per Labour Unit Indicator  
 
 
The differences in production per cow and labour unit were large and significant, especially 
between the group of resilient farms and the inefficient group. The differences between the 
groups found in absolute and statistical terms support the relevance of the selected physical 
indicators as KPIs in NZ. In a similar layout , financial KPIS will be  presented.  
Table 2 9 shows the summary on selected financial KPIs followed by the indicator- specific 
tables (Tables 30 to 34) .  
Class n
Inefficient 1944 342
Efficient 502 376
Resilient 142 391
MS/Cow
Class n
Inefficient 1944 48293
Efficient 502 59024
Resilient 141 59294
MS/FTE
Chapter V: DairyBase Results 
129 
Table 29  
Descriptive Statistics on Selected Financial KPIs 
 
Resilient 142 .081 .076 -.030 .495
Efficient 502 .072 .066 -.111 .664
Inefficient 1944 .041 .033 -.087 .287
Total 2588 .049 .047 -.111 .664
Resilient 142 2920 2024 -2654 10431
Efficient 502 2711 1788 -1207 8624
Inefficient 1944 1871 1493 -2498 8860
Total 2588 2092 1632 -2654 10431
Resilient 142 .341 .173 -.275 .715
Efficient 502 .328 .155 -.252 .616
Inefficient 1944 .251 .161 -.479 .714
Total 2588 .271 .164 -.479 .715
Resilient 142 4.49 1.15 2.33 8.43
Efficient 502 4.55 1.08 1.98 9.07
Inefficient 1944 4.97 1.09 2.54 10.72
Total 2588 4.86 1.11 1.98 10.72
Resilient 142 .268 .237 .068 1.444
Efficient 502 .246 .222 .046 1.802
Inefficient 1944 .151 .064 .020 .978
Total 2588 .176 .132 .020 1.802
Operating Profit/ha ($.ha)
Operating Profit margin (ratio)
MIN MAX
Dairy Return on Assets (ratio)
 
N M SD
Operating Expenses/kgMS 
($/kg)
Asset turnover (ratio)
Chapter V: DairyBase Results 
130 
For all the analysed financial variables, the resilient farms showed greater variability as 
indicated by standard deviations. They were also consistently more profitable than the other 
groups, and exhibited  lower operating expenses and better use of assets.  Efficient DMUs tend ed 
to be similar to the resilient group,  but sometimes exhibit ed lower minimum and higher 
maximum values. Differences between r esilient and inefficient observations were noticeable in 
all the indicators, but were more marked for  dairy ROA  where resilient farms had twice the 
return of inefficient observations. Furthermore, resilient farms had 77% higher  ATR and  56% 
greater profitability per hectare than the group of inefficient observations. Table 30 shows the 
means’ comparison for the profitability ratio Dairy return on assets (d ROA).  
Table 30  
Comparison of Means for Dairy ROA   
 
As expected because of the large gap in returns between the resilient farms and the groups of 
inefficient and efficient, the differences in ROA were significant between all groups. In absolute 
terms, the efficient observations were closer to the resilient than to the inefficient group, but still 
a difference of 0.009 can be considered large for this particular indicator. Table 31 shows the 
statistical comparison for another profitability ratio, operating profit margin . 
Table 31  
Comparison of Means for Operating Profit Margin 
 
In operating profit margin , both in absolute and statistical terms, e fficient observations behaved 
very similarly to the resilient farms. This measure indicates how well a farm generates profit 
from its revenue and/or how effective cost control is. Tabl e 31 clearly demonstrates that 
resilient and efficient groups are performing much better than the Inefficient DMUs. Taking this 
into consideration, Operating Profit per hectare was also expected to be significantly different , 
at least between two groups. Table 32 shows the comparisons carried out  on that indicator.  
 
Class n
Inefficient 1944 0.041
Efficient 502 0.072
Resilient 142 0.081
DairyreturnAssets
Class n
Inefficient 1944 0.25
Efficient 502 0.33
Resilient 142 0.34
OpProfitMargin
Chapter V: DairyBase Results 
131 
Table 32  
Comparison of Means for Operating Profit per Hectare  
 
As shown in Table 32, the Operating P rofit per hectare demonstrated that, in absolute and 
statistical terms, efficient observations behaved very similarly to the resilient farms, and both 
groups were again far away from the inefficient group. Two columns set these apart from 
efficient and resilient observations, which were clustered together by much higher levels of 
profit per hectare. These results may be in part explained by lower operating costs , as shown in 
Table 33.  
Table 33  
Comparison of Means for Operating Expenses per Kg of Output  
 
Resilient and efficient observations, as seen in Table 33 , had approximately 10% lower 
operating costs, which were statistically different than those for inefficient DMUs. Two 
columns were drawn from the means’ comparison for this variable, indicating that efficient and 
resilient observations had no significant differences in operating expenses per kg of output. The 
last indicator in the NZ benchmark set measures how well a farm generates dairy revenue from 
its assets. The results for this indicator are presented in Table 34.  
Table 34  
Comparison of Means for Assets Turnover Ratio in Different Classes 
 
Class n
Inefficient 1944 1871
Efficient 502 2711
Resilient 142 2920
Operating Profit/ha ($.ha)
Class n Operating Expenses/kgMS ($/kg)
Inefficient 1944 4.97
Efficient 502 4.55
Resilient 142 4.49
Class n
Inefficient 1944 0.15
Efficient 502 0.25
Resilient 142 0.27
Asseturnover
Chapter V: DairyBase Results 
132 
Table 34  supports to the considerable differences between resilient farms and inefficient 
observations. Inefficient DMUs had an ATR of 0.15 versus 0.25 and 0.27 for the efficient and 
resilient, respectively, which is a difference 40%  more favourable to these two groups 
( differences between them did not allow for statistical significance) . As has happened with 
physical indicators, the differences found in absolute and statistical terms support the relevance 
of the selected financial KPIs for  NZ  dairy farming. 
 
 
  
Chapter VI: TodoagroBase Results 
133 
6. Chilean Results. 
 
This chapter presents the Chilean results, consisting of the Data Envelopment Analyses (DEA) 
and a second stage DEA using  recursive partitioning analyses. It follows the same structure and 
order of Chapter 5 , but employs fewer tables given the smaller size of the Chilean sample.  
6.1 Data Envelopment Analyses (DEA)  
This section presents the Chilean data envelopment analysis results. Table 35 presents a 
summary for the five year period from 2007 to 2011, while Table 36  shows the results for both 
milk (L) and gross farm revenue (GFR) analyses for each  class22 . As in Chapter 5, results  per 
class were not strictly comparable since DEA were c arried out independently, so significance 
tests were not attempted at this stage. The considerations about milk and input prices made i n 
Chapter 5 over the period are  valid for the present section.  According to the Dirección 
Meteorological of Chile (2007, 2008, 2009, 2010, and 2011) , there was a diverse rainfall pattern 
for the period. In 2007, Valdivia had more than 1,200 mm of rain and Osorno had just less than 
900 mm. In 2008, rainfall was 1,995 mm and 1,028 mm, respectively, while in 2009, Valdivia 
had 1,950 mm and Osorno a maximum of 1,345 mm. In 2010, Valdivia had annual rainfall 
closer to the average of 1,491 mm, while  Osorno totalled 1,041 mm which was  similar to the 
results observed in 2011 which were  1,618 mm and 995 mm, respectively.  
Table 35  
Weighted Averages for DEA Results by Year  
 
                                                          
22 Note that the Chilean classification system has been explained in the Method chapter; the system 
comprises ten classes derived from splitting the sample into years and classifying observations into LOW 
or HIGH intensity, depending on the supplement feed to the cows, with a cut off value of 2,000 
kg/cow/year.  
Year n MILK GFR
2007 245 0.893 0.877
2008 231 0.895 0.885
2009 231 0.894 0.880
2010 197 0.898 0.894
2011 124 0.879 0.881
Period 1028 0.893 0.883
Weighted Average
Chapter VI: TodoagroBase Results 
134 
In the results where a score of 1 represents ‘full’ efficiency; the range from 0.95 to 1 is referred 
to as ‘virtually efficient’; the range from 0.90 to 0.95 is considered ‘very high efficiency’; and 
from 0.85 to 0.90 is ‘high efficiency’. In 2007 and 2008, t he weighted mean milk efficiency  was 
high (0.893, 0.895), and the c onversion of inputs into revenue (GFR) was lower but still high 
(0.877, 0.885). In 2009, t he mean physical efficiency was again high (0.894) , and the revenue 
efficiency was considerably lower (0.880). In 2010, both weighted averages increased, but GFR 
efficiency improved significantly with respect to the previous year; the mean efficiency 
converting inputs into physical product was 0.898, and the efficiency producing revenue was 
0.894. In 2011, there was a distinct fall in physical efficiency (0.879) , along with a drop in 
revenue efficiency to the level experienced in 2009 (0.881). In the whole period, 2007- 2011, the 
weighted mean efficiency was greater producing milk (0.893) than producing revenue (0.883). 
Overall, in 2007, 2008, 2009, and 2010, the revenue efficiency was below the physical 
efficiency, though this relationship was inverted in 2011. Table 36 shows more detailed 
information on the DEA outcomes for the period 2007- 11.   
Table 36  
DEA Results per Class for the Period 2007-11 
 
Table 36 shows mean efficiencies and standard deviations for the ten Chilean classes. The milk 
results ranged from 0.84 (moderately high) to 0.92 (very high), while the standard deviations 
n
Chilean class M SD M SD
2007-LOW use of supplements 131 0.87 0.11 0.85 0.12
2007-HIGH use of supplements 114 0.92 0.07 0.91 0.07
2008-LOW use of supplements 113 0.88 0.10 0.87 0.11
2008-HIGH use of supplements 118 0.91 0.08 0.90 0.08
2009-LOW use of supplements 143 0.87 0.10 0.86 0.11
2009-HIGH use of supplements 88 0.92 0.06 0.92 0.06
2010-LOW use of supplements 113 0.88 0.11 0.88 0.11
2010-HIGH use of supplements 84 0.92 0.07 0.91 0.08
2011-LOW use of supplements 66 0.84 0.14 0.85 0.14
2011-HIGH use of supplements 58 0.92 0.08 0.92 0.09
MILK efficiency GFR efficiency
Chapter VI: TodoagroBase Results 
135 
fluctuated between a relatively low 0.06 to a high 0.14. Similarly, the revenue results ranged 
from a high 0.85 to a very high 0.92, while the standard deviation fluctuated within the same 
range. Interestingly, the higher efficiency classes in both milk (0.92) and revenue analyses 
(0.92) were found in two contrasting years from the milk price stand point, 2009 and 2011, in 
classes with a HIGH use of supplements. These were follow ed by other very high efficiency 
classes in 2007 and 2010, again using HIGH levels of supplements (0.92 and 0.91). Results in 
all these four classes were tight and standard deviations were consistently low. Also of interest 
is the fact that the lowest milk  efficiency class (0.84) was found in 2011, comprising 
observations with a LOW use of supplements. This class was only slightly better at producing 
revenue from the same inputs (0.85). The other lower efficiency class producing revenue (0.85) 
was found in 2007, also using LOW levels of supplements and producing milk at a higher 
efficiency level (0.87). Despite a tendency to higher physical than revenue efficiencies, the 
Chilean milk and revenue scores were very close. The variability tended to be greater in  revenue 
than in milk efficiency and, consistently, all LOW classes showed higher standard deviation 
values in both measures of efficiency.  
6.2 Recursive Partitioning Tree results 
As explained in Chapter 5, both the milk and revenue efficiency scores obtained  in the Chilean 
DEA stage were attached to the complete original dataset and subsequently used to develop a 
series of country- specific RP trees using the efficiency scores as the response variable y. This 
section presents the milk trees first and later the revenue trees, followed by a summary with the 
identified indicators. 
6.2.1 Milk  results. 
Figure 17 shows the  milk efficiency tree  obtained from the regression of DEA estimates on all 
90 discretionary and non- discretionary inputs and ratios contained in the TodoagroBase (refer 
Appendix 2). The diagram shows 11 terminal nodes on two major branches split by the root 
variable, litres of milk per cow (L/Cow), at  a threshold value of 4,664 L.  
Chapter VI: TodoagroBase Results 
136 
 
Figure 17. MS DEA partitioning tree.
Chapter VI: TodoagroBase Results 
137 
The root variable, litres per cow, split the sample into right and left branches. The right branch 
accounted for more than two thirds of the sample, clustering those DMUs producing more than 
4,664 L/cow and year. One virtually efficient class (0.954, n = 235) was drawn as the result of a 
second split made by the ratio wages paid per litre of milk. The threshold value was 
CLP$16.22/ L, equivalent to NZ$0.41 23  per 10 L of milk 24 . The second best efficiency class 
(0.9375; n=236) in this branch had a higher wages ratio and used less than 295.7 g/L of 
supplements, but almost offset these with a gross farm revenue per cow greater than CLP$1.087 
million, equivalent to NZ$2,764 pe r cow. These two large classes made up almost half of the 
total sample.  
The root variable split just over one third of the Chilean sample into the left branch. The least 
efficient class (n = 56) was found on this branch at the end of the outer pathway. It  had an 
average efficiency of 0.7228, which resulted from the combination of lower production per cow, 
a higher than the threshold value for production costs and wages, and also a higher total cost, 
including rent and interests. The second split on this branch was the ratio production costs per 
litre of milk, which is a Chile - specific variable similar to the well- known farm working 
expenses in New Zealand  (see definition Chapter II) . The critical value was CLP$121.80/ L, 
equivalent to NZ$3.10 per 10 L of mil k. The third split was again the wages ratio at the 
threshold value of CLP$17.79/ L, equivalent to NZ$0.45 per 10 L of milk produced.  
Following the same reasoning as with the New Zealand sample, a modification was carried out 
to expose any hidden relationships among the less strong variables. Figure 18 shows a second 
tree generated by using the same DEA scores based on the milk output, with six milk- related 
ratios excluded from the dataset  (t he analysis regressed DEA estimates on 84 variables ) .  
                                                          
23 Based on an exchange rate of CLP$394.65 per NZ$. 
24 Based on the assumption that 10 litres of Chilean milk (3.4% Protein, 3.7% Fat) would be roughly 
equivalent to one kilogram of milk solids.  
Chapter VI: TodoagroBase Results 
138 
 
Figure 18. Modified MS DEA partitioning tree.
Chapter VI: TodoagroBase Results 
139 
Figure 18 revealed new relationships, but these were fewer than expected. Only two different 
variables came out compared to those in Figure 17 : operating profit  per cow, and litres produced 
per milked cow. It is worth clarifying that ‘milked cows’ are the average number of cows 
effectively milked throughout the year, which is different from ‘total cows’ which includes 
milked, pregnant, and empty cows on the farm. These new variables appeared at the third and 
fourth levels, respectively. The root variable changed from L/cow, which was exempt from 
available variables, to gross farm revenue per cow, confirming  a strong association between 
these two ratios.  
 
The right branch clustered 625 observations that had gross farm revenue per cow higher than 
CLP$956,500, equivalent to NZ$2,424. The second level split was made by the ratio 
Supplement fed per litre of milk, at the threshold value of 295.3 g/L. DMUs using lower le vels 
of supplement were further split to the outer side. Following this pathway, operating profit 
created the highest efficiency class (0.9572, n = 251) , exhibiting a profit per cow greater than 
CLP$303,200, equivalent to NZ$768/cow. The third highest effi ciency class (0.9254) 
comprised those 184 observations that had a lower profit than the threshold and also produced 
more than 6,574 L/milked cow.  
  
The root variable also split 403 observations to the left branch. Surprisingly, the second best 
efficiency class (0.9472) was located on this branch, at the end of a very different pathway than 
that followed to reach the best performing leaf. This second best class had a lower revenue per 
cow than the threshold, but this was offset by a lower than CLP$13.39 or NZ$0.34 per 10 L 
value for wages per litre of milk produced. In contrast, the lowest efficiency class also had a 
lower revenue per cow, but a much higher value of wages per litre at CLP$20.44 or NZ$0.52 
per 10 L; this infers that paid labour was 53% less productive on these low efficient 
observations. This class was also affected by higher direct costs of production,  and again, GFR 
CLP$/cow. Table 37 highlights the order and frequency of appearance of the variables exposed 
by both trees based on milk DEA  results,  from root to fourth level split as revealed by F igures 
17 and 18 . 
Chapter VI: TodoagroBase Results 
140 
Table 37   
Key Performance Indicators (KPIs) as Revealed by MS Trees 
 
Benchmarks, such as L/Cow, GFR/cow, Supplement g/L, and Wages/L, appeared as very strong 
variables at the first, second, third, and fourth levels, and repeatedly played a role in the 
physical–oriented trees. The structure of these trees is more or less sim ilar to that found for the 
New Zealand trees. The milk production ratio per cow consistently appeared on top fo the tree, 
and labour, expenditure, and profit also came out as strong variables.  
6.2.2 GFR results. 
The following results used DEA scores based on the economic output, GFR, instead of milk 
production. Figure 19 presents a tree based on an analysis that included all Chilean DMUs, 
ratios, and variables.  
 
KPI Level
L/cow 1ST
Wages CLP$/L 2ND R
Production cost CLP$/L 2NDL
Supplement g/L 3RD R
Wages CLP$/L 3RD L
GFR CLP$/cow 4TH R
Total cost CLP$/L 4TH L
GFR CLP$/cow 1ST
Supplement g/L 2ND R
Wages CLP$/L 2NDL
Operating profit CLP$/cow 3RD R
Production cost CLP$/L 3RD L
L/VO 4TH R
Total cost CLP$/L 4TH L
Figure 17
Figure 18
Chapter VI: TodoagroBase Results 
141 
 
Figure 19. GFR DEA partitioning tree.
Chapter VI: TodoagroBase Results 
142 
Figure 19 exposed 11 terminal nodes and four levels of partition. The top level variable was 
again gross farm revenue per cow at the critical value of CLP$952,400/cow, equivalent to 
NZ$2,413. The sample was split into the  right (n = 631) and the left branch (n = 394). The 631 
observations on the right branch had GFR per cow higher than the threshold value and were 
clustered into five leaves. The second split on supplement fed to the cows (g/L) led to the 
largest class (0.9535, n =266) , comprising the observ ations that combined lower than 293.9 g/L 
use of supplement, with an operating profit higher than CLP$289,200/cow, equivalent to 
NZ$733.  
 
On the left branch, six small classes were drawn, including the highest efficiency class (0.955, n 
= 58). DMUs at low er than the threshold value for the root variable were further split by direct 
production cost per litre at the critical value of CLP$121.50. The highest efficiency class 
comprised observations with lower production costs that paid less than CLP$8.014 mill ion or 
NZ$20,307 in total wages. In contrast, the lowest efficiency class (0.6954, n = 55) was found at 
the end of the outer pathway and had a higher production cost; it paid more than CLP$20.26/L 
on wages and exhibited a GFR lower than CLP$75,100 or NZ$1,964 per cow.  
 
In the search for new interactions and replicating the succession of steps taken with the New 
Zealand data, a modified tree was proposed based only on inefficient farms. This subsample, 
consisting of 807 DMUs with revenue efficiency score below 1, was set apart from 222 efficient 
farms (DEA score = 1).  Figure 2 0 presents a tree including all Chilean ratios, and variables but 
excluding all efficient (score = 1) observations.  
Chapter VI: TodoagroBase Results 
143 
 
 
Figure 20. Modified GFR DEA partitioning tree. 
Chapter VI: TodoagroBase Results 
144 
Figure 20 presents the fourth tree of the series. Similar to Figures 18 and 19, it had gross farm 
revenue per cow as the root node at a lower threshold of CLP$827,500, and it split the sample 
into 605 observations to the  right and 202 to the left. In fact, the variables were almost the same 
as those that appeared in the previous trees, but the critical values all tended to be lower. 
Interestingly, ‘Year’ came out as a fourth level variable, which never happened before in the 
previous trees. This suggests that in 2011 (b) and 2007 (d), there was a price component 
affecting revenue efficiency that could not be removed by the class- specific DEA. Table 38  
highlights the order and frequency of appearance of the variables expose d by both trees based 
on revenue DEA results, from root to fourth level split as revealed by Figures 19 and 20.  
 
Table 38  
Key Performance Indicators as Revealed by GFR Trees 
 
 
In summary, as in the NZ case the RPT approach  applied to TodoagroBase showed multiple 
pathways to achieve a high efficiency; revealed those important indicators in defining efficiency 
classes; presented a number of threshold or pivotal values that determine the pathway and the 
KPI Level
GFR CLP$/cow 1ST
Supplement g/L 2ND R
Production cost CLP$/L 2NDL
Operating profit CLP$/cow 3RD R
Wages CLP$ 3RD L
L/cow 4TH R
Wages CLP$/L 4TH L
GFR CLP$/cow 1ST
Operating profit CLP$/cow 2ND R
Subtotal Prod,Adm Costs CLP$ 2NDL
L/VO 3RD R
Wages CLP$/L 3RD L
Year 2008 and 2007 4TH R
Figure 19
Figure 20
Chapter VI: TodoagroBase Results 
145 
class where a particular DMU may sit ; and ultimately exposed the pathway or combination of 
indicators that identified homogenous groups accounting for different dairy performances in 
Chile.  
6.2.3 The fifth Chilean tree 
The striking absence of return on assets (ROA), and return on equity (ROE) in the Chilean trees 
was addressed by running a final tree that only included 862 observations whose asset values 
were considered reliable. The tree is presented in Appendix 5, and shows assets turnover (ATR) 
came out as the root variable at the pivotal value of 29.5%, accompanied by the ratio production 
costs per litre of milk and, again, ATR at the second level. The persistent absence of ROA 
coming out as a KPI was noticeably  and a possible explanation will be addressed in the 
discussion (Chapter VIII) . 
Chapter VI: TodoagroBase Results 
146 
6.3 Chilean Benchmarks Set 
The previous analyses suggested a set of benchmarks appropriate to describe the Chilean 
sample. These comprised variables and ratios that acted at the first, second, and third levels on 
the trees presented earlier. Table 39  shows this set divided into two groups depending on 
whether they were physical or financial indicators.  
Table 39  
Chilean Benchmarks Set Comprising Physical and Financial KPIs 
 
 
The tree presented in Appendix five  used efficiency scores estimated using backwards 
calculated assets as an input in DEA for a small group of observations for which there was more 
confidence in the reliability of the data. Appendix five  shows that ATR comes out as the root 
variable and again at the second level would suggest that the methodology was sensitive to the 
inclusion/exclusion of assets . It also meant that ATR could be included in the list of financial 
benchmarks. However it was not because there was little confidence in the asset values and, 
therefore, it could not be included as a key descriptor.   
The Chilean sample was grouped using the previously explained criteria to classify DMUs : 
inefficient, efficient, and resilient DMUs. Inefficient observations (score < 1 in any of two milk  
or revenue efficiencies) were set apart and made up the largest group (n = 752). The remaining 
276 efficient DMUs (score = 1 in both milk and revenue efficiencies) were clust ered by farm ID 
and split into efficient once or twice (n = 200) , and consistently efficient or resilient three or 
more times (n = 76, called) over the five year period. As happened in the NZ case, t he fact that 
TodoagroBase contains unbalanced panel data could have resulted in a resilient farm being 
classified as non- resilient just because it had  insufficient number of records (minimum of three). 
Results are presented as a summary  (Table 40)  containing descriptive statistics of physical 
indicators for each class, followed by indicator –specific significance tests. Statistical 
Physical KPIs Financial KPIs
L/cow GFR CLP$/cow
Supplement g/L Wages CLP$/L
Production cost CLP$/L
Operating profit CLP$/cow
Benchmarks set
Chapter VI: TodoagroBase Results 
147 
significance was reflected in the number of subsets of means drawn for a p value of 0.05. 
Significance Schaffe tests for unequal class sizes were conducted to compare the groups by 
using one way ANOVA. Financial indicators were then introduced in a similar layout. 
D escriptive statistics of physical variables of interest are shown in Table 40.  
Table 40  
Descriptive Statistics in Physical KPIs 
 
Table 40  shows that over the five year period, all groups produced, on average, almost the same 
output per cow. However, differences can be found looking at subsamples’ characteristics other 
than the mean: the groups of the resilient and efficient showed greater variability than inefficient 
DMUs, as indicated by the standard deviation values. Resilient DMUs also exhibited the highest 
minimum and a moderately high maximum. Interestingly, the e fficient DMUs exhibited the 
highest maximum and the lowest minimum, along with the greatest variability.  
All classes achieved these results by using variable levels of supplement per litre of milk 
produced. Again, the greatest variability is shown by the e fficient DMUs. These,  and the 
resilient farms, used less supplement than the sample mean (215 grams per litre), with som e 
DMUs not using supplements at all. In contrast, the Inefficient DMUs use d supplements at 
levels well above the mean with a minimum of 16 g/L. Also of interest were the maximum 
values for by each class: resilient farms used up to 330 g/L; e fficient DMUs u sed up to 465 g/L; 
and the group of inefficient observations exhibited a maximum of 527 g/L. Tables 41 and 42 
show the comparison of means for the indicators of interest. 
n M SD MIN MAX
L.cow (VM) Resilient 44 5472 1842 2029 8971
Efficient 154 5450 1886 1706 10552
Inefficient 830 5437 1327 1735 8910
Total 1028 5441 1447 1706 10552
Supplement (g/L) Resilient 44 183 98 0 330
Efficient 154 176 102 0 465
Inefficient 830 224 78 16 527
Total 1028 215 85 0 527
Chapter VI: TodoagroBase Results 
148 
Table 41  
Comparison of Means for Litres per Cow Indicator (L/cow) 
 
The comparison of means identified no significant differences between the groups for the 
physical indicator of production ‘L/cow’. Table 41 shows that the figures that looked very 
similar at a glance, were not significantly different one from another. Table 42 shows the 
ANOVA for the indicator ‘Supplement g/L’  
Table 42  
Comparison of Means for Supplement Used Per Litre of Milk produced (g/L)  
 
The mean use of supplement per litre of milk produced was significantly different between the 
resilient and efficient, and the inefficient groups. On average, i nefficient DMUs used 22.4% 
more supplements than the resilient farms and 27.3% more than the efficient observations. 
Considering that production per cow from the inefficient group averaged 5,437 L/cow, the extra 
supplement is equivalent to an extra 223 and 261 kilograms of feed per cow, respectively.  
Descriptive statistics on financial indicators of interest are shown in Table 43 . 
 
 
 
 
 
 
Chapter VI: TodoagroBase Results 
149 
Table 43  
Descriptive Statistics of Financial Indicators 
 
Similar to that observed for the physical indicator, ‘L/cow,’ differences in mean GFR  per cow 
were small between the groups. The comparison of means using the Schaffe test indicated that 
there were no statistical differences either. Again, only differences were found by looking at 
N M SD MIN MAX
GFR CLP$/cow Resilient 44 1057330 383865 343785 1772877
Efficient 154 1086500 466702 318046 3856439
Inefficient 830 1051410 286250 300090 1833900
Total 1028 1056920 323907 300090 3856439
Wages CLP$/L Resilient 44 17.5 7.3 5.7 35.9
Efficient 154 15.6 6.4 5.3 48.0
Inefficient 830 21.1 8.3 4.6 106.6
Total 1028 20.2 8.3 4.6 106.6
Production cost CLP$/L Resilient 44 124.0 25.9 66.5 185.0
Efficient 154 115.3 29.1 57.9 258.5
Inefficient 830 131.9 27.6 67.2 279.0
Total 1028 129.1 28.4 57.9 279.0
SubTot Prod,Adm CLP$/L Resilient 44 137.0 26.2 78.0 195.3
Efficient 154 127.9 29.1 72.6 261.2
Inefficient 830 146.0 28.7 88.0 356.6
Total 1028 142.9 29.4 72.6 356.6
OpProfit CLP$/Cow Resilient 44 264565 188129 -117156 589697
Efficient 154 332365 242094 -309328 945882
Inefficient 830 206434 208872 -1126574 814844
Total 1028 227787 217936 -1126574 945882
Chapter VI: TodoagroBase Results 
150 
other descriptive measures. The resilient and efficient observations showed much greater 
variability than the group of inefficient DMUs, as denoted by the standard deviation values. 
Interestingly, the resilient farms had the smallest range, with the highest minimum and the 
lowest maximum. Also of interest was that the efficient DMUs had  the highest maximum and 
the second lowest minimum, along with the greatest standard deviation.  
D ifferences were significant when the other indicators were analysed. The three classes spent 
very differently as demonstrated by results on three different cost ratios per litre of milk 
produced: wages, direct costs of production, and the sum of production and administration 
costs. The group of resilient farms showed an intermediate behaviour in all indicators 
considered, having the lowest standa rd deviation in the production, and production plus 
administration cost ratios. This group also showed the lowest variability in operating profit per 
cow. The efficient DMUs performed well overall: they exhibited the highest mean revenue and 
profit, but al so the lowest minimum cost ratios. Their weakness as a group is their high 
variability in all these parameters as demonstrated by they low minimums and high maximums, 
which in almost all cases resulted in high standard deviations.  
Table 4 4 shows the comparison of means for the wages ratio.  
Table 44 
Comparison of Means for the Wages ratio (CLP$/L) 
 
The wages paid per litre of milk produced were not s ignificantly different between the resilient 
and efficient DMUs , and both had significantly lower wage ratios than the i nefficient group. 
This pattern changed when more inclusive ratios were taken into account.  
Table 45 present the compar ison of means for the indicator of direct production costs.  
 
