The RAND Corporation
Hospital Market Structure and the Behavior of Not-for-Profit Hospitals
Author(s): Mark Duggan
Source: The RAND Journal of Economics, Vol. 33, No. 3 (Autumn, 2002), pp. 433-446
Published by: Blackwell Publishing on behalf of The RAND Corporation
Stable URL: http://www.jstor.org/stable/3087466
Accessed: 19/09/2010 23:58
Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at
http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unless
you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you
may use content in the JSTOR archive only for your personal, non-commercial use.
Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at
http://www.jstor.org/action/showPublisher?publisherCode=black.
Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed
page of such transmission.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of
content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms
of scholarship. For more information about JSTOR, please contact [email protected].
The RAND Corporation and Blackwell Publishing are collaborating with JSTOR to digitize, preserve and
extend access to The RAND Journal of Economics.
http://www.jstor.org
RAND Journalof Economics
Vol. 33, No. 3, Autumn2002
pp. 433-446
Hospital market structure and the
behavior of not-for-profithospitals
Mark Duggan*
I exploit a change in hospitalfinancial incentives to examine whether the behavior of private
not-for-profithospitals is systematically related to the share of nearby hospitals organized as
for-profitfirms.Myfindings demonstratethat not-for-profithospitals infor-profitintensiveareas
are significantlymoreresponsiveto the change thantheircounterpartsin areas served byfewforprofitproviders.Differencesinfinancial constraintsand other observablefactors correlatedwith
for-profithospital penetrationdo not explain the heterogeneousresponse. Thefindings suggest
that not-for-profithospitals mimicthe behaviorof privatefor-profitproviderswhen they actively
competewith them.
1. Introduction
*
Hospitals are the largest segment of the not-for-profitsector, accountingfor nearly40% of
all not-for-profitrevenuesin 1995. Unlike not-for-profiteducationalinstitutionsand religious organizations,manynot-for-profithospitalsactively compete with profit-maximizingfirms.Recent
researchsuggests thatthe presenceof one or more for-profithospitalsin a marketmay affect the
behavior of privatenot-for-profitproviders(Cutlerand Horwitz, 1999; Silvermanand Skinner,
2001). If this is the case, then the fractionof hospitalsorganizedas profit-maximizingfirmsmay
understatethe trueimpactthatfor-profithospitalshave on the marketsin which they operate.
Because each hospital's type of ownership is endogenous, determiningwhether it is the
presence of for-profithospitalsor some otherfactorthatdrives any observabledifferencein notfor-profitbehaviorormarketoutcomespresentsa difficultidentificationproblem.1Theveryfactors
that cause for-profitfirmsto enterparticularmarketsmay simultaneouslylead otherhospitalsto
behave differentlyfrom hospitals of the same ownershiptype in marketswith correspondingly
few for-profithospitals.
In this articleI deal with this identificationproblemby exploiting a significantandplausibly
exogenous change in hospital financialincentives.2Specifically, I use the change in incentives
* Universityof Chicago and NBER; [email protected].
I am gratefulto David Cutler,MartinFeldstein, David Scharfstein,Doug Staiger,an anonymousreferee, and to
seminar participantsat the 2000 NBER Conference on the IndustrialOrganizationof Medical Care for many helpful
comments. Thanks to the California Office of Statewide Health Planning and Development for supplying the data.
Financial supportfrom the National Instituteon Aging (grantno. P30 AG-12857-06) is gratefullyacknowledged.Any
errorsin the articleare my own.
l This point is stressedby Norton and Staiger(1994), who demonstratethatfor-profithospitalslocate in systematically differentareasthan do not-for-profitfacilities.
2 An alternative
approach,employedby CutlerandHorwitz(1999), is to examinethe effect of hospitalconversions
to for-profitstatus.
Copyright? 2002, RAND.
433
OFECONOMICS
434 / THERANDJOURNAL
createdby California'sDisproportionateShareprogram(DSH) to examine whetherinstitutional
responses and market outcomes vary systematically with the share of hospitals that are forprofit.DSH increasedthe reimbursementrate for patientsinsuredby the federal-stateMedicaid
program,leading both privatefor-profitand privatenot-for-profithospitals to "cream-skim"the
most profitableindigentpatientsfrom government-ownedproviders(Duggan, 2000a).
The results presentedin the first empiricalsection demonstratethat the shareof Medicaidinsuredpatientswithina countyreallocatedfrompublicto privatehospitalsis significantlyrelated
to the fraction of hospitals there organizedas profit-maximizingfirms. Public hospitals located
in counties served by relatively many for-profithospitals experiencedmuch greaterreductions
in their numbersof Medicaid-insuredpatients than did other government-ownedfacilities. For
example, the generalacute carehospitalsowned by the county of Los Angeles, a marketin which
half of the hospitals are for-profit,saw their share of Medicaid-insurednewborndeliveries fall
from 45% to 12% from 1990 to 1996. The county-ownedhospital in San Francisco,a county
with no for-profithospitals, was much less affected. Its shareof county Medicaid deliveries fell
by only 4% duringthe same time period.
To test whether the observed difference across marketareas is caused by the presence of
for-profitfirms or some other factor that is correlatedwith for-profithospital penetration,I next
control for other potentiallyimportantmarketcharacteristics.If, for example, the quality of the
government-ownedhospital in a county is significantlyrelated to the share of hospitals there
organizedas for-profitfirms, then the differencedescribedabove may actually be drivenby the
lower costs associatedwith skimmingpublic hospitalpatientsratherthanby the presenceof forprofitfirms.Controllingfor this and severalotherpotentiallyconfoundingfactorsdoes not change
the robustrelationshipbetween for-profithospital penetrationand the reallocationof Medicaid
patients.
In the next empiricalsection I take the hospital as the unit of observation,and test whether
part of the difference across market areas reflects not-for-profitbehavior that is significantly
relatedto the ownershiptype of nearbyhospitals.Consistentwith the hypothesisoutlinedabove,
I find that not-for-profithospitals in marketswith relatively many for-profithospitals respond
more aggressively to a change in the financialincentive to treatMedicaidpatientsthan do other
not-for-profitproviders.3There is no correspondingrelationshipbetween the ownershiptype of
competinghospitals and the behaviorof profit-maximizinghospitals.Public hospitalslocated in
for-profitintensiveareaslose a substantialshareof theirnewly profitablepatientsbothbecause of
the responseof for-profithospitalsto the changein incentivesandbecause of the moreaggressive
response by privatenot-for-profithospitalsin these markets.
