POLICY RESEARCH WORKING PAPER                3 05 8
Working for God?
Evaluating Service Delivery of Religious Not-for-Profit
Health Care Providers in Uganda
Ritva Reinikka
Jakob Svensson
The World Bank
Development Research Group
Public Services
May 2003



POLICY RESEARCH WORKING PAPER 3058
Abstract
Reinikka and Svensson exploit a unique micro-level data      counterparts, although government facilities have better
set on primary health care facilities in Uganda to address   equipment. These findings are consistent with the view
the question: What motivates religious not-for-profit        that RNP facilities are driven in part by altruistic
(RNP) health care providers? The authors use two             concerns and that these preferences matter
approaches to identify whether an altruistic (religious)     quantitatively. Second, the authors exploit a near natural
effect exists in the data. First, examining cross-section    experiment in which the government initiated a program
variation, they show that RNP facilities hire qualified      of financial aid for the RNP sector. They show that
medical staff below the market wage, are more likely to      financial aid leads to more laboratory testing of
provide propoor services and services with a public good     suspected malaria and intestinal worm cases, and hence
element, and charge lower prices for services than for-      higher quality of service and lower prices, but only in
profit facilities, although they provide a similar           RNP facilities. The findings suggest that working for
(observable) quality of care. RNP and for-profit facilities  God matters.
both provide better quality care than their government
This paper-a product of Public Services, Development Research Group-is part of a larger effort in the department to
evaluate frontline health care delivery in poor countries. Copies of the paper are available free from the World Bank, 1818
H Street NW, Washington, DC 20433. Please contact Hedy Sladovich, room MC3-607, telephone 202-473-7698, fax 202-
522-1154, email address hsladovich@worldbank.org. Policy Research Working Papers are also posted on the Web at http:/
/econ.worldbank.org. The authors may be contacted at rreinikka@worldbank.org or jsvensson@worldbank.org. May
2003. (49 pages)
The Policy Research Working Paper Senes disseminates the findings of work in progress to encourage the exchange of tdeas about
development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished The
papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this
paper are entirely those of the authors. They do not necessanly represent the view of the World Bank, its Executive Directors, or the
countnes they represent.
Produced by Partnerships, Capacity Building, and Outreach



Working for God?*
Evaluating Service Delivery of Religious Not-for-Profit Health Care
Providers in Uganda
Ritva Reinikkatand Jakob Svenssonl
*We are grateful for comments and suggestions by James Adams, Markus Goldstein,
Daniele Guisti, Michael Kremer, Magnus Lindelow, Peter Okwero, Tomas Philipson, Halsey
Rogers, Rob Yates, and Poverty and Applied Micro seminar participants at the World Bank.
We also thank Makerere Institute of Social Research for their assistance in survey implemen-
tation Financial support from the Japanese PHRD grant is gratefully acknowledged The
findings, interpretations, and conclusions expressed in this paper are entirely those of the au-
thors and do not necessarily represent the views of the World Bank, its Executive Directors,
or the countries they represent Working papers describe research in progress by the author(s)
and are published to elicit comments and to further debate.
IDevelopment Research Group, The World Bank. Email rreinikka@worldbank.org
lInstitute for International Economic Studies, Stockholm University, Development Re-
search Group, The World Bank, and CEPR. Email. jakob svensson©iies.su.se