 
 
 
Chapter VI: TodoagroBase Results 
151 
Table 45  
Comparison of Means for the Production Cost Indicator (CLP$/L) 
 
 
Production cost per litre of milk produced as well as the subtotal of production plus 
administration cost showed a similar behaviour. Tables 4 5 shows  that the resilient and efficient 
groups had no significant differences in terms of expenditure, with  only the efficient group 
having a mean significantly lower than the inefficient group.  
Table 46 shows the comparison of means for the operating profit per cow indic ator   
Table 46  
Comparison of Means for the Operating Profit per Cow Indicator (CL$/Cow) 
 
The resilient farms had no significant differences regarding mean operating profit per cow when 
compared against the efficient observations. However, neither had any differences related to the 
inefficient group. In turn, the mean operating profit exhibited for this last group was 
significantly lower when compared against the group of efficient DMUs.  
  
Chapter VI: TodoagroBase Results 
152 
 
Chapter VII: Metafrontier analyses Results 
153 
7. Metafrontier Analyses. 
 
Chapter 7 presents the individual New Zealand and Chilean metafrontier results, followed by a 
metafrontier analysis across both countries for two selected years of interest. A ‘good’ year was 
represented by the 2010/ 11 season in New Zealand and the 2011  year in Chile, while the 
contrasting ‘poor’ year was the 2008/ 09 season in New Zealand and the 2009  year in Chile. 
Inputs used in these analyses were only three:  herd size (cows), supplements (kg/cow), and 
CPI –adjusted operating expenses expressed in US$. The selection of inputs was reduced to the 
variables present in both databases that were reliable, and were selected  to keep consistency 
with the previous efficiency analyses. Assets were deliberately excluded from the metafron tier 
analyses because of the unreliability of the backwards calculation in TodoagroBase.  
Despite the default output in this type o f technical efficiency analysis is ‘milk’, expressed either 
in kilogram of milk solids or in litres, the metafrontier output was CPI - adjusted gross farm 
revenue, also expressed in US$. The use of this variable so affected by output price was the 
straightforward option given the lack of information about milk solids’ content in 
TodoagroBase. Alternatives to this could have been the use of several assumptions to correct 
litres of milk contained in TodoagroBase to energy corrected milk  (ECM)  using the IFCN 
methodology. However, the revenue option avoided using the assumptions, while price 
information meant the results could be contextualised in the discussion.  
Similar to Chapters 5 and 6, a  summary table per country shows the descriptive statistics for the 
metafrontier results and the inputs/outputs used, followed by a series of tables showing 
statistical differences revealed ANOVA and Schaffe tests.   
Chapter VII: Metafrontier analyses Results 
 
154 
7.1 New Zealand Metafrontier Results 
The New Zealand metafrontier used all the observations, across all years, totalling 2,588 DMUs. 
Table 47 presents the descriptive statistics for the metafrontier results first, then the output and 
the three inputs used in the New Zealand metafrontier efficiency analysis.  
Table 47  
Descriptive Statistics and Inputs/Output used for Metafrontier Efficiency in New Zealand 
 
n M SD MIN MAX
Metafrontier efficiency
200607 618 0.37 0.11 0.16 1
200809 496 0.37 0.11 0.19 1
200910 564 0.41 0.11 0.22 1
201011 296 0.48 0.12 0.28 1
200708 614 0.51 0.14 0.25 1
Period 2588 0.42 0.13 0.16 1
GFR CPI US$
200809 496 702,635 576,887 62,802 4,621,358
200910 564 850,997 653,930 90,615 5,116,878
200708 614 1,029,702 864,510 131,996 7,367,019
200607 618 569,488 426,677 74,268 3,981,424
201011 296 1,024,704 852,500 195,351 6,316,417
Period 2588 817,605 702,279 62,802 7,367,019
CPI OPEX US$
200809 496 560,311 492,399 57,822 4,168,995
200910 564 542,761 414,478 68,872 3,489,759
200708 614 562,436 450,873 57,245 3,975,797
200607 618 400,064 307,174 79,361 3,137,947
201011 296 573,093 465,332 113,133 3,058,534
Period 2588 520,186 428,666 57,245 4,168,995
Peakcows
200809 496 459 353 60 2,894
200910 564 461 318 55 2,800
200708 614 424 310 60 2,966
200607 618 408 281 99 2,800
201011 296 444 338 87 3,100
Period 2588 437 318 55 3,100
Supplement kg/cow
200809 496 313 232 0 2,308
200910 564 339 249 0 1,936
200708 614 329 258 0 2,029
200607 618 335 234 0 1,781
201011 296 317 222 0 1,545
Period 2588 328 242 0 2,308
Chapter VII: Metafrontier Analyses 
155 
As expected, the metafrontier scores were lower than the individual year results presented in 
Chapter 5, which were in the range of 0.8 to 0.9. The highest average efficiency was observed in 
2007/ 08, and the lowest in 2006/ 07. The standard deviation values were relatively low though, 
which suggests moderate variability and consistent results. The lowest minimum value (0 .16) 
was recorded in 2006/ 07, and the highest minimum value was estimated in 2010/11 (0.28). T he 
years with lower efficiencies had lower milk prices and, conversely, the higher efficiency years 
had higher milk prices.  
Table 48 shows the annual comparison of means for the metafrontier efficiency scores.  
Table 48  
Comparison of Means for the Metafrontier Scores 
 
Efficiency produci ng revenue over all the years were significantly different from each another, 
except for 2006/ 07, which was statistically similar to  2008/ 09. Efficiency was relatively low 
during 2006/ 07 (0.37) due to a comparatively low milk price, then efficiency peaked in 2007/ 08 
(0.51) and came back to the 2006/ 07 level in 2008/ 09 (0.37). In 2009/ 10, it was significantly  
higher again (0.41) , and continued to increase in 2010/ 11 (0.48). A similar pattern can be 
observed in Table 49 for  the metafrontier output, CPI - adjusted gross farm revenue.  
Class n
200607 618 0.37
200809 496 0.37
200910 564 0.41
201011 296 0.48
200708 614 0.51
Metafrontier efficiency
Chapter VII: Metafrontier analyses Results 
 
156 
Table 49  
Comparison of Means for the CPI -adjusted Gross Farm Revenue 
 
The comparison of means shows a similar pattern to the average metafrontier scores through the 
period, except for years 2007/ 08 and 2010/ 11 that were similar in statistica l terms. Interestingly, 
the CPI - adjusted GFR increased 81% from the beginning to the end of the period, whereas 
efficiency only improved by 38%. T able 50  presents the comparison of means for the CPI -
adjusted operating  expenses. 
Table 50  
 Comparison of Means for the CPI-Adjusted Indicator Operating Expenses 
 
Total operating expenditure, on average, was only statistically different from other years in 
2006/ 07. It is interesting that even when facing a bad milk price scenario in 2008/ 09, expenses 
were not significantly lower or higher than other more favourable years. This suggests a high 
share of fixed costs in total expenses in New Zealand. Throughout the period, and under 
ANOVA and Schaffe te sts, there were no significant differences in the average size of the herd 
or in the use of supplements per cow, which were two inputs also used in the metafrontier 
analysis. 
Class n
200607 618 569,488
200809 496 702,635
200910 564 850,997
201011 296 1,024,704
200708 614 1,029,702
CPI GFR US$
Class n
200607 618 400064
200910 564 542761
200809 496 560311
200708 614 562436
201011 296 573093
CPI OPEX US$
Chapter VII: Metafrontier Analyses 
157 
7.2 Chilean Metafrontier 
The Chilean metafrontier used the observations across all years, totalling 1,028 DMUs. Table 
51 presents descriptive statistics of results, output and inputs used .  
Table 51  
Descriptive Statistics and Inputs/Output used for Metafrontier Efficiency in Chile 
 
n M SD MIN MAX
Metafrontier efficiency
2007 245 0.68 0.11 0.47 1
2008 231 0.74 0.10 0.51 1
2009 231 0.56 0.12 0.34 1
2010 197 0.67 0.12 0.42 1
2011 124 0.72 0.11 0.47 1
Period 1028 0.67 0.13 0.34 1
CPI GFR US$
2007 245 569,645 509,866 37,934 4,079,934
2008 231 674,610 543,042 38,001 4,276,795
2009 231 403,872 371,571 18,120 2,591,149
2010 197 685,142 617,539 44,454 4,289,108
2011 124 904,760 919,340 50,684 5,972,281
Period 1028 618,536 595,916 18,120 5,972,281
CPI OPEX
2007 245 426,303 389,302 27,097 3,536,960
2008 231 477,375 386,868 26,105 2,998,412
2009 231 359,477 320,961 16,398 1,956,635
2010 197 554,249 483,154 40,191 2,990,723
2011 124 657,845 698,207 33,940 4,596,861
Period 1028 475,211 452,043 16,398 4,596,861
Cows
2007 245 250 182 25.3 1207
2008 231 251 167 22 1225
2009 231 270 199 29.2 1404
2010 197 292 221 32 1738
2011 124 369 306 24 2118
Period 1028 277 212 22 2118
Supplement Kg/cow
2007 245 1252 671 3 2697
2008 231 1323 660 2 2652
2009 231 1128 651 0 2680
2010 197 1152 591 0 2609
2011 124 1370 606 4 2683
Period 1028 1235 647 0 2697
Chapter VII: Metafrontier analyses Results 
 
158 
As expected, the metafrontier scores were lower than the individual year results presented in 
Chapter 6, with most of these in the range 0.8 to 0.9. The highest and lowest average 
efficiencies were recorded in 2008 (0.74) and 2009 (0.54). The results were consistent as shown 
by relatively low standard deviation values. The lowest minimum value (0.34) was recorded in 
2009, and the highest minimum value was estimated in 2011 (0.51). Table 52 shows the 
comparison of means for the metafrontier efficiency scores.  
Table 52  
Comparison of Means for the Metafrontier Scores 
 
The metafrontier efficiency in 2009 was significantly lower than the rest of the period. The 
years 2010 and 2007 were characterised by intermediate mean efficiencies, while the years 2011 
and 2008 stood out with higher mean efficiencies. Throughout the per iod, there were alternative 
highs and lows, which may reflect the influence of weather and price shocks. Overall, New 
Zealand and Chile behaved in similar ways: the New Zealand 2008/09 season can be compared 
to the Chilean 2009 year, even if not exact ly the same time period. Correspondingly, the New 
Zealand 2010/ 11 season can be compared to the Chilean 2011 year. Table 53 shows the 
comparison of means for the CPI - adjusted gross farm revenue.  
 
 
 
 
 
 
 
Class n
2009 231 0.56
2010 197 0.67
2007 245 0.68
2011 124 0.72
2008 231 0.74
Metafrontier efficiency
Chapter VII: Metafrontier Analyses 
159 
Table 53  
Comparison of Means for the CPI-Adjusted Gross Farm Revenue 
 
The mean GFRs in 2009 and 2007 were not statistically different. In addition, 2007 was found 
to be statistically no different from 2008 and 2010. The 2011 year was clearly different from the 
others, exhibiting very high mean revenue. Interestingly, the CPI - adjusted GFR was 59% higher 
in 2011 than at the beginning of the period (2007) , and more than doubled the US$403,872 
average in 2009. As in the New Zealand analysis, the changes in efficiency were smoother than 
the fluctuations observed in mean gross farm revenue; in addition, the greatest difference 
between maximum and minimum efficiencies was in the order of 32%. Table 5 4 presents the 
comparison of means for the CPI - adjusted  operating expenses.  
Table 54   
Comparison of Means for the CPI-Adjusted Operating Expenses 
 
From 2007 to 2011, despite fluctuations, there was an overall trend in increasing operating 
expenses over time. However, there were no statistical differences in operating expenditure in 
the first three years of the period. Similarly, operating expenses in 2007, 2008, and 2010 were 
Class n
2009 231 403,872
2007 245 569,645 569,645
2008 231 674,610
2010 197 685,142
2011 124 904,760
CPI GFR US$
Class n
2009 231 359,477
2007 245 426,303 426,303
2008 231 477,375 477,375
2010 197 554,249 554,249
2011 124 657,845
OPEX US$
Chapter VII: Metafrontier analyses Results 
 
160 
not statistically different, while the expenditure in 2010 was similar to that of 2011. The 
minimum mean expenditure was recorded in 2009, and the maximum in 2011. Table 55 
presents the comparison of means for the input variable, herd size.  
Table 55  
Comparison of Means for the Physical Input Cows 
 
Herd size showed an increasing trend. However, only the last year of the period was 
significantly different from the rest of the years. This trend to increasing numbers of cows in the 
herd was accompanied by a similar tendency towards intensification, as d enoted by the higher 
use of supplements per cow found in Table 56.  
Table 56  
Comparison of Means for the Physical Input Supplements per Cow 
 
The beginning of the period in 2007 showed a moderate use of supplements at 1,252 kg/cow on 
average. This was not statistically different from the supplements used in 2008, 2009, 2010, and 
2011. However, there were differences between supplement feed to the cows in 2011 relative to 
2009 and 2010, which were at a lower level in response to an unfavourable milk:feed price ratio.  
Class n
2007 245 250
2008 231 251
2009 231 270
2010 197 292
2011 124 369
Cows
Class n
2009 231 1128
2010 197 1152
2007 245 1252 1252
2008 231 1323 1323
2011 124 1370
Supplement Kg/cow
Chapter VII: Metafrontier Analyses 
161 
7.3 Metafrontier across countries 
The last section in this Chapter presents the metafrontier across countries, using CPI - adjusted 
gross farm revenue (US$) as the single output. CPI - adjusted  operating expenses (US$) , cows, 
and supplement (kg/cow) were used as inputs. Descriptive statistics for these output and inputs 
can be found in the previous sections.  
The cross- country metafrontier addresses the research question of which country is more  
efficient and why. F rom the previous analyses, two contrasting years were chosen and 
metafrontier analyses conducted separately for them. These years are considered contrasting 
because of both price and climate: in both countries, 2008/ 09 and 2009 represe nt a poor year for 
New Zealand and Chile, respectively, and 2010/ 11 and 2011 a good year. Note that a relatively 
less favourable weather scenario affected the New Zealand pastoral season in 2010/ 11, with 
officially declared areas of drought across the country (MAF, 2011). Therefore, t he metafrontier 
results must be interpreted with care. In addition, it must be remembered that the years do not 
refer to exactly the same period, since New Zealand uses the agricultural year (June- May ), 
while Chile uses a calendar year (January - December). In addition, results across years are not 
strictly comparable since different farms made up the samples each year. Figure 21 shows the 
descriptive statistics for the two New Zealand years, 2008/ 09 and 2010/ 11, and their Chilean  
counterparts, 2009 and 2011.     
 
Figure 21. Boxplot of metafrontier efficiency across countries. 
The horizontal lines across the boxes  in the box plot  represent the medians for each country–
year combination. Those in the top quartile (25% ) with the higher efficiency are shown by the 
top ‘whisker’ and dots. Dots usually represent outliers b ut, in this case, these are DMUs with 
distinctively high efficiency scores. Conversely, the bottom ‘whisker’ represents the lower 
quartile and, although they may occur, there were no low effici ency dots in any of the country-
Chapter VII: Metafrontier analyses Results 
 
162 
year combinations. The whiskers  ends show the minimum values present in each subset. The 
poor year, 2008/ 09, severely affected revenue efficiency in New Zealand (mean = 0.38) , which 
was consistently below the mean for the whole period (0.41), with a relatively low standard 
deviation (0.11) in a relatively large sample (n = 494). In the equivalent yea r in Chile (2009), 
the Chilean sample (n = 231) had  a lower efficiency (0.37) and also greater variability (0.18) 
than the New Zealand sample. D ifferences in average efficiencies between DairyBase and 
TodoagroBase samples during the poor year were not significant. Table 57 presents the 
comparison of means for the metafrontier across countries in the selected years of interest.  
Table 57  
Comparison of Means for the Metafrontier across Countries  
 
The good year, 2010/ 11 in New Zealand and 2011 in Chile, showed a recovery in efficiency for 
both countries. The mean efficiencies were both significantly higher than those observed in the 
poor year. New Zealand ’s  mean efficiency (0.49) was also significantly higher than the Chilean 
mean efficiency (0.41) and, despite the smaller samples, the variability in the results was greater 
on average, as shown by standard deviation values of 0.13 (New Zealand) and 0.15 (Chile). The 
minimum metafrontier efficiency score observed in New Zealand in  2008/09 was 0.19 and in 
Chile in 2009, it was 0.16. The minimum estimates in the good years were 0.28 and 0.18, 
respectively. It is clear that in these types of years, there are Chilean observations doing very 
well, and also very badly, as demonstrated by  standard deviations.  
Means comparison were also carried out with the inputs used in the construction of the 
metafrontier, which were the same as those used in the country - specific metafrontier (refer to 
Tables 47 and 51) . The significance tests showed that in the poor year, gross farm revenue was 
significantly lower in Chile compared to New Zealand, and both revenues were significantly 
lower than those observed in 2011, but not significantly different one from the other. However, 
in 2009, the mean expenditure was also significantly lower in Chile than the expenditure me ans 
for the other three country- year combinations (no difference between them). In addition, the use 
of supplements was significantly lower in New Zealand t han in Chile for all the years under 
review
Class n
CHILE 2009 231 0.37
NZ 0809 494 0.38
CHILE 2011 124 0.41
NZ 1011 293 0.49
Metafrontier efficiency
Chapter VIII: General Discussion 
163 
8. General Discussion. 
 
The purpose of this study was to describe and compare key drivers of NZ and Chilean pasture -
based dairy farming systems. These countries are natural benchmarking partners ,  since farms in 
both of them transform similar inputs or resources,  into similar outputs or products (Bogetoft & 
Otto,  2011) . The systems were assessed in terms of their internal resources, key performance 
indicators, and drivers of success. This chapter discusses the key findings presented in the 
results (Chapters 5, 6 and 7) and their implications. Three independent sources of data were 
used to assess and benchmark the NZ and Chilean pastoral systems. The International Farm 
Comparison Network  data was used at the individual farm businesses level,  while at the farming 
systems level, the research used longitudinal panel data from DairyBase  from NZ and 
TodoagroBase from Chile. 
The IFCN farm to farm comparison (Chapter 4) confirmed the expected international  level of 
competitiveness and the cost advantages in both countries. The typical farms in each country 
were able to profitably produce milk  at a lower cost than the world’ s average. These farms 
demonstrated Porter’s (1980) cost efficiencies, effectively outperforming other farms in the 
IFCN comparison  when their cost profiles were analysed, and exhibiting t echnically feasible 
input- output combinations. The cost advantages in agricultural systems are found in those 
countries or regions where particular climate, soil conditions, or labour supply allow more 
efficient farming than in other regions (Kay, Edwards & Duffy, 2012). Both countries exhibit ed 
all these characteristics in their non- mountainous territory ( approximately  half of the total area) 
which contributes to them being lower cost producers compared to the world’s average.   
The assumption of perfect competition was considered appropriate in the context of NZ and 
Chile, where farmers do not have any control over milk pay out, accept input market prices as a 
given, and seek to minimise expenditure in variable inputs . The NZ dairy industry in particular, 
has adopted a widely used strategy of cost leadership ( Akoorie & Scott - Kennel, 1999; Shadbolt, 
2011).  The IFCN data also showed that the farms’ returns for both countries were well below 
the world’s average due to milk prices that more (NZ) or less (Chile) closely track international 
prices and the absence of quotas and subsidies. E fficient- scale processes and an on- going cost 
minimisation throughout innovation can be assumed for both countries given the farm sizes and 
cost profiles. According to Spulber (2004) these  processes are critical in order to achieve and 
maintain a cost- leadership position. However, NZ and Chilean farming systems demonstrated 
that a good international positioning can be achieved by using a set of inputs in different 
quantities  or by approaching the cost- leadership strategy in different ways. For instance, while 
Chapter VIII: General Discussion 
 
164 
expenditure in fertiliser  on average is higher in NZ than in Chile , conversely, expenditure in 
supplementary feed on average is greater in Chile. Shadbolt (2011) explained two basic 
different basic approaches to the cost- leadership strategy, one is the low - input approach and the 
other is the high- output approach. The first one is typically associated with NZ farming systems 
and is self- explanatory . The latter is the Chilean approach and is based on reducing the cost per 
unit by producing larger amounts of product.  
The comparison carried out using IFCN data confirmed that larger samples, including data 
recorded over several years, were needed to really assess farming systems in the respective 
countries. Langemeier  (20 10) indicated that benchmarking over several seasons was required to 
determine competitive advantages given the evident difficulty in consistent farming 
performance. Therefore, j udgements about the relative performance of one or more farming 
systems may be better if based on large samples from different regions (Benson, 2008). In 
addition, i ndividual farm performance will naturally fluctuate over time and thus, snapshots of 
performance can be misleading. In any year, a downturn or upturn in performance may be 
found, which in part could be explained by a n industry- wide trend, a general economic 
circumstance, or weather related factors that are out of the control of the manager. According to 
Fleming et al. (2006) failure to identify causal relationships between performance and factors 
such as rainfall and input- output prices has often led analysts into interpretative errors.  
In this study, the relevanc e of weather and market - related shocks  in isolation, and in 
combination, was apparent depending on the magnitude of the event. For example, over the 
period 2006- 11 there were good and poor years, climatically and economically. both in NZ and 
Chile. The NZ results (Chapter 5)  show those years of high milk prices affected  milk efficiency 
and especially revenue efficiency. Conversely, a poor  climatic year was more likely to a ffect the 
physical efficiency than the revenue efficiency depending on the input:output price ratio. The 
summer 2007/08 was the driest in a centu ry in parts of the North Island and South Island of NZ 
( NIWA , 2012) . The following year ( 2008/ 09 )  was again dry: some regions in the North Island, 
and also the South Island had  near record low annual rainfalls. The 2009/ 10 season was also a 
difficult climatic year for most regions, mainly due to a late summer/autumn drought (Dairy 
NZ, 2011). However, 2007/ 08 had exceptionally high milk prices , only comparable to the high 
milk price observed in 2010/ 11, and this helped farmers to overcome the climatic shocks.  In 
contrast, t he 2008 / 09 season in NZ showed how a really poor year can result from the 
combination of both low yields and low product prices.  
DairyBase and TodoagroBase provided the data to characterise NZ and Chilean pasture- based 
dairy farming systems over the 2006/ 07- 2010/ 11 period. Country- specific key metrics were 
identified and used as operating statistics to identify performance gaps, and the competitive 
Chapter VIII: General Discussion 
165 
position of the systems and particular groups of businesses. Before getting into this discussion, 
it is worth reviewing the methodology and some fundamental steps to understand and interpret 
these results. Motivated by the need to compare only similar (peer)  farms the data in both 
databases were nested into several classes using country- specific classification systems. The NZ 
one (27 classes)  included a combination of year, region and three levels of intensification, while 
the Chilean one (10 classes) , limited by  dataset size, combined just year and two levels of 
intensification. DEA within these classes meant that data was normalised to remove differences 
between seasons, regions and intensification level  which were expected from reading the 
literature.  
8.1 The Non-financial Benchmarks 
The three key non- financial benchmarks identified in DairyBase were the milk solids 
production measure ( kgMS/cow ) , the labour input measure ( FTE)  and a productivity of labour 
measure ( MS/ FTE ) . In TodoagroBase ,  the key metrics identified were a milk production 
measure ( L/cow ) , the wages paid relative to production and the feed input measure, supplement 
feed to the cows also relative to production ( grams/L ) . Therefore, some physical benchmarks 
found for NZ , and Chile where similar while others were different. The countries shared similar 
indicators for milk production and labour, however, labour and labour productivity were very 
important in NZ  while in Chile supplementary feed was key  in defining efficiency. Table 58  
summarises the findings discussed in this section.  
 
