Abnormal returns and the regulation of nonproﬁt hospital ... · Journal of Health Economics 24 (2005) 113–135 Abnormal returns and the regulation of nonproﬁt hospital sales

Journal of Health Economics 24 (2005) 113–135

Abnormal returns and the regulation of nonprofithospital sales and conversions

Andrew J. Leone, R. Lawrence Van Horn∗, Gerard J. Wedig1

William E. Simon Graduate School of Business Administration,University of Rochester, Rochester, NY 14627, USA

Received 1 December 2003; received in revised form 1 April 2004; accepted 1 July 2004Available online 2 November 2004

Abstract

During the 1990s, concerns that nonprofit (NP) hospitals were being sold at below-market pricesto investor-owned (IO) chains helped to prompt the widespread adoption of state laws regulating thesale and conversion of nonprofits. In this paper, we provide a simple test of under-pricing using the IOacquirer’s abnormal stock market returns at the time of the acquisition. Prior to regulation, we find thatIO chains did not earn abnormal returns from their acquisitions of NPs and earned greater returns frompurchasing other IO and privately owned hospitals. In states that subsequently adopted regulations,acquisition activity slowed significantly and acquirer returns became negative. Efficient markets theorysuggests that, absent regulation, expected merger synergies were already being transferred to the NPtarget and that regulation may have reduced expected synergies or increased the costs of acquiringNP hospitals.© 2004 Elsevier B.V. All rights reserved.

JEL classification:G14; G34; I18; L31

Keywords:Agency costs; Acquisitions; Event studies; Nonprofit; Hospitals

∗ Corresponding author. Tel.: +1 585 273 4890; fax: +1 585 442 6323.E-mail addresses:[email protected] (R.L. Van Horn), [email protected]

(G.J. Wedig).1 Tel.: +1 585 273 1647; fax: +1 585 442 6323.

0167-6296/$ – see front matter © 2004 Elsevier B.V. All rights reserved.doi:10.1016/j.jhealeco.2004.07.004

114 A.J. Leone et al. / Journal of Health Economics 24 (2005) 113–135

1. Introduction

During the decade of the nineties, numerous nonprofit (NP) hospitals were acquiredby investor-owned (IO) hospital chains.1 In some of these transactions, regulators voicedconcerns that the NP targets were being sold at prices below their market value. For example,in 1996 the California Attorney General halted the sale of Sharp Memorial Hospital toColumbia/HCA. He threatened to hold the board liable, in part because they acceptedan offer from Columbia/HCA that was $200 million less than competing offers.2 Shortlythereafter, California passed a law regulating the sale and conversion of NP hospitals.Ultimately, 24 states and the District of Columbia enacted legislation to regulate the saleand conversion of nonprofit hospitals, primarily during the time period from 1997 to 1998.In virtually all cases, the legislation contained provisions to ensure that the acquirer paid afair market price for the NP target.

In this paper, we test the hypothesis that NP hospitals targeted for acquisition by 10chains were priced below their market value, especially prior to 1997. Specifically, wetest whether IO chains earned excess returns from NP acquisitions prior to 1997 andwhether these returns abated in the aftermath of the state legislation. The topic we pro-pose is important for at least two reasons. First, there is a long-standing concern that NPassets are frequently sold at below-market prices, resulting in transfers from donors andthe government to private parties.3 Second, under-pricing and the effects of regulationmay help to explain the cycle of acquisition activity in the hospital sector over the pastdecade.

The concerns of regulators are consistent with the view that nonprofit managers andboards lack incentives to bargain for a fair price because they do not share in the pro-ceeds from the sale. Parties such as the IRS are also concerned that financial promises ofhighly-paid employment further weaken the incentives of NP CEOs to bargain with theirIO acquirers.4 However, if there are many bidders for the target, an IO acquirer will beforced to offer a fair market price for the target’s assets to outbid its rivals. Moreover, NPboards realize benefits “in kind” from realizing higher bids. For example, a generous bid isfrequently used to fund a charitable foundation that board members manage. These benefitsprovide NP boards with their own incentives to demand fair market prices before approvinga sale.

We use event studies to test for fair market pricing both before and after the passage ofthe state regulations. We argue that if IO chains fail to realize abnormal returns from their

1 Our data, which are a virtual census of transactions involving publicly traded hospital companies, show 189IO purchases of domestic, acute care NP hospitals from 1990 to 2001 (we have valid stock returns for 135 of thesetransactions). It is worth noting that IO chain acquisitions of nonprofit hospitals are currently on the rise, onceagain. In 2001, there were 22 transactions of this type, up from 11 in 2000.

2 See, “Some states taking steps to protect not-for-profit hospitals,” 11/20/96, Gannett News Service.3 More recently, this concern has been raised over conversions of nonprofit health insurers to for-profit status.

For example, the sale and conversion of the Maryland Blue Cross plan was blocked amid concerns that its assetshad been under-priced. Our evidence in this paper is confined to NP hospital sales.

4 For example, the IRS put out a training manual that urged its agents to be on the lookout for instances ofimproper inducements to sell NP hospitals, such as overcompensated employment at the acquirer. In this paperwe, in effect, test whether this type of behavior existed on a widespread basis.

A.J. Leone et al. / Journal of Health Economics 24 (2005) 113–135 115

acquisitions of NP hospitals, then this is consistent with a rigorous standard of fair marketpricing.5 Based on this standard, our event study results provide no evidence that nonprofithospitals were systematically under-priced. Our estimates of average abnormal returns arevery close to zero during the time period 1990–2001. These tests also have a reasonabledegree of power – our 95% confidence intervals suggest that themaximumlevel of under-pricing was about 10% of the average NP’s asset value. We also find that between 1990and 2001, IO acquirers earned the same or greater returns by acquiring IO and privately-held hospitals rather than NP hospitals. Overall, these results are consistent with the viewthat IOs paid “fair market value” for NP hospital assets, even before the passage of stateregulations.6

After 1997, we find that acquirer abnormal returns became significantly negative instates that adopted regulations and yet remained close to zero in states that did not adoptregulations. IO acquisitions of NP hospitals also declined in regulated states but continuedat the same rate in unregulated states. These declines are consistent with the view that stateregulations placed a floor under NP acquisition prices and/or reduced the targets’ expectedsynergistic values to acquirers.

Abnormal return estimates do not reveal the mechanisms that drive returns to zero.To help shed light on this question, we also investigate operational changes in tar-get hospitals, after they were acquired. We find that, before 1996, IO chains tended toreduce employees in their NP targets. This is consistent with the findings ofPiconeet al. (2002). Interestingly, IO acquirers rarely eliminated selected unprofitable ser-vices (e.g., AIDs care). The fact that IOs streamlined operations in their NP targetsand yet still earned negligible merger returns suggests that transacted prices were flex-ible and accurately accounted for expected synergies. After 1997, the practice of re-ducing employees in NP targets largely abated, in both regulated and unregulatedstates.

Our paper fits into a broader literature that considers the consequences of nonprofithospital conversions (Gray, 1997). Sales of nonprofit assets to IO acquirers potentially af-fect a hospital’s operational efficiency, service provision, supply of charity care and itsquality of patient care (Picone et al., 2002; Sloan, 2002). As with previous studies, wedo not provide a complete assessment of these conversions. Our study contributes to oneimportant aspect of this debate by providing evidence that 10 acquisitions of NPs do notresult in net transfers of financial resources from the community to investors. Our evi-dence suggests that they are retained by the community, often in the form of charitablefoundations.7

The paper is organized as follows. In Section2, we discuss trends in hospital acquisi-tions over the past decade and discuss the event study methodology. Section3 discusses

5 There is a well-established literature in finance that relies on announcement date abnormal returns to assessthe value implications of acquisitions, including studies that test whether bidders pay less than the expected valueof the target. Examples include:Asquith et al. (1983), Bradley et al. (1988), Jarrell and Bradley (1980), andJarrellet al. (1988).

6 These findings are also consistent with case study analyses that indicate that for-profit buyers do not underpayfor nonprofit hospitals (Sloan et al., 2000).

7 This, in theory, allows the community to fund the same level of charitable health-related activity after theacquisition (Kane, 1997).


our data and methods, while results are provided in Section4. We conclude in Section5.