1    Introduction
In many developing countries the not-for-profit sector plays an important role
in the provision of social services. In the health and education sectors, re-
ligious organizations are particularly prevalent. What implications does this
have for the quantity, quality, and the price of services? Clearly, the answer
to this question requires knowledge about the not-for-profit actors' objectives
and behavior as service providers. Despite their importance, however, there is
currently little such information available from developing countries.' In this
paper, we fill this gap with unique data obtained from a quantitative survey of
government, private for-profit, and private not-for-profit (religious) providers
of primary health care in Uganda.
The literature provides two general explanations for what drives not-for-
profit actors. The first explanation is based on the premise that not-for-profit
health facilities are driven by altruistic (or religious) concerns. This is the very
reason why they have chosen to operate as a "not-for-profit." The alternative
hypothesis stresses the potential benefit of the regulatory not-for-profit status.
Specifically, not-for-profits (at least in the United States) enjoy exemption from
taxation and are the main beneficiaries of charitable donations. At the same
time, not-for-profits cannot directly appropriate profits. Thus, any surplus
must be used to finance perks (wages and perquisites) for the management
and/or staff. The differential tax treatment and differential access to grants
and donations between for-profit and not-for-profit may explain why it may be
optimal for an entrepreneur to choose a not-for-profit status. The regulatory
status, i.e., the nondistribution constraint, may also have an additional benefit
as hypothesized in Hansmann (1980) and formalized in Glaeser and Shleifer
(2000). Namely, the nondistribution constrain will make profits less valuable
1 A large literature attempts to identify the behavior of not-for-profit firms or organizations
in the developed world, especially in the United States. The theoretical work has mainly
evolved around three types of models; altruism models, which have quantity and quality
of output in the firm's objective function; physician cooperative models that are analogous
to earlier cooperative firm theories (Pauly and Redisch 1973), and non-contractible quality
models, where for-profit firms have an incentive to shirk on the quality of service to cut costs
(for a review, see Malani, Philipson, and David 2002; Lakdawalla and Philipson 2001) With
respect to the U.S. health sector, where most services are produced by the not-for-profit
sector, empirical evidence shows that not-for-profits are larger (for example, Rose-Ackerman
1996), produce the same or higher quality services than for-profit firms (Sloan and others
1998, McClellan and Staiger 2000), while the results oni costs are mixed. Furthermore, the
empirical literature finds that demand shocks increase the market share of for-profit firms;
tax increases either raise the share of not-for-profits, or does not affect it at all; and somewhat
surprisingly, that not-for-profit firms appear to be more profitable than for-profits However,
their prices are not significantly higher than for-profits (Malani, Philipson, and David 2002;
Philipson 2000, Sloan and others 1998)



for not-for-profit firms, which in turn will provide them with softer incentives
and thereby protect consumers from ex post appropriation. Since private for-
profit firms are more responsive to profits, they will have stronger incentives to
cut costs and pursue ex ante nonverifiable quality reductions on the service(s)
provided. This commitment problem may lead entrepreneurs to choose the
not-for-profit status as a means of committing not to reduce quality.
The evidence, almost exclusively based on data from the United States, on
distinguishing the different theories of non-profit firms is mixed (see footnote
1). One reason it may be difficult to disentangle the three theories is that the
type of ownership may be endogenous. For example, a non-altruistic entrepre-
neur may choose a not-for-profit status and locate in a poor neighborhood if
she expects to benefit from charitable donations as a consequence of this own-
ership/location choice. Thus, although the location choice will have adverse
financial consequences, higher expected donations will compensate for them
and make the ownership/location choice optimal (i.e., the total expected cash
value of perks is higher when taking donations into account).
In this paper we exploit data from primary health care providers in Uganda.
Studying a poor developing country has two clear advantages when assessing
the underlying preferences of service providers. First, as discussed in more
detail below, not-for-profit primary health care providers in Uganda are to a
large extent self-regulated. The nondistribution constraint may therefore not
be binding, in which case the Hansmann (1980) and Glaeser and Shleifer (2000)
theory is less relevant. Second, the not-for-profit facilities have no obvious tax
advantages over private-for-profit firms. In fact, for the mostly rural primary
health facilities we are considering, neither type of owner pays any direct or
indirect taxes. Moreover, until 1999/2000 (the fiscal year for which we have
data) donations play a marginal role in financing primary health care providers'
recurrent operations. Thus, at least prior to 1999/2000, there were no obvious
advantages for a nonaltruistic entrepreneur to choose a not-for-profit status.
However, the self-regulatory status raises another concern, namely that the
preferences of the owner, or the founder (for instance a Catholic parish) and
the manager may differ. In particular, the facility may be captured by a man-
ager with different objectives from the owner.2 Clearly, capture can take many
forms (and be of varying degrees) and some forms (degrees) cannot easily be
2 Glaeser (2002) argues that weak board control may be just as important as the differential
tax privileges/donations and nondistribution constraint in explaining the behavior of the not-
for-profit firms Thus capture is not specific to not-for-profits in poor developing countries,
although it seems plausible that boards in general have stronger control in the U.S. not-for-
profit sector than in the Ugandan primary health care sector (see discussion in section 2). The
capture argument is also close to the Pauly and Redisch (1973) view of hospitals as physicians'
cooperatives.
2