In the final empirical section I explore whetherthe heterogeneityin not-for-profithospital
behavior can be explained by differences in their financial constraints.If for-profitproviders
compete more aggressivelyon price thando otherhospitals,then theirnot-for-profitcompetitors
may be more financiallyconstrainedthanthe averagenot-for-profit.Even if the objectivefunction
of not-for-profithospitalsdoes not vary acrossmarketareas,this differencein the hospitalbudget
constraintcould lead to variationin the optimalresponse to a change in the marketenvironment.
My empiricalevidence indicates that not-for-profithospitalsin for-profitintensive areasare not
more financiallyconstrainedthan are other not-for-profitfacilities, suggesting that the presence
of one or more for-profithospitals in a marketchanges the objective function of not-for-profit
providersand leads them to behave more like profit-maximizingfirms.
The outline of the article is as follows. Section 2 provides backgroundon the California
hospitalmarketandthe DisproportionateShareprogram,anddescribesthe reallocationof patients
that resultedfrom the change in hospital financialincentives. The empiricalresults in Section 3
reveal thatthe market-leveleffect of the change in financialincentivesvariedsystematicallywith
the shareof hospitals organizedas privatefor-profitfirms. Section 4 demonstratesthat the more
aggressive response by not-for-profithospitals in for-profitintensive areasexplains much of the
3 This is consistent with the
findingsof Silvermanand Skinner(2001), who show that not-for-profithospitals in
marketswith one or more for-profithospitalsare significantlymore aggressiveaboutMedicareupcodingfor pneumonia.
? RAND 2002.
DUGGAN / 435
variationin market-leveloutcomes. In Section 5 I examine whethernot-for-profithospitals with
relatively many for-profitcompetitorsare more responsiveto financialincentives because these
organizationsare more financiallyconstrained.Section 6 concludes.
2. Dataand background
*
California'shospitalmarketis servedby privatefor-profit,privatenot-for-profit,andgovernment-ownedhospitals.4Virtuallyall of the state's large urbanareas have at least one publicly
owned safety-netprovider.The patientstreatedat these facilities are disproportionatelypoor and
tend to be withouthealthinsuranceor insuredby the Medicaidprogram(Weissmanand Epstein,
1994). For example, more than 90% of the patients treated at the four county-owned general
acutecare hospitalsin Los Angeles are Medicaid-insuredor uninsured,whereasonly 2% of their
patients are privatelyinsured.The mix of patientsat privatefor-profitand privatenot-for-profit
hospitals is similar, with not-for-profitsactually treatingfewer indigent patients, as a share of
theirtotal patientmix, than theirfor-profitcounterparts(16% and 18%,respectively,in 1990).
One factorthatpartiallyexplains the substantialdifferencebetween public and privatehospitals is location-hospitals owned by the governmentare located in relatively poorer areas.
Additionally,these facilities tend to offer services that are used differentiallyby the poor. But
even aftercontrollingfor these differences,public hospitalsin Californiatreatsignificantlymore
low-income patients than do privatehospitals. Cross-sectionalestimates of the corresponding
differencebetween the two types of privatehospitals suggest that privatenot-for-profitfacilities
do not providemore medical care to the poor thanfor-profithospitalsdo (Duggan, 2000b).
Recent workhas exploitedthe changein incentivescausedby the introductionof California's
DisproportionateShare(DSH) programto explore whethera hospital'stype of ownershipaffects
its responseto a change in the incentiveto treatlow-income patients(Duggan, 2000a). The DSH
programsubstantiallyincreasedhospitals' financialincentivesto treatMedicaidpatientsbut left
the incentive to treatindividualswithout health insuranceessentially unchanged.The nonlinear
incentives thatresultedare shown in Figure 1, which plots a hospital'sMedicaid DSH per-diem
as a function of that facility's low-income number.5Hospitals that served relatively many lowincome patientswhen this programwas first introducedhad a significantincentive to treatmore
Medicaid patients.So too did facilities close to but below the 25% threshold.
Duggan (2000a) shows that both types of private hospitals were similarly aggressive in
respondingto these incentives, leading to a substantialreallocationof the most profitableMedicaid patients from publicly owned hospitals to privateones. This reallocationwas especially
pronouncedfor pregnantwomen. From 1990 to 1996, the share of Medicaid-insurednewborns
deliveredat hospitals owned by the governmentfell from 43% to 23%, afterremainingroughly
constant prior to 1990. In the empirical sections that follow I focus primarilyon low-income
pregnantwomen, the group for which competitionintensifiedthe most after the new financial
incentives were introduced.6
3. The effect of for-profit hospital penetration on
market-level changes
In this section I use county-leveldatato investigatewhetherthe fractionof Medicaidpatients
*
reallocatedfrompublicto privatehospitalsafterthe changein financialincentivesis systematically
relatedto the share of hospitals in a marketthat are privatefor-profit.This would be the case if
4 The dataused in this study are obtainedfrom two sources,both of which are compiled by the CaliforniaOffice
of StatewideHealthPlanningandDevelopment(OSHPD).The hospital-leveldatasetis obtainedfrom OSHPD's Hospital
Disclosurereports,while the sourceof patient-leveldatais the OSHPDPatientDischargedataset.These dataaredescribed
extensively in Duggan (2000a).
5 The low-income numbermeasuresthe shareof a hospital'scosts thatare attributableto Medicaidand uninsured
patients(Duggan, 2000a).
6 See Duggan(2000a) for an explanationof why pregnantwomen were the most profitableof all Medicaidpatients
following the change in financialincentives.
? RAND 2002.
436
/
THE RAND JOURNAL OF ECONOMICS
FIGURE 1
DSH INCENTIVES FOR A PRIVATEHOSPITAL
1,200
1,000
E
QC
G)
I03
c,
0
800
600
400
200
0
0
10
20
30
40
50
60
70
80
90
100
Low-income number (percent)
profit-maximizinghospitalsdid, on average,respondmoreaggressivelythanprivatenot-for-profit
facilities to the change in incentives or if privatenot-for-profithospitals behavedmore like forprofithospitalswhen activelycompetingwith them.Recentworkhas providedempiricalevidence
consistent with this latter hypothesis, suggesting that for-profithospitals exert a peer effect on
their not-for-profitcompetitors(Cutlerand Horwitz, 1999; Silvermanand Skinner,2001).
