Publication: RAND Journal of Economics
Publication Date: 22-DEC-06
Delivery: Immediate Online Access
Author: Davis, Peter

Article Excerpt
Retail markets are extremely important, but economists have few practical tools for analyzing the way dispersed buyers and sellers affect the properties of markets. I develop an econometric model of retail demand in which products are location specific and consumers have preferences over both geographic proximity and other store and product characteristics. The model uses data on the observed geographic distribution of consumers within a market to (1) help explain observed variation in market shares and (2) affect predicted substitution patterns between stores. Using data from the U.S. cinema industry, I use the estimated model to evaluate the form of consumer transport costs, the effect of a theater's price and quality choices on rivals, the effects of geographic differentiation, and the nature and extent of market power.

"[B]ecause, once in place, most multi-screen theaters in new markets effectively satisfy market demand, the Company has no competition in over 67% of its theater locations." [Regal Cinema's 1995 Annual Report, p. 4.]

1. Introduction

* Retail markets constitute a large fraction of the U.S. economy, employ many millions of people, and provide the interface between consumers and producers for vast numbers of products. As the final stage in the distribution channel, retail markets are also inherently differentiated product markets, so retailers may benefit from market power. If they do, researchers in the tradition of Hotelling (1929) have established that the resulting differences in social and private incentives can potentially translate into suboptimal location, scale, and pricing decisions. Theoretically, then, there is scope for welfare-enhancing regulatory and/or merger control intervention in retail markets, and quotations such as the one above are not entirely reassuring for the view that laissez-faire is always the appropriate policy. In practice, multiple agencies regularly and directly affect actual retail market structure: zoning agencies control the form and structure of individual retail store locations, while competition authorities impose constraints on the extent of multiple store ownership.

In this article I build a model of demand directly suitable for undertaking policy analysis in retail markets. I use it to evaluate the extent of geographic markets and the nature of market power in a specific retail market, the motion picture exhibition market. I find that travel costs result in limited theater substitutability and localized markets. I also find that the main constraints on the exercise of market power by theaters, at least through admission prices, involve (i) consumer substitution to other activities and (ii) the incentives of distributors. Specifically, I find that the average theater owner would prefer to actually lower admission prices, if she could attract the same set of films. The reason is that while theaters share box office revenues with distributors, high-margin concessions revenues (e.g., from popcorn) are not shared. Thus, theater owners have strong incentives to provide lower admissions prices than distributors would otherwise like.

Geographers and marketing specialists have led in the effort to understand the spatial nature of retail markets empirically. Reilly (1931) developed his "law of retail gravitation" linking retail sales data to population and the proximity of rival stores or towns. Huff (1964) provides an early model of retail store choice within a single market that explicitly uses an underlying model of individual consumer choice behavior, though Neidercorn and Bechdolt (1972) show that Reilly's model can also be derived from a particular utility maximization model. Variants of the consumer preference approach, including applications of the multinomial logit (MNL) model, have subsequently dominated (see Nakanishi and Cooper, 1974; Theil, 1969; McFadden, 1973; Arnold, Roth, and Tigert, 1980; Craig, Ghosh, and McLafferty, 1984; Fotheringham, 1991; Thill, 1995; and Wrigley and Brouwer, 1986). More recently, Fotheringham (1983), Sheppard, Haining, and Plummer (1992), and Plummer, Haining, and Sheppard (1998) have suggested alternatives to MNL, while nonparametric approaches to mapping the area from which stores attract consumers from have also been explored; see, for example, Rust (1991).

I follow Berry (1994) and Berry, Levinsohn, and Pakes (1995) in estimating a differentiated product demand model with unobserved consumer heterogeneity to avoid the disadvantages of the MNL model. In contrast to most unobserved consumer heterogeneity, a fully nonparametric estimate of the distribution of consumers within each market is available from the decennial Census of Population. To paraphrase Spence (1976), the welfare consequences of any government policy or business strategy can only be evaluated relative to the distribution of purchasers, and we should clearly use our best estimates of the relevant one. Doing so provides an empirical model directly suitable for empirically evaluating objects of direct policy interest such as the geographic scope of actual retail markets.

The closest prior work is provided in de Palma et al. (1994), who calculate a model of retail demand incorporating the distribution of consumers provided by the census. In contrast, I use the actual observed store locations, prices, and characteristics to estimate the parameters in a retail demand model. Doing so means that questions that have pervaded the theoretical literature, such as whether transport costs are more appropriately treated as a linear or quadratic function of distance, become empirical questions about the structure of consumer preferences, informed by data from observed aggregate choice behavior. In contrast to the theoretical literature following d'Aspremont, Gabszewics, and Thisse (1979), which assumes that travel costs are convex, I conclude that the marginal cost of travel declines with distance.

The remainder of the article is organized as follows. In Section 2 I briefly describe the structure of the exhibition market. Section 3 describes the demand modelling framework, and Section 4 describes the dataset. In Section 5 I present the econometric model and identification assumptions, while Section 6 presents estimates and analysis. Section 7 presents evidence on the extent of geographic market definition. Section 8 concludes and outlines directions for future research.

2. The movie exhibition industry

* In 1997, North American consumers went to the movies approximately 1.4 billion times, generating box office revenues of $6.4 billion. (1) In contrast to the national distribution market, where the eight major players account for over 95% of sales, the largest five theater circuits (chains) operated just 41% of the approximately 32,000 screens.

Relative fragmentation at the national level masks high levels of regional concentration, and recent examples of antitrust geographic market definitions have included "Chicago" and "Manhattan" where there was concern over the Sony and Cineplex merger in 1996. (2) I will begin with broad potential market definitions, allowing the data to determine where narrower definitions are more appropriate. Since the famous Paramount antitrust case, (3) the relevant product market definition for antitrust purposes has involved the exhibition of "first-run" movies. First-run movies are those that are in their first theatrical exhibition run, and I follow that approach to product market definition.