Table 58  
Production, Labour and Feed Benchmarks 
25  
                                                          
25 DairyBase data was only at level one so did not have the level two indicator KgDM eaten per Kg MS. 
INDICATOR NZ CHILE SIMILAR
PRODUCTION MS/cow L/cow yes
LABOUR FTE Wages CLP$/L yes
Paid wages yes yes
Paid management yes no 
Family labour yes no 
Family management yes no 
Share of Operating Expenses 20% 14%
PRODUCTIVITY OF LABOUR MS/FTE WAGES/L yes
FEED no g/L no
Chapter VIII: General Discussion 
 
166 
Table 58  presents non- financial benchmarks, except the Chile an labour indicator presented here 
to facilitate the comparison with the physical NZ indicator full time equivalent (FTE). All the 
non- financial metrics are discussed in this section while the indicator Wages/L is discussed at 
the beginning of the financial benchmarks section.  
Since m ilk is both a suspension and a solution, it can be  measured in either volume or weight 
(Draaiyer et al., 2009). I n NZ, the driver for dairy production is milk solids for export, rather 
than fluid milk to supply the small domestic market. Therefore , the dominance of kilograms of 
milk solids rather than litres  (which also was available in the data)  in NZ  was not surprising 
given the country- specific payment system. In NZ , farmers are rewarded for milk solids content 
and penalised for volume (Bates, 1998). In contrast, in Chile milk is measured and paid for in 
litres and, thus,  this was the obvious production measure to be expected . The largely domestic 
market  for milk , results in a payment system based on the reverse of the NZ system , rewarding 
volume and providing bonuses for milk composition  (Prolesur, 2011) . This is why milk solids 
content is not even recorded in TodoagroBase. The threshold values at which these indicators 
came out of the trees were between 320 and  406 kg MS/cow in NZ , and between 4,700 and 
5,300 L/cow in Chile  ( roughly equivalent 26  to 470 and 530 kg MS) . These values suggested that 
the production per cow in Chile was greater than in NZ . However, the difference was not as 
large as the observed gap between the IFCN farms (discussed earlier in Chapter 4) , for which 
the Chilean farm has been shown here to be less typical than assumed. 
The dominance of the production measure ‘ per cow’  rather than ‘ per hectare’ , could have turned 
out to be a controversial factor, especially in NZ.  The indicators kg  MS/cow and  L/cow are 
measures traditionally associated more with supplement intensive systems than with pastoral 
farming (Hansen et al., 2005). In pasture - based systems the tendency is make the most out of 
the first limiting factor which usually is land, to the detriment of productivity per cow. In NZ, 
the ratio milk solids produced per hectare has been regarded as the ultimate measure of physical 
performance (Holmes et al., 2002). In Chile, the Department of Agricultural Economics at the 
Catholic University of Chile (1991) , quoted in Silva ( 1997) , identified the most reliable 
indicators of efficiency for a dairy farm as being production of milk per hectare and year . Other 
authors, however, have c hallenged these traditional views about pasture- based farming systems 
and their ultimate production benchmark. Ferreira et al. (2010) explained that L/cow is the 
industry standard benchmark for efficiency in Chile because the main breed is the American 
Hol stein and North American farm management principles predominate. Lerdon et al., (2008) 
and Lobos et al. (2001) do not mention productivity per hectare in the list of factors considered 
particularly important as determinants of results for Chilean dairy farms but included milk 
                                                          
26 Based on the assumption that 10 litres of Chilean milk (3.4% Protein, 3.7% Fat) would be roughly 
equivalent to one kg of MS 
Chapter VIII: General Discussion 
167 
production per cow among the measures. Neto, et al. (2012) , found in the largest milk-
producing region in Brazil that the productivity of the herd was more strongly correlated to 
return on capital than the productivity of the land. Ben son (2008)  suggested that relative 
indicators per cow, and per unit of land, were both necessary physical performance benchmarks 
for pasture- based farming systems regardless of the world’s setting .  
In this study, m ilk production per cow in both countries was strongly associated with technical 
efficiency but milk production per hectare did not come through as a KPI in the trees . Far from 
being controversial, t his was expected  considering the research method (Chapter 3) . Two factors 
were involved that may have had variable weight identifying these ‘per cow’  key benchmarks 
rather than ‘per hectare’ . Firstly, the classification system for each country  may have partially 
removed the land and the stocking rate effects  on production by nesting into year- region-
intensification combinations for analysis. O nce these factors were removed, the trees picked up 
the key variable from what was left. After the nesting, and from the literature, a similar 
production per hectare (within a range) was expected with in the classes and major variability 
could only be explained by individual cow productivity . The second factor is that efficiency 
scores are slightly biased towards the variables used as inputs or outputs, w hich has been 
pointed out as a weakness of the second stage DEA procedures by Simar and Wilson ( 2007 ) . 
Following their reasoning , our results would be biased towards ‘ cows’  rather than ‘hectares’ 
( the latter was excluded from the DEA inputs to avoid multi- collinearity) . However, evidence in 
this study suggested that the estimated efficiency scores and their ranking were  highly reliable,  
which indicates that the first methodological factor had the greater weight in determining the 
results. This particular point will be addressed further at the end of this chapter where methods 
are reviewed and discussed.  
The second important physical benchmark in NZ was total  labour measured as full time 
equivalents (FTE) , suggesting that after milk production, labour cost and labour productivity are 
the k ey drivers for efficiency in NZ . A weakness of this indicator is that despite including both 
paid and unpaid labour and management, it does not distinguish the quality of the input. 
Winkler (1999) concluded that high minimum wages might be  a weakness o f NZ dairy farming 
systems compared to lower- wages systems in developing countries. While t his high labour cost 
may be beneficial as part of an effective social welfare policy ( Bonoli, 2001)  and trigger 
correspondingly higher labour efficiency (IFCN, 2011)  than in other countries, it also  threatens 
NZ’s international competitiveness when competing against lower labour cost countries. The 
dairy industry repeatedly considers the labour issue when looking for options for dairy farming 
in NZ, with  farmers adopting labour saving systems such as once- a- day milking,  and investing 
in technology such as bigger rotary sheds, milking robots ( Robertson, 2010) , or looking at 
fence- less systems (Emerson & Rowarth, 2009) .  
Chapter VIII: General Discussion 
 
168 
O wner- operated family businesses have traditionally been the predominant structure in NZ  
farming systems, with the ownership of the assets combined with managerial control. This 
offers some advantages from the labour perspective. Typically in family businesses, the family 
provides a significant part of the labour requirements  ( Robbins & Wallace, 1992 ) . A clear 
advantage from the labour stand point is that greater confidence in the family working team  and 
elasticity in task allocation  can be expected . This is also advantageous because what is a 
typically fixed cost  can become a variable cost by employing family members as required . 
Fixed costs remain  constant for a business regardless of production levels and are mutually 
exclusive with variable costs unless the labour agreement is fl exib le which is possible in family 
teams. However, perhaps the most positive factor relates to the cash flow advantage: reliance on 
family labour means that wages do not have to be paid rigidly and can be reduced in poor 
income years (Gersick et al., 1997) . All these labour- related factors may be significant in 
defining the business and the labour structures that characterise the NZ  dairy industry.  
 
The importance of labour- related variables in NZ is seen by  the fact that labour came out in the 
analyses as an absolute measure (total labour)  and as a ratio (productivity per FTE) . The drivers 
of total FTE are size of the bus iness and intensification level. This suggested that different 
scales or labour intensities affect efficiency in a major  way. However, there was no clear pivotal 
value of FTE to predict high efficiency , with the indicator appearing at different levels in 
several trees and exhibiting threshold values ranging from 0.58 to 12.44  total FTEs  (most 
commonly between 1.305 and 1.705) suggesting that medium  size farms were on average more 
efficient. However, t his research found that business size was not a limitation to efficiency in 
NZ over the period 2006/ 07- 2010/ 11. Farms employing 12.44 FTE w ere as efficient as those 
employing only 0.58 FTE. Thus, scale should not be  a big concern for dairy business since what 
really matters is having the right combination of inputs. The farm management literature has 
traditionally reported the opposite, empha sising the importance of economies of scale in 
farming (Fox, Bergen, & Dickson, 1993; Olson & Vu, 2009; Purdy et al., 1997) . This apparent 
contradiction may be explained by the fact that most NZ dairy farms are already large enough to 
display significant economies of scale. Moreover, J iang’s results  (2011) show that the 
subsamples of dairy farms from both islands in NZ , on average, exhibited decreasing returns to 
scale. Earlier work by Rouse et al. (2010) found that in 2004/ 05 the herd size  for optimum 
performance was around 300 cows . In our s ample of dairy farms over the 2006/ 07- 2010/ 11 
period, the average herd size  was larger than 400 cows with an associated labour input over two 
FTEs, or 200 cows per labour unit in most regions . National data as repor ted by LIC and 
presented in Figure 3 also demonstrate s that, over the same period, herd size has increased from 
340 cows to approximately 390 on average , and that more than 50% of the herds are  already 
larger than 300 cows.  
Chapter VIII: General Discussion 
169 
The third and last non- financial benchmark of  importance in NZ  was a combination of a 
production and a labour measure, commonly known as labour productivity , expressed  as milk 
solids produced per labour unit (kg  MS/FTE). Over the last decade, labour productivity in NZ, 
measured as cows per full time equivalent, has increased at an annual  rate of 3.1% (Dairy NZ, 
2012). The main driver for this has been the progressive trend to increase herd size , since labour 
in most cases behaves like an indivisible factor of production. Across NZ , efficient farms were 
likely to have higher production per unit of FTE  than less efficient farms. Although the 
relevance of this indicator was found for the whole of the DairyBase sample, some regional 
differences were detected in previous efficiency analyses. Jiang (2011) looked at the regional 
efficiency performance of dairy farming in New Zealand over the period 1998/99 to 2006/ 07 ,  
and found that the different climate, soil, and farming history between the North and South 
Island of NZ resulted in markedl y different dairy farming technologies, and that l abour was 
relatively more important in determining technical efficiency in the North Island, while farming 
systems in the South Island relied more heavily on capital and fertiliser. Laca Vina (2010) 
suggested that given the high levels of technical efficiency of NZ farms in most regions, the 
industry could only expect to achieve major productivity gains through technical progress that 
increases labour and capital productivity. These concordant findings support the results of the 
present study.  
 
I n summary, t o perform efficiently a dairy farm in NZ needed  to produce a given amount of 
milk solids with as few labour units as possible. Production per cow within the range of 340 to 
400 kg MS/cow , combined with a value for labour efficiency of around 45,00 0 kg MS/ FTE ,  
ensured high technical efficiency in producing milk solids in  NZ dairy farms  from 2006/ 07 to 
2010/ 11. In countries such as Chile, where the level of wages is lower than in NZ , the expected 
importance of labour in the overall performance of a dairy farm was lower.  
 
The second important physical benchmark identified i n the TodoagroBase data was supplement 
fed to the cows measured as g/L of milk produced. The importance of supplements in defining 
efficiency in Chile is such that it could not be removed by the classification system. Pasture-
based dairy farming systems in Chile are characterized by grazing herds with a variable 
percentage of the diet being supplements of various types (Smith, Moreira, & Latrille, 2002). A  
Chilean pastoral farm may also house cows fed on concentrates, maize or grass silage and hay 
over the winter period, and rely to some extent o n the use of supplements all year round 
(Challies & Murray, 2006; Smith, et al., 2002).  
 
All of these facts suggest that most dairy farm businesses in Chile could be classified as System 
4, 5 or above  using the Dairy  NZ classification system , which was verified in this research . In 
Chapter VIII: General Discussion 
 
170 
addition, it  is well reported that the feed cost was the largest proportion of the total cost in most 
dairy farming systems in the world, including Chile, and total cost of feed has a direct and 
positive relationship with the use of supplements (IFCN, 2010, 2011 and 2012). T herefore, the 
importance of this benchmark in Chilean farming systems is well justified and was not a 
surprising finding. This raises the question of whether the Chilean agricultural systems behave 
more like confinement systems than like typical pa sture- based systems. However, this question 
was not considered a fundamental one since the answer depends on how relaxed the criteria 
used to define ‘pasture - based farming systems’ are.   
 
The Chilean indicator, grams of supplement fed to the cows per litre of milk produced,  allows 
for the calculation of response to the use of supplements. Milk production responses of grazing 
cows offered supplements reported in the literature are positive, and quadratic ,  varying between 
30 and 150 g of milk solids  per kg of dry matter intake (Auldist et al., 2013). Given this large 
variability and the cost of supplementary feed in NZ, i t may be important to consider the 
relevance of the Chilean indicator, grams used per unit of output, at high intensification levels 
(farming systems 4 and 5).  
8.2 The Financial Benchmarks 
Fleming et al. (2006)  warned that the use of partial benchmarks would not provide an accurate 
picture of whole- farm performance unless these are contextualised adequately . Whole farm 
measures would be valid indicators of economic performance while partial measures would not. 
The tree analyses showed that to predict an efficiency class, in most cases, physical indicators 
needed to be assisted by financial benchmarks. Two exceptions to this rule were MS/cow  alone, 
and in combination with MS/FTE , as shown in Figure 13. Thereafter, non- physical benchmarks 
which include expense and revenue measures will be referred to as financial benchmarks. The 
NZ trees revealed five financial benchmarks: return on assets (ROA), operating profit (OpP/ha), 
operating profit margin (OPM), operating expenses (OpEx/kgMS), and asset turnover (ATR). 
The Chilean trees also revealed five benchmarks as follows: gross farm revenue (GFR/cow), 
wages (CLP $/L), production costs  (CLP$/L), and operating profit (CLP$/ cow).   
 
Given that there are some similarities between the physical NZ benchmark ‘FTE’  and the 
Chilean benchmark for labour, ‘wages CLP$/L ’ (Table 57)  will be discussed first. The literature 
showed conflicting opinions about the relevance of this indicator. Neto et al., (2012) found that 
regardless of the operating scale, labour , and labour productivity, were only weakly correlated 
with return on capital in Brazilian dairy farms where most labour is provided on low wages 
similar to Chile (IFCN, 20 11, 2012). In contrast, l abour cost per litre of milk has been regarded 
as an important indicator in Chile (Bywater, 2010).  Along with this our findings show that 
Chapter VIII: General Discussion 
171 
labour input was also a very important indicator for Chilean farming systems. As in NZ, t he 
Chilean pasture- based dairy farm is usually family owned but the concepts discussed earlier 
about the family labour input are not particularly relevant to the Chilean context. Since the 
wages are low, the Chilean family far m frequently  has absentee land lords and thus, it is not 
owner- operated (Anrique & Latrille, 1999). For the same reason, in the majority of the cases t he 
contribution of the children and/ or the partner is negligible.  
The lack of a physical measure for labour  in TodoagroBase makes the cost of ‘wages’ the only 
recorded measure of labour input. According to the IFCN (2011) the family input as a 
percentage of total labour input on a typical Chilean dairy farm would be 7% , and thus, wages 
would be a sound approximation to total labour input. In  this research,  TodoagroBase revealed 
that the average share of wages in the direct cost of production was 16% , and in total cost only 
14%. Although this is not a high proportion, t his variable was second in relative importance 
after feed cost ( 41 %  of total cost on average including concentrate and pasture cost) . Therefore, 
the one indicator that was not to appear in the Chilean sample was necessary in explaining  an 
important part of the total variability in efficiency as it did with the NZ sample . Table 59  
summarises the findings discussed in this section.  
Table 59  
Financial Benchmarks and Similarities Across Countries 
 
** Fifth Chilean tree refer to Appendix five  
Dairy return on assets was the most important financial benchmark  revealed by the N Z trees 
which concurs with the dairy industry view of the farm business (DairyBase, 2009). Other 
research in NZ has indicated that the most important measure of profit is ROE , which according 
to the Du Pont model, is fundamentally similar and closely linked to ROA (Shadbolt, 2011) . In 
this research it was shown that ROA was the closest to an efficiency measure of all the financial 
indicators (including ROE )  because it has a relatively high focus on the capital used for 
INDICATOR NZ CHILE SIMILAR
PROFITABILITY ROA no no
OPM ATR** no
Operating Profit/ha Operating profit/cow yes
per unit input per unit input
EXPENDITURE Opex/kgMS Production cost/L yes
per unit output per unit output
REVENUE no GFR/cow no
per unit input
Chapter VIII: General Discussion 
 
172 
production, and captures multiple inputs and outputs in the same way that DEA did. Note that 
this benchmark was highly correlated with both milk  and revenue efficiency, and consistently 
came through with the threshold value of 4.5%  in the trees. The partitioning into the left and 
right branches where ROA acted as a root variable (Figures 14 and 15)  suggested that most NZ 
dairy farms are in a strong to medium position related to ROA 27 . ROA is widely used across 
many industries and types of businesses, and is recommended as the leading indicator of the 
earning capacity of the asset (Boehlje, 1994; Doehring, 2001). If assets are valued at market 
value, ROA can be looked at as the opportunity cost of farming relative to  alternate investments 
(Olson, 2010). In NZ, its calculation uses  the value of opening assets, while in Chile annual 
averaged values are used following American farm management principles ( Ferreira et al., 
2010) . The most common concern associated with the ROA measure is its reliance on asset 
valuations. However, this is not an issue  in NZ since the as set value is defined using market 
values and there is confidence in the valuation of the properties. 
The results presented in Chapter 5 also indicated that ROA, at the pivotal value of 4.5% ,  
interacted with physical indicators such as MS/cow and MS/FTE to create efficiency classes, as 
well as with other financial ratios such as rent and interest paid, and equity. ROA also interacted 
strongly with the operating expenses ratio at a critical value of NZ $ 4.39/ kg M S.  For instance, 
in Figure 14,  m ore than one- fifth of the DairyBase sample (582 observations) were found to be 
virtually efficient at producing milk  ( DEA efficiency score 0.96) following the pathway of ROA  
> 4.5%  and relatively low operating expen ses (< $4.39/ kg MS) . All these interactions were 
logical, as was the fact that ROA came through as a key performance indicator closely linked to 
efficiency. Two major strengths of ROA are firstly that it is a comprehensive indicator in which 
cash and non- cash expenses are truly represented, and secondly that it allows for comparison of 
farms which pay family wages with those than do not, or the comparison of farms which own 
run- offs to those that do not (DairyBase, 2009) .  
Operating profit per hectare was the second financial benchmark of importance revealed by the 
NZ trees. This indicator was not completely normalised by the classification system (as it 
happened with the kg MS/ha) which could suggest that it was stronger at defining efficiency 
than milk production per hectare (kg MS/ha). Its use is not as widely spread as the use of ROA 
or net farm income, but in NZ it has been postulated as being critical for dairy farming systems 
(DairyBase, 2009). In NZ, t he use of relative measures expressed on a per kilogram of milk 
solids, per cow or per hectare basis is encouraged to make comparisons between different sized  
businesses possible. Operating profit is a very comprehensive indicator, calculated after several 
adjustments have been made to the net farm  income including depreciation. The presence of 
                                                          
27 Based on Kohl and Wilson (1997) where strong > 5% when assets are mostly owned and > 12% if 
assets are leased; weak < 1% and < 3% respectively. 
Chapter VIII: General Discussion 
173 
Op erating Profit/ha among the key performance indicators suggests that it is strongly correlated 
with technical efficiency, and more precisely, with revenue efficiency. However, it has been 
stated that although operating profit is undoubtedly a measure of business performance, it is not 
always an indicator of operating efficiency (Shadbolt & Gardner, 2005).  
Operating profit margin was the third benchmark of importance revealed by the NZ trees. It is 
one of four major indicators recommended in NZ  for measuring profitability in farming ( along 
with net farm income, ROA, and return on equity)  (DairyBase, 2009) . OPM measures the firm's 
ability to generate revenues and control costs. Since its  major drivers are operating profit and 
gross farm revenue, OPM becomes the most comprehensive expression of the input/output ratio. 
When leasing costs are not deducted, OPM is an accurate indicator of the business’s operating 
efficiency in producing profit from its revenue (S hadbolt, 2011). In this research, OPM was 
strongly associated with kg  MS/cow  suggesting that farmers who get the most out of their cows 
are also financially more efficient. The pivotal values in the trees for OPM were between 22 .5 % 
and 26. 5%,  indicating a strong to stable position28  for most NZ farms through the 2006/ 07-
2010/ 11 period. F or each dollar of revenue earned, most NZ dairy farmers were above the 
pivotal values, and thus retained more than one fifth of returns as profit with the rest being 
expenses.  
Operating expenses ratio per kilogram of milk  solids was the fourth benchmark to come through 
in the NZ trees. The importance of a cost measure was not unexpected. The importance of 
assessing costs as well as production and income to optimise profit has been well highlighted in 
the literature (Anderson & Ridler, 2010) . A strong to stable position in OPM indicated  an ability 
to control costs effectively, and also suggested  a clear cost control focus. In general, information 
on farm production cost and detailed information about expenditure may be valuable for all 
farming systems. However, the importance of detailed expenditure may become higher at 
increasing intensification levels, scale and complexit y of the business.  
The predominance of operating expenses  (OpEx)  over farm working expenses  (FWE) was not a 
surprise after the importance shown by labour in determining efficiency. OpEx is a more 
comprehensive measure than FEW,  and may include unpaid labour, if relevant, in a manner 
similar to total economic costs,  which is the indicator suggested as the leading indicator of a 
sector’s abili ty to compete, adapt and expand since it includes all resource input costs, including 
family labour, equity capit al and owned land (Dillon et al., 2008; Lobos et al., 2001) . Both  FWE  
and OpEx are ,  nonetheless, leading indicators of farm expenditure in the NZ dairy industry  
(DairyBase, 2009 ; Shadbolt & Gardner, 2009). OpEx and FWE can be both expressed in 
                                                          
28 Kohl and Wilson (1997) indicated the reference for OPM reflecting a strong (0.25), stable (0.10-0.25) 
or weak position (<10%). 
Chapter VIII: General Discussion 
 
174 
absolute terms ($), or as ratio s per unit of output such as FWE/ kg MS or unit of input such as 
dairy grazing area, for example OpEx/ha. Fingleton (1995) and Thorne and Fingleton (2006) , in 
examining the c ompetitiveness of Irish milk producers , opted for expenses relative to kilograms 
of milk solids rather than area, to allow for differences in the potential of the land for milk 
production. 
Asset turnover ratio was the fifth and last benchmark revealed in the NZ trees. This came 
through twice, in two different trees (Figures 14 and 15) , at the threshold values of 25% and 
15%. In general , asset value has an inverse relationship with the ATR  and ROA  value, and a 
direct relationship with operating profit margin. Some authors suggest that a low profit margin 
can be tolerated if there is a high asset turnover ratio (Kohl & Wilson, 1997) . This supports our 
findings in Chapter 5 , where it was shown that a dairy farm in NZ may  have high operating 
expenses and still be very efficient, simply by making better use of the assets as indicated by the 
interaction with Assets Turnover ratio (A TR) at a threshold value of 25 %. This would be the 
case with irrigated, supplement - intensive farms in the South Island of NZ  where the land is less 
expensive than in most  NZ regions. However, when a lower ATR value was combined with a 
low profit margin, a low efficiency operation was always identified. Therefore, the pivotal value 
for ATR which defines a strong, stable or weak position  strongly depended on the type of 
operation, and whether assets were owned or leased. A key point to note from t he results in this 
study is that there was no one pathway to efficient performance, and that many variable 
combinations of good practices could result in high and very high efficiency scores. Also, these 
results demonstrate that strong, stable and weak threshold levels of measures cannot be taken as 
general truths.  
The relationship between ATR and OPM also relates to the competitive strategy. In this study, 
farms following the low- cost strategy are likely to have lower ATR and ROA ,  and a higher 
OPM. More intensive farms such those found in Chile  and some in NZ, however , are likely to 
behave in the opposite manner. For example,  the efficiency literature in NZ consistently found 
that technical efficiency was, on average, higher in the South I sland than in the North Island, 
which was also found in this research. Without exception, all efficiency analyses have included 
land capital as an input. We found that North Island farms were able to be as efficient as South 
Island farms through having a lower ROA b ut higher OPM. Therefore, when facing the 
dilemma of whether to look at overall efficiency (ROA)  or operating efficiency (OPM)  in 
choosing just one indicator , the answer would be that it depends on the values and goals in the 
farming business.  
The Chilean trees revealed three financial benchmarks, as shown in Table 59:  gross farm 
revenue per cow ( GFR CLP$ /cow) , production  costs per litre ( PC CLP$/L) , an d operating profit 
Chapter VIII: General Discussion 
175 
per cow (OpP CLP$/cow ). GFR per cow was highly correlated with net milk sales  per cow, and 
thus, it is ultimately correlated with physical milk production. The importance of this 
benchmark suggested that Chilean farmers are revenue maximising . This assumption has been 
found in the literature for those situations where the level of inputs, especially land, operator 
and family labour, is co nsidered fixed in the short run (Olson & Vu, 2009). Both t he literature 
reviewed and the evidence in this study all indicated that an important driver for milk 
production in all countries is milk price . This is perhaps more clear in Chile, where farmers may 
be more reactive to milk price  information than a NZ farmer. The reason for the strong 
influence of this variable is because it is easily quantifiable, resulting i n a direct incentive to 
improve production and increase revenue when milk price is high , or to reduce costs per unit of 
output when the price is low. This ability to respond rapidly to milk price is a fundamental 
difference between pastoral and supplement intensive farming systems. In this respect, Chilean 
farming systems are more likely to behave like confinement  systems than like pasture - based 
systems.   
Ler don et al., (2008) studied the interactions between  different social, productive, economic and 
financial variables in dairy farms in Chile during two consecutive years, 2000 and 2001. They 
found a strong correlation between milk production and  milk price in both p eriods, explaining 
more than 60% of the variation in milk production. The authors also found that milk price was 
of higher importance in planning and decision making than profitability and cost per litre. The 
physical information contained in DairyBase and TodoagroBase, and used in this study, clearly  
indicated that farmers responded to price information by feeding more supplements to the herd, 
which led to farming systems shifting towards greater use of imported feed in years of higher 
prices.  
In both countries , a price crisis would produce discontent and then uncertainty, which may 
stimulate variable cost reduction. This is likely to negatively affect production  in the same way 
that a high milk price will stimulate production. However, only in Chile would a price crisis 
encourage the search for new alternatives by diversifying the production portfolio. The literature 
reports that one basic difference between NZ and Chilean dairy farming systems is that NZ 
farmers have long opted for specialisation in dairying, while Chilean farmers have opted for 
diversification. For most specialized farming systems, any change from dairy to beef production 
is a long- term decision, however, for the Chilean dairy systems it can be either a long or a short-
term decision (W inkler, 1999). Chilean dairy farms often allocate land to a variety of 
agricultural activities such as beef or cropping. These non- dairying activities generally occupy 
more than 50% of the total land area (Anrique et al., 2004; Lerdon, Munoz & Moreira, 2010). In 
particular the beef and the dairy markets are negatively correlated , mutually influencing the 
decision- making process. While this can be advantageous for the farmer, it becomes a serious 
Chapter VIII: General Discussion 
 
176 
disadvantage for the Chilean dairy industry because it increases both price and volume 
instability.  
In Chile, m ilk price was  strongly related to physical milk production ( L/cow ), and this  to gross 
farm revenue per cow and farm efficiency. The second financial benchmark to emerge  from the 
trees was production costs which is closely linked to the other benchmark  yet to be discussed 
that is operating profit per cow. Production cost per litre of milk produced was 90% on average 
of the total production cost averaged in the TodoagroBase data over the 2007- 2011 period. The 
structure of the Chilean trees was similar to the NZ trees,  in that production, labour, cost and 
profit variables came out in almost the same order as in the NZ trees . Operating profit  per cow 
was a variable created as part of this study to test whether or not this NZ - specific variable was 
relevant for Chilean farming systems. No literature exists suggesting that th is might be the 
situation. In this particular case, the farmers’ focus on Operating profit  per cow can be discarded 
as a causal agent of the measure being so relevant. Therefore, the strong association between 
operating profit (per hectare and per cow) and technical efficiency revealed by the trees for both 
countries was reinforced. The fact that in Chile it is expressed as a relative measure on a per 
cow basis is also compatible with the industry view, where the preference is for cow rather than 
per hectare measures in Chile.  
8.3 The Benchmark Sets Applied to Different Performance Groups 
The corresponding benchmarks for each country were used to comprehensively  describe the 
five- year NZ and Chilean samples, which were divided into three sub- groups according to their 
efficiency scores. The ability to have an efficiency score of one in producing both milk , and 
revenue, for three or more years, as explained in Chapters 5 and 6 was recognised as resilient 
efficiency. Top performers were described as ‘resilient’ which implies robustness and the 
capability to adapt to alterations in the environment and to take advantage of opportunities, 
while maintaining productive capacity despite variability in financial, market and production 
related factors (Shadbolt, Rutsito & Gray, 2011). Aven (2011) pointed out that a key difference 
between robustness and resilience is that the first relates to a fixed threat , whereas resilience is 
relevant to any type of potential shock  and uncertainties given a certain environment. This is 
why the term ‘resilience’ was preferred over ‘robust ness’.  Scholz, Blumer , and Brand (2012) 
pointed out that resilience is linked to risk and vulnerability; it incorporates the capability of a 
system to cope with the adverse effects or risks that a system has been exposed to.   
The percentage of resilient farms in Chile ( 4.3 %  extracted from Table 40 ) was lower than in NZ  
( 5.6 %  extracted from Table 26 ) . This may be more likely ex plained by a higher percentage of 
the farms being in the NZ database for three or more years, as the criterion for defining resilient 
farms indicated, or/ and by a better overall management among NZ dairy farmers than among 
Chapter VIII: General Discussion 
177 
Chilean farmers. In Chile, along with a higher use of supplements there is a lower dependence 
on rainfall and pasture production to feed the herds. This may have the effect of increasing 
resilience since supplements are well known as a way to , simultaneously, improve milk yields  
and reduce the climatic risk associated with  pasture- based systems (Edwards & Parker, 1994) . 
According to the IFCN criteria (2011) presented in Table 9, and recent research by Anderson 
and Ridler (2010) , due to climatic risk NZ farms have a more risky profile than Chilean farms. 
Therefore, if resilience refers to the capability of a system to cope with risks, under comparable 
scenarios the ‘natural’ resilience of NZ dairy farms would be more challenged than under other 
less risky situations.  
Rodriguez, deVoil, Power, Cox, Crimp and Me inke (2011) claimed that flexible  farming 
systems express greater resilience than more rigid farming systems. Research suggests that this 
flexibility is  an important attribute for farming systems, likewise ecologists have long identifi ed 
that phenotypic plasticity as a key element in the functioning of organisms (DeWitt & 
Schneider, 2004). Rodriguez et al. (2011) explained that where farm management is hi ghly 
contingent on, for instance, environmental conditions, farm managers often vary crops and 
inputs based on the availability of limited and variable resources (land, water, finances, labour, 
machinery), and signals from its operating environment (rainfa ll, markets), with the objective of 
achieving a number of often competing objectives, such as reduce risk or increase profits. This 
is similar to what the Chilean farmers do with their diversification strategy by considering beef 
and cropping as complements but also as alternatives to dairying. Therefore, farmers in Chile  
with a higher proportion of variable costs may be more responsive to changing market 
conditions than NZ farmers. This is what is likely to b e expected from a supplement intensive 
farming systems versus a pasture- based system. However, given the current environment it may 
be advisable for NZ farming systems to find mechanisms to be more responsive to changing 
conditions in general and to milk price in particular.  
In NZ , the observed frequency of resilient farms in the 2010/ 11 season was significantly higher 
than the expected value. Similarly, the Otago - Southland region, farming system number five, 
irrigated farms, and dairy farms larger than 600 cows dairy farms  showed a higher freque ncy of 
resilient farms (Appendix 8) . In Chile there was no difference  between the observed and 
expected frequency of resilient farms by year and farming system (identified by use of 
supplements) but there were differences in region and scale. The observed frequency of resilient 
farms was significantly higher in the Bio - Bio  region, and in farms larger than 600 cows  
(Appendix 8 ) .  
Interestingly, f or many of the analysed variables in NZ and Chile , consistent top farms showed 
greater variability than poorer performers. This suggests that in both countries, the overall 
Chapter VIII: General Discussion 
 