2. Background and theory

2.1. Historical background

Sales of NP hospitals to IO chains were relatively rare prior to 1984. Starting in 1984,several IO chains approached financially healthy NPs with the intent of purchasing themand converting them to FP status.8 In the mid-1990s, the pace of IO acquisitions of NPhospitals accelerated. Our data indicate that IO acquisitions of NP hospitals grew from 33in the period from 1993 to 1994 to 65 in the period from 1995 to 1996. In 1996, Californiaand Nebraska passed laws regulating the sale and conversion of NP hospitals. Eventually 24states and the District of Columbia passed laws regulating nonprofit hospital and medicalservice corporation conversions (Community Catalyst, 2002). The vast majority of theselaws were passed between 1997 and 1998.

After 1997, the pace of acquisitions slowed. There were only 19 acquisitions of NPhospitals by IO chains during the period from 1998 to 1999. The role of state laws in thisdecline is one issue that we address in our analysis.

2.2. State regulations and under-pricing of NP assets

From a community perspective, the sale and conversion of nonprofit hospitals raisesseveral concerns, including whether a fair price is paid for the NP entity and whether the IOacquirer will continue to supply “needed” community services, either directly or indirectlythrough the sale proceeds.9 The state’s attorney general (AG) is charged with oversight ofthese transactions to guard against abuses and to ensure the satisfaction of various aspectsof the state’s laws concerning nonprofit entities.10

In spite of existing state AG oversight, the rapid growth in nonprofit hospital conversionsprompted calls for additional state regulations. These regulations typically: (1) extendedthe scope of reviewable transactions (e.g., to partial sales of NP assets); (2) provided foradvance notice to the state’s AG so that objections could be voiced in a timely fashion; (3)

8 Lutz and Gee (1998)cite the 1984 acquisition of the NP Wesley Medical Center by Hospital Corporationof America as a landmark acquisition in this respect. Some of the motivations behind these sales are explored inCutler and Horwitz (1998). Between 1984 and 1989, 22 charitable foundations were created from NP hospitalconversions (Grantmakers in Health, 2001).

9 In practice, the complete sale of NP hospitals in the 1990s often resulted in the creation of a charitablefoundation whose mission was to serve charitable purposes in the community, such as the general promotionof health. During the 1990s, 81 charitable foundations were created from the conversions of hospital and healthsystems (Grantmakers in Health, 2001). In some cases, distressed nonprofits were paid amounts sufficient to onlyretire their debt, with no funds left to fund a foundation.

10 For example, the AG may bring suit to apply the Cy Pres doctrine. Cy Pres requires that the proceeds fromthe sale of a nonprofit be dedicated to the original charitable purpose of the hospital or as close as possible to thispurpose.


mandated disclosure of all pertinent information to the public and the state’s AG; (4) guardedagainst private inurement of sale proceeds; (5) ensured clarity in the role of the state’s AGand guaranteed aggressive review of all such transactions; and (6) codified procedures thatwould ensure payment of a fair price for the target, such as formal appraisals (Butler, 1997;Shriber, 1997). The 1996 Nebraska law contained all of these provisions and became amodel for many state laws.

News publications frequently cited potential under-pricing of the NP entity as a moti-vation for the legislation.11 Yet, the existence of several bidders may make the bargainingefforts of NP managers and boards irrelevant.12 Even where competitive market conditionsdo not hold, an IO acquirer can increase the likelihood of a successful bid for a NP byincreasing the bid price. From the NP board’s perspective, a higher bid price can be usedto better fund a resulting charitable foundation that the board may control or point to as apositive byproduct of the merger. A higher price can also be used to pay off a financiallydistressed hospital’s bonds, so that the board can avoid the embarrassment of presidingover a bankruptcy. These considerations provide NP boards with their own incentives tonegotiate fair market prices.

2.3. Fair market prices and abnormal returns

2.3.1. Definition of fair market pricing“Fair market value” is defined by the American Society of Appraisers as “The amount

at which property would change hands between a willing seller and a willing buyer whenneither is acting under compulsion and when both have reasonable knowledge of the relevantfacts” (Pratt et al., 2000). A violation of fair market pricing occurs where, for example, theprice is influenced by “special motivations or characteristics of a typical buyer or seller”(Pratt et al., 2000). Thus, a community may reasonably expect its nonprofit board to negotiatea price corresponding to an anonymous auction of the asset to the highest bidder.

2.3.2. Abnormal returns and fair market pricingUnder certain reasonable assumptions, cumulative abnormal (stock) returns (“CARs”)

around the time of an acquisition measure the market’s assessment of the acquisition’snet effect on the acquirer’s market value.13 Where the acquirer pays less than the target’sexpected value, the abnormal return is positive. Given these assumptions acquirer CARs

11 For example, state senator Ken Hollis, who sponsored a conversion bill in Louisiana, was quoted as sayingthat the bill ensures that prospective buyers “don’t try and take control of not-for-profit hospitals at discountprices. . .” (New Orleans CityBusiness, August 11, 1997, p. 4).

12 Our interviews with the mergers and acquisitions staff at a privately owned hospital company indicate that intypical deal, a potential NP target will contact three to four potential IO acquirers and three to four potential NPacquirers, for a total of six to eight potential bidders.

13 A cumulative abnormal stock return (CAR) is a summation of the amount by which an acquiring firm’s stockreturns exceed the returns predicted by an accepted model, such as the capital asset pricing model. A formaldefinition is provided in Section3. The “reasonable assumptions” that we refer to include: (a) markets are efficientand (b) the value associated with the acquisition is not anticipated by the market prior to the measured eventwindow. In practice, CARs are subject to alternative interpretations that we consider in the paper’s empiricalsection.


may be either zero or positive.14 Where IO chains earn zero CARs from their purchases ofNP targets, we assert that this finding is consistent with the definition of fair market pricingdiscussed above. The price associated with a zero CAR must represent the highest price(net of transactions costs) that could be offered in an arm’s length auction of the target,assuming that no party would offer a price that yields a negative CAR. If the acquirer hassynergies with the target, an acquirer CAR of zero implies that the target has bargained allof the synergies into the bid price.

A positive acquirer CAR does not necessarily imply that a NP target has been sold atless than fair market value. A positive CAR may indicate that the acquirer creates uniquesynergies with the target. Even if IO chains earn systematically higher CARs by acquiringNP rather than IO targets, this may also be consistent with fair market pricing if uniquesynergies are greater among nonprofit targets. There is little that can be inferred about fairmarket prices in relation to positive acquirer CARs without making additional assumptions.

One useful assumption is that the structure of the market for acquisitions is the sameacross NP and IO targets.15 For convenience, we refer to this as the symmetry assumption.If this assumption holds, then where we observe zero CARs among the set of (IO–IO)transactions, we infer a competitive acquisition market and by the symmetry assumptionwe expect zero CARs in the (IO–NP) acquisition market as well. The symmetry assump-tion also helps to interpret the case where we observe positive acquirer CARs from bothIO and NP targets. Under this assumption, equal bargaining effort should result in equiv-alent, positive acquirer CARs across ownership classifications. The implication is that NPtargets are priced fairly if and only if acquirer CARs are no larger for NP targets than forIO targets.

2.3.3. Abnormal returns and regulationRegulations of the sale and conversion of NP hospitals have two potential effects on

the acquisition market. First, regulations may redistribute synergy value created from ac-quisitions. Second, regulations may also impose costs on acquirers and thereby reduce theexpected value of the target, ex post to the acquisition.16

If acquirer CARs are positive ex ante to regulation, then regulations may reduce bothCARs and the number of transactions. This will occur if regulations change the allocationof value and/or impose costs that reduce the expected value of the target. Where CARs arezero, ex ante, no change in observed CARs is expected post-regulation since, presumably,acquirers are not willing to accept negative CARs. However, the number of acquisitionsmay still fall if the regulations impose a price floor and/or reduce the target’s expected value,thus reducing the number of transactions that can earn a zero CAR.

14 That is, negative CARs are inconsistent with rational acquirer behavior.15 We define the structure of the acquisition market as the number of bidders as well as the distribution of target

synergy values across bidders. Where market structures are identical, this implies that the difference betweensynergy values to the top bidder and the second highest bidder are equal, for example.

16 State regulations cause delays related to notification of the state’s AG. This is one potential cost that mustbe weighed against their benefit. Some laws require the payment of fees to outside experts, as well as communityassessments and ongoing monitoring of the impact of the acquisition on community health. Assessments andmonitoring may lead to mandates to provide charity care after the merger is completed.