measured. In particular, as argued by Glaeser (2002), the workers' preferences
are often themselves altruistic and they often internalize the stated goals of the
firm or organization. What we can test is whether the not-for-profit service
facilities act as profit- or perquisites-maximizing firms, or whether they are
systematically different.
We take two approaches to identify whether religious affiliation (not-for-
profit status) matters. The first builds on the assumption that we can identify
the behavior of the not-for-profit providers by comparing their performance in
various dimensions with government and private for-profit providers. Specifi-
cally, we exploit the cross-section variation across types of ownership, control-
ling for other confounding observable characteristics and unobserved location-
specific effects. The idea is that since the behavior of private for-profit providers
(presumably driven by profit maximization) and government-operated units (to
a large extent regulated by central and local authorities to deliver a minimum
package of services) is generally quite well understood, by comparing outcomes,
we can learn about the objectives of the not-for-profit actors. To guide the
empirical work, we develop a simple model of the not-for-profit providers and
derive the implications for their choice of wages, prices, service mix, and quality
of care under the two alternative hypotheses laid out above.
The second approach relies on a near natural policy experiment of public
financial aid for the not-for-profit sector.3 In 1999, the year of the survey,
the government initiated a program in which each not-for-profit health unit
in a given category (e.g., dispensaries) was to receive an untied grant for the
fiscal year. As this was a new and unanticipated program and due to poor
communications on the government's part, some not-for-profit facilities did not
receive their grant until the following year. This de facto phasing-in of the
financial aid program provides a source of (what we argue to be) exogenous
variation that we can exploit to identify the effects of ownership.
In the cross section, we find that religious not-for-profit facilities hire qual-
ified medical staff below the market wage. Moreover, religious not-for-profit
facilities are more likely to provide pro-poor services and services with a public
good element, and charge strictly lower prices for services than for-profit units.
Religious not-for-profit and for-profit facilities both provide better quality care
than their government counterparts, although government facilities have better
equipment. These findings are consistent with a premium in working in a reli-
3Duggan (2000) also studies the diffential response of not-for-profit versus for-profit health
facilities (hospitals) to a natural experiment induced by a government subsidy program. He
examines hospitals affected by California's Disproportionate Share program and shows that
the behavior of not-for-profit hospitals varies with the share of nearby hospitals organized as
for-profit firms. increased for-profit penetration makes not-for-profit hospitals more profit-
oriented.
3



