Quantity aggregation and quality price adjustment : problems with measuring Indonesian food demand : a thesis presented in partial fulfilment of the requirements for the degree of Masters in Agricultural Economics at Massey University
One Asian country which is predicted to provide excellent opportunities for New Zealand exporters is Indonesia. To measure Indonesian demand responses requires the use of cross-sectional survey data. However, the use of such data produces aggregation errors, caused by quality effects, in the measurement of commodity price and quantity elasticities. Such aggregation errors produce biased elasticity estimates. The bias in the estimation of price elasticities is compounded by errors of measurement in measured quantities and expenditures. Adapting the theoretical model of Houthakker (1952) and Theil (1952), Deaton (1988) developed a methodology which accounts for both aggregation and measurement error. Using Deaton's methodology, the demand for five commodities - rice, meat, fruit, vegetables, and milk products - were computed. The expenditure elasticities are ordered much as would be expected, with rice and vegetables close to zero, and meat, fruit and milk products all having elasticities greater than one. Comparing the expenditure elasticities with respect to total expenditure with those of previous studies, the quantity elasticities are decreasing with time as expected. Thus while meat, fruit, and milk products are still considered luxury items, they will increasingly be within the average Indonesian consumer's reach in the near future. Also, the quality of foodstuffs consumed is increasing with incomes. The effects of quality, and measurement error especially, produced dubious price elasticity estimates. After adjusting for these influences, the price elasticity estimates are, with the exception of milk products, negative, although the rice price elasticity is larger than predicted. Meanwhile, the effects of quality in the estimation of quantity elasticities is relatively minor. While the model produced satisfactory results, it was considered that further exploration of the methodology was required, particularly with regard to the use of food expenditure. In the absence of data on total expenditure, Deaton (1988) assumed that food expenditure was a theoretically acceptable alternative explanatory variable. To measure if this was so, elasticities using both food and total expenditure were calculated. Although no formal non-nested tests were used, differences in the price elasticities between the two models casts doubt on the use of food expenditure in place of total expenditure. Nevertheless, estimation of a 'food share' elasticity provides a method for moving from food expenditure elasticities to total expenditure elasticities, with a proxy value for this elasticity providing encouraging results. Yet the estimation of such an elasticity requires information on total expenditure, providing limited empirical value for the researcher with just food expenditure data, and only a theoretical curiosity for the researcher with information on both. Other methodological problems included a larger than expected rice price elasticity and the estimation of biased OLS parameter estimates through the use of zero expenditures in the model. Consequently, without further exploration of these issues, the procedure should only be applied cautiously.