Antitrust policy and hospital mergers: recommendations for a new approach.

I. Introduction

During the 1990s, there were over 900 hospital mergers and acquisitions, many involving hospitals in the same metropolitan areas. (1) These transactions consolidated the hospital industry, dramatically concentrating the supply of hospital services. In principle, several purposes motivated this consolidation. First, consolidation might have facilitated the elimination of excessive beds and services. By the end of the 1980s, the average hospital capacity utilization rate had fallen to 60%. (2) Indeed, eliminating excess capacity (3) or duplicative services was a stated goal of many hospital mergers. (4) This consolidation also appears to be a direct response to the simultaneous growth of managed care (5) and the shift to outpatient care in the 1980s and 1990s.

Merging hospitals rarely mention a second plausible motive, namely, to enhance market power with respect to managed care organizations (MCOs). MCOs obtain discounts from hospitals' stated charges by threatening to steer patients to alternative hospitals offering more favorable pricing. To make this threat credible, preferred provider organizations (PPOs) generally charge enrollees higher co-payments if they visit a noncontracting hospital, while health maintenance organizations (HMOs) usually provide no coverage at all for nonemergency care at noncontracting providers. This threat enables MCOs to play hospitals against each other to extract larger discounts. By consolidating, hospitals can limit the ability of MCOs to steer patients, and thereby resist MCO demands for discounts.

Despite the potential for consolidation to enhance market power, the Federal Trade Commission (FTC) and Department of Justice (DOJ) have challenged only a handful of hospital mergers. (6) In the 1980s, the government prevailed in all but one case, in Roanoke, Virginia. (7) This situation dramatically reversed in the 1990s. After winning an injunction on appeal in circuit court in Augusta, Georgia. (8) the FTC and DOJ lost six successive cases. These losses accumulated in district court, circuit court, and in one instance, before an FTC administrative law judge. In all but one case, the definition of the relevant geographic market played a key role in the outcome. The various courts began accepting hospitals' claims that the relevant geographic market was quite large. In one case, hospitals over 80 miles away were ruled to be in the relevant market.

Hospitals justified the inclusion of distant hospitals in their geographic markets through analyses of patient flow data. Using an approach for geographic market definition first advocated in Elzinga and Hogarty, (9) merging hospitals presented evidence that a nontrivial proportion of local residents--usually in excess of 25%--traveled to distant hospitals. The courts reasoned that if many local patients traveled prior to a merger, then even more patients would travel if the merged hospitals raised prices. Thus, the courts surmised that merging hospitals could not profitably raise prices.

There are well-known problems inherent in using patient flows to define geographic markets. Generally, there is no theoretical link between patient flows and the presence or absence of market power. (10) But the courts have not always found these doubts persuasive. More to the point, the court decisions of the last decade indicate a lack of sympathy with the doubters, presuming instead that significant patient flows preclude the presence of market power. It is this presumption--heretofore untested on any data--that we question.

The basis for our critique is simple in principle: patient flow data do not come close to approximating generally accepted criteria for determining market boundaries. Both the standard merger guidelines (11) and the hospital-specific guidelines (12) advocate using the small but significant nontransitory increase in price (SSNIP) criterion. (13) Under this standard, a narrow market definition is initially proposed. If the hospitals in the narrowly defined market could, by acting as a joint monopolist, implement a SSNIP, then they constitute the relevant set of competitors. If they cannot do so, then it must be because the market is defined so narrowly as to exclude a close substitute. Thus the market definition should be expanded to include the next closest competitor. The process is iterated until the SSNIP question is answered in the affirmative.

Even though the SSNIP is theoretically appealing, patient flow analysis remains the main tool for defining geographic markets for hospital mergers. In part, this is due to historical precedent. By the time of the release of the hospital merger guidelines, there were at least a half dozen hospital merger cases in which courts had based geographic market definition on flow data, including two cases decided by appellate courts in different circuits. Another reason is the appeal of short cuts. It is easy to obtain and analyze patient flow data. Most states have regulatory agencies that collect patient-level hospitalization data, and most analysts can easily replicate the methods advocated by Elzinga and Hogarty. In contrast, the SSNIP standard is challenging to implement. It requires answering hypothetical questions that require knowing the entire demand system faced by all hospitals. This is difficult in any market, but is particularly so in hospital markets. The requisite data are often unavailable for two reasons: negotiated prices, as opposed to list prices, are secrets closely guarded by hospitals, and even where these prices are available, the prices faced by patients are not observed. (14)

We investigate whether inferences using Elzinga-Hogerty (E/H) flow data are close to the inferences using the ideal SSNIP criteria. We propose three alternative but related methodologies for doing so. Each combines the same data used in patient flow analysis--evidence on the zip codes of patients and hospitals--with readily available data on hospital characteristics. (15) Our methods require no a priori definition of the relevant geographic or product market, nor do they require patient-level price data. Yet, the methods provide defensible predictions of the price increase any given merger would generate. The three approaches are

