IN PRODUCT MARKETS

WHILE THE OVERALL IMPORTANCE of drawing a distinction between information concerning characteristics of goods which are directly observable and information concerning characteristics of goods which cannot be observed without actual "consumption" in some appropriate sense has just begun to be realized, labor economists have implicitly been making use of the distinction for some time. For example, researchers studying labor mobility in the early fifties such as Reynolds [9] emphasized that high job turnover by young workers could be explained in part as a kind of job-shopping. More recently, Pencavel [7] has used the notion that all characteristics of jobs cannot be observed by simply sampling firms as the basis for an equilibrium model of job quits, and Wilde [10] has explored optimal search strategies when jobs are characterized by multiple characteristics, some of which are directly observable and others of which cannot be observed without actually taking the job. Besides these applications to the quit-rate, the distinction between different kinds of information can be used to analyze a number of economic phenomenon, both in the labor market and in other markets characterized by imperfect and costly information. Stiglitz [8] has applied it to the economic theory of screening and education, and Nelson [6] has used it in a unique approach to consumer behavior in product markets. Nelson partitions goods into two classes: search goods and experience goods. The distinction between the two classes is drawn in terms of the consumer's preferred mode of evaluating the potential stream of utility yielded by purchasing a good. For search goods the evaluation occurs prior to purchase; that is, a decision to purchase the good is based only on price (and other directly observable characteristics), not on any unobservable qualitative features of the product. Experience goods are those for which the primary information process used to evaluate the potential utility of a purchase is actual consumption of the good. This paper is motivated in part by Nelson's work but, unlike it, the focus is directly on individual characteristics as opposed to total packages of characteristics. That is, two classes of characteristics are defined. Search characteristics are those which can be observed without actually consuming the good, and experience characteristics are those which cannot be observed without consumption. Investment in information about search characteristics will be called inspection, as suggested by Hirshleifer [1], and investment in information about experience characteristics will be called evaluation. To see the difference between this approach and Nelson's, suppose a good is described by a vector of characteristics (x1, x2, .. .,xn). Nelson implicitly assumes