8 - Causation and components in market share–performance models: the role of identities

Summary

The marketing literature contains several structural models, many of them based on the PIMS database, in which one variable is a definitional component of another, related to it through an identity. These definitional relationships have the potential for providing important insights into marketing phenomena, if they are appropriately modeled. On the other hand, they result in inconsistent parameter estimates if they are not separated from other, non-definitional, relationships in the model that need to be empirically estimated. This chapter first discusses the substantive information that can be obtained by studying the definitional components of a composite variable instead of the variable alone. Then, it examines each of the ways in which definitional relationships appear in marketing models, identifies those that are misspecified, analyzes the impact of the misspecification, and then provides the correct specification. It also disproves a commonly held belief that using instrumental variable estimation in a simultaneous equation system resolves the problems caused by mixing definitional relationships with structural ones. Thus, it provides a comprehensive view both of the potential benefits and of the pitfalls of definitional relationships in structural models. Much of the work reviewed in this chapter was inspired and enabled by the PIMS research database, which provides data not only on profitability, but also on each of its cost and revenue components for a variety of strategic business units over multiple years.