3. Data and methods

3.1. Data

We collect data on all acquisitions involving publicly traded IO hospitals between theyears 1990 and 2001, including both sales and purchases of single and multiple hospitalsby publicly traded hospital companies.17 We obtain a list of acquisitions for the years1994–2001, inclusive, from reports published byIrving Levin and Associates (1995–2002),which list all hospital mergers and acquisitions on an annual basis.18 These reports list thetransacting parties, their ownership status (publicly traded, private for-profit or nonprofit),the announcement date, the purchase price of the target (if disclosed) and other terms of thedeal. We subsequently verify the announcement dates.19

We use three additional data sources to extend the time frame of our study back to1990. First, to identify transactions among IO chains that occurred between 1990 and1993, we use the Securities Data Company (SDC) database, which provides records of allpublicly announced mergers and acquisitions involving publicly traded firms. SDC providesa comprehensive list of IO to IO transactions, but does not identify IO transactions withnonprofit or privately owned corporations. Second, we consult the January, 1994 publicationof Modern Healthcarewhich lists all hospital mergers during 1993. Finally, we check theAmerican Hospital Association’s annual surveys for all reported changes in ownershipstatus between 1990 and 1993. We then do a news media search of each reported change inownership to see if it is associated with an acquisition.

Using these methods, we are able to compile a near census of all hospital transactionsinvolving a publicly traded IO hospital between 1990 and 2001. We observe 394 suchtransactions in which at least one party is publicly traded.20 These 394 transactions giverise to 444 possible “events” where an “event” refers to a single party in a transaction forwhich we can observe a stock return.21

From this universe, we exclude events that coincide with an earnings announcementof the publicly traded firm, events that are announced on weekends, events for the samefirm whose event windows overlap and events for which returns data cannot be obtained.We also consolidate multiple transactions that occur on the same date into a single event.This leaves us with a final sample of 291 transactions comprising 319 events. Of thesetransactions, 40 are IO acquisitions of IO hospitals (IO–IO) (with 68 associated events),135 are IO acquisitions of one or more NP hospitals (IO–NP), 52 are NP acquisitions of IO

17 In transactions where the target is publicly traded, the sale typically represents only a fraction of the hospitalsowned by the target (e.g., Columbia/HCA sells a single hospital to Tenet). Where a NP is the target, the sale ismore likely to include all of the NP corporation’s assets and the creation of a charitable foundation.

18 These reports were first published in 1995 under the titleHospital AcquisitionReport. They are now publishedon an annual basis.

19 To verify the announcement date, we use the Factiva information service. In a small number of cases, ourFactiva search identifies a different announcement date than the one reported in the Levin and Associates reports.In these cases, we verify the correct announcement date by doing multiple media searches.

20 This number excludes swaps of hospitals and transactions involving hospitals outside of the United States.21 In (IO–IO) transactions, we observe two events associated with a single transaction, provided that we observe

the stock return of both the acquirer and target. In all other transactions, we observe a single event.


Table 1Summary of hospital acquisitions

Acquisitions Sales Total

IO–IO IO–NP IO–PR IO–IO NP–IO PR–IO

Panel A: summary of acquisitionsa

Columbia 6 37 4 5 16 7 75Tenet Healthcare 5 16 1 3 5 3 33Quorum 1 15 1 5 22Health Management Associates 2 13 4 1 1 21Vencor 6 7 7 20OrNda 2 4 3 2 3 2 16HCA 1 1 2 6 5 15Community Health Systems 2 9 4 15Province Healthcare 1 10 2 13Universal Health Services 3 8 1 12Healthtrust 1 5 2 1 9Triad 1 1 4 3 9American Medical Holdings 2 1 3 2 8Lifepoint 2 1 3 1 7Paracelsus Healthcare 1 1 2 3 7New American Healthcare 1 1 2 1 1 6Galen 4 4HealthSouth 2 1 1 4Humana 1 1 1 1 4National Medical Enterprises 3 1 4

IO–IO IO–NP IO–PR NP–IO PR–IO Total

Panel B: frequency of transaction type by yearb

1991 0 2 3 0 0 51992 4 2 2 0 1 91993 6 4 1 2 2 151994 6 5 4 1 4 201995 3 26 6 4 0 391996 5 36 8 4 5 581997 2 20 1 7 0 301998 3 9 1 8 1 221999 2 5 0 12 11 302000 5 5 3 10 3 262001 4 21 4 4 4 37

40 135 33 52 31 291a Panel A contains a list of the investor-owned hospitals in our sample with four or more transactions. These

firms account for over 95% of the transactions in our sample. Cell entries reflect the number of transactions ofeach type by each acquirer.

b Panel B presents the frequency of transaction types between 1991 and 2001.

hospitals (NP–IO), 33 are IO acquisitions of for-profit, privately owned hospitals (IO–PR),and 31 are privately owned company acquisitions of IO hospitals (PR–IO). The list of theincluded IO acquirers and targets, along with the number of acquisitions and sales, by year,is reported inTable 1.


We obtain financial and operating information for this sample, including daily stockreturns for the IO corporation, acquisition prices for the deals, as well as net income,staffed beds, employees and service profiles of the target hospital. Daily stock returns areobtained from the Center for Research on Security Prices (CRSP) files. Acquisition pricesare obtained from the Levin reports, the SDC data base, Modern Health Care or from a mediasearch, depending on the original data source for the merger. Net income for the target isobtained from the Medicare Cost Reports. Staffed beds, employees and service profiles forthe targets are collected from the Hospital Blue Books (Billian Publishing, 1995–2003) andthe AHA data base, depending on the date of the merger.

Finally, we compile a list of enactment dates for state legislation regulating the sale andconversion of nonprofit hospitals. From this list, we create a series of dummy variables toindicate whether each of our transactions occurred in a state where legislation had previouslybeen enacted.Table 2provides a list of the 24 states that enacted legislation, along withenactment dates, key provisions of the legislation and the number of (IO–NP) acquisitionsthat occurred before and after enactment.

3.2. Methodology

3.2.1. Tests using cumulative abnormal returnsWe use estimates of the cumulative abnormal returns (CARs) of IO acquirers to test

the null hypothesis of fair market pricing, before and after regulation. Abnormal returnsfrom (IO–NP) transactions that: (a) are significantly greater than zero and (b) exceed thereturns earned from transactions within other target ownership categories (i.e., (IO–IO) or(IO–PR)) allow us to reject the null hypothesis of fair market pricing. Similarly, reductionsin both CARs and the number of (IO–NP) transactions after the imposition of regulationsallow us to reject the null hypothesis that regulations had no effect on the returns availablein the (IO–NP) acquisition market.

We employ a standard market model to calculate cumulative abnormal returns. Giventhat our study is industry-specific, we add industry returns as a second factor to the marketmodel. We first estimate market model parameters for each acquiring firm using daily returnsover the periodt=−180 tot=−5.22 Firm i’s cumulative abnormal return is then computedby summing ARit over the event period.23 Standard errors of the means are computed usinga methodology described byMacKinlay (1997).

22 Abnormal returns are estimated as follows:

ARit = Rit − (ai + b1iRmt + b2iRHt)

where ARit is the abnormal returns of firmi on dayt,Rit the return of firmi on dayt, ai the estimated intercept termof firm i, b1i, b2i the estimated slope coefficients of market returns and healthcare industry returns, respectively,Rmt the daily return on the CRSP equally weighted index on dayt, andRHt is the daily return on an equallyweighted index of healthcare industry stocks on dayt.

23 As a robustness check, we examine CARs of multiple window lengths, ranging from 3 to 6 days. Arguably, a3-day window length provides the most accurate measure of information specific to the hospital acquisition. Longerwindow lengths, however, allow for more gradual transmission of information to the market, but are confoundedwith other non-transaction specific events.


