Costly buyer search in a differentiated products model: an experimental study.

• Author: Mago, Shakun Datta
• Position: Report

1. Introduction

   Does improved product information always cause a positive externality for the buyers in the form of lower prices? The answer, surprisingly enough, is no. Stiglitz (1979, p. 343) states "Improved information may affect the elasticity of demand facing the firm; if it increases the degree of monopoly power of each firm, it may lead to higher prices and lower welfare." In this article we use laboratory markets to investigate whether the content of information and the timing of its dissemination affect the sellers' market power. Our main goal is to investigate the impact of improved buyer information on sellers' pricing strategies. More precisely, we seek to attain empirical support for the prediction that improved product information may exert a negative externality on the buyers in the form of higher prices.

   We examine price competition in a horizontally differentiated duopoly where buyers can engage in costly search. In the basic structure, all buyers are initially unaware of the prices or their valuations (hereafter, referred to as match value) for products sold by the two sellers and must engage in costly search to obtain this information. Theory predicts that improved buyer information can have differing effects on sellers' market power, depending on its content--price or product attributes--and its timing--presearch and postsearch. Sellers are able to charge a higher price when the buyers are only informed about their match values (either pre- or postsearch) as compared with the case where buyers have no information or when they have information about prices alone. Furthermore, sellers are able to charge a higher price if information about the match values is provided to the buyer before she begins the search process, thereby signifying the influence of the "timing" of information dissemination. The intuition for these results is rather simple: A buyer who receives information about her match values from both sellers before she begins the search process will reveal this information by visiting the seller from whom she has a better match value (because she expects both sellers to set the same price). She will forego further search if price at the preferred seller exceeds the expected price by less than the search cost. This search behavior renders the demand less elastic and results in higher prices paid by all buyers. If, however, information about the match value of the second seller is provided after visiting a seller, then the seller--whom the buyer has randomly chosen to search first--can no longer deduce that he offers the buyer a higher match and, therefore, cannot exploit his advantageous position. This intensifies price competition because the seller will attempt to retain the buyer by offering her a better "price-match value" combination. Similar reasoning entails that price competition is even more intense if the additional information available upon visiting one seller pertains to the price of the second seller.

   The result that improved information may impart a negative externality is not new to the literature. Several articles have demonstrated that the value of information can be negative in multiagent settings. (1) Our results are in the same vein. However, our study contributes to the literature in two important ways: First, we provide a theoretical extension of Anderson and Renault's (2000) model, which is itself based on Wolinsky (1986). Wolinsky was the first to introduce costly buyer search to a differentiated goods market. He examined a market setting where the buyers are aware of the availability of all the brands in the market but do not know either the prices or their match value from them. He concluded that costly search limits the "effective" substitutability among brands and gives the seller more market power, which translates into supracompetitive prices. Anderson and Renault (2000) modify Wolinsky's basic model of product differentiation to allow for heterogeneity in buyer information and examine how improved product information can be a source of negative externality for the buyers. In this study, we extend their analysis by varying the informational conditions along two additional dimensions and analyzing its impact on seller pricing behavior. Our model yields a rich set of comparative statics predictions.

   Second, to the best of our knowledge, this is the first laboratory study of costly buyer search in a differentiated goods market. Prior literature on buyer search (Grether, Schwartz, and Wilde 1988; Abrams, Sefton, and Yavas 2000; Davis and Holt 1996; Cason and Friedman 2003; Cason and Datta 2006; Cason and Mago 2008) has focused on homogeneous goods environment where paradoxically enough, the buyers should not search in equilibrium. Introducing product heterogeneity is important because it provides "search model with search." The differentiated taste specification gives buyers a reason to search for a good that is a better match for their preferences, even when all sellers set the same price in equilibrium. Because buyer search is ubiquitous and truly homogenous goods as assumed in the theory are not, studying price competition and buyer search in a differentiated goods environment seems pertinent. (2)

   Our results also relate to the recent literature on advertising. Although we do not examine the mechanism by which the buyers receive information, our study can provide important insight into the role of advertising. For instance, in a recent article, Anderson and Renault (2006) evaluate a monopolist seller's advertising decision under various information disclosure policies. They find that it is not necessary to impose a full disclosure policy for prices and that the seller will choose to limit the information about its product attributes even when advertising is costless. Because forced disclosure of product attributes may be harmful to a seller, it gives the competitors an incentive to provide that information through comparative advertising. In a subsequent study, Anderson and Renault (2009) find that comparative advertising is used by weaker firms to target the market leaders and it reduces the overall social welfare. Our study contributes to this literature by providing an empirical analysis of market prices under exogenously specified informational conditions. A natural and straightforward extension of our experimental design is one that allows sellers to choose their advertising strategy pertaining to both price and match value information.

   Our choice of laboratory methodology is guided by constraints imposed by both theory and availability of field data. Theoretical search models are often quite sensitive to their assumptions regarding information acquisition and dissemination. These assumptions tend to be highly stylized, detailed, and often unrealistic and cannot be easily adapted to accommodate industry-specific characteristics (Grether, Schwartz, and Wilde 1988). Furthermore, their empirical assessment is impaired by lack of suitable field data. This problem is particularly evident when the goods are differentiated because field studies are limited in their ability to precisely measure consumers' valuations for different attributes. Such estimation problems make the laboratory methodology particularly attractive, wherein there are no measurement errors and the basic underlying structural and informational conditions are known. Laboratory data are also untainted by various confounding factors that afflict the field data such as reputation formation and brand recognition.

   To capture the essential features of the conventional retail markets in the laboratory, we employ the standard posted offer trading institution. We vary the number and type (human vs. automated) of buyers to provide both a rigorous test of the theory as well as outline its boundaries. We include treatments where robot buyers play a known equilibrium search strategy. This eliminates strategic uncertainty regarding buyers and allows the focus to be entirely on interseller competition. Incorporating human buyer population, on the other hand, provides a robustness check of the equilibrium price predictions to potential deviations from the optimal search behavior.

   We have a rich panel data set of pricing and search decisions of 200 subjects under different informational treatments across 20 sessions comprising 72 periods each. The data from 16 robot buyer sessions indicate that the model predicts seller pricing behavior rather well. Observed prices conform to the model's comparative statics and are broadly consistent with the predicted levels. The results from four additional sessions with human buyers, however, show systematic and substantial departures from the theory. We find that excessive search by human buyers instigates increased price competition, and sellers post prices that are significantly lower than predicted.

   The rest of the article is organized as follows: Section 2 describes the theoretical model, and provides examples to motivate the different informational treatments. Section 3 describes the experimental design and the empirical results when sellers face stable, equilibrium-playing robot buyers. Section 4 includes a similar analysis for markets where sellers face human buyers who may or may not engage in optimal search. We split our analysis for the two buyer populations because of the drastic difference in the results. Section 5 concludes.

2. Theoretical Model

   The theoretical structure underlying the various informational treatments has a common layout. For simplicity of exposition, we begin with the case of uninformed (UI) buyers, which is essentially Wolinsky's (1986) model in a duopoly setup. Figure 1 shows the game tree for this baseline model.

   Uninformed (UI)

   Consider a market where two sellers, 1 and 2, sell a variant of a horizontally differentiated product. Sellers face no capacity constraints and compete by simultaneously setting prices...