Competition among hospitals.

We examine competition in the hospital industry, in particular the effect of ownership type (for-profit, not-for-profit, government). We estimate a structural model of demand and pricing in the hospital industry in California, then use the estimates to simulate the effect of a merger. California hospitals in 1995 faced an average price elasticity of demand of -4.85. Not-for-profit hospitals faced less elastic demand and acted as if they have lower marginal costs. Their prices were lower than those of for-profits, but markups were higher. We simulate the effects of the 1997 merger of two hospital chains. In San Luis Obispo County, where the merger creates a near monopoly, prices rise by up to 53%, and the predicted price increase would not be substantially smaller were the chains not-for-profit.

1. Introduction

* One of the most important sectors of the U.S. economy is health care, accounting for over one trillion dollars in expenditure annually. This sector is also one in which competition is a real issue, given the extensive consolidation that has occurred in recent years (Gaynor and Haas-Wilson, 1999).

During the second half of the 1990s, a dramatic wave of hospital consolidation occurred in the United States. One source puts the total number of hospital mergers from 1994-2000 at over 900 deals (Jaklevic, 2002, and http://www.levinassociates.com), on a base of approximately 6,100 hospitals. Further, many local markets, including quite a few large cities such as Boston, Minneapolis, and San Francisco, have come to be dominated by two or three large hospital systems. Not surprisingly, many health plans have complained about rising prices as a result of this consolidation (Lesser and Ginsburg, 2001).

Hospital markets have been an active area of antitrust enforcement. Since 1984, the federal antitrust authorities have brought 11 suits seeking to block hospital mergers but have won only one (1) of the six cases brought since 1993. Not-for-profit status has played a key role in hospital antitrust cases. Not-for-profit hospitals wishing to merge have argued that they will not raise prices after merging because they are motivated by community interest rather than by profit. Court reactions to this have ranged from sympathetic--"The board of University Hospital is quite simply above collusion" (2)--to outright rejection--"no one has shown that [not-for-profit status] makes the enterprise unwilling to cooperate in reducing competition ... which most enterprises dislike and which nonprofit enterprises may dislike on ideological as well as selfish grounds." (3) On balance, however, the courts have been receptive to this line of argument, particularly in recent years, and the only recent case in which the government has prevailed involved two for-profit hospitals (4) (see Gaynor and Vogt, 2000).

Our goal in this article is to understand the nature of hospital competition and its implications for antitrust policy, in particular, differences in the exercise of market power between for-profit and not-for-profit hospitals. To that end, we estimate a structural model of hospital conduct that explicitly allows for differences between for-profits and not-for-profits, then use the estimates to simulate the effects of a merger. We simulate merger effects both for a merger between for-profits and for a merger between not-for-profits.

Using detailed microdata from California on patients and hospitals in 1995, we find that hospitals face a downward-sloping demand for their products, with an average price elasticity of demand of -4.85. Not-for-profit hospitals face less elastic demand and act as if they have lower marginal costs. Their prices are lower, but markups are higher (26%) than those of for-profits (20%). The merger simulation shows no difference in the tendencies of not-for-profits versus for-profits to exploit merger-created market power. The simulated merger results in postmerger price increases of up to 53%, and changing the firms' profit/nonprofit status has little impact on this figure.

The article is organized as follows. In Section 2 we briefly review relevant prior literature. Section 3 contains a description of the model. The data are discussed in Section 4. Section 5 describes the econometric specification. Section 6 contains the estimates of the structural model. In Section 7 we report a merger simulation highlighting the for-profit/not-for-profit distinction. The conclusion is contained in Section 8.

2. Prior literature

* To date, the hospital competition literature has consisted largely of structure-conduct-performance (SCP) studies. These studies have found that, at least during the 1990s, hospital prices are lower in less-concentrated markets. There are several reviews of this literature available (Gaynor and Vogt, 2000; Dranove and Satterthwaite, 2000; Dranove and White, 1994). There is more variation, however, in the results of the small number of studies that examine this relationship separately for not-for-profits and for-profits. Three of these studies find that both not-for-profit and for-profit hospitals set higher prices in more-concentrated markets (Dranove and Ludwick, 1999; Keeler, Melnick, and Zwanziger 1999; Simpson and Shin, 1998). Two others, however, find that not-for-profit hospitals set lower prices in more-concentrated markets, while for-profits set higher prices (Lynk, 1995; Lynk and Neumann, 1999). Although the results from this literature are interesting, SCP methods suffer from well-known deficiencies for testing hypotheses about competitive conduct. In addition, this type of modelling makes it extremely difficult to sort out the differences in results between the studies of not-for-profit pricing. These studies cover different time periods, use different geographic and product markets, and employ different functional forms. The reduced-form framework makes it difficult to assess the reasons for the different results across these studies, let alone evaluate their relative merits.

There is also an emerging structural hospital competition literature. In this literature, consumer-level data are used to estimate models of demand for hospital services, and then the information from the demand estimation is used to calculate the market power of various hospitals. Town and Vistnes (2001) and Capps, Dranove, and Satterthwaite (2003) each use their demand systems to calculate measures of the marginal value of adding a hospital to a network. Town and Vistnes (2001) then regress prices paid by health plans to hospitals on their measure of a hospital's marginal value and find that hospitals having a high marginal value, either because of isolation in product space or because of high average utility, receive higher payments. Capps, Dranove, and Satterthwaite (2003) regress their marginal-value measure on hospital profit margins and similarly find a positive relationship. Capps et al. (2001), in an approach similar to ours, use their demand estimates to simulate mergers and find that mergers of hospitals even in markets that look quite "competitive" by conventional antitrust methods would nevertheless lead to large price increases.

This article also contributes to the literature on differentiated-product oligopoly by providing evidence from a new market and by using data on individuals. By virtue of the data collected for the hospital industry, we can use microdata on individuals, which has not been commonly utilized in econometric studies of differentiated-product oligopoly (although this has been changing recently). The availability of detailed microdata allows us to flexibly model individual heterogeneity with a directness not possible with aggregate data. The availability of microdata also allows us to employ an instrumenting strategy that differs from those commonly used in the literature.

3. The model

* We model hospital markets as a differentiated-product oligopoly. Hospitals sell products that are differentiated on a number of dimensions. One of the most important dimensions is physical location. Hospitals have their physical plant in distinct locations, and consumers value proximity to their residences. (5) Hospitals also have different religious affiliations. They are differentiated in the breadth of product lines they offer, in the technological sophistication of their services, in the quality of the "hotel" services they offer, in their use and deployment of staffing, in their mortality rates, and probably in other dimensions as well. It seems reasonable, therefore, to model hospital competition using models of differentiated-product oligopoly.

An added complexity in the case of hospitals is that many hospitals are not-for-profit organizations. (6) A literature has grown up around the idea that not-for-profit hospitals, unlike other firms, do not maximize profits but rather some utility function, possibly reflecting the preferences of the board of trustees, the administrators, the employees more generally, or the physician staff (Newhouse, 1970; Pauly and Redisch, 1973; Lee, 1971; Lakdawalla and Philipson, 1998), and our model reflects the potential for preference differences in the different forms.

Since our goal is a structural, estimable model of demand and supply, we lay out the structure in terms of the demand and supply sides of the model. In what follows we first describe consumers, then producers. On the supply side, we specify a model that explicitly takes into account the differing objectives of not-for-profits and for-profits.