In: Economics
Product differentiation is pervasive in markets. It is at the heart of structural empiricism and it smoothes jagged behavior that causes paradoxical outcomes in several theoretical models. Firms differentiate their products to avoid ruinous price competition. Representative consumer, discrete choice, and location models are not necessarily inconsistent, but performance depends crucially on the degree of localization of competition. With (symmetric) global competition, rents are typically small and market variety near optimal. With local competition, profits may be protected because entrants must find profitable niches. These rents lead firms to competitively dissipative them, and performance may be poor.
There are three basic families of product differentiation models that are typically used for modeling equilibrium with free entry and comparing optimal to equilibrium diversity.
Representative consumer models start by positing a utility function intended to portray aggregate preferences. This preference ordering generates the demand system for differentiated products and it measures welfare for the optimality analysis. Such functions typically embody global competition insofar as demands for varieties of the differentiated product are symmetric substitutes. Models in this class include the often-used CES preference formulation and the quadratic utility that gives rise to a linear demand system. These are parameterized utility functional forms that embody taste for variety in that more variety raises welfare even when total consumption is fixed.
The discrete choice approach is founded in econometric and probabilistic models of consumer behavior. Each individual has an idiosyncratic taste (or "match value") for each product. Aggregating individual choices yields the demand function and aggregating the surpluses yields the welfare function. Any i.i.d. tastes yield global competition in that products are symmetric substitutes (e.g., the logit model).
Discrete choice models are not constrained to symmetric substitutability among variants. Models such as the nested logit embody closer substitutability between products within the same nest and the general probit model embodies quite elaborate substitutability patterns through the variance-covariance matrix of the match terms. These models are commonly used in the new structural empirical industrial organization literature.
Location models explicitly describe product specifications and consumer preferences as addresses and assume that consumers dislike distance "traveled" between ideal type and product. Location models may also be viewed as discrete choice models because individuals make discrete choices and have idiosyncratic match values. There is a difference in interpretation: location models typically assume the population of consumers to be given and deterministic, while discrete choice models suppose that an individual's taste is a realization from a probability distribution.
In models such as the circle model, the emphasis is on the number of products produced in equilibrium and exogenous symmetric locations are effectively imposed: however, the standard symmetric location pattern can be proved to be a location equilibrium under some circumstances.
One major benefit of discrete choice and location models is that the explicit micro foundations indicate how to introduce some economic phenomenon of interest. For example, network externalities may be incorporated into consumer utility and a consistent set of demands is then generated. Representative consumer models are less satisfactory since they do not start with a population of differentiated individuals.
The different approaches are not necessarily inconsistent with or substitutes for each other. Rather, they may frequently be twinned and one approach may be reinterpreted within the setting of the others. The CES model is a variant of the logit model, and a representative consumer does exist for the circle model and for probabilistic discrete choice models. Indeed, although global competition is typically generated from models such as the CES representative consumer or models of discrete choice with i.i.d. errors, it can also be derived from a spatial model if there are sufficiently many dimensions (so that each good can be a “neighbor” to each other one).
These models are also useful for comparative static analysis of changing patterns in industries in response to structural changes in cost structures, population growth, transportation costs, and consumer tastes. These descriptions are useful for Urban Economics, Industrial Organization, International Trade, and Economic Geography.