Table 2Summary of state regulations

State Yearadopted

Communityassessment/accessto care provisiona

Independentvaluationrequireda

Transactionsbefore lawpassed

Transactionsafter lawpassed

Arizona 1997 × 1 0California 1997 × 7 3Colorado 1998 1 0Connecticut 1997 × 0 0District of Columbia 1997 × × 2 0Georgia 1997 × 3 2Hawaii 1998 × 0 0Idaho 2000 0 0Louisiana 1997 × 2 3Maine 1997 × × 0 0Maryland 1998 × 0 0Massachusetts 1998 × 1 0Nebraska 1996 × 2 0New Hampshire 1997 0 0New Jersey 1997 × 1 1New Mexico 1999 0 0North Carolina 1998 × 2 0Ohio 1999 × 5 1Oregon 1997 × 0 0Rhode Island 1997 × 2 0South Dakota 1997 0 0Vermont 1997 × 0 0Virginia 1997 2 2Washington 1997 × 0 0Wisconsin 1997 0 1

Sum 31 13

This table summarizes the state laws passed to regulate the sale and conversion of nonprofit hospitals. We list, first,the state and the year of adoption. The “community assessment/access to care provision” column indicates whetherthe law requires the regulator or transacting parties to consider the health impact of the acquisition, the acquirer tosubmit a community benefit plan and/or requires the regulator to monitor the impact of the acquisition on healthcare. The independent valuation column indicates whether the regulator is required to obtain an independentvaluation of the target prior to granting approval. The “transactions before” and “transactions after” columnsindicate the number of (IO–NP) transactions that occurred in our data set before and after the state law was passed.

a The information in these two columns (columns 3 and 4) is taken from the Community Catalyst (2002).

Because univariate analyses of CARs provide our main tests of fair market pricing, itis useful to establish their statistical power given a stipulated level of under-pricing. To doso, we assess the relationship between any observed CAR and the level of under-pricing.In our sample, the mean value of NP target assets to IO acquirer market capitalization is6%. If each dollar of the NP target’s assets is worth one dollar of market capitalization (i.e.,equivalent to a Tobin’s “q” of 1 for NP assets), then the IO acquirer would realize a 6%CAR if it acquired the NP target for free and implemented no unique synergies.24 More

24 In this case, a “unique synergy” occurs where the acquirer implements a synergy that cannot be matched bya competing bidder. This allows the acquirer to capture some of the synergy in its CAR.


realistically, the acquirer would realize a CAR of 1% given 16.67% under-pricing and nounique synergies.

To establish an upper bound on the degree of under-pricing, ex post, we provide 95%confidence boundaries around the estimated CARs in each transaction class. For example,suppose that we estimate with 95% confidence that the upper bound on CARs from IOacquisitions of NPs is 1%. Then, again using the ratio of NP target assets to IO acquirermarket value of 6%, we would conclude with 95% confidence that the average extent ofunder-pricing wasno greaterthan 16.67% of NP asset value.25,26

We also provide tests and confidence bands for the hypothesis that (IO–NP) returns areno greater than (IO–IO) or (IO–PR) returns. Under the market symmetry assumption, testsof this hypothesis provide a second way to judge whether asset sales were fair.

3.2.2. Limitations of event studiesThe abnormal returns derived from event studies may be interpreted in a variety of ways.27

If we find that acquiring firms earn no abnormal returns at the time of their acquisitions,then this is consistent with the hypothesis that NP targets were priced fairly. However,zero abnormal returns may also occur where the market has already anticipated the target’sgains from its acquisition because the firm has announced an acquisition program and theterms of subsequent deals (both synergies and acquisition prices) conform to the priorexpectations of the market (Schipper and Thompson, 1983). The market may anticipate thegains if the acquisition and its terms are leaked prior to the announcement date. Second,abnormal returns may also reflect other information that the acquisition process reveals. Forexample, an acquisition of a NP hospital may provide a negative signal about the availabilityof internally-generated investment opportunities. This would also bias returns downwards.Moreover, the same critique applies to the analysis of CARs in the aftermath of regulation.

In Section4, we provide several tests of robustness in an attempt to rule out alterna-tive explanations for our results. To address Schipper–Thompson concerns, we re-estimateour results using only the first acquisition made by each acquirer. If Schipper–Thompsonconcerns are relevant, initial transactions should be more informative than later transactions.

3.3. Analyses of prices, operations, and service reductions

In order to provide additional information about the circumstances surrounding acquisi-tions, we also collect data on acquisition prices, as well as changes in the target hospital’s

25 The 16.67% figure in this example is an upper limit on the magnitude of under-pricing because it assumesthat the entire 1% CAR is due to under-pricing as opposed to, for example, unique synergies that are captured bythe bidder.

26 We can also assess the power of our univariate tests, ex ante, by consulting simulation results provided byMacKinlay (1997). For event windows with estimated standard errors similar to ours, the likelihood of detecting a2% CAR using a sample of 100 events is virtually 100%. There is a 96% likelihood of detecting a 1.5% CAR and71% likelihood of detecting a 1% CAR. Because we have in excess of 100 (IO–NP) events, we are confident thatour methodology is able to detect even reasonably small abnormal returns that resulted from IO chains acquiringNP entities.

27 SeeFama et al. (1969), Brown and Warner (1985), andMacKinlay (1997)for discussions of the methodolog-ical issues surrounding event studies.


service profile, staffed beds, and employees, before and after being acquired. To definea service profile, we focus on six distinct services that we judge to be among the leastprofitable services that hospitals supply.28

This information provides some insight into how acquiring firms create synergies withthe target. For example, we are interested in knowing whether IO acquisitions of NP targetsare characterized by significant reductions in beds, employees, and unprofitable services.Furthermore, observed reductions in services, beds, and employees may be of interest toa general audience that wishes to evaluate the distributional impact of the hospital merg-ers. Even where acquirers pay fair market value, service reductions in some areas (e.g.,AIDs) may disproportionately impact some vulnerable groups if the community (e.g., newfoundation) does not spend the sale proceeds on services for these groups.

The analysis of transaction prices and operational data also provides insights into howefficiently transaction prices are determined. If markets are efficient and buyers and sellers ofhospitals are well-informed, then where the acquirer reduces services, beds, and employees,these decisions will be capitalized into the price of the acquisition (e.g., the target willdemand and receive a higher price if it knows that the acquirer plans to strip out unprofitableservices). Thus, a finding of zero CARs where there are operational improvements suggeststhat transacted prices reflect these expected synergies and vice versa.

4. Results

4.1. Fair market pricing in the absence of regulation

4.1.1. Abnormal returnsTable 3presents univariate statistics for the cumulative abnormal returns of IO acquirers

around the time of acquisitions. Event windows ranging from 3 to 6 days are reported.Observations are limited to events that either occurred prior to the passage of state regu-lations or occurred in states that did not pass regulations. Panel A presents results poolingtransactions across all years of our study and panel B presents results divided into pre- andpost-1997 periods.

Examining the results in Panel A, we cannot reject the null hypothesis that IOs paidfair market prices for NP assets, based on the standard of a zero CAR. Across all years,IO chains’ returns from acquisitions of NPs are not statistically different from zero, andrange from−0.004% to−0.13% across all four-event windows (see Panel A, column (3)).Average CARs for IO acquisitions of other IO hospitals range from 0.45% to 0.95% and arealso indistinguishable from zero, while acquisitions of privately owned hospitals resultedin the greatest abnormal returns, ranging from 1.11% to 2.12%.

Our results allow us to place confidence intervals on themaximum levelof abnormalreturns that IOs may have earned. The size of our confidence intervals is an indication of theability of our test procedure to detect abnormal returns (and, by implication, under-pricing).

28 Specifically, we track the following services: (1) AIDs; (2) adult psychiatric services; (3) adolescent psychi-atric services; (4) alcohol and drug abuse services; (5) rehabilitation, and (6) trauma care services. We intervieweda hospital executive to help generate this list of services.