gious not-for-profit facility and that religious not-for-profits are driven (partly)
by altruistic concerns. The near natural experiment reveals that financial aid
leads to more testing of suspected malaria and intestinal worm cases and lower
prices in religious-not-for-profit facilities. Moreover, the estimated effects are
substantial. Thus, working for God appears to matter!
The rest of the paper is organized as follows. Section 2 describes the institu-
tional setting of health care in Uganda, including ownership and management
of health facilities (in particular dispensaries), the labor market and human
resources in health care, and the government financial aid program to not-
for-profit health providers. Sections 3 and 4 present a simple model of the
behavior of a religious not-for-profit health facility, develop two extensions of
the model, and lay out the inference procedure. Section 5 discusses the survey
data used in the empirical analysis. Section 6 presents the empirical evidence
from the cross-section regressions. Section 7 explores the impact of financial
aid on service delivery. Section 8 concludes.
2   Hlstitutional setting ian healt          care
It is commonly held that Uganda had a well-functioning public service delivery
system in the 1960s. Health care was provided free of charge, and access to care
was relatively good. Steady improvements were experienced in most health
indicators. However, as a result of the political and military turmoil of the
1970s and 1980s, the government de facto retreated from funding and providing
public services. In primary education, which was nationalized in the 1970s,
the vacuum was filled by parents who gradually took over running the public
schools (Reinikka and Svensson 2002). In health care the burden was taken
up by private for-profit sector and religious providers. The latter were able
to mobilize external resources to sustain activities during the turbulent times
(Republic of Uganda 2001a). Despite efforts by the private for-profit and not-
for-profit sectors, health indicators fell dramatically.
Following restoration of peace in the late-1980s and subsequent economic
recovery, government implemented a major program of infrastructure rehabili-
tation in the (public) health sector in the 1990s. This coincided with political,
administrative, and financial decentralization, which led to limited recurrent
funding for health facilities, as districts prioritized areas other than health care
(Jeppson 2001). As a result the quality of public services did not improve
at the same pace with infrastructure, which is reflected in the continued high
demand for privately provided care (Hutchinson 2001). Some key health in-
dicators did not improve as expected, despite a GDP growth rate of over 6
4



percent and a 40-percent reduction in consumption poverty in the 1990s (Ap-
pleton 2001). Specifically, the infant mortality rate stagnated during the latter
half of the 1990s at 88 deaths per 1,000 live births in 2001 (Republic of Uganda
2002). Mortality during the first month is particularly high. The underlying
reasons for the stagnant trend include high fertility, short birth intervals, and a
high proportion of home deliveries. These factors remained constant during the
period. The fall in immunization rates-partly attributed to decentralization-
made the situation worse (Moeller 2002). HIV/AIDS must play a role in this
stagnation, but malaria (51 percent) and diarrhoea (19 percent) are the most
important direct causes of infant mortality in Uganda. Recent trends in the
clinical causes of infant deaths are, however, not available.
Today the modern health sector in Uganda is composed of four types of
facilities: hospitals, health centers, dispensaries (health center III), and aid
posts or subdispensaries. These facilities can be government, private for-profit,
or private not-for-profit operated and owned. The health facility survey we
exploit in this paper has the dispensary (with or without a maternity unit) as
its unit of observation. Dispensaries are the most common health facility and
an important health service provider in Uganda. Most dispensaries are rural
(89 percent).
According to the government health sector strategic plan, the standard
for dispensaries includes preventive, promotional, outpatient care, maternity,
general ward, and laboratory services (Republic of Uganda 2000). An average
dispensary has eight beds for inpatient care and serves a population of 20,000.
Dispensaries usually do not have a medical doctor (although some do), and are
managed by a either clinical officer or a comprehensive or registered nurse.
Factors that have determined where dispensaries have been located over
time include population density and growth, accessibility, existence of other
units, health needs, and poverty rates. Similarly, national and local politics,
communities building their own units, and the preferences of not-for-profits,
donors and even the colonial authorities have had their impact. More re-
cently, both the government and the medical bureaux have attempted to provide
(sometimes conflicting) guidelines on where to put new not-for-profit facilities,
mostly centered around filling gaps in geographic coverage.
2.1 Ownership and management of health facilities
The private not-for-profit health sector in Uganda consists of religious and
nonreligious providers.4 The first ever census on the not-for-profit health care
4This section draws considerably on interviews and communications with health praction-
ers and public officials in Uganda, as many aspects of private health care provision are poorly
5



