This thesis evaluates some commonly used socio-economic classification systems. Some of the systems evaluated have been used for many years in the market research industry in New Zealand whilst others are recent additions or are more commonly used in the United Kingdom.
The main objective of this study was to test the ability of the systems to predict purchasing levels of consumer products and services. The second objective was to evaluate how well the various systems predict brand choice.
A sample of 1596 respondents was provided by AGB McNair from their media survey database. Multiple regression was used to predict the level of usage of each product, with the adjusted R valua of the equation as the measure of the power of the classification system. Nominal variables, such as brand last used, were crosstabulated against the classification categories, and Lambdas calculated. A further measure of the ability of the classification systems to predict brand choice was obtained by performing discriminant analysis, which generated classification tables. The percentage of cases correctly classified provided a further measure of performance.
The various classification systems were not very good at predicting purchasing behaviour. The better systems accounyted for about 2% or 3% of the variation in quantities purchased. The various classification systems were also not very good at predicting brand choice. Even though the various classification systems explained little of the variation in quantities purchased and brand choices, they are still very useful. The socio-economic classification systems can be used as a starting poing from which better preditors of purchasing behaviour can be developed.