A.J.L

eoneetal./Jo

urnalofH

ealth

Economics

24(2005)113–135

125Table 3Comparison of cumulative abnormal returns to IO acquirers by target type

Window Type 1, IO–IO (n= 34) Type 2, IO–PR (n= 29) Type 3, IO–NP (n= 122) Type 1− Type 3 (difference)

Panel A(−1,+1) 0.45% (−1.32%, 2.21%) 1.20% (−0.69%, 3.10%) −0.10% (−0.92%, 0.72%) 0.55% (−1.40%, 2.49%)(−1,+2) 0.95% (−1.12%, 3.03%) 2.12%* (−0.07%, 4.31%) −0.13% (−1.08%, 0.82%) 1.09% (−1.19%, 3.36%)(−1,+3) 0.52% (−1.76%, 2.80%) 1.11% (−1.25%, 3.47%) −0.004% (−1.05%, 1.05%) 0.52% (−1.99%, 3.03%)(−1,+4) 0.54% (−2.00%, 3.08%) 1.86% (−0.82%, 4.53%) −0.04% (−1.20%, 1.13%) 0.58% (−2.21%, 3.37%)

Type 1 IO–IO (n= 24) Type 2 IO–PR (n= 24) Type 3 IO–NP (n= 75)

Panel B1990–1996

(−1,+1) 1.63%* (−0.32%, 3.68%) 1.68% (−0.32%, 3.68%) 0.02% (−0.75%, 0.79%) 1.61% (−0.75%, 0.79%)(−1,+2) 2.16%* (−0.03%, 4.34%) 2.55%** (0.24%, 4.86%) −0.18% (−1.07%, 0.71%) 2.34%* (−0.02%, 4.70%)(−1,+3) 2.01% (−0.43%, 4.46%) 1.78% (−0.75%, 4.31%) −0.27% (−1.27%, 0.72%) 2.28%* (−0.35%, 4.92%)(−1,+4) 1.50% (−1.11%, 4.67%) 2.15% (−0.68%, 4.98%) −0.28% (−1.37%, 0.81%) 1.78% (−1.11%, 4.67%)

Type 1 IO–IO (n= 10) Type 2 IO–PR (n= 5) Type 3 IO–NP (n= 47)

Panel B1997–2001

(−1,+1) −2.44% (−7.20%, 1.52%) −1.55% (−6.99%, 3.90%) −0.30% (−2.05%, 1.46%) −2.14% (−6.47%, 2.18%)(−1,+2) −2.36% (−8.24%, 2.48%) −0.36% (−6.64%, 5.93%) −0.05% (−2.07%, 1.98%) −2.31% (−7.56%, 2.94%)(−1,+3) −3.14% (−8.39%, 4.18%) −2.10% (−8.39%, 4.18%) 0.42% (−1.79%, 2.64%) −3.56% (−9.13%, 2.00%)(−1,+4) −2.10% (−6.39%, 3.83%) 0.19% (−7.51%, 7.88%) 0.37% (−2.14%, 2.88%) −2.47% (−8.91%, 3.97%)

This table presents the mean cumulative abnormal returns (CARs) to the acquirer calculated over varying window lengths, ranging from 3 to 6 days, by target ownershiptype. The sample is restricted to those transactions occurring either in states without regulation or in states prior to regulation being passed. Panel A presents meanCARs for our complete sample while Panel B presents mean CARs separately for two time windows, before and after 1997. The naming convention used to identifythe type of transaction is abbreviated as “IO” for investor-owned for-profit organizations, “PR” for private for-profit organizations and “NP” for nonprofit organizations.The abbreviation before the – indicates the form of the acquirer and the abbreviation following the – gives the organizational form of the target. Column 4 shows thedifference between CARs where the target is IO vs. NP. Statistically significant results are in bold. 95% confidence intervals are given in parentheses.

∗ Statistically significant at the 0.10 level in two-tailed tests.∗∗ Statistically significant at the 0.05 level in two-tailed tests.


Focusing on (IO–NP) transactions in Panel A and using the 3-day window, there is a 95%likelihood that IO acquisition returns fell between−0.92% and 0.72%. If we consider aone-sided test of maximum positive returns, there is a 95% likelihood that the average(IO–NP) transaction earned less than a 0.58% abnormal return.29 Following our discussionin Section3, a 0.58% abnormal return corresponds to amaximumof 9.67% under-pricingfor an average transaction, using the strictest standard of fair market pricing that requires azero CAR.

Column (4) in panel A presents the differences between IO acquirer returns earned fromIO versus NP targets. While none of the differences is statistically significant, we are 95%confident that the difference in (IO–IO) and (IO–NP) returns lies between−1.40% and2.49%. Using the difference in reported returns, together with its estimated standard error,we compute a 71% likelihood that IOs earned a greater return by acquiring other IOs ratherthan NPs.30 Doing a similar computation, we compute an 89% likelihood that IOs earneda greater return by acquiring privately owned hospitals compared to NP hospitals. Theseresults also support the hypothesis of fair market pricing, using the standard ofrelativereturns (see Section2).

Panel B divides our results into pre- and post-1997 time periods. We divide the samplein this way to allow for structural differences in the pattern of returns over time.31 Panel Bshows that, prior to 1997, IO acquirers earned statistically significant positive CARs whenacquiring IO or privately owned targets, but still realized negligible returns when acquiringNPs.T-tests of the differences listed in column (4) are statistically significant for 4- and5-day windows. In this case, 95% confidence intervals for individual classes of transactionsare larger than in Panel A, given the reduction in observations.

After 1997, IO returns from acquiring IO or privately owned hospitals diminished andbecame negative but statistically indistinguishable from zero. Abnormal returns from pur-chasing NP targets remained close to and indistinguishable from zero in the post-1997period as well. As a result, after 1997, there is no statistical difference in the returns that 10chains earned from purchasing hospitals from any of the three ownership classifications.

These results suggest that NPs probably received fair market value for their assets, onaverage. While it is possible that individual instances of severe under-pricing occurred, ourtests indicate that the extent of under-pricing cited in the Sharp Memorial Hospital exampleappears not to have been widespread or systematic.

Table 4provides evidence on IO returns where the IO chainsellsone or more hospitals todifferent organizational types. We present these results as a check on the validity of acquirerCARs presented inTable 3. Consistent with prior literature, we wish to show that in periodswhere acquirers earned positive CARs, targets did as well. Panel B ofTable 4shows thisto be the case. In the period prior to 1997, where IO acquirers earned positive CARs from

29 That is, 95% of the statistical distribution of mean returns is in the region less than 0.58%.30 That is, the estimated (positive) difference between (IO–IO) and (IO–NP) returns is equal to 0.55 times the

standard error of the estimated difference. Using the normal distribution, this leaves about 29% of the distributionin the region less than zero.

31 There are several reasons to expect, a priori, differences in the pattern of returns over time. These include: (1)the drop in the number of acquisitions that occurred after 1997 that may have been due to reasons other than stateregulations and (2) the legal problems that Columbia/HCA had with the government starting in 1997 that causedit to halt its acquisition program.

A.J.L

eoneetal./Jo

urnalofH

ealth

Economics

24(2005)113–135

127Table 4Comparison of cumulative abnormal returns to IO targets by acquirer type

Window Type 1 IO–IO (n= 34) Type 2 PR–IO (n= 27) Type 3 NP–IO (n= 37) Type 1− Type 3 (difference)

Panel A(−1,+1) 2.85% (−1.06%, 6.76%) −0.79% (−3.37%, 1.78%) −1.20% (−2.86%, 0.46%) 4.05%* (−0.20%, 8.30%)(−1,+2) 2.07% (−2.45%, 6.59%) −0.43% (−3.41%, 2.54%) −0.86% (−2.80%, 1.09%) 2.92%* (−2.00%, 7.84%)(−1,+3) 2.82% (−2.23%, 7.87%) 0.28% (−2.98%, 3.54%) −1.79% (−3.90%, 0.33%) 4.61%* (−0.87%, 10.08%)(−1,+4) 2.82% (−2.83%, 8.47%) 0.70% (−3.01%, 4.42%) −1.45% (−3.83%, 0.93%) 4.27% (−1.86%, 10.40%)

Type 1 IO–IO (n= 24) Type 2 PR–IO (n= 12) Type 3 NP–IO (n= 11)

Panel B1990–1996

(−1,+1) 4.81%*** (2.07%, 7.55%) 0.94% (−1.28%, 3.16%) −2.86%** (−5.31%,−0.41%) 7.67%*** (4.00%, 11.35%)(−1,+2) 3.58%** (0.42%, 6.75%) 0.50% (−2.06%, 3.06%) −2.68% (−5.65%, 0.28%) 6.26%*** (1.93%, 10.60%)(−1,+3) 4.81%*** (1.27%, 8.35%) 2.45%* (−0.41%, 5.32%) −3.41%** (−6.57%,−0.25%) 8.22%*** (3.47%, 12.97%)(−1,+4) 4.50%** (0.62%, 8.37%) 2.07% (−1.07%, 5.21%) −2.57% (−6.20%, 1.06%) 7.06%*** (1.75%, 12.38%)

Type 1 IO–IO (n= 10) Type 2 PR–IO (n= 15) Type 3 NP–IO (n= 26)