An attractive feature of the exhibition market for studying retail competition is that distribution is a relatively simple task; each theater shows a small number of films. Moreover, the Paramount consent decrees resulted in the vertical disintegration of the industry, so distribution occurs largely at "arm's length" In particular, a controversial element of the consent decrees still essentially in place requires that films be licensed on a theater-by-theater basis, on the merits.

There are features of the exhibition market that are unusual within retail. One is that the price for admission by final customers is essentially set in the market for an input, the film licensing market. Prior to 1948, the distributors controlled admission prices primarily by owning the theaters, but since then prices have been chosen "exclusively" by the largely vertically separate exhibitors. Box office revenues are shared between the distributor and the exhibitor contractually, while revenues from concessions sales are not, so exhibitors typically prefer lower admission prices than do distributors. Licensing contracts, while technically offered by the exhibitors, generically specify the admission prices for the three key demographic groups: adults, children (under 12) and seniors (over 60), and a maximum percentage of admissions that will be discounted tickets (e.g., 2% in the first two weeks). Evidently a theater that promises to charge a very low admission price in the market for film licenses is fairly likely to find itself unable to license popular films. Minimum retail price maintenance is per se illegal in the United States, effectively making the retail sales price noncontractible in most other retail markets. Here, however, a distinction appears to be drawn because retail prices are offered to the distributors by the theaters rather than imposed by the distributor.

3. The demand model

* The model I present is a MNL model with unobserved consumer heterogeneity following Theil (1969), McFadden (1973), Boyd and Mellman (1980), Cardell and Dunbar (1980), and more directly Berry (1994) and Berry, Levinsohn, and Pakes (1995) (BLP). Space limitations mean that I must refer the reader to Nevo (2000a) and Davis (2000) for a discussion of other feasible approaches.

Suppose we observe t = 1,..., T days of sales data for each film f = 1,..., [F.sup.ht] in each theater (house) h = 1,..., [H.sup.m] in a set m = 1,..., M markets. A product is defined to be a film playing at a specific theater within a market at a given point of time. The conditional indirect utility that consumer i obtains from watching film f at theater h is assumed to be of the form

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII.] (1)

where ([beta], [lambda]) are parameters; [X.sub.fhmt] are observed product characteristics such as the number of screens, the type of sound system, whether the theater operates a consumer service line, and is handicap accessible or operates a system for the hearing impaired; and [P.sub.hmt] is the price of admission. [[tau].sub.f] denotes a film fixed effect, [[gamma].sub.c] is a circuit (chain) fixed effect, and [K.sub.t] is a time fixed effect. One element of consumer heterogeneity is her location, [L.sub.i], and its interaction with the theater's product characteristic, location [L.sub.h], determines her distance d([L.sub.i], [L.sub.h]) from each theater. The function g(.,.) is assumed to be a parametric function of d([L.sub.i], [L.sub.h]) known up to the vector of parameters [gamma], so the model nests a variety of transport cost specifications, including those used by Hotelling (1929) and d'Aspremont, Gabszewics, and Thisse (1979). I use Euclidean distance so that a consumer's indifference curves in geographic space are concentric circles about her location. Consumers then prefer to attend closer theaters, but two theaters that provide good substitutes are not necessarily physically close to one another. (4) Following Berry (1994), [[xi].sub.fhmt] represents an unobserved (by the econometrician) product characteristic, assumed observed by, and common to, all individuals so that ceteris paribus, all consumers prefer a high [[xi].sub.fhmt] product. (5) Finally, [[member of].sub.ifh] is a mean-zero individual and option specific stochastic term with a type-I extreme value distribution. Consumers do not just go to their closest store (Rushton, College, and Clark (1967), for example, found that even rural Iowans chose to go to their nearest store only 35% of the time), and so this model incorporates nondistance store characteristics and also allows for consumer heterogeneity.

Specification of the demand system is completed with the introduction of an "outside option"--some consumers decide not to attend any screening of a film. The conditional indirect utility from the outside option is

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII.] (2)

where [[member of].sub.i0] is a mean-zero individual market and time specific stochastic term and [v.sub.i] is an individual specific component reflecting heterogeneity in tastes for going to the movies across people. Unlike location, we do not know much about this element of the distribution of consumer heterogeneity, so I assume that [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII.] N(0, 1) and estimate the parameter [[sigma].sub.outside]. Without an outside option, a uniform price increase would not, for instance, result in lower market sales. I normalize [[xi].sub.0mt] to zero.

Surveys report (6) that 94% of patrons choose which film at which theater to attend before arriving at the theater. Hence a model of...



NOTE: All illustrations and photos have been removed from this article.



More articles from RAND Journal of Economics
Durable-goods oligopoly with secondary markets: the case of automobiles., 22-JUN-07
Nonlinear pricing in an oligopoly market: the case of specialty coffee, 22-JUN-07
Measuring consumer welfare in the CPU market: an application of the pure-characteristics demand model., 22-JUN-07

Looking for additional articles?
Click here to search our database of over 3 million articles.

Looking for more in-depth information on this industry?
Click here to search our complete database of Industry & Market reports by text, subject, publication name or publication date.

About Goliath
Whether you're looking for sales prospects, competitive information, company analysis or best practices in managing your organization, Goliath can help you meet your business needs.

Our extensive business information databases empower business professionals with both the breadth and depth of credible, authoritative information they need to support their business goals. Whether it be strategic planning, sales prospecting, company research or defining management best practices - Goliath is your leading source for accurate information.