For-profithospitalpenetrationvariessubstantiallyacrossmarketareasin California.Extreme
examples include Los Angeles and San Francisco,wherefor-profitfacilities accountfor 50% and
0%, respectively,of all generalacute care hospitals.A comparisonof these two large urbanareas
suggests thatthe presence of for-profithospitalsmay affect the behaviorof otherprovidersin the
same market.Table 1 reveals thatwhereasthe shareof Medicaidbirthsat public hospitalsin Los
Angeles fell substantiallyfrom 1990 to 1996, the correspondingdecline in San Franciscowas
much less marked.Although for-profithospitals accountedfor some of the reallocationwithin
Los Angeles, not-for-profithospitalsthereenjoyed much largerincreasesthantheircounterparts
in San Francisco.Specifically,the shareof Medicaidbirthsdeliveredat not-for-profithospitalsin
Los Angeles rose from 37% to 61%, a much largerincrease than the four-percentage-pointrise
at San Francisco'snot-for-profitproviders.
The county-level regressionssummarizedin Table 2 investigatewhetherthis patternholds
for the typical county in California.Summarystatisticsfor the explanatoryvariablesare included
in the /u, a column. The dependentvariablein this set of regressionsis equal to the change from
1990 to 1996 in the share of Medicaid-insuredbirthsthat are deliveredat public hospitals. The
TABLE 1
Reallocation from Los Angeles and San
Francisco Safety-Net Facilities
LAC Public Hospitals
SF GeneralHospital
Categoryof Patient
1990
1996
1990
1996
% Medicaid
44%
20%
38%
37%
% Medicaidbirths
45%
12%
42%
38%
% Births
21%
7%
14%
13%
Percentagesequal the shareof patientsdeliveredat hospitalsowned by Los Angeles
Countyor San FranciscoCountyin 1990 and 1996.
? RAND 2002.
DUGGAN
TABLE 2
/
437
For-Profit Hospital Penetration and the Reallocation of Medicaid Patients from
Public to Private Providers
%A MC-BIRTH-PUBLIC90-96
it, a
%FOR-PROFIT90
A%MEDICAID9o-96
(1)
(2)
%MANAGEDCARE90
(4)
.291
-.468***
-.427***
-.308***
-.322***
(.198)
(.074)
(.106)
(.108)
(.126)
%BLACK9o
%NOTCH9o
***
-.41
(.107)
-.318'**
-.423***
(.116)
(.108)
(8)
-.249*
(.139)
-1.89
(1.31)
.195
.202
.061
(.191)
(.129)
(.144)
.005
-.237
.207
(.294)
(.255)
.639
-.296***
(.245)
(.095)
.073
-.643 ".
(.077)
(.276)
-
29
R2
-
(.147)
-.543
(.354)
(.312)
-.045*
(.025)
Numberof observations
-.178
-.112
.065
(.069)
CONSTANT
(7)
(.96)
(.083)
%HISPANIC9o
(6)
(5)
-2.36**
.038
(.022)
%PRIV-AT-PUBLIC9o
(3)
-.064
(.037)
-.009
(.041)
-.136
(.058)
-.020
(.061)
.139
(.069)
-.058
(.041)
.063
(.317)
.089
(.104)
50
29
29
29
29
29
29
29
.456
.377
.494
.418
.397
.584
.380
.630
Dependent variable is the change from 1990 to 1996 in the share of Medicaid-insurednewborns delivered at public
hospitals within each county (mean = -.188 and standarddeviation= .138). Regressionsare weighted by the numberof
generalacute care hospitalsin the county.Standarderrorsare incluldedin parentheses.
*Significantat 10%.
**Significantat 5%.
***Significantat 1%.
explanatory variable of interest, %FOR-PROFIT90, represents the fraction of general acute care
hospitals in the county that are private for-profit firms.
The coefficient estimate for %FOR-PROFIT90 in the first specification demonstrates that
the share of Medicaid births delivered at public hospitals fell by significantly more in counties
with substantial for-profit hospital penetration. A 10-percentage-point increase in the share of
hospitals that are for-profit is associated with a 4.7-percentage-point increase in the share of
Medicaid-insured pregnant women switching from public to private facilities. This regression
result is consistent with the San Francisco-Los Angeles comparison described above, and it is
robust to the exclusion of these two counties.7 The next specification includes only those counties
that have at least one public hospital and at least one private provider, reducing the number of
counties in the sample to 29.8 The coefficient estimate for %FOR-PROFIT90 falls slightly but
remains statistically significant at -.427.
In the next several specifications I explore whether these results are robust to the inclusion
of other county-level control variables. If the share of hospitals organized as for-profit firms is
7 The estimatedoes fall, however,to -.247, with a standarderrorof .066. Giventhatmorethanhalf of California's
for-profithospitals are located in Los Angeles, eliminatingthis observationfrom the sample substantiallyreduces the
amountof variationin %FOR-PROFIT9o.
8 This excludes primarilyruralcounties.Of the 20 mostpopulousCaliforniacounties,only Solano (rankednumber
20) is not included in this sample of 29 counties. The share of hospitals located in the excluded counties is less than 25%.
? RAND 2002.
438 / THERANDJOURNAL
OFECONOMICS
correlatedwith some other factor that is actually driving the reallocationof Medicaid-insured
pregnantwomen from public to privatehospitals, then the precedingresults will be misleading.
One importantfactorto consideris the changein the characteristicsof Medicaid-insuredpregnant
women that was occurringduringthe first half of the 1990s. Specifically,Medicaid expansions
led to an increase in the share of newborndeliveries that were Medicaid-insured,from 39% in
1990 to more than47% by 1996.9 If these expansionsoccurredat differentrates across counties,
and if those made eligible were more likely to attend one type of hospital, then differences in
the dependentvariablecould result without any reallocationof patients.'0I thereforeintroduce
the variableA%MEDICAID90_96
to controlfor changes in the shareof pregnantwomen in each
with
Medicaid
The
coefficientestimateon this variableis significantlynegative,
county
coverage.
suggesting that the marginalMedicaid-eligiblewoman is more likely to attenda privatehospital
than is the average one.1l While controllingfor the growth in Medicaid eligibility does reduce
the magnitudeof the coefficientestimateon %FOR-PROFIT90,
it remainsstatisticallysignificant
at -.308.