Panel B1997–2001

(−1,+1) −2.19% (−13.4%, 9.0%) −2.27% (−6.6%, 2.1%) −0.47% (−2.6%, 1.7%) −1.73% (−13.1%, 9.68%)(−1,+2) −1.83% (−14.8%, 11.1%) −1.23% (−6.2%, 3.8%) −0.13% (−2.6%, 2.3%) −1.70% (−14.9%, 11.47%)(−1,+3) −2.29% (−16.7%, 12.2%) −1.46% (−6.9%, 4.0%) −1.10% (−3.8%, 16%) −1.19% (−15.9%, 13.52%)(−1,+4) −2.19% (−19.3%, 14.9%) −0.56% (−6.93%, 5.80%) −1.00% (−4.01%, 2.01%) −1.19% (−18.6%, 16.18%)

This table presents the mean cumulative abnormal returns (CARs) to the target calculated over varying window lengths, ranging from 3 to 6 days, by acquirer ownershiptype. The sample is restricted to those transactions occurring either in states without regulation or in states prior to regulation being passed. Panel A presents meanCARs for our complete sample while Panel B presents mean CARs separately for two time windows, before and after 1997. The naming convention used to identifythe type of transaction is abbreviated as “IO” for investor-owned for-profit organizations, “PR” for private for-profit organizations and “NP” for nonprofit organizations.The abbreviation before the – indicates the form of the acquirer and the abbreviation following the – gives the organizational form of the target. Column 4 shows thedifference between CARs where the acquirer is IO vs. NP. Statistically significant results are in bold. 95% confidence intervals are given in parentheses.


∗∗∗ Statistically significant at the 0.01 level in two-tailed tests.


acquiring IO targets, the IO targets earned positive residuals as well.Table 4also indicatesthat overall and especially before 1997, IO targets realized significantly higher returns wherethey were purchased by IO rather than NP acquirers.

4.1.2. Tests of robustness using event study dataWe also test the sensitivity of these findings by further dividing the sample in several

ways. Because Columbia provides 24% of our overall sample observations (75 out of 319events), we compute our results excluding the Columbia data. The results are very similarto Table 3. We also divide the sample according to whether the target is: (1) financiallydistressed (has negative net income); (2) is part of a multi-facility purchase (more thanone hospital purchased); (3) has more or less than $1 billion in assets; or (4) disclosed theacquisition price. In no case does the IO acquirer earn a statistically significant CAR byacquiring a NP facility, either before or after 1997, regardless of how the sample is divided.32

A second set of tests focuses on Schipper–Thompson concerns. As noted earlier, abnor-mal returns may have previously been capitalized into the acquirers’ share prices (Schipperand Thompson, 1983). As a result, it is important to know whether the first acquisition madeby each acquirer yields a higher CAR than later acquisitions. If so, this would bias the testsof all abnormal returns (including those for (IO–NP) acquisitions) towards zero.

To address these concerns we select only the first acquisition made by each acquirer andrepeat the analysis ofTable 3(now excluding transactions with privately owned targets).Table 5presents the average level of abnormal returns for an acquirer’s initial acquisition. InPanel A, we present results for only the first acquisition of each acquiring firm, regardless ofthe target’s ownership type (i.e., for some acquirers, the initial acquisition may be a NP whilefor others the first acquisition may be an IO). Each acquirer contributes one observationto panel A. In panel B, we show results for each acquirer’s first acquisition of both targettypes (i.e., the acquirer’s first acquisition of an 10 and its first acquisition of a NP). Eachacquirer may contribute up to two observations to panel B. These results allow us to testwhether the abnormal returns from early stage acquisitions are more positive and whetherthis effect differs by the ownership classification of the target.

Table 5indicates that there is some basis for Schipper–Thompson concerns, but thatthe effect is more pronounced for IO targets. ComparingTable 3with Table 5, we findthat initial (IO–IO) acquisitions have higher average CARs compared to the overall set of(IO–IO) acquisitions reported inTable 3. This is consistent with the Schipper–Thompsoncritique since these returns exceed the averages realized across all (IO–IO) transactions.

Initial (IO–NP) acquisitions still have average CARs that are indistinguishable from zero.However, they are larger, on average, than we found inTable 3. In panel A, the CARs rangefrom 0.59% to 1.59%, while in panel B they range from 0.64% to 1.48%. More importantly,the upper range of the confidence interval generally exceeds 3% and is as high as 6.52%(see panel A, 6-day window).

In light of these findings, it is worth reconsidering the question of whether IOs mayhave made abnormal returns from acquiring NPs. If we assume that the initial acquisitioncapitalizes the expected earnings from all subsequent transactions, then a maximum CARof 6.52% must be evaluated relative to a base of all hospital assets acquired subsequently.

32 Results of these additional analyses are available from the authors upon request.


Table 5Comparison of cumulative abnormal returns to IO acquirers for the first acquisition

Window Type 1 IO–IO (n= 10) Type 2 IO–NP (n= 11) Type 1− Type 2 (difference)

Panel A(−1,+1) 4.02%** (0.09%, 7.94%) 0.73% (−2.75%, 4.21%) 3.28% (−1.96%, 8.53%)(−1,+2) 4.58%** (0.05%, 9.11%) 0.59% (−3.43%, 4.61%) 3.99% (−2.07%, 10.0%)(−1,+3) 4.86%* (−0.20%, 9.93%) 1.42% (−3.07%, 5.92%) 3.44% (−3.33%, 10.2%)(−1,+4) 4.55% (−1.00%, 10.1%) 1.59% (−3.33%, 6.52%) 2.96% (−4.46%, 10.4%)

Type 1 IO–IO (n= 16) Type 2 IO–NP (n= 20)

Panel B: comparison of FP and NP first acquisition(−1,+1) 3.56%** (0.75%, 6.37%) 0.64% (−1.84%, 3.12%) 2.92% (−0.83%, 6.7%)(−1,+2) 3.88%** (0.63%, 7.13%) 0.86% (−2.01%, 3.72%) 3.02% (−1.31%, 7.4%)(−1,+3) 3.17%** (−0.46%, 6.80%) 1.48% (−1.72%, 4.69%) 1.69% (−3.15%, 6.5%)(−1,+4) 3.61%* (−0.36%, 7.59%) 1.45% (−2.06%, 4.96%) 2.16% (−3.14%, 7.5%)

This table presents the mean cumulative abnormal returns (CARs) to the acquirer calculated over varying windowlengths, ranging from 3 to 6 days, by target ownership type. The sample includes only the first acquisition ofeach acquirer or the first acquisition by ownership class of acquisition target. Panel A presents mean CARsfor each acquirer’s first acquisition while panel B presents the CARs for each acquirer’s first acquisition in eachownership class. The naming convention used to identify the type of transaction is abbreviated as “IO” for investor-owned for-profit organizations, “PR” for private for-profit organizations and “NP” for nonprofit organizations.The abbreviation before the – indicates the form of the acquirer and the abbreviation following the – gives theorganizational form of the target. Column 4 shows the difference in CARs where the target is IO vs. NP. Statisticallysignificant results are in bold. 95% confidence intervals are given in parentheses.


In our data base, there are 20 acquirers of NP assets. These acquirers made, on average, sixseparate acquisitions of NP hospitals. In this case, a 6.5% return, if realized from under-pricing of NP assets alone, would constitute an average CAR, per target, of about 1.1% pertransaction, which is similar to the upper limit estimated inTable 3, Panel A, column 3.Thus, in spite of the Schipper–Thompson effects, the conclusions drawn from the results inTable 3continue to hold.

4.1.3. Other tests of robustness using operational changesTable 6provides some additional insights into our findings. We investigate three cate-

gories of operational changes in target hospitals 1 year before and 2 years after the acqui-sition: changes in employees, changes in the number of staffed beds, and changes in thenumber of services offered from a set of six services that are believed to be less profitablethan most services. We are interested in whether IO acquirers made similar operationalchanges when acquiring IO or NP targets (columns 1 and 2). We are also interested in howthese changes compare to the changes measured in a control sample of non-acquired NPsduring the same time period (column 3).33

33 We construct a matched sample of NP hospitals that were not acquired by matching on the year, state andbed size of the acquired hospital sample (matched hospital bed sizes were within 25% of the acquired hospital bedsizes). In assessing the significance levels inTable 6, we also employ a Bonferoni correction to account for a lackof prior hypotheses. Bonferoni corrections are also applied inTables 7 and 8.