178 
ability to perform well consistently did not depend on the size of the business. This reinforces 
the discussion about economies of scale in NZ presented earlier . This finding has not been noted 
yet for Chilean dairy farming systems; on the contrary  the opposite was found by authors in 
several studies. Although traditionally Chilean dairy farming systems have shown considerable 
variability in efficiency, profitability , and other variables (Aichele, 1979; Campos et al., 1995 ;  
Silva , 1997) , generally increasing returns to scale have been found (Diaz &  Williamson , 1998 ; 
Moreira 1999 ;  Schilder and Bravo - Ureta , 1994).  The long time gap between these results and 
the findings presented in this research makes  it believable that the norm has changed for 
pasture- based dairy farms over this period.  
In NZ and Chile , top performers, on average, produced the same or more output per cow than 
the other groups. The resilient top performers in NZ  produced about 391 kg MS/cow on 
average, with a minimum of 2 13 kg  MS/cow and a maximum of 650 kg  MS/cow. The labour 
productivity of resilient farms was in the order of 60,000 kg MS/ FTE, which mean s that 
variable herd productivity needed to be accompanied by very high labour productivity to 
consolidate resilient farm efficiency. Labour productivity was crucial and showed a mean 
significant difference in the order of 15%  between resilient and poorer performers. In Chile, a 
resilient farm could also produce a very similar output per cow compared to an average farm 
(about  5,300 L/cow ) , but would do this with higher efficiency in the u se of supplements.  
Supplement fed to the cows was the key defining  efficiency and resilience among Chilean dairy 
farms. Resilient performers used less supplement than the mean (215 grams per litre), with 
some not using supplements at all to achieve resilient efficiency. In contrast, the poorer 
performers all used supplements and most were heavy users. Under Chilean condit ions, and 
with Chilean dairy breeds and strains, dry matter intake of pasture may often be insufficient for 
the high genetic merit cows to meet their potential, and therefore supplements are needed 
(Kolver & Muller, 1998). This implies dependence on the use of supplements which can result 
in inefficient farmers being cost- inefficient if the milk:feed prices ratio is unfavourable. O n 
average, the resilient performers used 22.4% less supplement per cow than the poorer 
performers. Milk production responses of grazing cows offered supplements varied and positive 
quadratic responses to incre asing amounts of supplement have been observed for yield of milk  
( Auldist et al., Wales, 2013). The resilient farms, may be achieving higher responses to 
supplement in milk production than poorer performers . They may also be more flexible in 
adjusting to unfavourable milk: feed prices ratio by using breeds and strains more suited to a 
pasture- based feeding strategy, or by better grazing managem ent . 
Interestingly, despite lower reliance on supplements, the direct cost of production expressed in 
CLP$/L achieved by the resilient performers in Chile on average was not significantly different 
Chapter VIII: General Discussion 
179 
from that achieved by the non- resilient observations. Most  Chilean farmers may be maintaining 
a relatively low- cost status by reducing cost per unit through the denominator effect by 
producing larger volumes of milk . However, resilient farms on average had significantly higher 
profitability per cow with the lowest variability, and also had wages per litre of milk produced 
which were significantly lower than the inefficient performers. This is consistent with the fact 
that generally higher intensification levels (exhibited by inefficient observations) require  higher 
labour input.  
I n NZ , resilient farms were consistently more profitable than non- resilient farms, exhibiting 
10% lower operating expenses per kilogram of milk solids on average, and better use of assets, 
as reflected by dairy ROA and ATR  results. Profitability per hectare was also markedly higher 
among resilient performers, and OPM in absolute and statistical terms was also higher,  
indicating that these generate more profit from their revenue by effectively controlling costs. 
Shadbolt, Rutsito an d Gray (2011) proposed that efficiency measured as profitability could be 
equated to  resistance, one desirable attribute clearly exhibited by  resilient farms in this study.  In 
NZ and Chile , resilient performers remain production-  and cost- focussed simultaneously, 
achieving greater profitability than non- resilient performers no matter what milk price, feed 
price or rainfall received. Resilient farms are flexible enough to take advantage of upside risk 
and cope with downside risk.  
8.4 The Metafrontier Analyses 
The purpose of the metafrontier analyses within countries was to compare efficiencies scores 
across a number of years, regions and farming systems in NZ and Chile, separately. In addition, 
the metafrontier analysis across countries compared efficiencies of dairy farms across countries 
for two selected years of interest.  
The country- specific analyses found that the countries benefited from high technical efficiencies 
relative to their own frontiers,  with efficiency estimates averaging around 0.90. Thus, in each 
country there were a number of farms operating relatively close to the technical efficiency 
frontier. However, when compared to a metafrontier, broadly representing the pasture- based 
technology set shared by both countries averages showed very low technical efficiency scores. 
This reinforces the idea that within each country, a range of diverse and sometimes contrasting 
farming systems can coexist.  
All metafrontier analyses used herd size (cows), supplements (kg/cow) and CPI adjusted 
operating expenses expressed in US$ as inputs. The selection of inputs was reduced to the 
reliable variables present in both databases that were reliable, with the set chosen to keep 
consistency with the previous efficiency analyses. Assets were deliberately excluded from the 
metafrontier analyses because of the unreliability of the backwards calculation of  return on 
Chapter VIII: General Discussion 
 
180 
assets in the TodoagroBase. The default output in this type of technical efficiency analyses is 
‘milk’, expressed either in kilog ram of milk solids or in litres. Ho wever, the metafrontier output 
was CPI adjusted gross farm revenue, also expressed in US$. The use of this variable affected 
by production and milk  price was the straightforward option given the lack of information about 
milk solids content in  the TodoagroBase. 
The fact that some inputs and outputs are expressed in US$  (because physical measures were 
recorded in different units in DairyBase and TodoagroBase ) , deflated by average price indices 
per country may introduce problems in comparison between countries. In addition to issues 
related to relative prices, the movements of  the respective exchange rate can affect the 
conclusions (Sharples, 1990) . Real movements in exchange rates tend to be associated with 
changes in the competitive position of a country. In particular, real appreciation of local 
currency means, ceteris paribus , that the goods produced are comparatively more expensive 
domestically, which is a situation associated with a loss of competitiveness (Harrison & 
Kennedy, 1997). For example, the NZ$ has appreciated against the US$ following an increasing 
trend since 1998, reaching its historical maximum value in 2011 (Figure 6) while the C hilean 
peso has been erratic behaviour over the same period (Figure 7) . This might mean data 
limitations might have weakened the metafrontier results, and therefore, some interpretative 
caution is required.  
The NZ metafrontier used all the observations, ac ross all years, a total of  2,588 DMUs. Not 
surprisingly, the years with  lower efficiencies had lower milk price, and conversely, the higher 
efficiency years had higher milk  prices. Overall, efficiency tracked  fluctuations in GFR which 
suggests a similar consumption of inputs throughout the period. Operating expenditure on 
average was only statistically different in 2006/ 07  from the rest of the period. It is interesting to 
note that even when facing an average milk price scenario in 20 09/ 10, expenses were not 
significantly lower than in other more favourable years. This suggests a high fixed costs 
proportion in total expenses in NZ. In  2008/ 09 expenses were not significantly different from 
other more favourable milk price years because milk price dropped half way through the season 
when the majority of the costs were already locked in.   
Throughout the period, there were no significant differences  between years in the average herd 
size or in the use of supplement  per cow. These suggest that NZ dairy farming systems may 
benefit from being more reactive to milk:feed prices ratio which  could be achieved by 
increasing or decreasing expenditure per unit of output as a funct ion of the price received for 
milk . Despite feed being more expensive in NZ than the world’s average, it would make sense 
to have higher use of supplements at higher milk prices.  This may be very challenging for 
isolated farmers, but may possibly be achieved by using a complete systems approach involving 
Chapter VIII: General Discussion 
181 
the entire value chain, encouraging more cropping under contract and other arrangements that 
make more feed available to farmers.     
The Chilean metafrontier used the observations across all years, totalling 1,028 DMUs. Through 
the period there were highs and lows alternatively, which may reflect the influence of weather 
and price shocks. The metafrontier efficiency in 2009 was significantly lower than the rest of 
the period, explained by an observed milk price well below the expectations tha t farmers had at 
the beginning of the production year. Average years, such as 2010 and 2007, showed 
intermediate mean efficiencies, while the years 2011 and 2008 , had noticeably higher mean 
efficiencies due to record high milk prices . Overall, NZ and Chile  behaved in a similar way; the 
NZ 2008/ 09 season can be compared to the Chilean 2009 year, even if it was not  exactly the 
same time period. Correspondingly, the NZ 2010/ 11 season can be compared to the Chilean 
2011 year.  
In both countries ,  the changes in efficiency were smoother than the fluctuations observed in 
mean gross farm revenue. In Chile , the greatest difference between maximum and minimum 
mean efficiencies were in the order of 32% , while the maximum mean revenue was almost 
double the minimum. From 2007 to 2011, despite fluctuations there was an overall trend in 
increasing operating expenses over time. The minimum mean expenditure was recorded in 
2009, and the maximum recorded in 2011. This suggests that Chilean farmers are  milk pric e 
focused and react to it by increasing or decreasing variable costs. Herd size showed an 
increasing trend over the period, but only the last year of the period was significantly different 
from the rest of the years. This trend to increase numbers of cows in the herd was accompanied 
by a similar tendency towards intensification, as denoted by the higher use of supplements per 
cow. The beginning of the period ( 2007 )  had a moderate use of supplements of 1,252 kg/cow on 
average. This was not statistically different from the supplement used in 2008, 2009, 2010 and 
2011. However, there were di fferences between supplement fed to the cows in 2011, relative to 
2009 and 2010, in response to more favourable milk:  feed ratio. 
8.5 The Metafrontier Across Countries 
The metafrontier across countries for two selected years of interest addressed the research 
question ‘What  country is more technically efficient regarding dairy farming and why? ’  From 
the previous analyses, two contrasting years were identified by both price and climate. A ‘good’ 
year was represented by 2010/ 11 in NZ and 2011 in Chile, while the contrasting ‘poor’ year was 
2008/ 09 in NZ and 2009 in Chile. Metafrontier analyses were conducted separately for those 
years. Note that a relatively less favourable weather scenario affected the NZ pastoral season in 
2010/ 11, with officially declared areas of drought across the territory (MAF, 2011). In addition, 
it must be remembered that the years do not refer to the same period, since NZ uses the 
Chapter VIII: General Discussion 
 
182 
agricultural year (Ju ne- May ) while Chile uses a c alendar year (January - December). It also is 
worth noting that results across years within countries are not strictly comparable since different 
farms made up the samples each year.  
S ignificance tests showed that in the poor year, gross farm revenue was significantly lower in 
Chile compared to NZ. B oth countries’ poor year’s revenues were significantly lower than those 
observed in 2011 , but were not significantly different from each other. In contrast, mean 
expenditure for Chile in 2009 was also significantly lower than the expenditure means for the 
rest of the country- year combinations ( but not significantly different between them) . The use of 
supplements was significantly lower in NZ than in Chile for  both years under review. In both 
countries, efficiency was highly affected by input/output prices and climate . Highly  correlated 
world market prices for milk, feed and oil , as well as rainfall , are important drivers of efficiency 
of milk production in both NZ and Chile. Particularly for milk price, e xceptionally high values 
occurred at the end of 2007 and the beginning of 2011, while low values were observed in June 
2009 and 2012 (IFCN, 201 2).  
The low milk price and poor  climatic year, 2008/09, severely affected revenue efficiency in NZ 
(mean = 0.38) , which was consistently below the m ean for the whole period (0.41) . The 
equivalent year for the Chilean sample ( 2009 )  demonstrated a lower efficiency (0.37) , and also 
greater variability (0 .18) , than the NZ sample. However, d ifferences on average efficiencies 
between DairyBase and TodoagroBase samples for the poor year were not significant. The 
IFCN (2012) and Tozer, Bargo, and Muller ( 2003)  both remark  that a low milk:feed price s ratio,  
as observed in 2008/09 (below 1.5) , may benefit low input farming systems with moderate milk 
yield levels. In NZ, the same pattern was found by Jensen, Clark, and Macdonald (2005) who 
analysed different levels of intensification within pasture- based systems. At a higher milk 
payout ( NZ$4.5/ kg MS ) , the higher input systems showed higher profitability. The metafrontier 
showed that in NZ and Chile, a good year (2010/ 11 in NZ and 2011 in Chile) allowed for a 
recovery in efficiency. The mean efficiencies in both countries were significantly higher than 
those observed in 2009. The difference between NZ mean efficiency (0.49) and the Chilean 
mean efficiency (0.41)  was also significant. 
The metafrontier across countries results indicated that there is a very large variability within 
countries, and even larger variability between them. To the best of our knowledge , this is the 
only possible explanation for  the fact that the metafrontier scores were very low compared to 
the class- specific scores. The metafrontier across countries also suggests that both NZ and 
Chilean samples had similar mean efficiency during the poor year,  but differences were 
expressed in the good 2011 year. This would suggest that NZ farmers had greater ability to 
capitalise a high milk price as revenu e, by using fewer inputs that their Chilean counterparts.  
Chapter VIII: General Discussion 
183 
8.6 Review of Method 
This section provides a review of data envelopment analysis with recursive partitioning as a 
second stage approach. Despite criticism of  DEA models applied to agriculture, this research 
relied to a large extent on DEA estimates of efficiency. Major criticism has been that DEA does 
not account for random errors, is sensitive to nesting and input/output selection, and it usually 
ignores undesirable outputs (Coelli et al ., 2005).  
Random factors that affect production , such as rainfall, can have a large impact and are 
particularly important in pasture- based farming systems. These random factors can affect the 
relative performance scores calculated by DEA, which attributes all movement away from the 
frontier of the scores to inefficiency rather than providing some allowance for random effects. 
Typically, operating conditions, and environmental and contextual variables  such as climate or 
land attributes, are assumed to be non- discretionary inputs and are rarely included in a given 
dataset. If the environmental impact component can be identified, its separation from 
managerial inefficiency is advised. Otherwise the DMUs can be penalised for the unco ntrollable 
conditions they face. Several efforts to include non- discretionary inputs in the DEA are 
testament not only to their importance, but also to the difficulty of satisfactorily incorporating 
their impact in any evaluation of performance.  
In this r esearch, this limitation was addressed and its undesirable effect reduced through the 
country- specific classification system. Nesting DMUs is one way researchers have attempted to 
adjust for heterogeneity in operating conditions when estimating DEA efficie ncy scores, since 
the discriminatory power of DEA, and to some extent its validity, relate to the homogeneity of 
the sample (Semoneiko & Osei - Bryson, 2008). However, DEA typically has a high level focus, 
and the use of aggregated information is actually one of its strengths (Rouse, Harrison & 
Turner, 2011). When there are multiple ways to nest the observations , DEA c an yield very 
different results which is undesirable. This study found heterogeneity within the DairyBase and 
TodoagroBase data rather than homogeneity, and used a nesting process to group naturally 
occurring peers based on the literature, normalising the sample against the larger sources of 
heterogeneity. This was done at the year, region and feeding system levels. This disaggregation  
proved to be of benefit in estimating efficiency scores and was the most important step in the 
process because all the successive steps relied on it.  
DEA is known to be sensitive to the selection of inputs, outputs and to the presence of outliers. 
In thi s research, we determined the relative importance of competing explanatory factors 
without a built- in test based on their appropriateness and reliability and avoiding multi-
collinearity. This is an issue when inputs are highly collinear, which can lead DEA into the 
second error type. In contrast, when inputs have very low correlations , DEA is more likely to 
Chapter VIII: General Discussion 
 
184 
view each DMU as being unique , and therefore  fully efficient,  especially when there are limited 
observations resulting in the loss of discriminatory power. Moreover, the flexibility of the  
envelope underlying normal DEA  may be a cause of upward bias, especially for limited sample 
sizes (Pedraja- Chaparro et al., 1997). This research acknowledged this limitation and tested the 
robustness of DEA estimates fo r the benefit of this discussion, evaluating the influence of 
including the input ‘hectares’ , rather than ‘cows’ , in the DEA  applied to the largest class in the 
study ‘Rest of North Island’ (n = 189; NZ, season 2006/ 07). The i nfluence of this on the 
efficiency scores was minor as reflected on a Spearman correlation coefficient of 0.884, 
significant at the 0.001. A second test evaluated the impact of including hectares and cows 
simultaneously in the same NZ class. The impact of this on the efficiency scores was  also 
negligible, as reflected by a Spearman correlation coefficient of 0.968, significant at the 0.001.  
These mean that the reliability of efficiency estimators calculated in the first place was high, as 
was the validity of our results.  
Another major limitation of DEA is the existence of undesirable outcomes, often not measured 
or included in the databases, and therefore, excluded from the analyses. Lack of information 
about undesirable outputs in this study may have led to a misclassification of a DMU  as 
efficient when it is actually inefficient because it may produce large quantities  of undesirable 
outcomes. According to Seiford and Zhu (2002) , these undesirable outcomes should be included 
as outputs when performing efficiency analyses. Frequently ign ored undesirable outputs could 
include pollutants and wastes. However, these are often difficult to quantify  which is why they 
are not included. For instance, FAO (2006) attributed to agriculture 18% of the world’s total 
anthropogenic greenhouse gases, and  the IFCN (2012) estimated that dairy farming systems 
worldwide have on average a carbon footprint of 1.36 kg CO2 equivalent per kg of energy 
corrected milk. In the last few years, there has been an increasing tendency for people to be 
more conscious of undesirable outputs and other issues derived from agricultural activities. This 
trend should help to increase work i n this area, which, in the near future may generate measures 
and identify techniques to quantify undesirable outputs, making the information m ore available 
for the debate on how to deal with them.  
I n this research bias was expected to vary between the classes within the country - specific 
classification systems. It was expected to be lower in larger classes where there were enough 
efficient or near- efficient observations to delineate the frontier adequately, and to be higher 
when the class had fewer DMUs. Banker (1993) pointed out that i n DEA two possible types of 
error may occur: firstly, falsely classifying an efficient unit as inefficient, and secondly, falsely 
classifying an inefficient unit as efficient. At smaller sample sizes, the frontier is thinly 
populated and DEA is particularly liable to fall into the latter error  (Banker, 1993) . This means 
that the frontier estimator is biased below the theoretical frontier for a finite sample size and that 
Chapter VIII: General Discussion 
185 
the bias approaches zero for large samples. Kittelsen (1995) addressed the problem of bias in 
DEA and concluded that even though bias may be an intrinsic limitation of the methodology, it 
is at least predictable such that it increases with dimensionality and decreases at higher average 
efficiency, sample size, and concentration of observations adjacent to the frontier. Hawley 
(2010) in contrast , suggested that the use of a large sample would not necessarily prevent bias, 
which is an intrinsic property of, and a limitation,  in DEA. In this research , bias was not only 
expected , but also tolerated, as is the measurement error in the explanatory var iables.  
This research explicitly ignored other potential applications of DEA, such as slacks, peers or 
scale efficiency scores. These, although acknowledged valuable information, were out of the 
scope of this study. For time constraint reasons, individual  farm analyses were not feasible,  
while peer information using DEA efficient DMUs  was neither analysed and nor taken into 
account. DEA provides a formal method of performance measurement that addresses some of 
the problems encountered in the use of conventional accounting methods of such measurement. 
In particular, it enables comparisons to be made of relative performance where multiple 
performance measures exist, but the appropriate weightings to be applied to those measures are 
not known or cannot be agre ed upon (that is, where the production function is unknown), and 
where nonfinancial, as well as financial, measures are important.  
Finally, second stage DEA in general, including the analysis presented  in this study using 
regression trees, has many support ers, but also critics, because of its limitations. Many authors 
have used DEA efficiency scores in a two - stage analysis for various purposes (Banker & 
Natarajan, 2008; Olson & Vu, 2009; Simar & Wilson, 2007). The first , and stronger, argument 
against the second stage DEA  approach states that if the variables in the second stage were 
expected to affect the performance, they should have been included in the original DEA model 
(Grosskopf, 1996). Grosskopf also draws attention to the likelihood of bias and inconsistency in 
the second stage due to the correlation with the factors included from the first stage. Other 
detractors of this approach also claim that efficiency scores are inevitably biased, and often 
strongly correlated with the variables used as inputs or outputs (Siman & Wilson, 2007). For 
example, peak cows milked w as used in DEA as an input , rather than the effective dairy area 
(ha) , and both were expected to affect performance since they both represent dairy farm size and  
have a very high correlation, so one had to be discarded. The second stage revealed that the ratio 
per cow was in fact dominant over the ratio per hectare, although it is uncertain how much of 
this is due to the nesting process or to the correlation between DEA i nputs and the trees’ results. 
We also addressed this limitation by applying RPT to DairyBase and TodoagroBase using the 
first split variables ROA and GFR/cow , respectively , as response variables to re- test consistency 
and relevance of the respective benchmark sets. These trees, which are b riefly discussed in the 
Chapter VIII: General Discussion 
 
186 
appendices, showed the desired consistency and relevance. These regression trees could be used 
as prediction tool for ROA  without the need to estimate efficiency scores with similar results.  
In addition,  and related to the recursive partitioning trees (RPT), Strobl et al., ( 2009) explain 
that an undesirable side effect of the trees is that an entire tree structure could be altered if the 
root variable or its cut- off point was chosen differently because of a small change in the dataset. 
A solution to this instability problem was also provided by these authors, who suggest the use of 
a series of trees instead of a single tree, where there are reliability concerns that could affect the 
consistency of the results. In this research , this limitation was addressed by using multiple trees 
for each country, and only selecting the benchmarks that were consistent across the trees.
Chapter IX: Conclusions 
187 
9. Conclusions. 
 
This research focused on New Zealand an d Chilean dairy farming systems and particularly on 
the interactions between inputs, outputs, and the macro- environment. It was in response to the 
need for greater understanding of the farming systems and demand for information- sharing, 
triggered by both globalisation and increased market risk.  
In New Zealand, well - delineated pastoral systems following the cost- leadership strategy were 
confirmed regardless of the intensification level. Herd and labour productivity played key roles  
in defining efficiency. The operating cost per kilogram of milk solids, return on assets, 
operating profit margin, operating profit per hectare, and asset turnover, joined the production 
measures in the final set of K PIs that adequately describe d and explain ed the variability in 
efficiency in New Zealand farmin g systems. In Chile, supplement - intensive systems based on 
pasture were identified. Herd productivity and su pplement fed to the cows played the most 
important roles in defining efficiency. Chile benefits from temperate grasslands just like New 
Zealand, and has the advantage of a low cost labour supply and land, both of which place Chile 
in a very good competitive position internationally as a milk producer. The absence of return on 
assets in Chile was noticeable, but may be justified by the fact that the main asset, land, is not 
actively marketed since land belongs to families for sucesive generations and there is little 
chance in disinvesting from farming to invest in any other activity. From these findings, it may 
be concluded that operating profit should be included as an indicator in Chile since it was the 
only profitability indicator that came out from the regression trees (operating profit was  not 
originally available in TodoagroBase ).  Similarly, in New Zeal and it may be interesting to study 
the relevance of including a similar KPI to that found in Chile (g supplement/L milk produced) ,  
to evaluate performance in the use of supplements in Farming systems 4 and 5.   
The fact that different KPIs were identified for each country indicates that metrics should n ot be 
used to benchmark across systems without a previous evaluation. It has been proven here that 
system- specific benchmarks can objectively be identified . Other conclusions also extracted from 
this work are that New Zealand farming system s are more dependent on weather- related factors 
and may have a riskier profile than Chilean farming systems due to higher stocking rates and 
greater reliance on pasture as feed. Furthermore, New Zealand farmers may be technologically 
more advanced at managing pasture, while Chilean farmers may be better at managing 
supplements since they have traditionally used them more intensively. Chilean farmers are more 
revenue focused and respond to milk:feed ratio and response to the use of supplement, while 
New Zealand farmers are profitability  and cost- focused looking at both the OPM and capital 
invested. This suggests a greater orientation towards entrepreneurship in New Zealand than in 
Chile, as well as differences in values and goals.   
Chapter IX: Conclusions 
 
188 
New Zealand and Chile are natural benchmarking partners for pasture - based dairy systems since 
they produce similar outputs, trade under similar market conditions, and consume similar inputs 
although in different proportions, and  efficiently and profitably produce milk at a lower cost 
than the world’s average. Their  systems are also evolving in similar ways through efficient- scale 
processes, innovation, and gradual increases in average intensification levels and specialisation. 
Other similarities are that, despite being heterogeneous groups, the consistent top performers in 
both countries were more likely to be bigger, to produce on average the same or more output per 
cow, and to have higher labour productivity than poorer performers . Interestingly, the 
consistently top performers in Chile, on average, used significantly less supplements per litre of 
milk produced  than the poorer performers. Similarly, New Zealand’s top performing  farmers 
had significantly lower operating expenses ratios than the other groups. They also used assets 
better, and had higher operating profit margins and profitability per hectare. This is important in 
terms of sustainability because dairy farms with a higher profitability have a lower risk of being  
unsuccessful when faced with a turbulent environment. 
Benchmarking has been criticised as being ‘just’ backward looking. The innovative second 
stage DEA and RPT approach could be used as a prediction tool in each country using actual 
pivotal values found in this research. For extension and education purposes RPT may be also 
useful to graphically show relationships between k ey variables of interest rather than using the 
standard multivariate regression method.  
A natural follow up for this research could be further exploring  the outcomes of this study. 
Peers and slacks from the data envelopment analyses, surrogate splits from  the trees and 
metatechnology ratios from the metafrontier results which were out of the scope of this work , 
could be used to gain more insight into the studied farming systems and the virtues of the 
method. Another direction for future investigation based on this work would be qualitative 
research both in New Zealand and Chile, to identify the key management factors and the 
different pathways that lead to consistent efficiency.  
The identification of a few KPIs among all the measures currently available  may permit a better 
focus on the really key metrics. This is important for at least two reasons. Firstly, because all the 
measures and available information at times might be overwhelming and secondly, because a 
better focus on just few KPIs could allow t ime and resources to identify and measure other 
variables needed for improved farming systems such as undesirable outputs that threaten 
sustainability and resilience. If that  were possible the differences in results when undesirable 
outputs, such us greenhouse gases or soil degradation, would need to be included in future 
estimations of efficiency in agriculture. 
 
Chapter X: References 
189 
10. References. 
 
