Asymmetric competition on commuter routes: the case of gasoline pricing.

1. Introduction

In many industries, retailers enjoy the benefits of product differentiation simply because they are scattered about the city. Even if the firms sell identical products, their different locations make certain stores more accessible or convenient for some customers. Consumers incur a smaller travel cost when they go to a nearby store; hence, the geographic distribution of stores creates a consumer preference to buy from stores that are closer. The key question is, closer to what? The standard approach has considered proximity to the consumer's home as the critical measure of closeness. The real issue, however, is how far the consumer must go out of her way to purchase from a particular store. The retailer for which this incremental travel cost is smallest is the seller that is closest to a consumer.

If consumers must make a separate trip to purchase each product, these two views of closeness yield the same results. The distinction between these definitions of "close" becomes relevant, however, if consumers also travel for other reasons. Consumers who commute to work, for example, consider a store close if it is near the worker's regular route to work. As Claycombe (1991) points out, stores located along a consumer's commuting route are effectively undifferentiated (spatially) because the consumer does not have to go out of her way at all to make a purchase from any of these firms. Even though some of these stores may be far from the consumer's home, there is no incremental travel cost for buying from any of them; thus, the apparent spatial differentiation is actually irrelevant. For this reason, the competitive environment along a commuter route is more intense than it appears at first glance.

The purpose of this article was to examine the pricing behavior of gasoline retailers located on commuter routes. To address this question, we estimated a price function for gas stations on commuter arteries in Lexington, Kentucky. Focusing on routes where commuters constitute a large portion of the drivers affects our study in two important ways. First, the absence of spatial differentiation (in the eyes of the commuters) creates a "narrow market" (Claycombe and Mahan 1993) where the firms on each route compete primarily with other firms on that route. We incorporate this insight into our work by treating each route as a separate market instead of defining the entire city as a single market. Second, most commuters in Lexington drive to a downtown Central Business District (CBD), which means more consumers commute past a firm near the CBD than past a firm far from the CBD. This feature creates a potential asymmetry that we exploit in our empirical analysis.

For the price function, we focus on location and the number of competitors as the primary structural determinants of price. To accommodate the potential asymmetry, however, we split each variable into two pieces, an inner component (between the firm and the CBD) and an outer component (between the firm and the outer end of the market). The empirical results show that a firm's price is an increasing function of the distance variables and a decreasing function of the number of competitors. We also find that the directional flow of commuting seems to matter, for each type of variable has asymmetric effects, with the inner variable having a stronger impact on the price. Because these effects are unequal for a given type of variable (distance or competitors), our regressions yield a predictable pattern of relative prices on a route, a finding that seems to accord well with experience.

In addition to the broad inner-outer asymmetry that is our primary result, we find two other respects in which competition on these routes is not uniform. First, interbrand competition is much more effective than intrabrand competition. Firms that sell the same brand exert almost no influence over each other's prices, but a firm sets a lower price when it faces more competitors selling different brands. Second, even though all firms on the route affect the price, the nearest neighbor seems to be the most important competitor. For a given number of competitors and length of route, a firm's price is higher when the distance to the nearest neighbor is greater. Because this one distance relaxes the competitive pressure from the closest firm only, it appears that this neighbor influences price more than the other firms.

Several studies have investigated whether pricing in the retail gasoline industry is competitive, both for policy and academic purposes. Frequently, consumer advocates and government officials have claimed that rapid price increases arise because sellers (retailers or big oil companies) engage in monopolistic or collusive behavior. In response, the U.S. Federal Trade Commission and the Canadian Competition Bureau have each conducted several investigations to determine whether anticompetitive behavior caused the gasoline price increases in question, but they have found no evidence of illegal behavior. (1) Academic studies, however, have found that there are differentiating factors that confer some market power on firms. Sources of differentiation include whether a station offers full-serve, self-serve, or both (Barron, Taylor, and Umbeck 2001); whether a station sells a major brand (Hastings 2004); and other station characteristics like location or carwash (Eckert and West 2005). Because these factors blunt direct competition, Sen's (2003) finding that market concentration affects gas prices (although cost is the primary determinant of price) is not surprising.

[FIGURE 1 OMITTED]

Additional studies have examined data to test alternative theories of firm behavior. By means of very different approaches, Slade (1986) and Eckert and West (2005) both reject the hypothesis of independent Nash price-setting behavior in the retail gasoline industry. Whereas Slade (1986) also rejects several standard oligopoly models, Eckert and West's (2005) estimates are consistent with tacit collusion. In theoretical models, tacit collusion produces patterns of price variation over time (see Friedman 1971; Rotemberg and Saloner 1986; Maskin and Tirole 1988); thus, several studies have tested for tacit collusion in the richer context of supergame models that generate price dynamics. By a variety of theoretical models, Slade (1987), Castanias and Johnson (1993), Borenstein and Shepard (1996), and Eckert (2002) all find their data are consistent with a theoretical model of tacit collusion.

Our article adds to the empirical efforts to characterize gasoline pricing in a static model by incorporating key insights from the commuting literature. We find that there is interaction between traditional structural determinants of price and commuting patterns, which generates asymmetric impacts of market structure on price. In addition, we show that interbrand competition is more important than intrabrand competition on these routes. To develop these results, in the next section, we discuss the commuting literature and explain some implications of our decision to focus exclusively on commuter routes. The following section contains an illustrative Hotelling (1929) model adapted for commuter routes and derives empirical predictions by exploring some examples. The empirical model and results constitute the fourth section. Finally, there is a concluding discussion.

2. Commuter Routes as Markets

Gas prices on commuter routes are the focus of this study. We relied on a city map (see the Appendix), observed traffic patterns, and broadcast traffic reports to identify several commuter routes within Lexington. (2) The stylized representation of a commuter route in Figure 1 helps illustrate important features of this type of market. In the figure, everyone works in the CBD, but the workers live at various points along the road leading to the CBD. The linear structure of the market orders the firms and consumers relative to each other.

Relative location is important in both our example and our empirical work; hence, let us pause briefly to define terms that we will use throughout the article to describe one position in comparison to another's position. We say that an agent (a firm or consumer) or a market segment is located inside agent i if that agent or segment is nearer the CBD than agent L In our example, firm A is inside firm B, and commuter 1 lives inside firm B. The inner or inside market (for firm B) is the segment between firm B and the CBD. Similarly, we refer to agents or market segments as outside agent i if they are farther from the CBD than agent i. In Figure 1, commuter 2 and firm C are both outside firm A and firm B.