130A.J.L

eoneetal./Jo

urnalofH

ealth

Economics

24(2005)113–135

Table 6Change in hospital operations betweenT− 1 andT+ 2 by acquisition type

Description [1], IO–IO [2], IO–NP [3], NP [4], [1]− [2] [5], [2] − [3]

N Mean N Mean N Mean Difference Difference

Panel AChange in employees 17 46.65 (1.02) 105 −43.18 (−1.99) 115 54.42*** (3.00) 89.83 (1.78) −97.60*** (−3.50)Change in number of beds 20 2.75 (0.33) 115 −7.60* (−2.13) 115 1.27 (0.19) 10.35 (1.14) −8.87 (−1.20)Change in count of selected services 20 0.00 (0.00) 115 −0.03 (−0.23) 42 0.10 (1.92) 0.03 (−0.08) −0.13 (1.43)

Description [1], IO–IO [2], IO–NP [3], NP [4], [1]− [2] [5], [2] − [3]

N Mean N Mean N Mean Difference Difference

Panel B1990–1996

Change in employees 6 16.50 (0.43) 57 −71.56 (−2.02) 64 53.59 (1.91) 88.06 (1.69) −125.16** (−2.83)Change in number of beds 7−9.43 (−0.57) 64 −10.06 (−1.80) 64 0.25 (0.02) 0.63 (0.04) −10.31 (0.63)Change in count of selected services 7−0.43 (−0.63) 64 0.05 (0.26) 64 0.06 (0.73) −0.48 (−0.67) −0.02 (−0.80)

Panel B1997–2001

Change in employees 11 63.09 (0.92) 48 −9.48 (−0.44) 51 55.45** (2.63) 72.57 (1.01) −64.93* (−2.14)Change in number of beds 13 9.31 (1.01) 51 −4.51 (−1.14) 51 2.55 (0.68) 13.82 (1.38) −7.06 (−1.36)Change in count of selected services 13 0.23 (0.71) 51 −0.12 (−0.97) 51 0.16** (2.68) 0.35 (1.01) −0.27 (−2.04)

This table presents mean changes in the number of hospital employees, the number of hospital beds, and in a select number of services offered by the target hospital,measured 1 year before and 2 years after the acquisition. The data are obtained from The Hospital Blue Book and the AHA data base. Panel A presents mean differencesfor our complete sample while Panel B presents mean differences separately for two time windows, before and after 1997. The naming convention used to identify thetype of transaction is abbreviated as “IO” for investor-owned for-profit organizations, “PR” for private for-profit organizations and “NP” for nonprofit organizations.The abbreviation before the – indicates the form of the acquirer and the abbreviation following the – gives the organizational form of the target. Column 4 shows thedifferences between IO purchases of IO and NP hospitals, while column 5 shows the difference between IO purchases of NP and a sample of NP control hospitals.T-statistics are in parentheses. Statistically significant results are in bold. Bonferoni adjustments are made to critical values for tests of statistical significance.

∗ P< 0.10 (two-tailed test), for test of difference from 0.∗∗ P< 0.05 (two-tailed test), for test of difference from 0.

∗∗∗ P< 0.01 (two-tailed test), for test of difference from 0.


Panel A shows the results for the set of all unregulated acquisitions, across all years. Mostof the results are insignificant. Among our significant results, we find that the control sampleof NP hospitals added employees during this period, leading to a significant difference indifference comparison with the NP targets (column 5). This suggests that acquisitions ofNP targets by IO acquirers leads to reductions in employees, relative to a control groupof non-acquired NPs. Second, IOs generally reduced beds when acquiring NP targets. Inthis case, however, the differences in differences comparisons listed in columns 4 and 5 areinsignificant.

When these results are divided across time period (Panel B), the difference in differenceresult for change in employees persists across both time intervals. These finding providesome evidence that IO acquirers of NP hospitals employ fewer individuals than do NPhospitals. Moreover, the difference in difference in employment practices is reasonablylarge (about 98 employees in Panel A, column 5). The patterns of employment changethat we observe may indicate that IOs have different standards of efficiency than NPs.This may be necessary to compensate for their differential tax burden. Previous results, inTables 3–5, indicate that any systematic difference in employment practices does not resultin positive CARs for (IO–NP) acquisitions. If IOs are able to generate efficiencies and profitsfrom reducing employees in NP targets, then our results suggest that these efficiencies arereflected in acquisition prices.34

4.2. Abnormal returns following the passage of regulations

4.2.1. Acquirer abnormal returns for nonprofit targetsTable 7presents acquirer CARs and the number of (IO–NP) transactions where the

sample is divided according to whether states adopted regulations. Given that acquirer CARswere close to zero prior to regulation, we do not anticipate significant reductions in acquirerCARs after regulation. Instead, we use changes in the number of acquisitions as a morereliable indicator of the effect of regulation. We find that the number of NP acquisitionsfell significantly in regulated states, from 31 to 13. On the other hand, the number ofNP acquisitions in unregulated states showed only a small decline, falling from 48 to 43after 1997. It appears that the decline in NP acquisitions after 1997 was largely confinedto states that adopted regulations. The fall in the number of acquisitions suggests thatregulations may have increased the price of NP targets and/or reduced their expected synergyvalue.35

We also find that CARs fell in states that adopted regulation, becoming statisticallynegative within the 3-day window. If we take into account the fact that 3-day windowreturns were also falling slightly in non-regulated states, the inferred effect of regulationis slightly less negative. Column 8 provides this difference in difference result. The results

34 We also have statistics on acquisition prices paid for IO and NP targets by IO acquirers (not shown here). Ourdata show no statistically significant difference in these prices, based on nonparametric tests of median transactedprices.

35 Of course, this is not the only possible explanation. It is conceivable that the supply of “viable” targetsdeclined disproportionately in regulated states. We revisit this issue in Section4.2.3.


Table 7Analysis of acquisitions and acquirer CARs in states with and without regulation (NP targets only)

Window Non-regulated states Regulated states

Pre-1997 Post-1997 Before After[1], n= 48 [2],n= 43 [1],n= 31 [4],n= 13

Mean values of acquirer CARs by state regulatory status(−1,+1) 0.039% (0.08) −0.308% (−0.32) −0.039% (−0.07) −3.063%* (−2.36)(−1,+2) −0.231% (−0.39) 0.036% (−0.03) −0.201% (−0.32) −1.902% (−1.27)(−1,+3) −0.288% (−0.43) 0.297% (−0.24) −0.231% (−0.52) −0.231% (−0.52)(−1,+4) −0.128% (−0.18) 0.503% (−0.36) −0.601% (−0.77) −1.099% (−0.60)

[5], [3] − [1] [6], [4] − [2] [7], [4] − [3] [8], [6] − [5]

Differences across regulated and non-regulated states(−1,+1) −0.078% (−0.10) −2.755% (−1.69) −3.024% (−2.15) −2.678% (−1.50)(−1,+2) 0.030% (−0.03) −1.938% (−1.03) −1.701% (−1.04) −1.968% (−0.95)(−1,+3) 0.057% (−0.06) −1.128% (−0.55) −0.600% (−0.34) −1.185% (−0.52)(−1,+4) −0.473% (−0.44) −1.603% (−0.70) −0.498% (−0.25) −1.130% (−0.44)

This table presents the number of acquisitions and the mean cumulative abnormal returns (CARs) for acquirers ofnonprofit target hospitals. The sample is divided into states that did and did not pass regulations. For transactionsthat occurred in states without any regulation, the sample is further divided into pre- and post-1997 periods, whichcorresponds with the introduction of laws in states, which chose to adopt them. For transactions that occurred instates that passed regulations, the sample is divided into pre- and post-regulation periods. Differences in CARsare provided in the second half of the table. Statistically significant results are in bold.T statistics are given inparentheses. Bonferoni adjustments are made to critical values for tests of statistical significance.

∗ P< 0.10 (two-tailed test), for test of difference from 0.

in column 8 are insignificant across all event windows but still range from−1.13% to−2.67%.36

4.2.2. Changes in services provided before and after regulationWe also check whether IO acquirers made different operational changes to NP targets

before and after regulation. This allows us to test whether regulation forced IO acquirersto scale back reductions in employees, beds and services in their NP targets. Constraints ofthis type may account for reductions in acquisition activity in regulated states.