sector in Uganda carried out in 2001 indicated that autonomous dioceses and
parishes own 70 percent of all private not-for-profit health facilities, which
total 450 lower-level units and 42 hospitals (Republic of Uganda 2001b). The
rest are owned by nongovernmental organizations (16 percent), some of which
are also religious, community-based organizations (6 percent), and by district
councils, mosques, and individuals (8 percent). The census also shows that
most private not-for-profit health facilities (82 percent) are coordinated by one
of three national umbrella organizations: Catholic, Protestant, and Muslim
medical bureaux.
The first religious not-for-profit health unit was established by missionar-
ies in 1897 (Republic of Uganda 2001a). Thereafter churches and missionar-
ies have set up hospital and health centers throughout the country.5 At their
departure, missionaries handed over the management to the local church (dio-
cese or parish). In the last three decades, as new parishes were established,
they (usually initiated by the parish priest) routinely set up their own social
services, particularly health services. In many cases parishioners contributed
to the investment cost of these facilities, sometimes aided by donations from
the respective medial bureau or outside sources. The majority of dispensaries
owned by religious providers were built between 1960 and 1990. In our sample,
the median year of establishment is 1983.
Not-for-profit health care providers are self-regulated and self-governing. At
the time of our survey, there was no certification for not-for-profit status (either
by a medical bureau or by the government). Hence, the in-charge of the not-
for-profit health unit together with the unit-specific management committee
were free to decide on the mix and price of services provided by the facility.
The Uganda Catholic Medical Bureau recently initiated an accreditation
process and issued guidelines for lower-level health units affiliated with it stating
the following (2001, p.1):
"Diocesan health units are different, they do not aim at profit. The
money left over is to be used for improving the services, lowering
the fees, or increasing the salaries. No funds should go to the owner
or the parish. These health units aim to offer health services as
good and as cheap as possible, from a healing mission. They try
to be well organized, aim at integration in the District, and are
community-oriented."
documented
5Since expatriates were not allowed to own fixed assets, missionaries established the units
in the name of the diocese or paxish
6



The accreditation procedure was initiated in response to government's financial
aid program to non-profits (see details in section 2.3). But in 1999/2000 (the
year of the survey), if any of these rules were followed, compliance by the
facilities was voluntary.
It is worth noting that the institutional structure of the not-for-profit sector
is considerably different from the government's institutional framework. Most
importantly, the medical bureaux operated by various religious denominations
do not have administrative authority over the individual units or owners (that
is, dioceses or parishes).
In the publzc sector, the Health Sector Strategic Plan determines facility
standards and the mix of services to be provided at each level (Republic of
Uganda 2000). Both central and local government authorities attempt to en-
force these standards by controlling inputs, say, by setting staffing norms, by
supplying pharmaceuticals, vaccines, and equipment, and by direct transfers
and investment funding. In addition, they issue management and technical
guidelines and supervise health facilities. Public health facilities have also a
unit-specific management committee to represent the local community and pa-
tients. Due to a variety of factors, such as difficulties in recruitment of qualified
medical staff, and the availability or absence of funding, the actual picture on
the ground may vary from the set standards.
Prvate for-profit practice began decades ago with a few medical practition-
ers in urban areas. Their numbers grew dramatically during the economic and
political turmoil of the 1970s and 1980s (Republic of Uganda 2001a). Private
health care was provided by a mixture of licensed and unlicensed private clinics,
pharmacies, drug shops, and home providers,6 and little systematic informa-
tion is available on these providers. Many medical professionals working in the
public sector are believed to also have a private practice to earn extra income
(McPake and others 1999), but factual evidence of this practice, particularly
in the case of dispensaries, is limited.
Finally, while all health service providers are exempt from the value added
tax, private for-profit provider are, in principle, expected to pay income tax,
as well as the pay-as-you-earn tax for their employees. But there are major
problems in compliance. Thus, apart from a few private clinics in Kampala (the
capital), the private for-profit dispensaries are de facto tax exempt. Religious
providers are also exempt from the income tax.7
6In principle, the national professional councils are supposed to regulate both private for-
profit and not-for-profit facilities (but not government facilities) by licensing them, setting
standards, and monitoring their premises. This regulatory system is not working in practice.
7Actually, the current tax code in Uganda does not even recognize the not-for-profit status
Still (or as a result of this), there have been attempts by local authorities to impose taxation on
the not-for-profit facilities' revenues, but the facilities and medical bureaux have successfully
7