The second factor that I consider is the quality of the public hospital(s)in a county.All else
equal, as the quality of the government-ownedfacility declines, the ease with which both private
for-profitand private not-for-profithospitals can skim newly profitableindigent patients will
increase. If for-profithospitals tend to be located in counties with lower-qualitypublic hospitals,
then the significantestimate describedabove may be due to this lower cost of respondingto the
DSH incentivesandnotto thepresenceof theprofit-maximizingfacilities.As thequalityof a public
hospital increases, the share of its patients with privateinsurance,who presumablyhave more
choices thando Medicaid-insuredor uninsuredpatients,should also increase.Because pregnant
women and newbornchildrenare the focus of this analysis, I use the fractionof public hospital
deliveries that are privatelyinsuredas my proxy for public hospitalquality.12This measureis no
doubt imperfect,but it should to some extent capturevariationin the quality of public hospitals
across counties. The coefficientestimatefor the %PRIV-AT-PUBLIC90
variablehas the expected
sign-higher-quality public hospitalsexperiencesmallerreductionsin theirnumberof Medicaid
newborns-but this does not substantiallyaffect the estimate on the %FOR-PROFIT90
variable,
which remainssignificantlynegative.
Because of the nonlinearnatureof the incentivesthat were introducedby the DSH program,
certain hospitals had a particularlystrong incentive to admit more Medicaid patients. Those
hospitals with low-income numbersabove 25% enjoyed an immediateincreasein theirmarginal
revenuefor Medicaidpatients.If for-profithospitalswere locateddisproportionatelyin areaswith
relativelymanyprivatehospitalslocatedabovethis thresholdwhenDSH was firstintroduced,then
the coefficient estimate for %FOR-PROFIT90
may actually be capturingthis average-incentive
effect ratherthana for-profiteffect. To controlfor this potentiallyconfoundingfactor,I introducea
variable%NOTCH90,which equals the fractionof privatehospitalswithineach county with lowincome numbersof 25%or more when DSH was firstintroduced.The coefficientestimatehas the
expected sign-public hospitalslocated in counties with more notch hospitalsdo appearto have
lost a largershare of their Medicaid-insuredpregnantwomen-but is statisticallyinsignificant.
As was truein the previoustwo cases, the introductionof this variabledoes not significantlyalter
the coefficient estimate for the %FOR-PROFIT90
variable.
Another importantfactor concerns the natureof the privateinsurancemarketwithin each
county. Specifically,if managedcare was more or less prevalentin counties with relativelymany
9 Cutler and Gruber(1996) show that these expansions substantiallycrowded out privateinsurancecoverage,
implying thatthe sample of Medicaid beneficiarieschangedsubstantiallyduringthe time period of interest.
10Individualsmade
eligible for Medicaidin the early 1990s had higherincomes thanthe averagerecipient(Currie
and Gruber,1996). Thus one would potentiallyobserve a decline in the shareof Medicaid deliveriesoccurringat public
hospitals even with no reallocation.
11
Additionally,as Medicaidbecomes a largershareof the market,privatehospitalswill have an increasedfinancial
incentive to admit more of them.
12McClellanand
Staiger(1999) use measuresof qualitybased on healthoutcomes.Because I do not have detailed
clinical informationabout individualpatients,it is not possible in this case to reliably separatequality differencesfrom
the effect of differencesacross facilities in averagepatienthealth.I thereforeuse this alternativemeasure.
? RAND 2002.
DUGGAN /
439
for-profit providers when DSH was first introduced, then private hospitals may have been actively
seeking out new sources of revenue in response to reduced inpatient demand. Controlling for the
share of newborn deliveries that were insured by managed care in 1990 does not, however, affect
the coefficient estimate of interest, and the estimate for this %MANAGED CARE90 variable is
not statistically significant.13
The characteristics of Medicaid-insured pregnant women may also affect a private hospital's
decision to respond to the change in financial incentives. Pregnant women on Medicaid are more
than twice as likely as privately insured pregnant women to be black or of Hispanic origin. If
hospitals are more inclined to admit low-income individuals from certain demographic groups than
from others, and if for-profit hospitals are located in geographic areas in which the demographics
of the indigent are systematically different from other areas, then the estimate on the %FORPROFIT90 coefficient may be capturing this other effect.
I therefore control for the share of Medicaid-insured pregnant women within each county
that are black or of Hispanic origin in the seventh specification. Interestingly, the estimates for
both variables are significantly negative,14 suggesting that more reallocation of Medicaid-insured
patients occurred in counties with relatively more minorities. This may suggest that, prior to
DSH, private hospitals were less inclined to admit black or Hispanic Medicaid patients, but that
the stronger financial incentives introduced by DSH led them to open their doors to these minority
groups. As was true in all of the previous cases, the coefficient estimate on the %FOR-PROFIT90
variable remains significant.
In the final specification, I include all the control variables in one regression. Because there
are only 29 observations and 7 explanatory variables, it is not surprising that every standard error
increases substantially. The only variable to remain significant, though, is the share of hospitals in
the county organized as for-profit firms, %FOR-PROFIT90.The next section investigates whether
the variation in market-level outcomes shown here is to some extent driven by differences in the
behavior of private not-for-profit firms across different market environments.
4. The effect of for-profit hospital penetration on
responses to a change in incentives
In this section I explore whether the behavior of individual hospitals is influenced by the
*
ownership type of nearby providers. Specifically, I test whether the share of hospitals organized as
for-profit firms within ten miles of each facility is significantly related to the average response by
each type of hospital to the DSH financial incentives.15 For each hospital, I define the variable A
MC-BIRTH90_96 to be the change in the number of Medicaid births delivered at the facility from
1990 to 1996, and test whether the observed change is significantly related to the share of nearby
hospitals organized as for-profit firms, FOR-FRAC90.16 I interact this variable with three separate
ownership dummies-NOT-FOR-PROFIT, FOR-PROFIT, and PUBLIC-to separately identify
the effect of for-profit hospital penetration by ownership type. The basic estimating equation is
AMC-BIRTHjt = a + Pf* OWNjt + X * (OWNji * FOR-FRACjt)
+ It * AMC-BIRTHj,t_i + yXjt + Ejt,
13Controllingfor the change in the share of births insuredby managedcare has a similarly small effect on the
coefficient estimatefor %FOR-PROFIT9o.
14These estimatesare similarif I insteaddefine the %BLACK9oand %HISPANIC9o
variablesto be the difference
in the shareof Medicaidandprivatelyinsuredpatientsin each demographicgroup.Additionally,controllingfor the change
in the share of Medicaid-insurednewbornswho are black or of Hispanic origin does not affect the %FOR-PROFIT9o
coefficient estimate.