The results of this analysis are given inTable 8. Table 8is restricted to (IO–NP) ac-quisitions and shows operational changes in regulated and non-regulated states, before andafter regulation. The estimated differences between regulated and unregulated states shownin Table 8are not statistically significant. Based on this evidence, we cannot attribute thefall in acquisitions and CARs that occurred in regulated states to regulatory constraints that

36 Negative CARs imply value decreasing behavior on the part of acquirers if they are the direct result ofan acquisition. More plausibly, negative CARs may reflect other information that is revealed at the time of theacquisition. For example, it is possible that acquisitions of hospitals in regulated states signaled that the acquirerlacked good targets in unregulated states or other “bad news”. To investigate this possibility, we identify the listof acquirers of hospitals in regulated states and check to see if they made subsequent acquisitions in unregulatedstates. We find that nine separate acquirers made acquisitions in regulated states. Of these nine, six also madeacquisitions in unregulated states subsequent to their regulated acquisition. Thus, regulated state acquisitions didnot reliably signal that other unregulated opportunities were unavailable.


Table 8Analysis of changes in operating characteristics before and after regulation

Description Non-regulated states Regulated states

Pre-1997 Post-1997 Before After[1], n= 48 [2],n= 43 [1],n= 31 [4],n= 13

Changes in mean values of target operations by state regulatory statusChange in employees −85.16 (−1.53) 6.59 (0.25) −64.14 (−1.75) −17.30 (−0.76)Change in number of beds −15.71 (−1.94) −4.19 (−0.79) −3.90 (−0.63) −0.50 (−0.24)Change in count of selected services−0.03 (−0.10) −0.08 (−0.62) 0.07 (0.27) −0.10 (−0.43)

[5], [3] − [1] [6], [4] − [2] [7], [4] − [3] [8], [6] − [5]

Differences across regulated and non-regulated statesChange in employees 21.02 (0.32) −23.89 (−0.68) 46.84 (1.09) −44.91 (−0.60)Change in number of beds 11.81 (1.16) 3.69 (0.65) 3.40 (0.52) −8.12 (−0.70)Change in count of selected services 0.09 (0.26) −0.02 (−0.07) −0.17 (−0.49) −0.11 (−0.25)

This table presents mean changes in the number of hospital employees, the number of hospital beds, and in aselected number of services offered by the target hospital, measured 1 year before and 2 years after the acquisition.The sample is divided into states that did and did not pass regulations. In regulated states the sample is furtherdivided into pre- and post-regulatory periods. For transactions that occurred in states without any regulation, thesample is further divided into pre- and post-1997 periods, which corresponds to the introduction of laws in states,which chose to adopt them. Differences between regulated and unregulated states are provided in the second halfof the table. Significant results are in bold.T statistics are in parentheses. Bonferoni adjustments are made tocritical values for tests of statistical significance.

were imposed on the ex post operations of NP targets. We speculate, instead, that regula-tions were associated with negative future expectations about the operating environmentsof regulated states.

4.2.3. Acquirer abnormal returns for IO and private targetsAs a final check on our regulatory results, we also analyze the CARs associated with

IO acquisitions of IO and privately owned hospitals in states that passed regulations.37 Wefind no decline in acquirer CARs within this sample. This result is expected if the declinein CARs earned on NP targets was due to state-level regulations as opposed to a spuriouscause.

5. Discussion

In this paper, we focus on one important component of the redistributive impact of non-profit acquisitions – whether (IO–NP) acquisitions were transacted at fair market value. Wedevelop two important findings in this regard. First, our evidence shows that sales of NPhospitals to IO chains met simple definitions of fairness based on zero cumulative abnormalreturns and similar returns across target ownership categories. Both before and after the pas-sage of state laws, CARs from NP purchases were non-positive and no greater than the CARs

37 These results are available from the authors upon request.


earned on IO and private targets. This suggests that any economic surplus that was createdwas transferred to the NP target. Second, state laws regulating the sale and conversion ofnonprofit hospitals were associated with reductions in the number of NP hospital acquisi-tions in these states. It may be that financial markets interpreted regulations as a negativesignal about the future regulatory environment of states that adopted such regulations.

Taken together, our results may ease concerns that potential agency problems betweencommunities and hospital boards resulted in below-market sales of NP hospitals to IOacquirers. State laws may still be desirable for a number of reasons, including ensuring thecontinuation of services and preventing random instances of under-pricing.

While we are unable to say whether such acquisitions create surplus, our evidence sug-gests that any financial value that was created was transferred to the NP target. Moreover,our interviews with a privately-held hospital company CEO indicate that the transactionscosts associated with these transactions are minimal, and unlikely to consume a significantpercentage of any surplus that is created.38 Finally, it is also relevant to point out that mostNP target hospitals retained services and beds. Services such as AIDs care were retainedby the IO acquirer. This is not surprising, given that communities can require the reten-tion of such services in sale agreements and that CON laws may also prohibit their beingdiscontinued.

Acknowledgements

We thank workshop participants at INSEAD, Northwestern and Rochester. We alsothank Mike Barclay, Jim Brickley, Hud Connery, Ken Lehn, Steve Rock, Mike Ryall, BillSchwert, Cliff Smith, Jerry Warner, Jerry Zimmerman, the referees and the editor for helpfulsuggestions and comments on earlier drafts of this paper. All remaining errors are our own.We appreciate the financial support of the Bradley Policy Research Center at the SimonSchool and the John M. Olin Foundation.

References

Asquith, P., Bruner, R., Mullins, D., 1983. The gains to bidding firms from merger. Journal of Financial Economics11, 121–140.

Billian Publishing, 1995–2003. The Hospital Blue Book, Official National ed. Atlanta.Bradley, M., Desai, A., Kim, E., 1988. Synergistic gains from corporate acquisitions and their division between

the stockholders of target and acquiring firms. Journal of Financial Economics 21, 3–40.Brown, J., Warner, J.B., 1985. Using daily stock returns: the case of event studies. Journal of Financial Economics

14, 3–32.Butler, P., 1997. State policy issues in nonprofit conversions. Health Affairs 16, 69–84.Community Catalyst, 2002. All Conversion Laws: Table. Internet Posting.Cutler, D., Horwitz, J., 1998. Converting Hospitals From Not-For-Profit to For-Profit Status: Why and What

Effects? NBER working paper, 6672.

38 Our interviews reveal that financial due diligence is typically performed by the internal staff of the acquirerand not amortized into the cost of the acquisition. Legal due diligence is typically outsourced. Our interviewsindicate that legal due diligence may cost from $250,000 to $750,000 per deal.


Fama, E., et al., 1969. The adjustment of stock prices to new information. International Economics Review 10,1–21.

Grantmakers in Health, 2001. A Profile of New Health Foundations. Washington, DC.Gray, B., 1997. Conversion of HMOs and hospitals: what’s at stake? Health Affairs 16, 29–47.Irving Levin Associates, 1995–2002. The Hospital Acquisition Report, 1995–2002 ed. New Canaan.Jarrell, G., Bradley, M., 1980. The economic effects of federal and state regulations of cash tender offers. Journal

of Law and Economics 23, 371–407.Jarrell, G., Brickley, J., Netter, J., 1988. The market for corporate control: the empirical evidence since 1980.

Journal of Economic Perspectives 2, 49–68.Kane, N., 1997. Managing charitable assets from conversions. Health Affairs 16, 229–237.Lutz, S., Gee, E.P., 1998. Columbia HCA: Healthcare on Over Drive. McGraw Hill.MacKinlay, A.C., 1997. Event studies in economics and finance. Journal of Economic Literature 35, 13–39.Picone, G., Chou, S., Sloan, F., 2002. Are for-profit hospital conversions harmful to patients and to medicare?

Rand Journal of Economics 33, 507–523.Pratt, S., Reilly, R., Schweighs, R., 2000. Valuing a Business. McGraw Hill.Schipper, K., Thompson, R., 1983. Evidence on the capitalized value of merger activity for acquiring firms. Journal

of Financial Economics 11, 85–119.Shriber, D., 1997. State experience in regulating a changing health care system. Health Affairs 16, 48–68.Sloan, F., 2002. Hospital ownership conversions: defining the appropriate public oversight role. In: Garber, A.

(Ed.), Frontiers in Health Policy Research, vol. 5, pp. 123–166.

Abnormal returns and the regulation of nonproﬁt hospital ... · Journal of Health Economics 24 (2005) 113–135 Abnormal returns and the regulation of nonproﬁt hospital sales

Documents