Adams, D. C., Clark, R. T., Klopfenstein, A. & Volesky C ., (1996). Matching  the cow with 
forage resources. Rangelands 18 (2): 57- 62. 
Adler, N. &  Golany , B.,  (2001). Evaluation of deregulated airline networks using data 
envelopment analysis combined with principal component analysis with an application 
to Western  Europe. European Journal of Operational Research 132 ( 2) : 260 - 273.  
Agosin, M., & Bravo Ortega, C., (2009). The Emergence of New Successful Export Activities 
in Latin America: The Case of Chile. Research Network Working Paper, 552, Banco 
Interamericano de Desarrollo [Interamerican Bank of Development] (BID) .  
Aichele, O. , ( 1979) . Estudio de casos en predios lecheros ubicados entre Valparaíso y 
Llanquihue [ Case studies in dairy farms located between Valparaiso and Llanquihue] . 
Tesis de Licenciado en Ciencias Agrarias [ Bachelor Thesis in Agricultural Sciences]. 
163p. Universidad Austral de Chile, Facultad de Ciencias Agrarias, Valdivia, Chile.         
Akoorie M. &  Scott - Kennel J. , (1999).  The New Zealand Dairy Board: A Case of Gr oup-
Internalization or a Monopolistic Anomaly in a Deregulated Free Market Economy . 
Asia Pacific Journal of Management 16( 1): 127- 156. 
Alston, J. M., (2002). Spillovers. The Australian Journal of Agricultural and Resource 
Economics 46( 3): 315- 346. 
Amtmann, C . & Blanco, G., (2003). Expansión transnacional y nueva ruralidad: Conflictos del 
sector lechero en el sur de Chile [Transnational expansion and new rurality: Conflicts 
in the dairy sector in southern Chile]. In Bendini M. &  Steimbreger N., Eds Cuaderno 
del GESA 4: Los espacios sociales y la organización de la agricultura. Buenos Aires: La 
Colmena. 
Anderson, K. & McAdam, R. (2007). Reconceptualising benchmarking development in UK 
organisations: the effects of size and sector.  International Journal of Productivity and 
Performance Management 56 (7): 538 - 58.  
Anderson, W. J. & Ridler, B. J., (2010). Application of resource allocation optimisation to 
provide profit options for dairy production systems. Proceedings of the New Zealand 
Society of animal production 70 (1): 291 - 295. 
Chapter X: References 
 
190 
Anrique, R., & Latrille, L. (1999). Caracterización del Chile lechero. [Characterisation of 
Chilean Dairy sector].  Producción animal 1: 140- 157. 
Anrique, R., Latrille, L, Balocchi, O., Alomar, D., Moreira, V., Smith, R., Pinochet, D. & 
Vargas, G., (2004). La produccion de leche en Chile: caracterizacion tecnica a nivel 
predial [Milk production in Chile: technical characterization at the farm level]. 
Valdivia: Facultad de ciencias Agrarias, Universidad austral de Chile .  
Auldist, M.., Marret, A., Greenwood, Hannah, Jacobs, Wales, (2013). Effects of different 
strategies for feeding supplements on milk production responses in cows grazing a 
restricted pasture allowance. Journal of Dairy Science 96 (2): 1218- 1231. 
Aven, T., (2011). On s ome recent definitions and analysis frameworks for risk, vulnerability, 
and resilience. Risk Analysis 31( 4) :515- 522. 
Balocchi, O., &  López , I., (2001) . Rol de las especies pratenses nativas y naturalizadas en las 
praderas permanentes del Sur de Chile  [Role of native and naturalized pasture species in 
permanent pasture of Southern Chile] . Proceedings, Simposio Internacional en 
Producción Animal y Medio Ambiente. Santiago: Departamento de Zootecnia, Facultad 
de Agronomía, Pontificia Universidad Católica de Chile: 285- 299.  
Banker, R. D. (1993). Maximum likelihood, consistency and Data Envelopment Analysis -  A 
statistical foundation. Management Science 39(10) : 1265- 1273. 
Banker, R. D., Charnes, A. & Cooper, W. W., (1984). Some models for est imating technical and 
scale inefficiencies in Data Envelopment Analysis. Management scie nce 30: 1078- 1092.  
Barnard, F. L., & Boehlje, M., (1998- 99). The financial troubleshooting of farm businesses: A 
diagnostic and evaluation system. Journal of the American Society of Farm Managers 
and Rural Appraisers: 6- 14. 
Barney, J. B., & Hesterly, W. S. (2006). Strategic management and competitive 
advantage.Upper Saddle River, N ew J ersey: Pearson Education.  
Barry, P. J., Ellinger, P. N., Hopkin, J. A. &  Baker, C. B., (1995). Financial management in 
agriculture. Illinois, USA: Interstate publishers, Inc. 
Barry, P. J., Ellinger, P. N., Hopkin, J. A. & Baker, C. B., (2000). Financial management in 
agriculture. Illinois, USA: Interstate publishers, Inc.  
Chapter X: References 
191 
Bart on, J. R., Gwynne, R. N., & Murray, W. E. (2007). Competition and co- operation in the 
semi- periphery: closer economic partnership and sectoral transformations in Chile and 
New Zealand. Geographical Journal 173 : 224- 241. 
Bates, S. A. E. (1998). Issues arising from the milk pricing mechanisms of the New Zealand 
dairy sector. Farm Management 10( 2): 55- 62. 
Battese, G. E., & Prasada Rao, D. S. P., (2002). Technology gap, efficiency and a stochastic 
metafrontier function. International Journal Business Economy, 1( 1): 87- 93. 
Battese, G. E., Prasada Rao, D. S. P. & O’Donnell, C. J.,  (2004). A metafrontier production 
function for estimation of technical efficiencies and technology gaps for firms operating 
under different technologies. Journal of Productivity Analysis, 21 ( 1) : 91 - 103. 
BCCL Banco Central de Chile, (2012). Official Reports serie, Retrieved March 2012 
from: http://www.bcentral.cl/eng/press/releases/index.htm . 
Bennett, E. M., Cumming, G. S., & Peterson, G. D., (2005). A system models approach to 
determining resilience surrogates for case studies. Ecosystems 8: 945- 947. 
Bennett, J. W., (1982). Of time and enterprise: North American family farm management in a 
context of resource marginality. Minneapolis: University of Minnesota Press.  
Benson, G., (2008). Pasture - based and confinement dairy farming in the United States: An 
assessment. Journal of International farm management 4( 2) .  
Bernard, J., Le Gal, P. Y., Triomphe, B., Hostiou, N., & Moulin, C. H. (2011). Involvement of 
small- scale dairy farms in an industrial supply chain: when production standards meet 
farm diversity. Animal, 5(6) , 961- 971.  
Bierlen, R., & Featherstone, A. M., (1998). Fundamental q, Cash Flow, and Investment: 
Evidenc e from panel data. The Review of Economics and Statistics 80: 427 - 435. 
Blasko, B., (2011). World importance and present tendencies of dairy sector: a thesis presented 
in partial fulfilment of the requirements for Doctor of Philosophy (Applied Economics 
and Rural Development) University of Debrecen, Budapest .  
Boehlje, M., (1973). The Entry- Growth - Exit Processes in Agriculture.  Southern Journal of 
Agricultural Economics, 5( 1) , 23- 36.   
Chapter X: References 
 
192 
Boehlje, M., (1992). Alternative models of structural change in agriculture and related 
industries. Agribusiness 8( 1 ) , 219- 231. 
Boehlje, M., (1994). Evaluating farm financial performance. Journal of the American society of 
farm managers and rural appraisers, 57, 4- 9 
Boehlje, M. & Schiek, W. (1998). Critical Success Factors  in a Competitive Dairy Market.  
Journal of dairy science 81( 6): 1753 – 1760.  
B oehlje, M. D., Dobblins C. & Miller, A., (2001). Are your farm business management skills 
ready for the 21st Century?: Self - assessment checklists to help you tell. Retrieved in 
March 2012 from:  http://ceplacer.ucdavis.edu/files/47701.pdf  
Bogan, C. E., & English, M. J., (1994). Benchmarking for best practices: winning  through 
innovative adaptation. New York: McGraw - Hill. 
Bogetoft, P. & Otto, L., (2011). Benchmarking with DEA, SFA, and R. International series in 
Operation research and management science 157 (1).  
Bravo - Ureta, B. E., (1986). Technical efficiency measures for dairy farms base don a 
probabilistic frontier function model. Canadian Journal of Agricultural Economics 
34( 3): 399- 415. 
Bravo - Ureta, B. E., & Rieger, L., (1991). Dairy farm efficiency measurement using stochastic 
frontiers and neoclassical duality. American Journal of Agricultural Economics 73: 421-
428.  
Bureau, J. C., Butault, J. P., Hoque, A., (1992). International comparisons of costs of wheat 
production in the EC and United States. Staff Report -  Economic Research Service, 
United States Department of Agriculture  (9222) : 21p.  
Bywater, A. C., (2010). Analisis de indicadores: Proyecto M1P2, Identificacion y monitoreo de 
sistemas de produccion de leche por macrozonas lecheras en Chile [Indicator analysis : 
Project M1P1, Identification and monitoring of dairy farming systems in Chile] . Dairy 
Consortium Report.  
Camp, R. C., (1989). Benchmarking: the search for industry best practices that lead to superior 
performance. Wisconsin, USA: ASQC Quality press.  
Chapter X: References 
193 
Campbell, B., (2008 ). Benchmarking medieval economic development: England, Wales, 
Scotland, and Ireland. Economic historic review 61(4): 896- 945. 
Campos, A., Beck, A., & Hausdorf, S., (1995). Análisis económico de la producción lechera en 
predios de la zona Central de Chile [ Economic analysis of dairy farming in the Chilean 
Central área]. Agricultura técnica 55: 140- 146. 
Carrillo, B. L., Rosas D.,  Moreira V. H., Lerdon, J. (2 010). Milk payment system in southern 
Chile: minimum and maximum milk price to farmer Esquemas de pago de leche en el 
sur de Chile: prec io minimo y maximo a productor. Idesia 28( 3) : 61- 67. 
Carrillo, B., Moreira, V. & Gonzalez, J., (2011). Caracterizacion y tipificacion de sistemas 
productivos de leche en la zona centro- sur de Chile: un analisis multivariable . 
[Characterization and classification of milk production systems in the central - southern 
Chile: a multivariate analysis]. IDESIA Chile 29( 1): 71- 81. 
Carter, O., Murphy, A. M. & Cheal, D., (2003). Natural Temperate Grassland. Flora Ecology 
Research. Arthur Rylah Institute for Environmental Research Department of Natural 
Resources & Environment (Victoria) . 
Central Bank of Chile [BCCL], (2012). Exchange rate CLP$:US$. Retrieved July 2012 
from: http://si3.bcentral.cl/Indicadoressiete/secure/Indicadoresdiarios.aspx . 
Central Intelligence Agency (CIA), (2012). The world factbook. Retrieved March 2012 
from: https://www.cia.gov/library/publications/the - world- factbook . 
Challies, E. & Murray, W., (2006). Productive transformations and bilateralism in the semi -
periphery: A comp. political economy of the dairy c omplexes of NZ and Chile. Asia 
Pacific viewpoint 47 ( 3 ): 351 - 365.  
Charnes A, Cooper W ., Rhodes E., (1978). Measuring the efficiency of Decision Making Units. 
European Journal of Operational Research 2: 429- 444. 
Chatterjee, S. & Wernerfelt, B. (1991). The link between resources and type of diversification: 
Theory and evidence. Strategic management journal 12 ( 1): 33- 48. 
Chavas, J . P., &  Aliber, M., (1993). An Analysis of Economic - Efficiency in Agricul ture –  a 
Non - parametric approach. Journal of Agricultural and Resource Economics 1: 1 - 1 6. 
Chapter X: References 
 
194 
Chopra, S. & Meindl, P. (2001). Suppy chain management: Strategy, Planning and Operation. 
Prentice Hall, Upper Saddle River: 457 p.  
Cinca, C. S. & Molinero, C. M., (2004). Selecting DEA Specifications and Ranking  Units Via 
PCA. Journal of RS 55( 5): 521- 528.  
Cochet, H. & Devienne, S., (2006). Operation and economic performance of farming systems: 
A regional approach. Cahiers Agricultures 15 (6): 578- 583.  
Coelli, T. J., (1997). A guide to DEAP Version 2.1: A Data envelopment analysis (computer) 
program. CEPA Working paper 96(8). Department of Economics, University of New 
England, Armidale . 
Coelli, T. J., Prasada Rao, D., O’Donnell, C. & Battese, G., (2005). An introduction to efficiency 
and productivity analysis (2 nd Ed.). Springer, USA . 
Conforte, D., Garnevska, E., Kilgoure, M., Locke, S. & Scrimgeour, F., (2008). Diary: The key 
elements of success and failure in the NZ Dairy industry, December 2008, The 
Agribusiness Research and Education Network (AREN). Retrieved 1 st of March 2012 
from: http://www.agmardt.org.nz  . 
Conner, J . R., Hamilton, W . T., Sheehy, D . P., Stuth, J . W. & Kreuter, U . P., (1998). Grassland-
based livestock production in temperate zones. World animal review 90:  6- 13.   
Consorcio Lechero, (2012 ). Estrategia del desarrollo competitivo del sector lácteo chileno. 
[Dairy Consortium. Competitive development strategy for the Chilean dairy sector]. 
Retrieved from: http://www.consorciolechero.cl/chile/docs/Estrategia - Desarrollo -
Sectorial - 2010- 2020.pdf .  
Cooper, W.W., Seiford, L. M., Tone, K., (2006). Introduction to Data envelopment analysis and 
its uses. USA, Spring er Science+Business media, Inc.  
Covey, S. R., (1989). The seven habits of highly effective people : restoring the character ethic. 
New York: Fireside Book .  
Curran, W. , Scott, C. B., Horton, J., N . James, Weinstein, A. S., Fischbach, J., Chang, C. H., 
Rotman, M., Asbell,  S.,  Krisch, R. &  Nelson , D., (1993). Recursive Partitioning 
Analysis of Prognostic Factors in Three Radiation Therapy Oncology Group Malignant 
Glioma Trials . Journal National Cancer Institute 85( 9) : 70 4- 710.  
Chapter X: References 
195 
DairyBase, (201 0). Report de scription handbook  (l ast update) . Retrieved January 2011 
from: http://www.dairybase.co.nz/page/pageid/2145841079/ Reports . 
DairyNZ, (2009). Strategy for New Zealand dairy farming 2009/ 20. Retrieved April 2012 
from: http://www.dairynz.co.nz/file/fileid/28814 . 
DairyNZ, (20 10). Annual report 2009/ 10. Retrieved April  2012 
from: http://www.dairynz.co.nz/Publications/AnnualReport/2009.  
DairyNZ, (2011). Annual repo rt 2010/ 11. Retrieved March 2012  
from: http://www.dairynz.co.nz/page/pageid/2145866853/ Dairy_Industry . 
DairyNZ, (2012). Economic survey 2010/ 11. Retrieved August 2012 
from http://www.dairynz.co.nz . 
Dartt, B. A., Lloyd, J. W., Radke, B. R., Black, J. R. & Kaneene, J. B., (1999). A comparison of 
profitability and economic efficiencies between management- intensive grazing and 
conventionally managed dairies in Michigan. Journal of dairy science 82: 2412– 2420.  
de Lauwere, C. C. (2005). The role of agricultural entrepreneurship in Dutch agriculture of 
today. Agricultural Economics 33( 2): 229- 238.  
de Vries, A. & Risco, C.A., (2005). Trends and seasonality of reproductive performance in 
Florida and Georgia dairy herds  from 1976 to 2002. Journal of dairy science 88( 1):  
3155–3165 . 
Decimavilla, E., San Juan, C., & Sperlich, S., (2008). Land prices with urban sprawl: a model 
for Spain . Economia Agraria y Recursos Naturales 8 (1) : 3 - 20.  
Derpsch, R., & Moriya, K., (1998). Im plications of no- tillage versus soil preparation on 
sustainability of agricultural production. Advances in Geoecology 31:1179- 1186. 
DeWitt, T. J., & Schneider, S. M., (2004). Phenotypic plasticity: Functional and conceptual 
approaches. New York: Oxford Uni versity Press.  
Diaz, C., & Williamson, C., (1998). Trade agreements and competitiveness: evidence from the 
Chilean dairy sector. Pontificia Universidad Catolica de Chile, School of Management. 
Studies in Business. ABANTE Journal 1: 58- 88. 
Chapter X: References 
 
196 
Diewert, E., &  Lawrence, D., (1990). Measuring New Zealand’s productivity. Treasury working 
paper 99/5, report prepared for the Department of Labour, Reserve Bank of New 
Zealand and The Treasury.  
Dillon, P., Hennessy, T., Shalloo, L., Thorne, F., & Horan, B. (2008). Future outlook for the 
Irish dairy industry: a study of international competitiveness, influence of international 
trade reform and requirement for change. Proceedings paper I nternational Journal of 
Dairy Technology 61(1): 16 - 29.  
Dillon, P., Roche, J. R., Sha lloo, L. & Horan, B., (2005). Optimizing financial return from 
grazing in temperate pastures. Utilization of grazed grass in temperate animal systems. 
Proceedings of a satellite workshop of the XXth International Grassland Congress, 
Cork, Ireland, July 2005: 31 - 147.  
Dirección  Meteorológica  de Chile, (2008). Anuario climatologico 2007. Retrieved February 
2013 from http://164.77.222.61/climatologia/ . 
Dirección  Meteorológica  de Chile, (2008). Anuario climatologico 2008. Retrieved February 
2013 from http://164.77.222.61/climatologia/ . 
Dirección  Meteorológica  de Chile, (2010). Anuario climatologico 2009. Retrieved February 
2013 from http://164.77.222.61/climatologia/ . 
Dirección  Meteorológica  de Chile, (2011). Anuario climatologico 2010. Retrieved February 
2013 from http://164.77.222.61/climatologia/ . 
Dirección  Meteorológica  de Chile, (2012). Anuario climatologico 2011. Retrieved February 
2013 from http://164.77.222.61/climatologia/ . 
Dobson, W., (2003). Developments in the Dairy Industries of Mexico, Ce ntral America, 
Argentina, and Brazil: Implications for the U.S. Dairy Sector. Babcock Institute for 
International Dairy Research and Development, Discussion Paper, 37616.  
Doehring, T. A., ( 2001 ). Analyzing the profitability of your operation. Retrieved March 2012 
from: http://www.centrec.com/Articles/FinAnalysisFarmRanches/Profitability.pdf . 
Dooley, A. E., (2002). Milk segregation on dairy farms : a thesis presented in partial fulfilment 
of the requirements for the degree of Doctor of Philosophy in Agricultural systems and 
Management at Massey University, Palmerston North, New Zealand .   
Chapter X: References 
197 
Draaiyer, J., Dugdill, B., Bennett, A. & Mounsey, J., (2009). Milk testing and payment systems. 
Food and Agriculture Organization, Resource book. Rome: Publishing and Multimedia 
service. 
Emerson, A. & Rowarth, J., (2009). Future food farming: New Zealand Inc. meeting 
tomorrow’s markets. Wellington, New Zealand: NZX Ltd.  
Escobar, G., Mladinic, C., Sanhueza, R., & Diaz, O. (2006). Rural territorial development: The 
milk territory in Southern Chila. Rimisip. Retrieved September 2012 
from www.rimisip.org. 
Ethier, W. J., (1982). National and international returns to scale in the modern theory of 
international trade. American Economic Review 72( 3) :389- 405. 
Evans, D.S., (1987). Tests of Alternative Theories of Firm Growth. Journal of Political 
Economy 95: 657- 7.  
Fairweather, J. & Keating, N.  (1990). Management styles of Canterbury farmers. Canterbury 
NZ: Lincoln University Agrigusiness and Economics research unit, Research report 
205.   
Fare, R., Grosskopf, S. & Lovell, C. A. K., (1994). Production Frontiers . Cambridge: 
Cambridge University Press. 
Farm Financial Standards Council (FFSC), (1997). Financial guidelines for agricultural 
producers: Recommendations of the Farm Financial Standards Council. Retrieved 
December 2011 from: http://agmarketing.extension.psu.edu/Business/PDFs/ 
FinGuidAgProd.pdf  
Farrell, M. J., (1957). The measurement of productive efficiency. Journal of Royal Statistical 
Society, Series A, CXX, 253 - 281.  
F erreira, O., Mujica, F., Uribe, H., Lanuza, F., Quinteros, G., & Concha, C., (2010). Chilean 
milk recording and its importance for the national herd genetic improvement. Agro Sur 
38( 3): 178- 193. 
Ferris, C. P., (2007). Sustainable pasture- based dairy systems -  meeting the challenges. 
Canadian Journal of Plant Science 87( 4) : 723 - 738.  
Chapter X: References 
 
198 
Fingleton, W. A., (1995). Comparative costs and returns for milk production in EU countries. 
Annual conference of the agricultural economics society of Ir eland, Dublin, October 
1995.  
Fischer, G., Shah, M., & van Velthuizen, H., (2002). Climate change and agricultural 
vulnerability: A special report, prepared by the International Institute for Applied 
Systems Analysis. United Nations Institutional Contract Agreement No. 1113 on 
Climate Change and Agricultural Vulnerability. World Summit on Sustainable 
Development, Johannesburg 2002. 
Fleming , E., Farrell, T., Villano, R. & Fleming, P., (2006). Is farm benchmarking the new 
acceptable face of comparative analysis?. A ustralasian Agribusiness Review 14(1). 
Folke, C., (2006). Resilience: The emergence of a perspective for social - ecological systems 
analyses. Global Environmental Change-Human and Policy Dimensions 16( 3) : 253 - 267.  
Fonterra, (2012). Fonterra annual report 2011. Retrieved August 2012 
from: http://www.fonterra.com/wps/wcm/connect/fonterracom/fonterra.com/Home/Sear
ch+Results?search_query=report+201 1&x= 0&y=0 . 
 Food and Agriculture Organisation of the United Nations (FAO), (2006). Greenhouse gas 
emissions from the dairy sector: A life cycle assessment. Retrieved December 2012 
from: http://www.fao.org/docrep/012/ k7930e/k7930e00.pdf  
Food Chain Centre (FCC), (2007). Farm Benchmarking: Lessons from the past five years. 
Retrieved March 2012 from: www.foodchaincentre.com . 
Fox, G., Bergen, P. A ., & Dickson, E., (1993). Why are some far ms more successful than 
others? A review. In Size, Structure, and the Changing Face of American Agriculture , 
Ed., Hallam Boulder CO: Westview Press.  
Frangley, A. G., & Engelbrecht, R. (1998). Economic deregulation and the adjustments of New 
Zealand farmers. Choices First Quarter : 14- 18. 
Fraser, I., & Cordina, D., (1999). An application of data envelopment analysis to irrigated dairy 
farms in Northern Victoria, Australia. Agricultural Systems 59 : 267- 282 
Gale, H. F., (199 4). Longitudinal analysis of farm size over the farmer's life cycle. R eview of 
agricultural economics 16 (1) : 113 - 123.  
Chapter X: References 
199 
Garcia, S. C., Cayzer, F. J., Holmes, C. W. & MacDonald, A., (1998). The effect of calving 
season on milk production: A system study . Proceedings of the NZ Society of animal 
production 58. 
Garcia, S. C. , & Holmes, C. W., (1999). Effects of time of calving on the productivity of 
pasture‐based dairy systems: A review . New Zealand Journal of Agricultural Research 
42( 3) :  347 - 362. 
Gardner, J. W. & Hargreaves, B., (1987). The economics of leasing far m land. NZ Farmer, 
August 13: 60- 61 
Gardner, J. &  Shadbolt, N.,  (2005) . Wealth creation. In Farm Management in New Zealand, 
edited by N. Shadbolt and S. Martin, Oxford University Press, Melbourne.  
Gasson, R., (1973). Goals and values of farmers. Journal of Agricultural Economics 24 ( 3): 521-
542.  
Gasson, R. & Errington, A. (1993).  The farm family business . Wall ingford, Oxon: CAB 
International  
Gattoufi S, Oral M, Reisman A. (2002). Data Envelopment Analysis literature: A Bibliography 
Update (1951- 2001). Working Paper SUGSM –  02- 08, Sabanci Uni versity, Istanbul, 
Turkey.  
Gaul, P., & Hughes, T., (1996). Canterbury diary farming – fitting feed supply to herd demand. 
Dairy farming annual 48: 4- 9 
Gazzarin,  C., Frey, H. J., Petermann, R. &  Hoeltschi, M., (2011). Pasture feeding or cowshed 
feeding -  which is more economical? A grarforschung schweiz, 2, (9) , 418- 423 
Gersick, K. E., Davis, J., McCollom Hampton, M. & Lansberg, I., (1997). Generat ion to 
generation : life cycles of the family business. Boston : Harvard Business School Press   
Gloy, B. A., & LaDue, E. L., (2003). Financial management practices and farm profitability. 
Agricultural Finance Review 63, (2) , 157- 174. 
Godfrey, J. (2011). Statistics advice on principal component analysis and regression 
partitioning. In J. Bensemann (Ed.). Palmerston North.   
G riliches, Z., (1987). Productivity measurement problems. In The New Palgrave: A Dictionary 
of Economics, Ed. By J. Eatwell, M. Milgate , and P. Newman. New York: Macmillan.  
Chapter X: References 
 
200 
Hadrich, J. C., &  Olson , F., (2011). Joint measurement of farm size and farm performance: a 
confirmatory factor analysis. Agricultural Finance Review 71( 3): 295- 309. 
Haghiri, M., Nolan, J. F. & Tran, K. C., (2004). Ass essing the impact of economic liberalisation 
across countries: A comparison of dairy industry efficiency in Canada and the United 
States. Applied Economics 36( 11) : 1233- 1243. 
Hallam, D., & Machado, F., (1996). Efficiency Analysis with Panel Data: A Study of  
Portuguese Dairy Farms.  Foundation for the European Review of Agricultural 
Economics 23( 1): 79- 93. 
Handley D., Shankar, B, Thirtle, C., &  Coelli, T., (2001) . Financial Exposure and Farm 
Efficiency: Evidence from the England and Wales dairy Sector. Paper Presented at the 
annual meetings of the American Agricultural Association, Chicago, August 5- 8, 2001.  
Hansen, L.B., (2000 ) . Consequences of selection for milk yield from a geneticist's viewpoint . 
Journal dairy science 83( 1): 1145– 1150.  
Hansen, B. G., Stokstad, G., Hegrenes, A., Sehested, E. & Larsen, S. (2005). Key performance 
indicators on dairy farms. Journal of international farm management 3( 1 ) .  
Hanson, G. D., Cunningham, L. C., Ford, S. A., Muller, L. D., & Parsons, R. L., (1998). 
Increasing Intensity of  Pasture Use with Dairy Cattle: An Economic Analysis. Journal 
Production Agriculture 11(2) :  175 –179.  
Harris, B.L., ( 2005 ).  Breeding dairy cows for the future in New Zealand. New Zealand 
veterinary journal (53) : 384- 389.  
Harris, B. L., Kolver, E. S., (2001).  Review of Holsteinization on intensive pastoral dairy 
farming in New Zealand. Journal of dairy science 84: E56 - E61.  
Harris, B. L. & Winkelman, A. M., (2000). Influence of North American Holstein genetics on 
dairy cattle performance in New Zealand. Proceedi ngs of the Large Herds Australia 
Conference: 122- 136.  
Harrison, W. R., & Kennedy, L. P., (1997). A neoclassical economic and strategic management 
approach to evaluating global agribusiness competitiveness. Competitive review 7( 1) , 
14- 27.  
Chapter X: References 
201 
Hicks, C. & Dietmar, R. (2007). Improving cleaner production through the application of 
environmental management tools in China. Journal of Cleaner Production 15( 5) : 395-
408.  
Hodgson R., & Chestnut K., (1999). Northern Wairoa experimental and demonstration farm 
calving systems trial results. Dairy farm ing Annual, Massey University, 159 p.  
Holmes, C. W.. (1986). Calving dates and their importance to the dairy farmer. Dairy farming 
annual 1: 55- 67. 
Holmes, C. W. , ( 2001).  Managing fertility in the New Zealand dairy herd. Proceedings of the 
New Zealand Society of Animal Production 61: 135 - 140.  
Holmes, C. W., (2003). Low cost production of milk from grazed pastures: An outline of dairy 
production systems in NZ. Palmerston North, NZ: Massey University .  
Holmes, C. W., Brookes , I. M., Garrick, D., Mackenzie, D.S., Parkinson, T.J., & Wilson, G.F., 
(2003). Milk production from pasture. Principles and practices. Massey University , 
Palmerston North, New Zealand . 
International farm comparison network (IFCN), (2010). Dairy report 2009 : For a better 
understanding of milk production world- wide. IFCN Dairy research Center, Germany.  
International farm comparison network  (IFCN) , (2011). Dairy report 2010 : For a better 
understanding of milk production world- wide. IFCN Dairy research Center,  Germany.  
International farm comparison network (IFCN), (2012). Dairy report 2011: For a better 
understanding of milk production world- wide. IFCN Dairy research Center, Germany.  
INDEX Mundi, (2012). GDP composition by sector, Retrieved March, 2012 
from: http://www.indexmundicom/chile/gdpcomposition_by_sector.html.September  
Jack, L. & Boone, J., (2009). Sustainable change and benchmarking in the food supply chain. In 
L. Ja ck (Ed.), Benchmarking in food and farming (p. 132). Surrey, England: Gow er 
Publishing Limited.  
Jaforullah, M. & Devlin, N. (1996). Technical efficiency in the NZ dairy industry: A frontier 
production approach. New Zealand Economic Papers 30( 1): 1 - 17. 
Chapter X: References 
 