15I exclude own hospital in the marketdefinitionand define all hospitalswithin ten miles in the marketarea.The
main results presentedbelow are robust to alternativemarketdefinitions, including the share within five miles of the
hospitalor the sharewithin each hospital'scounty.
16The dependentvariableis definedin levels ratherthan logs because nearly 30% of the hospitals did not deliver
any Medicaid-insurednewbornsin either 1990 or 1996 andbecauseone-thirdof these 117 facilities had a nonzeroamount
in one of the two years.
? RAND 2002.
440
/
THE RAND JOURNAL OF ECONOMICS
TABLE 3
Summary Statistics for Sample Hospitals
VariableName
Standard
Deviation
N
Mean
A MC-BIRTH90-96
401
43
968
A MC-BIRTH88-90
401
179
439
A MC-DISCH90-96
401
266
2248
A MC-DISCH88-90
401
477
1164
NOT-FOR-PROFIT
401
.529
.500
PUBLIC
401
.212
.409
FOR-PROFIT
401
.259
.439
FOR-FRAC90
401
.233
.275
NFP * FOR-FRAC90
401
.117
.225
FP * FOR-FRAC90
401
.092
.202
PUBLIC * FOR-FRAC90
401
.023
.126
NFP-FRAC90
401
.437
.343
NFP * NFP-FRAC90
401
.261
.346
FP * NFP-FRAC90
401
.118
.242
PUBLIC * NFP-FRAC90
401
.058
.212
BEDS90
401
190
152
ONLY-HOSP90
401
.217
.413
DEBT/ASSET90
376
.377
.305
DEBT/ASSET90
367
.012
.212
NUM-HOSPITALS-CLOSE
401
9.63
10.98
PUBLIC-HOSP-CLOSE90
401
.494
.501
NOTCH-HOSPITAL90
401
.232
.423
with OWNjt representing the three dummy variables for the hospital's type of ownership. The
specifications also include a control for the pre-existing trend in hospital-specific admission rates
for Medicaid-insured newborn deliveries A MC-BIRTH88_90 and for the number of beds available
at each facility in 1990. Summary statistics for these variables and for each of the additional
explanatory variables defined below are provided in Table 3.
The first set of regression results are summarized in Table 4. The significantly positive
estimate of 672 on the NFP * FOR-FRAC90 variable in the first specification reveals that not-forprofit hospitals with relatively many for-profit competitors enjoyed substantially greater increases
in their Medicaid caseloads from 1990 to 1996 than did other not-for-profit providers.17 This
estimate implies that a one-standard-deviation increase in the share of hospitals organized as
for-profit hospitals is associated with a .33-standard-deviation increase in the change in Medicaid
deliveries from 1990 to 1996 at private not-for-profit hospitals.18
17
Standarderrorsin Table 4 are correctedto accountfor the fact that the FOR-FRAC90variablevaries at the zip
code level and thus the residualin the estimatingequationis not independentacross hospitalswithin a zip code cell. See
Moulton (1990) for a discussion of this correction.
18
Among not-for-profithospitals, the averageincreasein Medicaid deliveries from 1990 to 1996 was 210 with a
standarddeviation of 555. The correspondingmean and standarddeviation for the FOR-FRAC90measureare .222 and
? RAND 2002.
DUGGAN / 441
TABLE 4
The Effect of For-Profit Penetration on Hospital Responses to Financial
Incentives
A MC-DISCH90-96
A MC-BIRTH90-96
(1)
NFP * FOR-FRAC90
672***
(244)
FP * FOR-FRAC90
PUBLIC * FOR-FRAC9o
FOR-PROFIT90
PUBLIC9o
A MC-BIRTH88-90
(2)
827***
(298)
(3)
756***
(253)
(4)
633**
(259)
(5)
693**
(295)
(7)
1,768***
1,780**
(644)
46
134
38
-68
-35
235
182
(213)
(217)
(243)
(235)
(487)
(519)
-1,421*
-1,493**
-1,610**
-934*
-1,758**
-3,296**
-3,488**
(759)
(721)
(769)
(584)
(744)
(1,650)
(1,593)
-196
-121
-133
-174
-68
(104)
(205)
(135)
(143)
(161)
-498***
-352**
-308***
-465***
(175)
(139)
(109)
(158)
(73)
-.583***
-.594***
-.600***
-.632***
-.558***
(.192)
(.186)
(.188)
(.206)
(.185)
-1.76
(1.58)
-1.49
(1.40)
-1.65
(1.53)
-186
ONLY-HOSP90
(180)
-1.61
(1.58)
-178
(181)
NFP * NFP-FRAC90
-64
-1.55
(1.54)
-291
(176)
58
-280
(323)
-601***
PUBLIC * NFP-FRAC9o
Numberof observations
-102
(159)
-.540**
-.502**
(.230)
(.230)
-2.72
-2.60
(3.49)
(3.52)
-465
-547
(439)
(399)
307
(358)
10
-44
FP * NFP-FRAC90
-97
(230)
(241)
(161)
CONSTANT
(735)
(260)
A MC-DISCH88-90
BEDS90
(6)
(149)
(344)
-670**
-1,136**
(249)
(584)
623
463
550
617
523
1,189
1,352
(369)
(268)
(344)
(373)
(307)
(809)
(872)
314
401
401
401
401
401
401
.325
.309
.313
.275
.326
.272
.238
Sample includes all general acute care hospitals in operationin Californiain 1990 and 1996. Specification(1) excludes
facilities that have zero hospitals within ten miles. In specifications (2) through (7), these facilities are included and
FOR-FRAC90is set equalto zero for them.Standarderrorsarecorrectedfor groupeffects andareincludedin parentheses.
*Significantat 10%.
**Significantat 5%.
***Significantat 1%.
Inferringa causal effect of for-profithospitalson the behaviorof not-for-profithospitalsfrom
this coefficientestimatewould be problematicif the extent of for-profithospitalpenetrationwere
systematicallycorrelatedwith some otherfactorthataffectedthe incentiveto admitmoreMedicaid
patients.One way to test for the importanceof such an omittedvariableis to investigatewhether
.270. The summarystatisticsin Table 3 provide means and standarddeviationsfor the full sample and do not breakout
these values by type of ownership.
? RAND 2002.
