Discussion papers

CPM-95-05 - 8 February 1995

A Formal Preference-State Model with Qualitative Market Judgments

Scott Moss and Bruce Edmonds

An updated version published as: Moss, S. and Edmonds, B. (1997): A Knowledge-based Model of Context-Dependent Attribute Preferences for Fast Moving Consumer Goods, Omega, 25(2), 155-169.

Formal models of market demands derived from the economic theory of choice have four deficiencies for the modelling of some FMCG markets. For many FMCGs customer preferences depend on the purposes for which purchases are made and each individual can have different purposes in mind when purchasing at different times. A theory of household demand (with one preference function for each agent) is therefore an inappropriate construct for the analysis of these markets. The theory has it that, income effects aside, cross-price elasticities of demand are symmetric as between pairs of products. In practice, asymmetries are present and important. Although brand data in these markets is typically much better than customer data, the theory implies a great deal about the customers and these implications dominate the predictions and forecasts generated by the models predicated on the theory. Marketing experts typically have a qualitative understanding of important aspects of their markets and these cannot be represented by means of utility functions or partial orderings based thereupon. In this paper, we set out an alternative modelling procedure for markets in which demand asymmetries are possible, qualitative aspects of demand can dominate, preference-states determine the preferences of individuals and relatively poor customer data is not a bar to modelling demand when there is good brand and product data. We demonstrate that models entailing these conditions can be used to explain observed qualitative characteristics of products, markets and sources of demands as well as good, numerical data such as EPOS data. Indeed, we will argue that models that rely on qualitative judgments and closely track actual numerical data series are more useful than conventional models which rely on numerical data to track numerical data.

Downloads