Home | Business News | Browse by Publication | M | Marketing Science

Removing the boundary between structural and reduced-form models.

Article Excerpt
The authors provide a comprehensive and useful review of structural models in marketing (Chintagunta et al. 2006). They evaluate the strengths and limitations of a structural approach and present examples of recent advancements in the development and application of structural models. Their discussion on the directions for future studies is very insightful. Despite the growing trend in the literature to claim contribution by offering structural treatments to research questions, it remains arbitrary and sometimes mysterious as to what constitutes a "structural" model. I agree with the authors that artificial boundaries are commonly imposed in our field between structural and reduced-form models. I believe it is more constructive to classify research studies along a structural/reduced-form continuum than to draw a distinctive line between, or even develop different research paradigms around, them.

My objective in this discussion is to provide further insights into the structural/reduced-form continuum. I first argue that this continuum is usually multidimensional, and one can move up the continuum in two directions. Following the argument are some examples of how the studies in the literature can be positioned along the continuum. I also discuss the role of data in developing more structural models. I argue that limited data quality can be a "double-edged sword" for the use of more structural models and that data integration and researcher creativity can be substitutes for limited data quality. Directions for future research on moving up the structural/reduced-form continuum are also identified. The discussion intends to support the complementarity between "structural" and "reduced-form" models, and the need to break the boundary between them.

1. The Structural/Reduced-Form Continuum

Empirical models are often compared in their efficiency to extract more information from data (Shugan 2002). The data of interest to researchers in marketing are often generated from some latent decision process. (1) The information extracted is usually used for policy evaluation. Policy-oriented models can be classified along a structural/reduced-form continuum. At one extreme are models that completely ignore the underlying decision process, but focus on the statistical relationships between variables (e.g., marketing-mix inputs and sales outputs). (2) There are two directions along which we can depart from this extreme and move up the structural/reduced-form continuum.

The first one lies in explicitly investigating the implicit assumptions, regarding the scope of the system of decision processes, that are necessary for statistical reduced-form models to yield consistent estimates. A noteworthy point, maybe the most important one in marketing that is essentially devoted to understanding the interactions among economic agents, is the recognition of the interactive roles of different decision makers in a market (e.g., consumers and firms). For instance, if we permit marketing mix (e.g., price) to...

View this article FREE - Now for a Limited Time, try Goliath Business News
Free for 3 Days!