442 / THERANDJOURNAL
OFECONOMICS
the behaviorof for-profithospitalsalso variessystematicallywith the FOR-FRAC90measure.If it
did, then one might reasonablyconclude thatfor-profithospitalpenetrationis proxyingfor some
other factor that affects the incentive of both types of privatehospitals to admit more indigent
patients following the introductionof DSH.19The small and statisticallyinsignificantestimate
on the FP * FOR-FRAC90coefficient suggests thatomittedvariablesare not drivingthe observed
heterogeneity in the responsiveness of not-for-profithospitals to the change in incentives, as
profit-maximizingfirmsare not affectedby the ownershiptype of theircompetitors.
The coefficientestimateof -498 on the PUBLICvariableshows that,consistentwith Table2,
the averagepublichospitalexperienceda substantialreductionin the numberof Medicaid-insured
newborns delivered at its facility from 1990 to 1996. Additionally,the significant coefficient
estimate for the PUBLIC * FOR-FRAC90variable implies that this loss was much greaterin
increasein the
marketsserved by relativelymany for-profitproviders.A one-standard-deviation
fractionof nearbyhospitalsthatarefor-profitis associatedwith a .21-standard-deviation
decrease
in AMC-BIRTH90-96for public hospitals.20
The significantlynegativeestimatefor AMC-BIRTH88-90
implies thattherewas a substantial
breakin trendin hospital-specificMedicaid admissionsin 1990, the year that the DSH program
was introduced.The hospitalswith the largestincreasesin Medicaiddeliveriesfrom 1988 to 1990
experiencedthe smallestincreases(or largestdecreases)duringthe next six years.The coefficient
estimate for the numberof hospitalbeds is statisticallyinsignificant.
In this firstspecificationI exclude those facilities thathave zero competitorswithinten miles.
In the next one I include these 87 facilities and set FOR-FRAC90equal to zero for them. The
coefficient estimates are largely unaffectedby this adjustment,or by the inclusion of a dummy
variableONLYHOSP90in the thirdspecificationthatis equalto one if a facilityhas no competitors
within ten miles and zero otherwise. In the fourthspecificationI define FOR-FRAC90to equal
the share of availablehospital beds within ten miles that are located in for-profitfacilities, thus
giving greater weight to larger competitors.This adjustmentdoes not affect the NFP * FORFRAC90estimate appreciably,though it does reduce the estimate for PUBLIC * FOR-FRAC90.
This estimate is still statisticallysignificantat the 10%level, while the estimate for FP * FORFRAC90remainsclose to zero.
One possible explanationfor this set of findings is that the estimates on the FOR-FRAC90
interactionsare picking up an effect of privatehospitalsratherthanan effect of privatefor-profit
hospitals specifically.To investigatethis possibility,in the fifth specificationI introducecontrols
for the share of competing hospitals that are privatenot-for-profits,and interactthis with the
ownershipdummies.The estimateson both the NFP * NFP-FRAC90and the FP * NFP-FRAC90
coefficientsare small and statisticallyinsignificant,providingfurtherevidence thatthe significant
estimate on the NFP * FOR-FRAC90coefficient is capturinga causal effect of privatefor-profit
hospitals on the behaviorof not-for-profithospitals.
It is interestingto comparethe effect of privatefor-profitand privatenot-for-profithospital
penetrationon the loss of Medicaidpatientsat public hospitals.The significantestimateof -670
on the PUBLIC* NFP-FRAC90coefficientsuggeststhatpublichospitalsin not-for-profitintensive
areasdid lose a significantnumberof theirMedicaidpatients.But this estimateis less thanhalf as
large as the correspondingone for for-profithospitalpenetration,implying that public hospitals
in marketswith relativelymany for-profithospitalswere the ones most affectedby the change in
financialincentives and the resultingreallocationof patients.Includingboth of these interactions
causes the estimateon thePUBLICdummyvariableto become smallandinsignificant.Thuspublic
19Even if the financialconstraintsof
privatefor-profithospitals do vary with theircompetitors'ownershiptypes,
theorypredictsthat such "incomeeffects" do not influencethe profit-maximizingresponseto a change in incentives.But
this type of variationcould influencethe optimalresponse of a not-for-profitfirm.
20Given thatthereare twice as
many not-for-profithospitalsas public ones, it is not surprisingthatthe magnitude
of the NFP * FOR-FRAC90estimate is only half as large as the correspondingone for PUBLIC * FOR-FRAC90.A
public hospitalcould, for example, lose 1,400 Medicaiddeliveriesbecause two privatefacilities skim 700 moreMedicaid
patientseach.
? RAND 2002.
DUGGAN / 443
hospitalsthatcompetedonly with othergovernment-ownedfacilities or hadno competitorswithin
ten miles did not experiencelarge declines in theirnumbersof Medicaiddeliveries.
Columns(6) and(7) summarizethe resultsfor specificationsanalogousto those in columns(3)
and (5), with the dependentvariablenow set equal to total Medicaid admissions (not newborns
only). Consistent with the above results, the coefficient estimates demonstratethat for-profit
hospital penetrationhas a significant effect on the behavior of not-for-profithospitals but no
correspondingeffect on the behaviorof profit-maximizingfacilities. Moreover,aftercontrolling
for for-profitpenetration,the share of hospitals that are privatenot-for-profitdoes not exert any
additionaleffect on either type of privatehospital. Public hospitals in marketsserved primarily
by privatenot-for-profithospitals are much less affected than those located in areas with a large
fraction of for-profitproviders.Public hospitals that do not face competition from the private
sector are unaffectedby the change in hospitalfinancialincentivescaused by the DSH program.
The resultspresentedin this section stronglysuggestthatthe behaviorof privatenot-for-profit
hospitals is affected by the presence of for-profitcompetitors,and that this largely explains the
observedvariationacrossmarketareasin the effect of the DSH program.The next sectionexplores
whetherthis variationis drivenby an impact of for-profithospitals on the financialcondition of
privatenot-for-profitproviders.
5. Do differences in financial constraints explain the
heterogeneous response?
*
Because a not-for-profithospitalmay havean objectivefunctionthatpositivelyvaluesfactors
other than profits, its optimal response to a change in financialincentives could plausibly vary
with its financial constraints.21A cash-strappedfacility that is struggling to break even may
respondmuch more aggressivelyto a profitableopportunitythanone with an identicalobjective
function that is in good financialshape. If this is the case, then the results presentedabove may
be drivenby the effect of for-profithospitalson the financialconditionof not-for-profitproviders.
Theoretically,one would not expect financialconstraintsto have any effect on the behaviorof forprofithospitals, as these facilities would respondto the change in incentives if it were profitable
to do so.