 
   The time-elasticity approach, in which the observed premerger travel 
   patterns of actual patients are used to infer postmerger willingness 
   to travel; 
 
   The competitor share approach, which hones in on the differentiated 
   product nature of hospital services and patient heterogeneity; 
 
   The option demand approach, which is slightly more complicated, but 
   is tailor-made for analyzing the effects of hospital mergers in a 
   managed care setting.(16) 

All three approaches generate consistent results. They indicate that mergers in markets with significant outflows of patients--30% or more--can generate price increases of 10% or higher. In other words, mergers that might easily pass muster using flow analysis may easily fail using the SSNIP criterion. We surmise that, for a wide range of plausible situations, patient flow data provide a highly inaccurate view of the appropriate market boundaries. It is only reliable in extreme situations, when flows are very large or nearly nonexistent--that is, flows very near zero likely support exclusion of an area from a proposed market and flows near 100% support inclusion. Unfortunately, such extreme situations do not present especially interesting problems for courts. Hence, we advocate eliminating the use of flow data and E/H criteria in any situation where their inferences are ambiguous, which is, practically speaking, all courtroom proceedings.

This article summarizes these new approaches. Before doing so, however, we first discuss several illustrative merger cases, highlighting the divergence between the reasoning of the courts and the policy outlined in the Merger Guidelines. Section II discusses these cases; section III discusses empirical evidence on mergers, including the inconsistent evidence on merger efficiencies. Section IV describes our new approaches and the results thereof, and section V concludes.

II. Illustrative cases

Two cases that illustrate the thinking of the courts in hospital merger cases both occurred in Missouri, the first in Joplin (17) and the second in Poplar Bluff.(18) In the Freeman case, the second and third largest hospitals in Joplin proposed a merger in 1995. In a harshly worded ruling denying the government a temporary restraining order, the district judge stated "I don't see how the Federal Trade Commission can claim there is lack of competition when there [are] four or five hospitals in the area, and reducing it by one is not going to wipe out competition." (19) On remand, (20) the district court faced conflicting testimony regarding the size of the relevant geographic market. The FTC expert testified that the relevant market was roughly a 54-mile diameter circle around the merging hospitals while the defendant's expert argued that the relevant market was a 13 county area, roughly 100 miles in diameter, and included 17 hospitals.(21) Under the latter scenario, which the court found more compelling, the merger would have only a small effect on concentration in the purported market. The hospital's expert used Elzinga/Hogarty patient flow analysis(22) to arrive at this conclusion.

Under the Elzinga/Hogarty criterion, the geographic market is expanded until two criteria are satisfied: Little Out From the Inside (LOFI), and Little in From the Outside (LIFO). This process defines a market by determining the smallest geographic such that (1) the portion of patients who leave the proposed market for care (LOFI) and (2) the proportion of patients from outside the boundaries who receive care within the market (LIFO), are both below a critical threshold--in this case, the defendants' expert used 10%. As discussed in detail in section IV, when applied to hospital markets this approach generally leads to very large geographic markets.

In finding for the merging hospitals, the district court accepted their E/H style approach to market definition. The court further criticized the FTC on the grounds that their evidence only presented a static picture of the market; it failed to address the key counterfactual question of whether patients could bypass the merging hospitals if they raised prices postmerger. (23) Exacerbating the FTC's troubles was testimony from local insurance payers, who stated that they could steer patients to outlying hospitals in response to significant price increases. (24)

On appeal, the Eight Circuit affirmed the majority of the district court's finding, going so far as to cite competition from Kansas City, Missouri and St. Louis, Missouri, as well as Tulsa, Oklahoma--cities all 70 miles or more from Joplin. (25) In the same year, the Department of Justice also lost a case in Dubuque, Iowa on nearly identical grounds. (26) In particular, the court noted that doctors did send patients beyond the borders of the market proposed by the DOJ. The court further noted the need to answer not just the question of where patients go, but where they would go in response to a price increase, (27) offering the opinion that patients would travel in response to such increases. (28) Previewing an upcoming case, the court did state that, had the government prevailed on market definition, neither of the alternative defenses offered by the hospitals were compelling. The first alternative defense was an efficiencies argument, which the court dismissed as highly speculative and likely overstated by the defendants. (29) The second rejected defense was that, even if they had market power, the hospitals would not increase price because both hospitals are nonprofits. (30)

In the second Missouri case, FTC v. Tenet, the commission prevailed in district court. The court accepted the FTC's proposed geographic market, stating, "at some point, a hospital ceases to become a practical alternative for general acute care because of distance." (31) Again, using the LOFI and LIFO criteria with a 10% threshold, the defendants argued instead for a much larger geographic market, including areas as …