Competition and market power in option demand markets.

We call markets in which intermediaries sell networks of suppliers to consumers who are uncertain about their needs "option demand markets." In these markets, suppliers may grant the intermediaries discounts in order to be admitted to their networks. We derive a measure of each supplier's market power within the network; the measure is based on the additional ex ante expected utility consumers obtain from the supplier's inclusion. We empirically validate the WTP measure by considering managed care purchases of hospital services in the San Diego market. Finally, we present three applications, including an analysis of hospital mergers in San Diego.

1. Introduction

* Some important markets feature intermediaries that offer a network of upstream suppliers to downstream consumers. Examples include general contractors, who assemble networks of skilled craftsmen and subcontractors; business-to-business web sites, which assemble networks of parts suppliers; and managed care organizations, which assemble networks of hospitals and physicians. These intermediaries take advantage of their expertise and purchasing economies to identify superior suppliers and extract better terms than could consumers shopping on their own. In some cases, such as managed care, they also provide insurance against the risk of needing the network's services.

Sometimes, consumers may know their specific needs at the time they select their intermediary. For example, homeowners may have detailed architectural plans at the time they select their building contractors. In other situations, consumers may select their intermediary before knowing their specific needs. Insurance markets are an important example. Automobile owners often commit to a network of auto repair shops at the time they purchase collision insurance, even though they do not know in advance what kinds of repairs their cars might require. Similarly, patients commit to a network of medical providers at the time they purchase their health insurance, but before they know their specific medical needs. Noninsurance examples include manufacturers who sign long-term contracts with suppliers, who in turn outsource specific manufacturing tasks as the need arises. Following Dranove and White (1996), we call these "option demand markets" (or OD markets). In OD markets, consumers commit to a potentially restricted network of sellers prior to fully knowing their needs, but retain the option to visit any seller in the network once their needs are known. The value that any one consumer places on a given network depends on his expectation of how well the network's members will be able to meet his needs. This contrasts with direct-purchase markets, in which consumers do not eliminate any potential sellers prior to learning their needs.

The article's goal is to develop and validate an index of the market power of suppliers in OD markets. We consider a competitive intermediary assembling a network of suppliers on behalf of many consumers. We assume that consumers' preferences follow the logit model of demand. We also suppose the intermediary knows the underlying logit utility function, the distribution of consumer characteristics, and the distribution of the possible states of the world that affect demand, but not the upcoming demand realizations. A straightforward calculation based on the properties of logit demand gives an estimate, for each supplier, of how much consumers in aggregate are willing to pay ex ante to retain it in the network. (1) A simple reduced-form bargaining model between the supplier and the intermediary suggests that a portion of this willingness-to-pay (WTP) is captured by the supplier. The WTP associated with a supplier is therefore a measure of its market power: a supplier for which WTP is high secures higher prices from the intermediary than does a supplier for which WTP is low.

After deriving the formula for WTP, we validate the measure by examining hospitals in the San Diego, California, metropolitan area. Following our discussion, managed care organizations (MCOs) in this market are the intermediaries and hospitals are the suppliers. Each MCO negotiates bilaterally with each hospital for inclusion in its network. Its goal is to come to agreement on rates with a set of hospitals such that the resulting network maximizes the difference between consumers' ex ante WTP for that network and its expected payments to the hospitals that provide the services. To the extent that the MCO succeeds in doing this it can offer employers a competitive price on a health plan that employees value highly. Once the network is formed, consumers realize their health state and, for those who need hospitalization, select the hospital they most prefer from among those included in their network.

Using 1991 data on inpatient hospital services in San Diego, we estimate a multinomial hospital-choice model for patients who have a free choice of hospital. This provides estimates of the parameters of patients' logit utility functions. Based on these parameter values and the empirical distributions of patient characteristics and health states we compute, for each hospital, the consumers' aggregate ex ante WTP to retain it in the network. (2) These WTP measures are denominated in "utils" because in estimating the logit function we intentionally select consumers who did not face prices that vary across hospitals. To convert utils to dollars, we regress each hospital's actual profits from inpatient services provided to managed care patients onto our estimates of consumers' ex ante WTE The results of this regression are consistent with the validity of the WTP measure: WTP is a highly significant predictor of hospital profits.

The last section of the article offers several applications of the WTP measure. Of greatest interest, we show how to use WTP to define geographic markets under current federal merger guidelines. In this application, we estimate how much, if any, additional profit two or more hospitals are able to extract from MCOs after increasing their WTP through a merger. We then infer the effect of the merger on prices. When we conduct this exercise for a merger among three geographically close hospitals in the suburbs of San Diego, we estimate that prices would increase by more than 10%, under the assumption of zero cost changes. Under the merger guidelines, this implies that the south suburbs are a well-defined geographic market. As a byproduct of this analysis, the courts could compare the predicted price increase, under the assumption of zero cost changes, against any asserted cost savings. This is essential to any rule-of-reason assessment of a proposed merger.

2. Background on antitrust

* In the 1990s, the U.S antitrust agencies lost virtually every challenge to hospital mergers. (3) In most of the contested merger cases, the court's ruling turned on geographic market definition. To define geographic markets, the antitrust agencies recommend using the small but significant nontransitory increase in price (SSNIP) criterion. (4) Under SSNIP, a narrow trial market definition is initially proposed. If the firms in the trial market could collectively implement a SSNIE then they constitute the relevant set of competitors. If they cannot do so, then it must be because a relevant competitor was excluded. In this case, the proposed market is expanded until the SSNIP criterion is met.

Rather than directly observing or attempting to predict price effects, the courts often rely on proxy measures. A popular proxy measure in hospital merger cases is derived from methods introduced by Elzinga and Hogarty (1973). Elzinga and Hogarty and related approaches use aggregate inflows and outflows of patients (or imports and exports of goods) to determine market boundaries. Given the propensity of some patients to travel substantial distances for care, this standard has led to large market boundaries and, consequently, permissive merger rulings.

Our results indicate that this may be a serious error. The south suburbs of San Diego would not be a well-defined market under Elzinga and Hogarty, yet the WTP approach indicates that they are. We conclude that the willingness of some patients to travel does not eliminate the market power hospitals may have in their local neighborhood. Many patients, especially those with conditions that are relatively straightforward to treat, have a strong preference to go to a convenient, nearby hospital. These preferences give hospitals with no nearby competitors a strong bargaining position.

3. Related literature

* There is an extensive literature on hospital pricing. (5) One branch of that literature, which includes Noether (1988), Dranove, Shanley, and White (1993), Lynk (1995), and Keeler, Melnick, and Zwanziger (1999), consists of traditional price-concentration studies using average hospital prices across a large cross-section of markets. Another branch that includes Staten, Umbeck, and Dunkelberg (1988) and Melnick et al. (1992) examines the hospital prices paid by specific health insurers. Both of these branches share two difficulties that our article addresses.