The firstspecificationsummarizedin Table5 exploreswhetherhospitalsin for-profitintensive
areas are more financiallyconstrainedthan other providers.The estimatingequation is similar
to the one presentedabove, and the dependentvariable(DEBT9o/ASSET9o)
is equal to hospital
debt as a fractionof total assets in 1990. The insignificantestimateon the NFP * FOR-FRAC90
coefficientsuggestsno systematicrelationship,thoughthe positivesign is consistentwith a modest
effect of for-profithospitals on the financial stress of not-for-profitfacilities. The estimates for
FP * FOR-FRAC90and PUBLIC * FOR-FRAC90are also positive, but neither is statistically
significant.22The next specification tests whether this debt-to-assetratio had been increasing
differentiallyin for-profitintensive areas prior to the introductionof DSH. The insignificant
estimates for all three of the FOR-FRAC90coefficients suggest that this was not the case.
Alternativedefinitionsof hospitalfinancialconditionsyield a similarresult.
The third specificationexamines whetherthe areas served by for-profithospitals are more
contested. If this is the case, then not-for-profithospitals in these areas may respond more
aggressively to a change in financial incentives because of the threat of potential competition
for theirown patientsandbecause therearemorepotentialpatientsto admit.To test this, I use the
numberof hospitals within ten miles of each facility NUM-HOSPS-CLOSE90as the dependent
variableand find thatfor-profithospitalpenetrationis stronglyrelatedto the numberof hospitals
in an area. While the typical not-for-profithospital has only 9.4 hospitals within ten miles, a
21A substantialbody of previousworkhas consideredthe interactionof preferencesandconstraintswhen modelling
the behaviorof privatenot-for-profitfirms.See, for example,Newhouse (1970), Feldstein(1971), Weisbrod(1988), Frank
and Salkever(1991), Glaeserand Shleifer (2001), and Eggleston, Miller, and Zeckhauser(2001).
22The results are qualitativelysimilar if I utilize alternativemeasuresof hospital financialconditions, including
net income or total equity per availablehospitalbed.
? RAND 2002.
444
/
THE RAND JOURNAL OF ECONOMICS
TABLE 5
The Relationship Between For-Profit
Penetration and Hospital Characteristics
DEBT/ASSET A DEBT/ASSET NUM-HOSPSRATIO90
RATIO88-90
CLOSE90
NFP * FOR-FRAC90
FP * FOR-FRAC90
PUBLIC * FOR-FRAC90
FOR-PROFIT
PUBLIC
CONSTANT
.040
(.068)
.135
.019
(.036)
-.086
20.5***
(3.7)
31.0***
(.153)
(.092)
(4.4)
.030
(.137)
.074
(.074)
7.5*
(4.1)
-.026
.088
-.55
(.062)
(.041)
(1.43)
-.119***
(.037)
-.014
(.036)
-2.68***
(.70)
.392
(.026)
-.004
(.016)
4.90
(.059)
# OBSERVATIONS
376
367
401
R-SQUARED
.036
.020
.418
Sample includes all general acute care hospitalsin operationin Californiain 1990
and 1996. Standarderrors are corrected for group effects and are included in
parentheses.
*Significantat 10%.
*Significant at 5%.
***Significantat 1%.
one-standard-deviation increase in the fraction of hospitals that are for-profit is associated with
an increase of 5.5 in the number of nearby competitors.
In Table 6, I focus exclusively on not-for-profit hospitals and examine whether differences
in financial constraints, market competitiveness, or other factors eliminate the estimated effect of
for-profit competitors on not-for-profit hospitals. The first specification summarized in Table 6
explains the change in Medicaid births as a function of the share of competing hospitals that are
for-profit, FOR-FRAC90.23 As in Table 4, the coefficient estimate is significantly positive, albeit
smaller in magnitude. A one-standard-deviation increase in FOR-FRAC90 is associated with a
.2-standard-deviation change in A MC-BIRTH90-96.
To test whether financially constrained not-for-profit hospitals respond more aggressively to
the change in incentives caused by the DSH program, I include the variable DEBT90/ASSET90
in specification two. This coefficient estimate has the expected sign-more debt leads to a larger
increase in Medicaid deliveries-but the effect is statistically insignificant. Including this variable
in the regression has virtually no impact on the FOR-FRAC90 coefficient, suggesting that the effect
of for-profit hospitals is not operating through an effect on not-for-profit balance sheets.
The next specification controls for the number of competing hospitals within ten miles of the
not-for-profit hospital. As was shown in the previous section, for-profit hospitals tend to be located
in densely populated areas with relatively many hospitals in a given geographic area. Although
the inclusion of this variable reduces the FOR-FRAC90 estimate by approximately one-third, it
remains significantly related with the change in Medicaid deliveries from 1990 to 1996.
In specifications four and five I control for two additional factors that may affect the
responsiveness of private not-for-profit hospitals to the DSH financial incentives. The first variable,
23 Thereis no interactionin this case because I
include only not-for-profithospitals in the Table6 regressions.
? RAND 2002.
DUGGAN
TABLE 6
/
445
For-Profit Hospital Penetration and the Behavior of Not-for-Profit Hospitals
A MC-BIRTH90-96
(1)
FOR-FRAC90
404***
(148)
A MC-BIRTH88-90
BEDS90
-.253
(2)
431***
(159)
-.285
(.191)
(.188)
.689***
(.231)
.793***
(.241)
(3)
269*
(4)
443***
(154)
(148)
-.287
(.192)
-.321*
(.190)
.533**
(.238)
.770***
(.234)
413***
(148)
-.255
NUM-HOSPITALS-CLOSE90
(6)
326*
(180)
-.368*
(.191)
(.188)
.639**
(.267)
.686**
(.290)
166
(124)
DEBT/ASSET90
121
(125)
8.21
8.29
(5.61)
(7.00)
234**
(97)
NOTCH-HOSPITAL90
171*
(96)
42
PUBLIC-HOSPITAL-CLOSE90
CONSTANT
(5)
5
(81)
(115)
7
(43)
-50
(74)
(44)
-43
(48)
-7
(43)
-79
(70)
Numberof observations
212
187
212
212
212
187
R2
.106
.129
.120
.126
.107
.156
0
Sample includes all private not-for-profitgeneral acute care hospitals in operation in California in 1990 and 1996.
Dependentvariableis the change in the numberof Medicaid-insuredbirthsat each hospital,andeach explanatoryvariable
is definedin the text. Standarderrorsare correctedfor groupeffects and are includedin parentheses.
*Significantat 10%.