202 
Jensen, R. N. ; Clark, D. A. ; Macdonald, K. A.(2005). Resource efficient dairying trial: 
measurement criteria for farm systems over a range of resource use. Proceedings of the 
New Zealand Grassland Association 67: 47- 52. 
Jiang, N. (2011). Efficiency analys is of NZ dairy farming and the issue of climate change 
policy. A stochastic frontier analysis. A thesis submitted for the degree of Doctor of 
Philosophy at the University of Auckland, New Zealand.  
Johnson, R. W., & Frengley, G. A., (1994). Economic adjustm ents and changes in the financial 
viability of the farming sector: The NZ experience. American Journal of Agricultural 
Economics 76( 4): 1034- 1040. 
Journeaux, P., (2002). Dairy farming – Trends over time. Proceedings of the 5th Annual 
conference of the New Zealand Agricultural and Resource Economics Society: 1 - 13. 
Jovanovic, B., (1982). Selection and Evolution of Industry. Econometrica 50:  649 - 70.  
Kay, R. D. & Edwards, W. M., (1999). Farm management . Seventh edition. New York: 
McGraw - Hill.  
Kay, R. D., Edwards, W. M. & Duffy, P. A. (2012). Farm management . Fourth  edition. New 
York: McGraw - Hill. 
Kay, R. D., Edwards, W. M. & Duffy, P. A. (2012). Farm management. Seventh edition. New 
York: McGraw - Hill.  
Kohl, D. & Wilson, T., (1997). Understanding key financial ratios and benchmarks. Northwest 
Farm Credit Services, Business Tools Bulletin, Spokane, Washington.  
Kristensen, E. &  Jakobsen, E.  B., (2011). Challenging the myth of the irrational dairy farmer; 
understanding decision- making related to herd health. New Zealand Veterinary Journal 
59: 1- 7.  
Kolver, E.S., &  Muller, L.D., (1998) .Performance and nutrient intake of high producing 
Holstein cows consuming pasture or a total mixed ration. Journal of Dairy Science  
81: 1403 –1411.  
Kraus, S., Rigtering , C. & H uges, M ., (2012). Entrepreneurial orientation and the business 
performance of SME’s: a quantitative study from the Neatherlands. Rev  Management 
science 6: 161- 182.  
Chapter X: References 
203 
Laca- Viña, H. R., (2010 ). A n application of Malmquist productivity index to compare 
technological and growth differences between traditional and non- traditional dairy 
regions in New Zealand : a thesis presented in partial fulfilment of the requirements for 
the degree of PhD in Agribusiness at Massey University, Palmerston North, New 
Zealand . 
Langemeier , M., ( 2010 ). Persistence in financial performance. Journal of international farm 
management, 5 (2) . 
Lapple, D., Hennessy, T., & O'Donovan, M. (2012). Extended grazing: A detailed analysis  of 
Irish dairy farms . Journal of Dairy Science 95( 1): 188- 195.  
Latruf fe, L. (2010). Competitiveness, productivity and efficiency in the agricultural and agri -
food sectors. OECD Food, Agriculture and Fisheries Working Papers 30: 63 p.  
Le Marechal, C., Thiery, R., Vautor, E., Le Loir, Y., (2011). Mastitis impact on technologi cal 
properties of milk and quality of milk products - a review. Dairy Science & Technology 
91( 3): 247- 282. 
Lerdon, J., Báez, A., & Azócar, G., (2008). Relation ship between social, productive and 
economical variables of sixteen small dairy farms of the Valdivia Province, Chile. Arch 
Med Vet 40: 179- 185.  
Livestock Improvement Corporation (LIC)  & DairyNZ, (2011). New Zealand dairy statistics 
2010- 11. Retrieved April 2012 from:  www.lic.co.nz/pdf/DAIRY% 20STATISTICS  
%2010- 11- W EB.pdf . 
Lobos, G., Soto R., Zenteno, N., & Prizant, A. (2001). Analisis de eficiencia y rentabilidad 
economica en dos lecherias de la Region del Maule, Chile [Efficiency and profitability 
analisis for two dairy farms in the Maule Region, Chile]. Agricultura tecnica  61: 367-
378.  
Lopez Andreu, M., Featherstone , A. M., Langemeier , M. R., & Grunewald, O., (2006). Impact 
of Financial Variables on Production in Kansas Farms Efficiencies. Selected Paper 
prepared for presentation at the American Agricultural Economics Association Annual 
Meeting, Long Beach, California, July 23- 26, 2006.  
Macdonald, J. M., (2007). Experience counts -  farm business survival in the U.S. Retrieved June 
2012 from: http://www.ers.usda.gov/AmberWaves/Apri l07/PDF/ April07.pdf  
Chapter X: References 
 
204 
MacDonald, M., Korb, P & Hoppe, R., (2007). Experience counts: Farm business survival in the 
US. Economic Research Service USDA  5(2): 10 - 15. 
Macmillan, K. L. and K. Moller (1977). Aspects of reproduction in New Zealand dairy herds. 2. 
Calving interval, breeding period and non- pregnancy rates. New Zealand veterinary 
journal 25(8) : 220 - 224. 
MacDonald, K. A. & Penno, J. W., (1998). Management decision rules to optimise milksolids 
production on dairy farms. Proceedings of the New Zealand Society of Animal 
Production 58:132- 135. 
MacDonald, K. A., Penno, J. W., Nicholas, P. K., Lile, J. A., Coulter, M., & Lancaster, J. A., 
(2001). Farm systems -  Impact of stocking r ate on dairy farm efficiency. Proceedings of 
the NZ Grassland Association: 223- 227. 
MacDonald, K. A., Penno, J. W., Lancaster, J. A. S., Roche, J. R., (2008). Effect of stocking 
rate on pasture production, milk production, and reproduction of dairy cows in pasture-
based systems. Journal of Dairy Science 91:2151 - 2163. 
Makeham, J . P,  & Malcom, L . R. , (1993). The farming game now . Cambridge, England: 
Cambridge University press.  
McCall, D. G. & Clark, D. A., (1990). Optimised dairy grazing systems in the Northeast United 
States and NZ. System Analysis. Journal of Dairy Science 82 : 1795- 1807.  
Malcom, B., (2004). Where is the economics?. Australian Journal of Agricultural and Resource 
Economics , 48 (3):  395 - 417. 
Marin, S. R., Cavalheiro, A. G., Anschau, D., ( 2011). Sazonalidade do preço do leite no Rio 
Grande do Sul (1986- 2009) [Milk Price seasonality in Rio Grande do Sul] . Ciencia 
Rural 41( 2 ) : 361 - 364.  
Martin, S., Zwart, T., Gardner, J. & Parker, W., (2005). Introduction. In N. Shadbolt & S. 
Martin (Eds.). Fa rm Management in New Zealand (pp. 95 - 122). Melbourne, VIC: 
Oxford University Press.  
Mattiauda, D. A., S. Tamminga, S., Gibb, M. J., Soca, P., Bentancur, O., & Chilibroste, P., 
(2013). Restricting access time at pasture and time of grazing allocation for Ho lstein 
dairy cows: ingestive behaviour, dry matter intake and milk production. Livestock 
Science 152( 1) : 53 - 62. 
Chapter X: References 
205 
McCarthy, S., Horan, B., Dillon, P., O’Connor, P., Rath, M. & Shalloo, L., (2007). Economic 
comparison of divergent strains of Holstein- Friesian cows in various pasture- based 
production systems. Journal of Dairy Science 90( 3): 1493- 1505. 
McCarthy, B.,  Delaby, L., Pierce, K . M., Journot, F., Horan, B., (2011). Meta - analysis of the 
impact of stocking rate on the productivity of pasture - based milk p roduction systems. 
ANIMAL 5 (5): 784- 794.  
McGuckian, (1996). Benchmarking sustainable farming systems. Victoria: Agricultural and 
management consultants. 
Ministry of Agriculture and Forestry (MAF ), (2011). Situation and outlook for New Zealand 
agriculture and forestry. Wellington: MAF publisher.  
Montgomerie, W. A., ( 2004).  Future genetic progress of dairy cattle in New Zealand. 
Proceedings of the New Zealand Society of Animal Production 64: 96- 100.  
Moot D, Mills A, Lucas R and Scott W (2010). New Zealand: Country forage 
profiles. http://www.fao.org/ag/AGP/AGPC/doc/Counprof/newzealand/newzealand1.ht
m. 
Moreira, V., (1999). Analisis del costo de produccion de leche en los sistemas productivos 
lecheros en Chile. [Production costs analysis in Chilean dairy farming systems]. Estudio 
de casos. Competitividad de la produccion lechera nacional,  p.221- 258. Uni versidad 
Austral de Chile, Facultad de Ciencias agrarias, Santiago de Chile.  
Moreira, V. H. and B. E. Bravo - Ureta (2010). Technical efficiency and metatechnology ratios 
for dairy farms in three southern cone countries: a stochastic meta - frontier model. 
Journal of Productivity Analysis 33( 1) : 33 - 45.  
Nahuelhual, L., Engler, A., Carrillo, B., Moreira, V. H., & Castro, I. (2009). Adoption of 
cleaner production practices by dairy farmers in southern Chile. Ciencia e Investigacion 
Agraria 36( 1): 97- 106.  
Nars, R., Barry, P., & Ellinger, P., (1998). Financial structure and efficiency of grain farms. 
Agricultural Finance Review 58: 33- 48. 
National research council, (2001). Nutrient requirements of dairy cattle, 10th rev. Edited  
National Academy of Sciences, Washington D. C. USA . 
Chapter X: References 
 
206 
National research council (NRC) , (2010). Toward sustainable agriculture systems in the 21 st 
century. Washington, DC: National Academy Press.  
Nell, W. T. & Napier, R. J., (2005). Strategic approach to farming success . South Africa: Will 
Nell Agricultural management consultant . 
Neto, M. C., Campos, J. M. D., Soares de Oliveira, A., Texeira Gomes, S., (2012). Identification 
and quantification of benchmarks of milk production systems in Minas Gerais.  Revista 
Brasileira De Zootecnia [Brazilian Journal of Animal Science] 41( 10): 2279- 2288. 
New Zealand Institute of Economic Research, (2011). New Zealand exports: more diverse than 
commonly thought. Retri ved April 2012  
from:  http://nzier.org.nz/sites/nzier.org.nz/files . 
New Zealand  Statistics , (2012). New Zealand in profile: 2012. Retrieved  August 2012  
from: http://www.stats.govt.nz/browse_for_stats/snapshots - of- nz/nz - in- profile-
2012.aspx . 
NIWA , (2010). State of the climate 2010: A snapshot of recent climate in New Zealand. NIWA 
Science &  Technology Series N o56. Retrieved October 2012 
from: http://www.niwa.co.nz . 
Nuthall, P. L., ( 2010 ). Farm business management: The core skills. Oxforshire, UK: CABI 
Head Office . 
O'Donnell, C. J., Prasada Rao, D. S., & Battese, G. E., (2008). Metafrontier frameworks for the 
study of firm- level efficiencies and technology ratios. Empirical Economics 34( 2) : 231-
255.  
Oficina de Estudios y Politicas Agrarias (ODEPA) [Office of studies and agrarian policies] , 
(2012). Leche: produccion, recepcion, precios y comercio exterior. Retrieved March, 
2012 from:  http://www.odepa.gob.cl/odepaweb/serviciosinformacion/Lacteos/informe_  
lacteo_1211.pdf  
Olson, K. D., (2010). Economics of farm management in a global setting. USA: John Willey &  
sons, Inc. 
Olsson, R. (1988). Management for success in modern agriculture. European Review of 
Agricultural Economics 15( 2/ 3) : 239 - 259. 
Chapter X: References 
207 
Olson, K. & Vu, L. (2009). Economic efficiency in farm households: Trends, explanatory 
factors, and estimation methods. Agricultural Economics 40 ( 5): 587- 599. 
Oltmans, A. W., (2007). A new approach in farm business analysis to fit a changing farmland 
investment market. Proceedings of the 16 th International farm management congress. 
Cork, Ireland.  
Organization of Economic Cooperation and Development (OECD), (2011). Fostering 
Productivity and Competitiveness in Agriculture. Retrieved December 2012 
from: http://www.oecd.org/agriculture/agricultural - policies/  . 
 Organization of Economic Cooperation and Development (OECD), (2011). OECD - FAO 
Agricultural Outlook 2011- 2020. Retrieved March, 2012 from: http://www.agri-
outlook.org/dataoecd/13/ 2/48186214.pdf . 
Painter, M. (2007). A comparison of the dairy industries in Canada and New Zealand  Journal 
of International Farm Management, 4( 1) .   
Parker, W. J., Muller, L. D., & Buckmaster, D. R., (1992). Managem ent and Economic 
Implications of Intensive Grazing on Dairy Farms in the Northeastern States. Journal of 
Dairy Science 75: 2587–2597.  
Parmenter, D. (2010). Key Performance Indicators: Developing, implementing and using 
winning KPIs [ S econd E dition]  New Jersey, US: John Wiley & Sons, Inc.  
Patton, D., Shalloo, L., Pierce, K ., M., & Horan, B ., (2012). A biological and economic 
comparison of two pasture- based production systems on a wetland drumlin soil in the 
northern region of Ireland. Journal of Dairy Science 95( 1):  484- 495. 
Pedernera, M., Garcia, S. C., Horagadoga, A., Barchia, I. & Fulkerson, W. J., (2008).  Energy 
balance and reproduction on dairy cows fed to achieve low or high milk production on a 
pasture- based system. Journal of dairy science 91(10) : 3896- 3907.          
Penno, J. (1999). Stocking rate for optimum profit. Proceedings of the South Island dairy 
Event: 25- 43. 
Penno, J. (2000). Profitable milk production from pasture. Irish Grassland and Animal 
Production Association Journal 34: 78- 86. 
Chapter X: References 
 
208 
Penno, J., McGrath, J. M. , Macdonald, K. A., Coulter, M., Lancaster, J. A., (1999). Increasing 
milksolids production with supplementary feeds. Proceedings of the New Zealand 
Society of Animal Production 59( 0) : 188- 191. 
Perkin, P., (1992). An investigation i nto the relationship between farm and farmer 
characteristics and objectives among a sample of Berkshire farmers. University of 
reading, Department of Agriculture, unpublished PHD thesis . 
Pertusa- Ortega, E. M., Molina - Azorin, J. F., Claver - Cortes, E., (2010 ). Competitive strategy, 
structure and firm performance: A comparison of the resource- based view and the 
contingency approach. Manage decis 48( 7 - 8) : 1282- 1303. 
Piccand V., Cutullic, E., Schori, F., Keckeis, K., Gazzarin, C., Wanner, M. & Thomet, P. 
(2011) . Which cow for pasture - based production systems?: production, reproduction 
and health. Recherche Agronomique Suisse 6: 252- 257. 
Popp, J. (2010). Which kind of milk will we drink in 2018? Haszon Agrar Magazin, Retrieved 
March 2012 from: http://www.haszon.hu/agrar/allattenyesztes/390- milyentejet- iszunk-
2018- ban.html. 
Porter, M. E., (1980). Competitive strategy : Techniques for analysing industries and 
competitors.  New York : Free Press.  
Porter, M. E., (1985). Competitive Advantage : Creating and sustaining superior performance. 
New York: Free Press.   
Porter, V., (1991). Cattle: A handbook to the breeds of the world. XX: Christopher Helm Ltd.  
PROLESUR ( Productores de Leche del Sur ) , ( 2011). Resumen pauta de pago de leche  [Pattern 
of milk payment abstract]. Retrieved June 2012 from:  http://www.prolesur.cl . 
Purdy, B. M., Langemeier, M. R., & Featherston e, A. M. (1997). Financial performance, risk, 
and specialization. Journal of Agricultural and Applied Economics 29 ( 1): 149- 161. 
Robobank, (2012). Exchange rate NZ$: US$. Retrieved July 2012 
from: http://www.rbnz.govt.nz/statistics/exandint/b1/download.html  
Rahelizatovo, N. C., & Gillespie, J. M., (2004). Factors influencing the implementation of best 
management practices in the dairy industry. Journal of Soil and Water Conservation 
59( 4): 166- 175. 
Chapter X: References 
209 
Rawlings, K. M., (1999). Key performance indicators for goal attainment in Dairy Farming: 
essential elements for monitoring farm business performance: a thesis presented in 
partial fulfilment of the requirements for Master’s Degree at Massey University, 
Palmerston North, New Zealand.  
Resilience alliance (2009). About RA. Retrieved August 2012 
from: http://www.resalliance.o rg/index.php/about_ra  
Robbis, B. & Wallace, D., (1992). The family business: How to successfully manage a family 
business. Melbourne: Business Library . 
Robison, L., & Barry, P. (1987). The Competitive Firm's Response to Risk. New York: 
Macmillan Publishing  Company. 
Robertson, M. (2010) 2010 Food Securi ty from Sustainable Agriculture . Proceedings of the 
15th ASA Conference, 15- 19 November 2010, Lincoln, New Zealand.  
Rodriguez, D., deVoil, P., Power, B., Cox, H., Crimp, S. & Meinke, H., (2011). The intrinsic 
plasticity of farm businesses and their resilience to change. An Australian example. 
Field Crops Research 124(2) : 157- 170. 
Ronan, G., & Clear y, G. (2000). Best practice benchmarking in Australia: issues and challenges, 
Agribusiness perspectives 2000. Retrie ved from http://www.agrifood.info/Review/ 
Perspectives/2000_Ronan/2000Ronan.htm . 
Rouse, P., Harrison, J. & Chen, L. (2010). Data Envelopment Analysis: A Practical Tool  to 
Measure Performance. Australian Accounting Review 20( 2): 165- 177. 
Rumelt, R. P. (1982). Diversification strategy and profitability. Strategic Management Journal, 
3( 4): 359- 369. 
Rust, J. W., Sheaffer, C. C., Eidman, V. R., Moon, R. D., Mathison, R. D., (1995). Intensive 
rotational grazing for dairy cattle feeding.  American Journal of Alternative Agriculture 
10( 4): 147- 151.  
Salas, L. 1995. Current status and prospects of the dairy sector. Border Agricultural Journal 3 
(1): 81- 88. 
Chapter X: References 
 
210 
Sankaran, J. K. & Luxton, P., (2003). Logistics in relation to strategy in dairying: The case of 
New Zealand dairy. International Journal of Ope rations & Production Management 
23: 522 – 545.  
Sauvant, D. & J. M. Perez (2010). Robustness, hardiness, flexibility, plasticity, resilienc e ... new 
quality criteria for animals and farming systems Foreword. P roductions Animales 23( 1).  
Schnettler B., López P. & Barchiesi, P. et al. (2004). Rentabilidad económica de sistemas 
lecheros con distinta inversión y financiamiento en la región de La Araucania, Chile: 
Estudio de casos. Agricultura Técnica  64: 192 - 204. 
Scholz, R. W., Blumer, Y. & Brand, F. (2012). Risk, vulnerability, robustness, and r esilience 
from a decision theoretic perspective. Journal of Risk Research 15( 3): 313- 330. 
Shadbolt, N. (2009). DairyBase: building a Best Practice Benchmarking System. In L. Jack 
(Ed.), Benchmarking in food and farming (pp. 39- 48). Surrey, England: Gower 
Publishing Limited.  
Shadbolt, N. ( 2011).  Competitive strategy analysis of NZ pastoral farming systems. 
www.ifmaonline.org: 18th International Farm Management Congress  [Conference 
Paper in Published Proceedings] : 122 - 132. 
Shadbolt, N. M. & Gardner, J., (2005). Financial management. In N. Shadbolt & S. Martin 
(Eds.). Farm Management in New Zealand (pp. 95 - 122). Melbourne, VIC: Oxford 
University Press.   
Shadbolt, N., Rutsito, B. & Gray, D., (2011). Defining and Measuring Resilience for New 
Zealand Dairy Farmers. Retrieved March 2012 
from: https://www.ifama.org/events/conferences/2011/cmsdocs/2011SymposiumDocs/3
11Symposium %20Paper.pdf  . 
Shallo o, L., Dillon, P., O’Loughlin, J ., Rath, M. & Wallace, M. (2004). Comparison of a 
pasture- based system of milk production on a high rainfall, heavy- clay soil with that on 
a lower rainfall, free draining soil. Grass Forage Science 59 : 157– 168.  
Shapiro, R. D . and J. L. Heskett (1985). Logistics strategy : cases and concepts. St. Paul, 
Minnesota, West Publishers.  
Siddique, F. & Gulfraz, M. (2009). Quality Assessment of Fresh Milk Obtained From Different 
Sources. Pakistan Journal of Zoology 41( 4) : 316 - 319. 
Chapter X: References 
211 
Silva, R. 1997. Estudio de rentabilidad de sistemas mixtos con predominancia lechera de la VII 
Región. [ Profitability study on mixed farming systems with prevalence of dairying 
within the VII Region] . Engineer Agronomist thesis . Universidad de Talca, Facul tad de 
Ciencias Agrarias, Talca, Chile.  
Smit, H.  J., Taweel,  H.  Z.,  Tas,  B.  M., Tamminga,  S.,  Elgersma,  A.,  ( 2005) . Comparison of 
techniques for estimating herbage intake of grazing dairy cows. Journal of Dairy 
Science 88: 1827–1836.  
Smith, R.; Moreira, V.;  Latrille, L., (2002). Caracterización de sistemas productivos lecheros en 
la X Región de Chile mediante análisis multivariable. Agricultura Técnica 62 : 375 - 395.    
Smyth, P., Butler, A. M., & Hennessy, T., (2008). Analysing the Economic Performance of Iri sh 
Dairy Farmers 1998 - 2006. The 82nd Annual Conference of the Agricultural Economics 
Society Royal Agricultural College, 31st March to 2nd April 2008  
Spulber, D. F., (2004). Management strategy. New York: McGraw – Hill . 
Statistics New Zealand, (2012).  New Zealand in profile: 2012. Retrieved 
from: http://www.stats.govt.nz/browse_for_stats/snapshots - of- nz/nz - in- profile-
2012.aspx . 
Steiner, G. A. (1997). Strategic planning : what every manager must know. New York, Free 
Press Paperbacks.  
Steinwidder, A., Starz, W., Podstatzky, L., Gasteiner, J., Pfister, R., Rohrer, H. & Gallnbock, 
M., (2011). Effects of time of calving on pasture - based dairy systems in mountainous 
regions. Zuchtungskunde 83( 3): 203- 215. 
Strobl, C., Malley, J. & Tutz, G., (2009). An introduction to recursive partitioning. Technical 
report 55, Department of Statistics, University of Munich . 
Sumner, D. A., & Wolf, C. A., (2002). Diversification, vertical integration, and the regional 
patter of dairy farm size. Review of Agricultural Economics 24 ( 2): 442- 457.  
Surinder Saggar, N. S. Bolan, R. Bhandra l, C. B. Hedley & J. Luo (2004).  A review of 
emissions of methane, ammonia, and nitrous oxide from ani mal excreta deposition and 
farm effluent application in grazed pastures.  New Zealand Journal of Agricultural 
Research 47( 4): 513- 544. 
Chapter X: References 
 
212 
Swinnen, J., & Squicciarini, P. (2012). Mixed Messages on Prices and Food Security. [Editorial 
Material]. Science 335 ( 6067) : 405 - 406.  
Taylor, J., & Foltz, J. (2006). Grazing in the Dairy State: Pasture Use in the Wisconsin Dairy 
Industry, 1993– 2003. Department of Agricultural and Applied Economics, University 
of Wisconsin, Madison US .  
Teuber, N. (1996) La pradera en el llano longitudinal de la X región (Valdivia - Chiloé). In  I. 
Ruiz, ed,. Praderas para Chile, 2 nd edition. Santiago: Instituto de Investigaciones 
Agropecuarias.  
Thomson, N. A., Turner, S. A. , Lopez - Villalobos, N. &  Glassey A.,  (2005). The effect of 
genotype on milk composition, milk value, and dairy farm profitability. Proceedings of 
the New Zealand Society of Animal Production 65: 276 - 282.  
Thorne, F. S., & Fingleton, W. (2006). Examining the relative competitiveness of milk 
production: An Irish case study (1996 –  2004). Journal of International Farm 
Management, 3( 4) .   
Tozer, P. R., Bargo, F., & Muller, L. D., (2003). Economic analyses of feeding systems 
combining pasture and total mixed ration. Journal of Dairy Science 86(3): 808- 818.  
Tozer, P. R., Bargo, F., Muller, L. D., (2004). The effect of pasture allowance and 
supplementation on feed efficiency and profitability of dairy systems. Journal of Dairy 
Science 87( 9) : 2902- 2911.  
Tripp, R., ( 1991 ). The farming systems research movement and on - farm research. In T ripp, R. 
Ed. Planned change in farming systems: Progress in on-farm research. John Widely 
and sons, West Sussex: 3:16 
Tripp, R., Anandajayasekeram, P., Byerleee, D. & Harrington, L., ( 1990). Farming systems 
research revisited. In Eicher C. and Staatz J. Ed s. Agricultural development in the third 
world. The Johns Hopkins University press, Baltimore and London, 384- 99  
Valentine, I. & Kemp, P. D., (2007). Pasture and supplement resource. In Pasture and 
supplements for grazing animals , edited by P. V. Rattray, I. M. Brookes and A. M. 
Nicol.  New Zealand Society of animal production  Ocassional publication 14.  
van Eijk, T., (1998). Farming research and spirituality; an analysis of the foundations of 
professionalisim in developing sustainable farming systems: a t hesis presented in partial 
Chapter X: References 
213 
fulfilment of the requirements for Doctor of Philosophy at Wageningen Agricultural 
University, Wageningen, The Neatherlands.  
van Praag, C.M., ( 1999 ) . Some classic views on entrepreneurship. De Economist 147 ( 3): 311-
335.  
van Vuure n, A. M., & Pol - van Dasselaar , A.,  (2006). Grazing systems and feed 
supplementation. Fresh herbage for dairy cattle: the key to a sustainable food chain: 
85- 101.   
Vargas, G., (2001). Cambio estructural en el sector lechero Chileno: Potencial exportador y 
desafíos, [Structural Change in the Chilean dairy industry: export potential and 
challenges]. Ciencia e Investigacion Agraria 28 ( 3): 117 – 129.  
Vera, R. (1996). Country pasture: Chile. Forage resource profiles FAO.  Retrieved May 2012 
from: www.fao.org/ag/AGP/AGPC/doc/Counprof/Chile/cile.htm . 
Verissimo, A. & Woodford, K., (2005). Top performing farmers are information rich: case 
studies of sheep and cattle farmers in the South Island of New Zealand. Proceedings of 
the Fifteenth International Farm Management Association Congress, Campinas, Brazil, 
August 2005( 1) :  365- 368. 
Villano ,  R.,  Boshrabadi , M . H. , &  Fleming , E., (2010)  When Is Metafrontier Analysis 
Appropriate? An Example of Varietal Differences in Pistachio Production in Iran. 
Journal of Agricultural Science and Technology 12( 4): 379- 389. 
Waghorn, G. C., Burke, J. L. & Kolver, E. S., (2007). Principles of feeding value. In Pasture 
and supplements for grazing animals, edited by P. V. Rattray, I. M. Brookes and A. M. 
Nicol. New Zealand Society of animal produc tion, Ocassional publication 14. 
Walker, B., Hollin, C. S., Carpenter, S . R ., &  Kinzig, A ., (2004). Resilience, adaptability and 
transformability in social- ecological systems. Ecology and Society 9( 2) . 
Washburn, S. P., Silvia, W.J., Brown, C.H., McDaniel, B.T. & McAllister, A.J., (2002). Trends 
in reproductive performance in Southeas tern Holstein and Jersey DHI herds. Journal of 
dairy science 85( 1): 244 –251 . 
Watson, G. H. (1993). Strategic benchmarking: how to rate your company’s performance 
against the world’s best. New York: John Wiley and sons, Inc.  
Chapter X: References 
 
214 
Weersink A ., Turvey  C. G., & Godah, A., (1990). Decomposition Measures of Technical 
Effi ciency for Ontario Dairy Farms.  Canadian Journal of Agricultural Economics 
38( 3): 439- 456. 
White, S. L., Benson, G. A., Washburn, S. P., & Green, J. T., (2002). Milk Production and 
Economic Measures in Confinement or Pasture Systems Using Seasonally Calved 
Holstein and Jersey Cows. Journal of Dairy Science 85(1) : 95–104.  
Whitehead, P., (2008). Farm benchmarking in the UK: An assessment by the Food chain 
Centre, 2002-7. In Benchmarking in food and farming (p. 132). Surrey, England: 
Gower Publishing Limited.  
Wilson, K. & Morren, G. E., (1990). System approaches for improvement in agriculture and 
resource management. New York: Macmillan Publishing Company.  
Williams, T., (2006). 101 ingenious Kiwis : How New Zealanders changed the world. Reed 
Publishing Ltd.: New Zealand. 
Willock, J., Dearyb, I. J., McGregor, M. M., Sutherlanda, A., Edwards - Jones, G., Morgana, O., 
Dentc, B., Grieved, R., Gibsone, G. &  Austine, E., (1999). Farmers' attitudes, 
objectives, behaviors, and personality traits: The Edinburgh Study of Decision Making 
on Farms. Journal of Vocational Behavior  54(1): 5 - 36. 
Winkler, D . K ., ( 1999). Comparison of the Chilean and Ne w Zealand milk production costs. A 
thesis presented in partial fulfilment of the requirements for the degree of Masters in 
Applied Economics in Agribusiness, Massey University, Palmerston North, New 
Zealand .   
Wyss, U., Mauer, J., Frey, H., Reinhard, T., Bernet, A. & Hofste tter, P., (2011). Comparative 
milk production system of Hohenrain. Aspects of milk quality and seasonality of milk 
suppliers. AGRARFORSCHUNG SCHWEIZ 2 ( 9) : 412- 4 17. 
Zegers, C., & von Baer, H. (1978). The dairy industry in Chile [ Die Milchwirtschaft in Chile ]. 
Deutsche Molkerei-Zeitung 99( 25) : 862 - 863. 
Zwanenberg , A., (2001) . Consolidation in the Dairy Industry: Industry Note. Rababank 
International , Netherlands.  
 