**Significantat 5%.
***Significantat 1%.
is set equal to one if the privatehospital had a low-income numberof
NOTCH-HOSPITAL90,
15% or more when DSH was first introduced,and thus a large incentive to respondto the new
incentives, and zero otherwise. The statisticallysignificantestimate for this coefficient has the
predictedsign-private not-for-profithospitalsclose to or abovethe DSH qualifyingthresholddo
respondmore aggressivelyto the change in financialincentives.But the inclusion of this variable
does not reducethe FOR-FRAC90coefficientestimate.
I then investigate whether being close to a public hospital is significantly related to the
responseof privatenot-for-profithospitalsto the DSH financialincentives.The variablePUBLICHOSPITAL-CLOSE90takes on a value of one if the not-for-profitis within ten miles of a
government-ownedhospital, and zero otherwise. The estimate has the predicted sign but is
statisticallyinsignificantanddoes not affectthe coefficientestimateon the FOR-FRAC90variable.
Inthe finalspecificationI controlfor all of thesefactorssimultaneously.Therobustrelationship
between for-profithospital penetrationand the response of not-for-profithospitals to the DSH
financialincentivesremains.The resultspresentedin this section andin the previousones suggest
thatfor-profithospitalsexert a "peereffect"on theirnot-for-profitcounterparts,as these facilities
mimic the behaviorof profit-maximizingfirmswhen they actively compete with them.24
24 See
Amould, Bertrand,and Hallock (2000) for evidence suggestingthat HMOs affect the objectivefunctionof
not-for-profithospitals.
? RAND 2002.
446
/
THE RAND JOURNALOF ECONOMICS
6. Conclusion
The findings presented in this article indicate that the behavior of not-for-profit hospitals does
vary systematically with the share of competing hospitals that are organized as profit-maximizing
firms. Not-for-profits in for-profit intensive areas are much more responsive to financial incentives
than are other not-for-profit hospitals. In the case of California's DSH program, this greater
responsiveness partially explains the substantial variation across market areas in the impact of a
significant change in hospital financing that was intended to improve medical care for the poor.
While not-for-profit hospitals in Los Angeles and similarly for-profit intensive areas responded
aggressively to the DSH incentives, their counterparts in San Francisco and other areas with
relatively few for-profit providers did not.
Markets served by relatively many for-profit hospitals are different from the average hospital
market in California. But controlling for these other factors does not eliminate the for-profit effect
described above. While the evidence suggests that for-profit hospitals have a modest effect on
the financial constraints of not-for-profit hospitals, the likely explanation for the heterogeneity
in not-for-profit behavior is a peer effect. Not-for-profit hospitals mimic the behavior of profit
maximizers when actively competing with them, suggesting that the share of hospitals that are
private for-profit in an area will understate the extent to which providers in a market are profitoriented. Whether private for-profit firms have a similar impact on the objective function of their
not-for-profit competitors in other sectors is an important topic for future research.
*
References
K. "Does Managed Care Change the Mission of NonprofitHospitals?
ARNOULD,
R., BERTRAND,
M., ANDHALLOCK,
Evidence from the ManagerialLaborMarket."NBER WorkingPaperno. W7924, 2000.
J. ANDGRUBER,J. "Saving Babies: The Efficacy and Cost of Recent Changes in the Medicaid Eligibility of
CURRIE,
PregnantWomen."Journal of Political Economy,Vol. 104 (1996), pp. 1263-1296.
D.M. AND HORWITZ,J.R. "ConvertingHospitals from Not-For-Profitto For-ProfitStatus: Why and What
CUTLER,
Effects?"In D. Cutler(ed.) The ChangingHospitalIndustry:ComparingNot-For-Profitand For-ProfitInstitutions.
Chicago: Universityof Chicago Press, 1999.
ANDGRUBER,
J. "Does Public InsuranceCrowd Out PrivateInsurance?"QuarterlyJournal of Economics, Vol.
111 (1996), pp. 391-430.
M. "HospitalOwnershipand Public Medical Spending."QuarterlyJournal of Economics, Vol. 115 (2000a),
DUGGAN,
pp. 1343-1374.
. "HospitalMarketStructureandthe Behaviorof Not-For-ProfitHospitals:EvidencefromResponsesto California's
DisproportionateShareProgram."NBER WorkingPaperno. W7966, 2000b.
EGGLESTON,
K., MILLER,
N., AND ZECKHAUSER,R. "OwnershipStructureand ProviderBehavior."Mimeo, 2001.
M.S. "HospitalCost Inflation:A Study of NonprofitPrice Dynamics."AmericanEconomicReview, Vol. 61
FELDSTEIN,
(1971), pp. 853-872.
D.S. "The Supply of Charity Services by Nonprofit Hospitals: Motives and Market
FRANK,R.G. ANDSALKEVER,
Structure."RANDJournalof Economics, Vol. 22 (1991), pp. 430-445.
E. ANDSHLEIFER,
A. "Not-for-ProfitEntrepeneurs."
Journalof Public Economics,Vol. 81 (2001), pp. 99-115.
GLAESER,
M. ANDSTAIGER,D. "ComparingHospitalQualityat For-ProfitandNot-For-ProfitHospitals."In D. Cutler,
MCCLELLAN,
ed., The ChangingHospital Industry:ComparingNot-For-Profitand For-ProfitInstitutions.Chicago:Universityof
Chicago Press and NBER, 1999.
B.R. "AnIllustrationof a Pitfall in Estimatingthe Effects of AggregateVariableson Micro Units."Reviewof
MOULTON,
Economicsand Statistics, Vol. 72 (1990), pp. 334-338.
NEWHOUSE,J.P. "Towardan Economic Theory of NonprofitInstitutions:An Economic Model of a Hospital."American
EconomicReview,Vol. 60 (1970), pp. 64-74.
E.C. ANDSTAIGER,D.O."HowHospital OwnershipAffects Access to Carefor the Uninsured."RANDJournal
NORTON,
of Economics, Vol. 25 (1994), pp. 171-185.
SILVERMAN,E. AND SKINNER,J. "AreFor-ProfitHospitalsReally Different?Medicare'Upcoding' andMarketStructure."
NBER WorkingPaperno. W8133, 2001.
B. TheNonprofitEconomy.Cambridge,Mass.: HarvardUniversityPress, 1988.
WEISBROD,
J.S. ANDEPSTEIN,
A.M. Falling Throughthe SafetyNet. Baltimore:Johns HopkinsUniversityPress, 1994.
WEISSMAN,
? RAND 2002.