Appendices 
215 
APPENDICES
Appendices 
216 
Appendix One: DairyBase variables and ratios   
( Source: DairyBase handbook)  
Table A-1  
List of Variables DairyBase Analyses 
DairyBase Variable status 
ACC ORIGINAL 
All Dairy Sheds ORIGINAL 
Animal Health ORIGINAL 
Asseturnover ORIGINAL 
BREED ORIGINAL 
Breeding & Herd Testing ORIGINAL 
Business profit before tax ORIGINAL 
CalfRearing(Excluding Labour) ORIGINAL 
Calving Season ORIGINAL 
Cash Operating Surplus ORIGINAL 
Cash surplus/deficit ORIGINAL 
Change in Dairy Livestock Value ORIGINAL 
Change in Non-dairy livestock Value ORIGINAL 
ChangeDairyLivestock ORIGINAL 
Close Liabilities.kg MS ORIGINAL 
Closing Dairy Investments ORIGINAL 
Closing.Equity ORIGINAL 
ClosingDairy Livestock value ORIGINAL 
ClosingDairyAssets ORIGINAL 
ClosingDairyLandBuildings ORIGINAL 
ClosingTotalAssets$ ORIGINAL 
Cows.ha ORIGINAL 
CowsMilked.FTE ORIGINAL 
Dairy Gross Farm Revenue ORIGINAL 
Dairy Operating Profit ORIGINAL 
DairyreturnAssets ORIGINAL 
DebttoAsset  % BM ORIGINAL 
Decision making unit (Farm ID+Season) ORIGINAL 
Discretiory cash ORIGINAL 
Drawings ORIGINAL 
EffDAIRYArea (Ha) ORIGINAL 
Effective Area : Non-Dairy (Ha) ORIGINAL 
Effective Area : Run-off Area (Ha) ORIGINAL 
EFs introduced funds ORIGINAL 
EFS: Dairy Operating Profit (EFS) ORIGINAL 
EqGrowthCapital ORIGINAL 
EqGrowthProfit ORIGINAL 
Equity Growth from Profit ORIGINAL 
Estimated $.kg MS ORIGINAL 
Extraordiry expenses ORIGINAL 
Farm ID ORIGINAL 
Farm System ORIGINAL 
Farm Working Expenses ORIGINAL 
Appendices 
217 
FarmDairyExpenses ORIGINAL 
Feed Inventory Adjustment ORIGINAL 
FeedInvAdj ORIGINAL 
Fertiliser ORIGINAL 
FEW ORIGINAL 
FEW.kg MS ORIGINAL 
FreightgeneralFWE ORIGINAL 
Fuel & Oil ORIGINAL 
GFR.ha ORIGINAL 
GFR/kg MS ORIGINAL 
Grazing and Run-off Expenses ORIGINAL 
GrowthEquity ORIGINAL 
GrowthEquity %  ORIGINAL 
IL CREATED 
Income Equalisation ORIGINAL 
Interest ORIGINAL 
Interest & Rent.GFR ORIGINAL 
Interest & Rent.kg MS ORIGINAL 
IRRIGATION ORIGINAL 
LabAdj-Management ORIGINAL 
LabAdj-Non-paid ORIGINAL 
Labour Adjustment ORIGINAL 
Labour Expenses (adjusted) ORIGINAL 
Liabilities.kg MS ORIGINAL 
Milk Fat composition ORIGINAL 
Milk fat/cow ORIGINAL 
Milk fat/ha ORIGINAL 
Milk litres/cow ORIGINAL 
Milk litres/ha ORIGINAL 
Milking interval ORIGINAL 
MS.cow ORIGINAL 
MS.FTE ORIGINAL 
MS.ha ORIGINAL 
MSComposition ORIGINAL 
Net Adjustments ORIGINAL 
Net Capital Expenditure ORIGINAL 
Net Cash Income ORIGINAL 
Net cost of leased runoff land ORIGINAL 
Net Debt ORIGINAL 
Net heifer/General grazing ORIGINAL 
Net Livestock (adjusted) ORIGINAL 
Net Milk Sales ORIGINAL 
Net Non-Dairy Cash Income ORIGINAL 
Net Off-farm income ORIGINAL 
Net revenue from dairy livestock ORIGINAL 
Net winter grazing ORIGINAL 
NetMilkSales ORIGINAL 
NetRDairyLivestock ORIGINAL 
Nitrogen ORIGINAL 
Non-Dairy (Ha) ORIGINAL 
Non-dairy operating profit ORIGINAL 
Appendices 
218 
Non-DairyCashIncome ORIGINAL 
Non-DairyOpEx ORIGINAL 
Non-dairyOpProfit ORIGINAL 
Number of replacement calves reared ORIGINAL 
Off Farm Income ORIGINAL 
Opening Dairy Investments ORIGINAL 
OpeningDairyLandBuildings ORIGINAL 
OpeningDairyLivestockValue ORIGINAL 
OpeningTotalAssets ORIGINAL 
OpEquity ORIGINAL 
OpEx.ha ORIGINAL 
OpEx.kg MS ORIGINAL 
OpProfit(EFS).ha ORIGINAL 
OpProfit(EFS).kg MS ORIGINAL 
OpProfitMargin ORIGINAL 
Overheads ORIGINAL 
Owned Run-off Adjustment ORIGINAL 
Owned Run-off Adjustment ORIGINAL 
Paid FTEs ORIGINAL 
PEAKCOWS ORIGINAL 
Protein Composition ORIGINAL 
Protein/cow ORIGINAL 
Protein/ha ORIGINAL 
RatesLandwater ORIGINAL 
Region ORIGINAL 
Rent ORIGINAL 
Repairsmaintenance ORIGINAL 
repairsmaintenanceLand&Buildings ORIGINAL 
ReturnEquity ORIGINAL 
Run-off (Ha) ORIGINAL 
Season ORIGINAL 
suppl.ha ORIGINAL 
Suppl.PurchasedMadeCropped ORIGINAL 
Supplementary Feed Expenses ORIGINAL 
Tax ORIGINAL 
Tax/ha ORIGINAL 
TotAdmEx ORIGINAL 
Total Area : Milking Platform (Ha) ORIGINAL 
Total Dairy Operating Expenses ORIGINAL 
Total Depreciation ORIGINAL 
Total fat for KPIs ORIGINAL 
Total FTEs  ORIGINAL 
Total Grazing & Run Off Expenses ORIGINAL 
Total milk for KPIs ORIGINAL 
Total Other Dairy Farm cash revenue ORIGINAL 
Total Other Working Expenses ORIGINAL 
Total Overheads ORIGINAL 
Total protein for KPIs ORIGINAL 
Total Stock Expenses ORIGINAL 
Total Supplement Expenses ORIGINAL 
TotalArea (Ha) ORIGINAL 
Appendices 
219 
TotalFixedAssetsClosing ORIGINAL 
TotalMSProduced ORIGINAL 
TotalrrigationEx ORIGINAL 
TotCostPastureRenovation ORIGINAL 
TotCurrentAssetsClosing ORIGINAL 
TotCurrentLiabilitiesClosing ORIGINAL 
TotDairyOpEx ORIGINAL 
TotDepreciation ORIGINAL 
TotFarmElectricity ORIGINAL 
TotFarmInsurance ORIGINAL 
TotFarmVehExexcludingfuel ORIGINAL 
TotFeedEx ORIGINAL 
TotFixedAssets-Opening ORIGINAL 
TotGraz&RunOffEx ORIGINAL 
TotLabExpenses ORIGINAL 
TotOpEx ORIGINAL 
 TOTOpPROFIT  ORIGINAL 
TotOtherDairyFarmcashR ORIGINAL 
TotOtherFWE ORIGINAL 
TotOverheads ORIGINAL 
TotreturnAssets  ORIGINAL 
TotreturnEquity  ORIGINAL 
TotStockExpenses ORIGINAL 
TotSupplExpenses ORIGINAL 
TotTerm LiabilitiesClosing ORIGINAL 
Ungrazeable area ORIGINAL 
UnpFamilyLabour ORIGINAL 
UnpFamilyManag ORIGINAL 
Wages (incl ACC, less subsidies) ORIGINAL 
Wages(inclACClesssubsidies) ORIGINAL 
Weed and Pest ORIGINAL 
Winter milk ORIGINAL 
Youngstock area ORIGINAL 
 
Production System: The type of production system as defined by the amount of imported feed 
and/or off farm dry cow grazing (irrespective of young stock grazing ) . 
Calving Season: The calving season for the herd. 1 Spring  only. 2 Autumn only. 3 Spring and 
Autumn. 4 Other, including year round and extended lactation.  
Milking Interval: Describes milking frequency: 1 TAD Twice a day milking for majority of 
milking season. 2 OAD Once a day milking for majority of milking season. 3 Mixed – OAD for 
between 16 and 30 weeks of milking season. 4 Other eg 3 times in 48 hours for majority of 
milking season.  
Winter Milk: Describes whether the herd is milked during the winter months of June and July.  
Region: DairyBase has 8 regions acr oss the country: 1 Northland. 2 Waikato. 3 Bay of Plenty. 4 
Taranaki. 5 Lower North Island. 6 West Coast - Tasman. 7 Marlborough- Canterbury. 8 Otago -
Southland (refer to the website www.dairybase.co.nz for a map showing regions ).  
Appendices 
220 
 % Milking Area Irrigated: De scribes the proportion of the Milking Area (C31) that is irrigated. 
1 No irrigation. 2 Some irrigation – but less than 30% of the Milking Area is irrigated. 3 More 
than 30% of the Milking Area is irrigated.  
Farm Dairy Type: The farm dairy type followed by the number of sets of cups. H = Herringbone 
R = Rotary. O = Other  
Predominant Dairy Breed: The main (predominant) breed of the herd. 1 Friesian. 2 Crossbred. 
(Under 70% of 1 specific breed). 3 Jersey. 4 Other (Ayrshire, Brown Swiss, Guernsey, Milking 
Short horn etc.)  
Peak Cows Milked: The highest number of cows milked at any time during the season.  
Stocking Rate (Cows/ha): The number of Peak Cows Milked divided by Milking area .  
Replacement Calves Reared: The total number of heifer calves reared as herd repl acements for 
the year. 
Non - replacement Calves Reared: The total number of beef (non- replacement) calves reared in 
both the spring and autumn ( includes dairy heifers raised for sale) . 
Full Time paid labour Equivalents: 1 FTE = 2,400 hours of work a year. Al l paid farm 
employees hours as a proportion of the working year (2,400 hours). Includes labour for calf 
rearing, relief milking and casual workers as well as paid farm managers. Excludes any specific 
contract work such as cultivation or fencing etc which i s attributed to another category. 
Full Time unpaid labour Equivalents: 1 FTE = 2,400 hours of work a year. All unpaid (usually 
family labour) farm labour hours as a proportion of theworking year (2,400 hours). Includes 
unpaid labour for calf rearing and relief milking. A maximum of 0.5 FTEs will be added to 
unpaid labour from unpaid management due to different treatments to calculator labour 
adjustment.  
FTE unpaid Management: 1 FTE = 2,400 hours of work a year. The Full Time Equivalent 
(FTE) of all unpaid (usually family) farm management. The number of FTEs displayed for 
unpaid management could be up to a maximum of 1.0. Management FTEs over 1.0 will be 
added to unpaid labour (C23) up to a maximum of 0.5 FTEs. Note: it is possible to have less 
than 1.0 FTE unpaid management. 
Total FTEs: Total paid and unpaid labour hours as a proportion of the working year (2,400 
hours).  
Total Dairy area: Total surveyed dairy farm area (freehold and leasehold) excluding run - off. 
Ungrazeable area: The area on the Total Farm t hat can not be grazed. This includes waste areas, 
waterways, races, fences, drains, buildings and forestry.  
Effective Dairy area: Total Dairy area less Ungrazeable area. Essentially this is the area 
available for grazing and/or cropping.  
Defined Young Stoc k area: The Effective area assigned to young stock over 3 months of age. 
(based on a 25% replacement rate). For young stock raised to 10 months; 4% of Effective area 
For young stock raised from 10 to 22 months, 11% of Effective area. For young stock raised  
from 3 to 22 months, 15% of Effective area  
Milking area: Effective area less Defined Young Stock area . It is the pasture and/or cropping 
area available for milking cows. This figure is used to calculate per ha KPI's.  
Dairy Run - off effective area: Effectiv e run- off area (freehold and leasehold) used to support the 
milking area. Note: this does not include the area used for non- dairy related production. 
Appendices 
221 
Non - dairy effective area: Any effective run - off area (freehold and leasehold) not used for the 
dairy cows within the business. Include: Effective area used for grazing non - dairying livestock, 
grazing dairy stock on a contract basis, forestry, horticulture, viticulture and cash cropping, 
Total – Total for the farm 
Per ha – Total divided by Milking area  
Per cow – Total divided by Peak Cows Milked  
Composition – Kg of milk solids (fat, protein, total MS) divided by Milk Litres. This shows the 
milk composition proportion per litre of milk supplied  
Milk litres: Total volume of milk supplied to dairy company plus milk fed to non - replacement 
calves. 
Fat Kg: Total milk fat supplied to dairy company plus milk fat fed to non- replacement calves. 
Protein Kg: Total milk protein supplied to dairy company plus milk protein fed to non-
replacement calves. 
Financial year -  Milk solids Kg: Total Fat Kg produced plus total Protein Kg produced Total 
milk solids supplied to dairy company plus milk solids fed to non- replacement calves. This is 
for the financial year e.g. May June. This figure is used to calculate per Kg Milk solid KPI ’s in 
the financial reports 
Non - replacement Calf Milk (litres): Milk (excluding colostrum) from the vat fed to non-
replacement calves. 
Non - replacement Calf Milk (Kg): Milk litres for Non - Replacement Calves converted to Kgs 
MS using the total Kg MS supplied to the dairy company for the season divided by total litres 
supplied to the dairy company for the season. 
Stocking Rate: Peak Cows Milked divided by Milking area.  
Kg Milk solids/ha: Milk solids Kg divided by Milking area . 
Kg Milk solids/cow: Milk solids Kg divided by Peak Cows Milked  
Cows/FTE: P eak Cows Milked divided by Total FTEs  
Kg MS/FTE: Total Milk solids Kg produced divided by Total FTEs. 
Gross Farm Revenue/ha: Dairy Gross Farm Revenue (GFR) -  cash and non cash from net milk, 
net livestock (adjust ed) and other dairy farm related revenue divided by milking area . 
Operating Expenses/ha: Total Dairy Operating Expenses (E29) -  cash and non cash-  divided by 
Milking area.  
Operating Profit (EFS)/ha: Dairy Gross Farm Revenue per ha (K06) less Total Dairy  Op erating 
Expenses per ha . 
Gross Farm Revenue/Kg MS: Dairy GFR divided by Milk solids Kg.  
Operating Expenses/Kg MS: Total Dairy Operating Expenses -  cash and non cash divided by 
Milk solids Kg . 
Operating Profit (EFS) /Kg MS: Dairy Gross Farm Revenue per Kg MS  less Total Dai ry 
Operating Expenses per Kg Ms.  
Appendices 
222 
FW E/Kg MS: Farm Working Expenses divided by Milk solids Kg.  
Operating Profit Margin %: Dairy Operating Profit as a percent age of Dairy Gross Farm 
Revenue . This measures how well a farm generates profit from its revenue. 
Asset Turnover %: Dairy Gross Farm Revenue as a percentage of Opening Dairy Assets. Note:  
Opening assets are Closing assets from the previous year . The rate of asset turnover measures 
how well a farm generates dairy revenue from its assets. 
Op erating Return on Dairy Assets %: Dairy Operating Profit plus owned run- off adjustment less 
rent as a percentage of Opening Dairy Assets. Note: Opening Dairy Assets are Closing Dairy 
Assets from the previous year . The RoDA measures the profit generated fr om the dairy assets 
employed 
Interest and Rent/Total Revenue: Interest and Rent (excluding run - off rent) paid as a percentage 
of Total Revenue: Total GFR + Net off - farm income. 
Interest and Rent/Kg MS: Interest and Rent (excluding run - off rent) paid  divided by Milk solids 
Kg . 
Total Return on Assets %: Total Operating Profit plus owned run- off adjustment less rent plus 
change in capital value divided by Opening Total Assets. Note: Opening Total Assets are 
Closing Total Assets from the previous year. The TRoA  is the profit generated by the assets 
employed plus capital gains and losses. It measures the overall financial performance of the 
business. 
Change in Capital Value: Closing Total Assets less Opening Total Assets less Net Capital 
Purchases less Value of C hange in Dairy Livestock. Note: Opening Total Assets are Closing 
Total Assets from the previous year  
Return on Equity % (excluding change in capital value): Total Operating Profit plus owned run-
off adjustment plus net off - farm income less rent less interest as a percentage of Opening 
Equity. Note: Opening Equity is Closing Equity from the previous year. The RoE measures the 
return on the funds of the owner but does not include the change in capital value. 
Total Return on Equity %: Operating Profit plus ow ned run- off adjustment plus net off  farm 
income less rent less interest plus change in capital value as a percentage of Opening Equity. 
Note: Opening Equity is Closing Equity from the previous year (Y - 1). The TRoE measures the 
return on thefunds of the owner including the change in capital value. 
Net Cash Income: Net Cash income from milk sales; net (sales- purchases) dairy livestock sales 
and other dairy farm related revenue. Note: this is net cash income from dairy operations only 
and does not include the value of the change in dairy livestock numbers.  
Farm Working Expenses: Total dairy farm cash expenditure, including labour, stock, feed, other 
working expenses and overheads.  
Cash Operating Surplus: Net Cash Income  less Farm Working Expenses. It is the cash available 
from dairying after paying for farm work ing expenses.  
Discretionary Cash: This is the cash available from dairy, non - dairy and off- farm operations to 
meet capital purchases, debt repayments, drawings, and extraordinary expenses (discretionary  
items). The calculation is Cash Operating Surplus less rent, interest and tax plus net non- dairy 
cash income, change in income equalisation and net off- farm income. 
Cash Surplus/Deficit: The cash surplus from dairy, non- dairy and off- farm operations over the 
year. It is measured by total discretionary cash plus introduced funds less net capital pur chases, 
net change in debt, drawings and extr aordinary expenses .  
Appendices 
223 
Closing Dairy Assets $: The value of all dairy related assets at close of the financial year. 1  
Closing land and buildings (includes all improvements used for dairying eg fences, races, dairy 
shed and the house value) valued at current market value by applying an adjustment based on 
Quotable Value sales to the latest Rateable Capital Value. Alternat ively an assessed market 
value can be entered for open and close. 2 plus plant, machinery and vehicles at book value, 3 
plus dairy livestock valued at NAMV, 4 plus any dairy related farm investments eg dairy 
company shares and shares in dairying related companies eg fertiliser, meat, animal breeding 
and trading companies) at market value. Note: Closing Dairy Assets does not include current 
assets eg bank, savings accounts, sundry debtors.  
Closing Total Assets $: The total value of all business assets at clo sen Closing Dairy Assets 2 
plus any non- dairy farm land and buildings 3 plus the book value of any plant, machinery and 
vehicles not included in Dairy Assets, 4 plus all non- dairy livestock valued at NAMV, 5 plus 
the market value of any off - farm assets and investments eg off- farm houses, commercial 
buildings, boats, aeroplanes Note: Closing Total Assets does not include current assets eg bank, 
savings accounts, sundry debtors.  
Closing Total Liabilities $: Closing long - term liabilities + (current liabilities  – current assets). 
Note: debt to family is included in liabilities if there is an expectation that it will be repaid. A 
long- term liability is one where the term for repayment is longer than 1 year. This also includes 
hire purchases for more than 1 year.  
Closing Total Equity $: This is sometimes referred to as Net Worth as it is the value of the 
owner’s share of the assets. Closing Assets–Closing Liabilities.  
Growth in Equity $: The change in owner’s equity during the year.Closing Equity – Opening 
Equity. Note: Opening Equity is Closing Equity from the previous year.  
Equity Growth from Profit $: Dairy operating profit plus the adjustment for non- paid family 
labour and management plus the adjustment for owned run- off, plus non dairy operating profit 
plus net off- farm income less extraordinary expenses, rent (excluding run - off), interest, 
drawings and tax .  
Growth from Capital $: Growth in Equity less Growth from Profit . It includes capital gain and 
introduced funds. 
Growth in Equity %: Closing Equity – Opening Equity as a percentage of Opening Equity. 
Note: Opening Equity is Closing Equity  from the previous year. 
Debt to Assets %: Closing Total Liabilities as a per centage of Closing Total Assets. This 
measures the proportion of the business value that is borrowed by the owners. Note: the Equity 
to Asset ratio is the inverse of the Debt to asset ratio and thus  the 2 measures sum to 100%.  
Liabilities per Kg MS: Opening: Total Liabilities divided by Milk solids Kg. Note: Opening 
Total Liabilities is Closing Total Liabilities from the previous year (Y - 1) . 
  
Appendices 
224 
Appendix Two: TodoagroBase variables and ratios  
 
Table A-2 
List of Variables TodoagroBase Analyses  
TodoagroBase Variable status 
Adm, freights, others US$ CREATED 
Administration, freights and others ORIGINAL 
Animal health CLP$ CREATED 
Animal health CLP$.L ORIGINAL 
Arriendo $.L ORIGINAL 
Assets CLP$ CREATED 
AssetsUSD CREATED 
AssetsUSD.ha CREATED 
ATR CREATED 
Cash I.ha US$ CREATED 
Cash income CLP$ CREATED 
Cash income CLP$.L ORIGINAL 
Cash Income.ha ORIGINAL 
Cash Op Surplus US$ CREATED 
Cash Op Surplus.ha US$ CREATED 
Cash operating surplus CREATED 
CL$/US$ CREATED 
Contribuciones CLP$.L ORIGINAL 
Contribution CLP$ CREATED 
Cow.ha CREATED 
Cows ORIGINAL 
Decision making unit (Farm ID+Year) CREATED 
Depreciation CREATED 
Depreciation $.L ORIGINAL 
Discretionary cash plus tax CREATED 
Effdairy area (ha) ORIGINAL 
Electricity CLP$ CREATED 
Electricity CLP$.L ORIGINAL 
Farm ID ORIGINAL 
FEW (PROD+ADM-DEPR) CREATED 
FEW US$ CREATED 
FEW.Cow CREATED 
FEW.cow US$ CREATED 
FEW.ha CREATED 
FEW.ha US$ CREATED 
Fuel & oil  CLP$.L ORIGINAL 
Fuel & oil CLP$ CREATED 
Gastosadmin $.L ORIGINAL 
Gastosfinancieros CREATED 
Appendices 
225 
GFR CLP$ ORIGINAL 
GFR CLP$.cow ORIGINAL 
GFR US$ CREATED 
GFR. US$ cow CREATED 
GFR.ha US$ CREATED 
Inseminación CREATED 
Insemination $.L ORIGINAL 
L.cow  ORIGINAL 
L.ha ORIGINAL 
L.milked cow ORIGINAL 
Livestock sold (heads) ORIGINAL 
MantencionCosts CLP$ CREATED 
Margen Neto CLP$ ORIGINAL 
Meat price CLP$/kg ORIGINAL 
Milk control CLP$ CREATED 
Milk control CLP$.L ORIGINAL 
Milk price (CLP$/L) ORIGINAL 
Milk price (US$/L) CREATED 
Milk:Feed prices ratio CREATED 
Milked cow/Total cow ORIGINAL 
Milked cows ORIGINAL 
Net milk sales CLP$ ORIGINAL 
Net MS US$ CREATED 
Net MS US$.ha CREATED 
Off farm forage ORIGINAL 
Operating Profit $ CREATED 
Operating Profit US$ CREATED 
OpEx (Total-Int-Rent) CREATED 
OpEX US$ CREATED 
OpEX US$.ha CREATED 
OpProf Margin CREATED 
OpProfit $.Cow CREATED 
OpProfit $.ha CREATED 
OpProfit.cow US$ CREATED 
OpProfit.ha US$ CREATED 
Pasture renov.&others CLP$ CREATED 
Pasture renov.&others CLP$.L ORIGINAL 
Prod+Aminsitration $ CREATED 
Production cost CLP$.L ORIGINAL 
Profitability % ORIGINAL 
Region ORIGINAL 
Rent CLP$ CREATED 
Resultado CLP$.L ORIGINAL 
ROA % CREATED 
SubTotCostprodGastosAdmin $.L ORIGINAL 
SubTotMantencion CLP$.L ORIGINAL 
Appendices 
226 
Supplement (g.L) ORIGINAL 
Supplement CLP$/kg ORIGINAL 
Supplement kg/cow ORIGINAL 
Supplements CLP$  CREATED 
Supplements CLP$.L ORIGINAL 
Total Cost $ CREATED 
Total cost CLP$.L ORIGINAL 
Total Interest CLP$ CREATED 
Total production cost $ CREATED 
TotProd US$ CREATED 
TotProd US$.L ORIGINAL 
UAxHa ORIGINAL 
Wages CLP$ CREATED 
Wages CLP$.L  ORIGINAL 
Year ORIGINAL 
 
Season: period between the 1st of January each year until the 31st of December.  
Dairy Area (ha): indicates the area under dairy, including run- off and crops.  
Milk production (L): milk produced annually by each farm.  
Cows: the average obtained between beginning and ending inventory of dairy cows (heads) per 
year.  
Annual production per cow (L/cow): milk production in the period in relation to the mass 
number of cows.  
Annual production per milked cow (L/milked cow): milk produced per cow milked.  
Milk equivalent (L): Milk plus the meat produced in the Dairy area “converted” to L of milk. 
Meat is obtained from the sum of income from meat (calves and cull cows sol d or transferred to 
fattening area, inventory difference, and new born calves). This is converted to ‘milk’ by 
dividing into the average price received per L of milk. ‘Litres’ resulting from this calculation are 
added to the milk produced on the farm. 
Milk  per hectare (L/ha): is equivalent milk production divided by dairy area.  
Stocking (AU/ha): is the number of cows (expressed in animal unit (AU)) for the Dairy area.  
Use of concentrate: variable measured by the consumption of concentrate per cow (kg 
concentrate/cow) and also by the grams of concentrate used per liter of milk produced (g of 
concentrate/L of milk).  
Production costs (CLP$/L) : includes wages, energy (electricity, fuels and lubricants), 
contributions and leases, maintenance (machinery, building s, parlor, fences, roads and other), 
grassland and fodder at (fertilizers, chemicals, seeds, forage conservation, among others), 
concentrates and forages external health and milk control, insemination, replacement and 
purchase of cows, depreciation and adm inistration and expenditure item general, which includes 
overhead, personal retreats, compensation manager, freight and sales commissions of animals;  
Direct Costs (CLP$/L): involves costs related to the management of the herd, such as animal 
health, input s to the parlour, insemination, own fodder production, fodder and concentrates 
bought, taxes, rent and labor costs, which involve direct wages and social security;  
Appendices 
227 
Indirect Costs (CLP$/L) : includes energy costs as defined above, indirect wages, water, 
telephone, repairs of buildings and machinery of the property, depreciation, overheads, freight 
and commissions sold animals, administration and personal retreats (two criteria were used for 
allocating indirect costs: a) sales allowance, which is to distribute indirect costs according to the 
relative importance of dairy item sales in relation to total sales of the company, b) animal unit 
allowance assigns the relative costs of the item dairy animal units, relative to the total units of 
the venture animals.  
Gross Farm Revenue (CLP$): product sales plus the value of milk consumed by calves plus 
inventory variation.  
Total Capital (CLP$): is the total value of investments in land, machinery, equipment, buildings, 
animals and working capital (only considers the investment in pastures).  
Net income: revenue less production costs and administrative expenses including bank  interest 
payments as the debt level of the producers is not related to production efficiency.  
Profitability (%): Different from return on assets. This indicator is obtained by dividing net 
income between total assets (land, machinery, equipment, building s, animals and operating 
capital)  
(Source: Lerdon, Baez & Azocar, 2008)  
 
 
 
  
Appendices 
228 
Appendix Three: R codes 
 
Data Envelopment Analysis (*requires lpSolve package to be installed). 
source("CRS.r")  
source("VRSInput.r")  
source("VRSOutput.r")  
source("Additive.r")  
Dat a = read.csv(" EXAMPLE.csv",sep="," ,header=T)  
Outputs= Data[,1]  
Inputs = Data[,2:6]  
EXAMPLECRS  = CRS(InputMat = Inputs, OutputMat = Outputs, ResultsFilename = 
" EXAMPLEResultsCRS", ResultsFileType="csv")  
EXAMPLE VRSI  = VRSInput(InputMat = Inputs, OutputMat  = Outputs, ResultsFilename = 
"EXAMPLEVRSI" , ResultsFileType="csv")  
EXAMPLE VRSO = VRSOutput(InputMat = Inputs, OutputMat = Outputs, ResultsFilename = 
" EXAMPLEVRSO" , ResultsFileType="csv")  
EXAMPLEAdd = Additive(InputMat = Inputs, OutputMat = Outputs, Results Filename = 
" EXAMPLEResultsAdd", ResultsFileType="csv")  
 
Recursive Partitioning Trees (requires the rpart package to be installed).  
Data = read.csv("EXAMPLE.csv",sep="," ,header=T)  
str(Data)  
library(rpart)  
Model = rpart(RESPONSE VARIABLE ~.,data=Data,control=c(minbucket=50))  
sink("RPartOutput.txt",append = FALSE)  
print(Model)  
summary(Model)  
sink()  
 
 
Appendices 
229 
Appendix Four: The NZ classification system, an iterative process  
This section describes the iterative process leading to the final classification system in NZ, 
which in turn, guided the selection of the Chilean classification system. A fter reviewing the 
literature it was evident that a combined approach was required. To split the sample according 
to year was the first straightforward decision. S ome particular, very responsive variables were 
certainly likely to show both intra and inter - annual variation. Examples of these are pasture 
growth and use of imported feed both highly responsive to rainfall which shows high variability 
in inter- annual comparisons. Issues around categorisation by DairyNZ farming systems were  
that previous empirical work has been done where differences between systems in efficiency 
terms were not recognized. The sample was split into three intensification levels (IL) , created by 
re- coding DairyNZ farming systems such that farming systems 1 and 2 composed LOW, 
farming system 3 represented MEDIUM and farming systems 4 and 5 composed HIGH.   
A preliminar analysis was conducted to evaluate the variability across regions. The regional 
analysis included the whole dataset, which comprises 2,614 observations across the period 2006 
to 2011. The dataset was  highly unbalanced in the size of sample per region as it is shown in 
Table 1 by the n value. The variables analysed were herd size, stocking rate, FTEs, MS/ha, 
MS/cow, Operating expenses and Operating profit per hectare. A  second preliminar analysis 
was then conducted to evaluate variability across seasons. The seasonal analysis included the 
whole dataset or pooled sample (n =2614 ) . Data was available for seasons 2006 / 07, 2007/ 08, 
2008/ 09, 2009/ 10 and 2010/ 11. A third preliminar analysis was conducted to evaluate whether 
or not the DairyNZ’s farming systems approach based on the use of imported feed was relevant. 
The analysis included the whole dataset, which comprises n observations across all regions and 
seasons (n = 2611 ) . More than 65% of the sample was composed of Farming systems 2 and 3 . 
Appendices 
230 
Table A-3: Descriptive Regional Analysis of Selected KPIs 
 
1 268 349.0 218.2 13.3 84.0 1850.0
2 864 384.4 280.7 9.5 55.0 3100.0
3 215 357.3 162.6 11.1 99.0 1030.0
4 292 295.2 138.1 8.1 93.0 1160.0
5 182 394.0 254.8 18.9 106.0 1800.0
6 226 392.7 247.5 16.5 114.0 1710.0
7 324 819.5 455.1 25.3 155.0 2966.0
8 243 527.2 264.8 17.0 126.0 1989.0
Total 2614 437.2 317.4 6.2 55.0 3100.0
1 268 2.5 .5 .0 1.3 4.1
2 864 3.0 .5 .0 1.5 5.3
3 215 3.0 .5 .0 1.7 5.3
4 292 2.9 .5 .0 1.8 4.8
5 182 2.9 .4 .0 1.9 4.5
6 226 2.5 .6 .0 1.2 3.8
7 324 3.5 .5 .0 2.2 6.3
8 243 2.8 .4 .0 1.6 4.2
Total 2614 2.9 .6 .0 1.2 6.3
1 268 2.7 1.4 .1 .6 9.7
2 864 2.6 1.5 .1 .4 13.2
3 215 2.5 1.1 .1 .7 7.7
4 292 2.3 1.0 .1 .7 6.0
5 182 2.7 1.5 .1 .8 9.8
6 226 3.0 1.6 .1 .9 12.4
7 324 5.0 2.5 .1 1.5 17.6
8 243 3.4 1.5 .1 .7 11.2
Total 2614 3.0 1.8 .0 .4 17.6
1 268 773.9 211.5 12.9 341.7 1631.0
2 864 1025.7 295.6 10.1 348.8 2888.8
3 215 1026.9 237.7 16.2 523.6 2195.7
4 292 981.8 239.8 14.0 7.7 1743.1
5 182 1006.3 287.1 21.3 275.8 2239.6
6 226 867.9 225.0 15.0 9.6 1634.6
7 324 1443.6 294.3 16.3 6.1 3028.2
8 243 1123.0 208.0 13.3 513.3 1760.0
Total 2614 1040.9 316.9 6.2 6.1 3028.2
1 268 309.5 50.9 3.1 183.0 451.8
2 864 334.4 59.1 2.0 178.1 649.5
3 215 344.1 41.8 2.9 200.3 472.2
4 292 338.0 49.1 2.9 2.8 563.9
5 182 349.0 70.2 5.2 118.2 603.4
6 226 348.7 54.2 3.6 4.2 507.8
7 324 408.7 54.4 3.0 2.0 539.5
8 243 398.4 52.1 3.3 222.2 550.9
Total 2614 350.4 62.9 1.2 2.0 649.5
1 268 3936.9 1258.8 76.9 1515.0 8852.1
2 864 4599.2 1631.3 55.5 1259.1 15354.4
3 215 4652.3 1498.2 102.2 1815.0 13854.3
4 292 4246.5 1174.7 68.7 1818.5 9321.7
5 182 4530.9 1536.4 113.9 1774.8 11379.5
6 226 4083.7 1451.7 96.6 1463.8 9873.4
7 324 6441.3 1647.8 91.5 3291.7 14871.1
8 243 4762.6 1408.0 90.3 1979.9 13650.7
Total 2614 4690.5 1655.3 32.4 1259.1 15354.4
1 268 1186.8 1270.1 77.6 -1463.2 5258.6
2 864 1904.4 1350.8 46.0 -2491.7 9794.0
3 215 2016.7 1276.1 87.0 -1096.5 6067.6
4 292 2017.5 1317.4 77.1 -1487.6 6725.8
5 182 1876.2 1406.5 104.3 -1295.0 6920.1
6 226 1398.2 1148.5 76.4 -1282.2 4598.6
7 324 2972.7 2167.6 120.4 -2345.8 8319.3
8 243 2287.7 1502.1 96.4 -1258.2 6912.6
Total 2614 1975.0 1536.9 30.1 -2491.7 9794.0
Operating 
Profit 
(EFS)/ha
Peak 
cows 
milked
Cows/ha
FTEs 
including 
Unpaid 
Labour
Milksolids/
ha 
Production 
year
Milksolids/
cow 
Production 
year
Operating 
Expenses/
ha
Descriptives
 
N Mean Std. Deviation Std. Error Minimum Maximum
Appendices 
231 
Beyond the differences in the absolute value of the variables presented in Table 1, the regional 
analysis of means using one way ANOVA and Schaffe tests, showed statistical differences 
between the majority  of the regions in every studied variable. Herd size presented significant 
differences in 16 out of the 28 pairs compared. Stocking rate and MS/ha was significantly 
different for 21 out of the same 28 pairs. FTEs and Operating profit/ha presented statistic al 
differences in 19 out of 28, while Operating Expenses/ha were significantly different in 18 out 
of 28 paired comparisons. Interestingly, Regions 2 and 3 performed statistically equal in all 
analysed variables, while region 3 and 5 statistically performed identical. Similarly, Region 3 
showed to have similarities with 4, which did not present significant differences for all variables 
except for OpEx/ha that was significant different and higher in Region 3 than in 4. Regions 1 
and 6 performed not significa ntly different in five out of seven variables. The variables that did 
show differences were related to MS production; both per hectare and cow, demonstrated to be 
statistically different and higher in Region 6.  In contrast, Regions 7 and 8 proved to be uni que 
performing different in all variables and paired comparisons.  
When efficiency scores were studied, Region 2 performed with the minimum efficiency and it 
showed to be different from all other regions. The higher efficiencies were shown by Regions 3, 
4, 7 and 8 and proved to have not significant differences. Region 1 was found to be different to 
all Regions except for number 5. Regions 1 and 5 behaved somewhere in between close to the 
averaged efficiency for all region and the whole period. Region 6 prov ed to be different from 
regions 1, 2, 3, 7 and 8.  
Appendices 
232 
Table A-4  
Descriptive Seasonal Analysis of Selected KPIs 
 
Unbalance is clearly reflected in the following: firstly the last season comprises only 217 
DMU’s while previous years’ numbers were between 497 and 628 and secondly, just 39 firms 
are present across all seasons. The latter in particular prevents from truly meaningful 
comparison between seasons. There are minimum absolute differences in herd size across the 
period reflected in few significant differences when pairs where tested. The 06/07 season 
presented differences when compared to season 08/09 and 09/10. Stocking rate was not 
affected, without significant differences across the period. Similarly, Labour did not vary much 
67 625.00 408.28 280.98 11.24 99.00 2800.00
78 628.00 424.51 308.83 12.32 60.00 2966.00
89 497.00 458.91 353.11 15.84 60.00 2894.00
910 567.00 460.60 317.99 13.35 55.00 2800.00
1011 297.00 443.51 338.11 19.62 87.00 3100.00
Total 2614.00 437.16 317.35 6.21 55.00 3100.00
67 625.00 2.90 0.56 0.02 1.20 5.32
78 628.00 2.94 0.57 0.02 1.30 5.27
89 497.00 2.96 0.58 0.03 1.25 5.22
910 567.00 2.96 0.58 0.02 1.43 6.25
1011 297.00 2.98 0.59 0.03 1.50 5.13
Total 2614.00 2.94 0.58 0.01 1.20 6.25
67 625.00 2.85 1.71 0.07 0.69 17.56
78 628.00 2.90 1.73 0.07 0.38 15.16
89 497.00 3.10 1.96 0.09 0.69 15.53
910 567.00 3.13 1.76 0.07 0.58 13.75
1011 297.00 3.01 1.75 0.10 0.77 13.24
Total 2614.00 2.99 1.78 0.03 0.38 17.56
67 625.00 1056.21 287.56 11.50 341.67 2385.71
78 628.00 999.76 317.04 12.65 275.84 2888.85
89 497.00 1052.12 306.56 13.75 375.00 2362.24
910 567.00 1031.89 339.91 14.27 6.10 3028.21
1011 297.00 1094.37 337.25 19.57 434.41 2494.45
Total 2614.00 1040.93 316.93 6.20 6.10 3028.21
67 625.00 361.26 54.49 2.18 181.67 591.55
78 628.00 336.58 63.22 2.52 118.22 649.50
89 497.00 353.17 57.93 2.60 190.21 593.21
910 567.00 344.51 71.32 2.99 2.01 603.38
1011 297.00 363.67 62.63 3.63 220.36 578.39
Total 2614.00 350.43 62.88 1.23 2.01 649.50
67 625.00 3754.47 1085.29 43.41 1259.08 8612.31
78 628.00 4866.05 1660.82 66.27 1699.92 14469.50
89 497.00 5119.70 1793.74 80.46 1604.97 15354.37
910 567.00 4801.93 1546.98 64.97 1514.97 14871.10
1011 297.00 5357.71 1788.52 103.78 2269.00 13910.59
Total 2614.00 4690.46 1655.28 32.38 1259.08 15354.37
67 625.00 1117.59 711.51 28.46 -1283.22 3459.02
78 628.00 3156.67 1523.31 60.79 -792.85 9793.97
89 497.00 781.06 1016.46 45.59 -2491.72 4173.01
910 567.00 2002.72 1209.85 50.81 -1237.95 6595.95
1011 297.00 3225.77 1363.97 79.15 -315.60 6725.77
Total 2614.00 1975.01 1536.94 30.06 -2491.72 9793.97
Operating 
Profit 
(EFS)/ha
Peak 
cows 
milked
Cows/ha
FTEs 
including 
Unpaid 
Labour
Milksolids/
ha 
Production 
year
Milksolids/
cow 
Production 
year
Operating 
Expenses/
ha
Descriptives
 
N Mean Std. Deviation Std. Error Minimum Maximum
Appendices 
233 
and showed just minor differences in absolute value only being significantly different when 
season 06/07 was matched to season 09/10.  
The comparison of means yielded four out of 10 pairs of seasons that behaved statistically 
different from each other when Milk solids, both per hectare and per cow, were analysed. 
Season 06/07 was significantly different from 07/08, which showed differences when compared 
to 08/09 and 10/11. Season 09/10 was also different from 10/11. Despite of these, differences 
did exist for 90% of the pairs when Operating Expenses and Operating profit as ratios were 
considered. Efficiency was also calculated on a seasonal basis using DEA. Results showed that 
mean efficiency for dairy farms across all New Zealand were statistically not different in season 
06/07, 08/ 09 and 10/11. In addition, season 07/08 was not significantly different from 09/10. In 
contrast, mean efficiency scores for season 06/07 proved to be different from seasons 07/08 and 
09/10. Similarly, season 07/08 also demonstrate differences with season 08/09 and 10/11, which 
were both different from seasons 09/10.  
Table A-5  
Descriptive Farming System Analysis of Selected KPIs 
 
1 270 279.85 138.65 8.44 55.00 980.00
2 803 354.50 202.96 7.16 93.00 1620.00
3 921 481.26 347.01 11.43 87.00 3100.00
4 476 561.88 393.79 18.05 76.00 2966.00
5 141 499.54 363.61 30.62 128.00 2800.00
Total 2611 437.13 317.48 6.21 55.00 3100.00
1 270 2.56 .48 .03 1.25 4.45
2 803 2.76 .49 .02 1.20 4.40
3 921 2.99 .52 .02 1.47 5.16
4 476 3.21 .53 .02 1.59 5.27
5 141 3.56 .72 .06 2.21 6.25
Total 2611 2.94 .58 .01 1.20 6.25
1 270 813.37 217.86 13.26 7.73 1764.51
2 803 919.89 213.06 7.52 352.33 1626.00
3 921 1053.88 280.89 9.26 9.57 2494.45
4 476 1225.71 294.89 13.52 411.47 2361.89
5 141 1453.82 485.21 40.86 6.10 3028.21
Total 2611 1040.72 316.96 6.20 6.10 3028.21
1 270 317.10 56.20 3.42 2.83 469.80
2 803 334.05 53.07 1.87 182.86 525.00
3 921 350.54 57.47 1.89 4.23 550.68
4 476 380.69 58.11 2.66 205.74 550.87
5 141 403.69 93.45 7.87 2.01 649.50
Total 2611 350.38 62.90 1.23 2.01 649.50
1 270 3523.66 1032.91 62.86 1259.08 7585.96
2 803 3970.78 1043.36 36.82 1463.83 9886.88
3 921 4787.19 1378.57 45.43 1774.83 13910.59
4 476 5659.44 1627.04 74.58 1979.87 13854.28
5 141 7099.48 2560.50 215.63 2256.47 15354.37
Total 2611 4689.33 1654.10 32.37 1259.08 15354.37
1 270 1632.56 1230.95 74.91 -1237.95 5593.53
2 803 1753.87 1262.42 44.55 -1463.19 6721.47
3 921 1995.23 1533.73 50.54 -2007.98 7541.46
4 476 2336.89 1781.27 81.64 -2345.84 8319.31
5 141 2525.90 2129.26 179.32 -2491.72 9793.97
Total 2611 1974.44 1537.30 30.09 -2491.72 9793.97
Descriptives
 
N Mean Std. Deviation Std. Error Minimum Maximum
Operating 
Profit 
(EFS)/ha
Peak 
cows 
milked
Cows/ha
Milksolids/
ha 
Production 
year
Milksolids/
cow 
Production 
year
Operating 
Expenses/
ha
Appendices 
234 
Herd size’s means showed not significant differences when FS 5 was compared to 3 and 4 
although statistical differences between 3 and 4 did exist. Operating Profit/ha proved to be 
significantly not different for FS 1 and 2; equal ly, FS 4 proved to behave similarly than FS 5.  
When efficiency scores were studied, FS 5 presented the highest averaged efficiency score and 
it was found to be different to all FS except for number 4. FS 4 performed with the second best 
mean efficiency and it showed to be different from FS 2 and 3 which showed the lowest 
averaged efficiency scores. Interestingly, FS 1 showed not significant differences with FS 4 
despite of being somehow contrasting farming systems.  
These results are essentially different to those found by Dake, Shadbolt and Dooley (2011 
unpublished) who analysed all five farming systems in the Waikato and the Canterbury regions 
for the 2010/ 11 season. They found all systems performed similarly in terms of efficiency 
scores, estimated using DEA under VRS. Differences can be explained by sample size, regional 
distribution, the use of a pooled sample including five seasons instead of one and the inclusion 
of different inputs.  
A combination of region (1 to 8) and intensification level (LOW, MED IUM , HIGH) based on 
the DairyNZ’s farming systems’ classification was evaluated. This yielded 2614 observations 
classified in 2 7 groups. The variables analysed were herd size, stocking rate, FTEs, MS/ha, 
MS/cow, Operating expenses, Operating profit per hec tare, Operating profit margin. There is a 
clear trend relating to level of intensification (LI) and herd size as shown in Graph 1.  
 
Figure A-1: Herd size by region and intensification level  
LI tends to increase with herd si ze or the other way round, at increased herd size, the use of 
imported feed increase. Region 1 shows the exception to this rule among all other regions 
explained by the existence of outliers. The variable stocking rate measured in cows/ha 
consistently responds to intensification level in all regions as presented in Graph 2.  
Appendices 
235 
 
Figure A-2: Stocking rate by region and intensification level 
The lowest values are exhibited by regions 1 and 6 while region 7 shows the highest mean and 
the maximum value for a particular observation: 6.25 cows/ha at 7- H IGH.  There is an 
observable trend between input labour and intensification level as shown in Graph 3.  
 
Figure A-3: FTEs per hectare by region and intensification level 
However, Regions 1 and 4 escape to the rule showing maximum labour input at MEDIUM 
levels of intensification. The highest, both mean and individual, observations are found at 7-
H IGH. This is consistent with the high correlation observed in t he same dataset, between 
stocking rate and FTEs. Production measured in MS/ha consistently responds to intensification 
level in all regions as presented in Graph 4. The response shows a very similar shape than the 
variable stocking rate, consistent with th e high correlation between these two variables.  
Appendices 
236 
 
Figure A-4: Milk solids per hectare by region and intensification level. 
Production measured in MS/cow also responds to intensification level in all regions but some 
differences are not significant. Regions 3,5, 6 and 7 show very similar means when levels LOW 
and MEDIUM are compared.  
 
Figure A-5: Milk solis per cow by region and intensification level. 
Operating Expenses per hectare consistently r esponds to intensification level in all regions as 
revealed by Graph 5.  
Appendices 
237 
 
Figure A-6: Operating expenses per hectare by region and intensification level. 
The lowest mean is found in region 1 closely followed by mean OpEx/ha in region 4, 5, and 6 at 
the LOW intensification level. Region 2 shows the minimum and maximum individual 
observations at the LOW and HIGH level respectively, and thus the greatest stand ard deviation 
as a region.  
There is an observable trend between Operating profit per hectare and IL as shown in Graph 6, 
where the OpP/ha increases with intensification.  
 
Figure A-7: Operating profit per hectare by region and intensification level. 
However, regions 3, 5, 6 and 7 present mixed results as shown in Graph. In region 3 the highest 
likelihood to maximise the result is found at the LOW IL. In regions 6 and 7 maximum values 
appear associated to MEDIUM IL.  
Appendices 
238 
There is another observable but opposed trend between Operating profit margin and IL as 
shown in Graph 7.  
 
Figure A-8: Operating profit margin (%) by region and intensification level.  
Differences in operating profit margin are not significant between IL in region 1. Otherwise, the 
consistent trend is to observe lower margin at the higher intensification levels. The highest mean 
is found in region 3- LOW and the lowest is found at 6- H IG H.  JUSTIFICATION FOR USIN G 
MILK SOL IDS AS AN OUTPUT There is little divergence in the use of outputs in the DEA 
literature. Candidate outputs are usually GFR, Dairy and non- dairy cash income, Milk solids or 
financial performance measures. This research proposes the use of physical output in the form 
of MS/ha to keep the milk price effect out. To be consistent, whenever possible, inputs such as 
labour are also expressed in physical units.  
 
  
Appendices 
239 
Appendix Five: IFCN results 
 
Table A-6   
IFCN Approach New Zealand Versus Chile 
 
Appendices 
240 
Appendix Six: 5th Chilean tree including assets (input variable) 
 
 
 
 
 
 
 
  
Appendices 
241 
Appendix Seven: 5th NZ tree using ROA as response variable 
 
 
Appendices 
242 
Appendix Eight: Frequency of consistently efficient farms  
 
Table A-7  
Frequency of Consistently Efficient Farms Across Years, Region, Farming systems, Irrigation and 
Size  
 
YEAR TOTAL RESILIENT NUMBER RESILIENT FREQUENCY YEAR TOTAL RESILIENT NUMBER RESILIENT FREQUENCY
2008/09 496 102 0.21 2007 245 68 0.28
2009/10 564 124 0.22 2008 231 68 0.29
2007/08 614 118 0.19 2009 231 55 0.24
2006/07 618 113 0.18 2010 197 51 0.26
2010/11 295 93 0.32 2011 124 35 0.28
REGION TOTAL RESILIENT NUMBER RESILIENT FREQUENCY REGION TOTAL RESILIENT NUMBER RESILIENT FREQUENCY
Bay of Plenty 214 40 0.19 Llano Central Sur 379 99 0.26
Lower North Island 178 43 0.24 Ribera de los Lagos 299 78 0.26
Marl-Canterbury 322 77 0.24 Litoral Sur 73 16 0.22
Northland 268 49 0.18 IX Sur and XIV Norte 215 53 0.25
Otago-Southland 240 70 0.29 Los Angeles-Angol 59 28 0.47
Taranaki 289 43 0.15
Waikato 853 172 0.20 COWS TOTAL RESILIENT NUMBER RESILIENT FREQUENCY
West Coast-Tasman 224 55 0.25 < 300 673 187 0.28
300-600 299 59 0.20
FARMING SYSTEM TOTAL RESILIENT NUMBER RESILIENT FREQUENCY > 600 56 30 0.54
1 268 52 0.19
2 796 152 0.19 SUPPLEMENT kg/Cow TOTAL RESILIENT NUMBER RESILIENT FREQUENCY
3 910 191 0.21 < 1300 555 146 0.26
4 474 109 0.23 > 1300 473 130 0.27
5 140 47 0.34
IRRIGATION TOTAL RESILIENT NUMBER RESILIENT FREQUENCY
< 30% 58 10 0.17
>30% 446 112 0.25
NOT IRRIGATED 2084 429 0.21
PEAKCOWS TOTAL RESILIENT NUMBER RESILIENT FREQUENCY
< 300 1040 250 0.24
300-600 1030 143 0.14
> 600 518 158 0.31
DAIRYBASE TODOAGROBASE