2.2   Labor market and human resources in health care
While the total number of medical personnel has increased since 1972, it has
not kept pace with population growth (Hutchinson 2001). Many medical stu-
dents leave the country after completing their education to work in neighboring
countries or go further afield where salaries are higher. In 1996 one medical
person was available for every 2,350 people, while in 1972 one medical per-
son was available for half that many people. The absolute number of doctors
actually declined during the same period (by 18 percent).
The distribution of medical personnel is uneven. The three largest cities
with less than 10 percent of the country's population account for nearly 60
percent of its doctors. Since almost half of government health sector employees
work in hospitals, many careseekers bypass the lower-level units and use hos-
pitals instead: in 1998 only 15 percent of hospital patients had been referred
from lower-level facilities (Okello and others 1998).
The quality of health personnel is also a major problem. In the public
and not-for profit sectors, about 40 percent of established health positions are
currently filled by staff with medical training. The remaining positions are
filled by nursing aides-with no medical training-hired by local governments
or facilities themselves, or are left vacant.
2.3   Financial aid for non-profits
The two umbrella organizations for not-for-profit health providers-the Uganda
Protestant Medical Bureau and the Uganda Catholic Medical Bureau-were
established in the 1950s to coordinate disbursement of government grants to
religious health care providers. While public subsidies continued after indepen-
dence, over time the relations between religious providers and the government
deteriorated, as there was competition and a perceived difference in pay and
privileges (Republic of Uganda 2001a). During the decline in public services in
the 1970s and 1980s, subsidies to not-for-profits dwindled and eventually ceased
altogether. In response to the disappearing public support, not-for-profits had
to resort to user fees and external donations. The two bureaux also established
a joint medical store to supply their affiliated facilities with drugs and other
medical consumables and equipment.8 In the early 1970s the Uganda Muslim
Supreme Council also established a similar permanent coordinating structure.
Over time, the importance of external donations declined. In our sample
lobbied against the tax claims.
8Today all types of health-care providers can purchase drugs from the joint medical store
and hence take advantage of its bulk purchase prices.
8



of (religious) not-for-profit facilities, only 3 out of 44 not-for-profit dispensaries
received donations from private sources and only 2 out of 44 facilities received
funds from the donor community in fiscal year 1999/2000.9
In 1997/98 the government reinstated financial aid to hospitals. In 1999/2000
a new program extended a similar subsidy to lower-level health units. The fi-
nancial aid program prescribed that each not-for-profit unit in a given category
(e.g., dispensaries) was to receive a fixed-amount grant for the fiscal year. Each
dispensary was to receive the same amount, namely 2.5 million shillings ($US
1,400) a year. Each dispensary with a maternity unit was to receive 3.4 million
Ush ($US 1,900). In subsequent years the allocation system was to become
more refined, to include both performance and needs-based criteria. The ad-
ministrative problems getting program off the ground are discussed in section
7.
3 Conceptual framework
Next we lay out a simple framework to analyze not-for-profit behavior. We
consider three models. The first two models implicitly assume that the (altru-
istic owner's) not-for-profit facility is captured by a nonaltruistic manager(s).
In the first model, the nondistribution constraint is not binding, so the not-
for-profit provider acts as a profit-maximizer. In the second version, we as-
sume the constraint binds (i.e., it is enforced), in which case the not-for-profit
provider maximizes perquisites instead. The third model assumes that (reli-
gious) not-for-profit facilities maximize the total health impact of its activities,
here conceptualized as the number of patients treated.10
We start by solving the simplest version of the three models, and then
consider two extensions: endogenous quality and costs.
3.1 Basics
A manager for a not-for-profit health facility i must hire workers to work in the
facility and agree on wages w,,. Each worker j can perform one task or service.
9As stressed above, donations were more important in the 1970s and 1980s, as well as at
the start-up phase, when raising funds for construction We have some indirect evidence for
the latter. Of the 29 not-for-profit facilities that had renovated their facility in the past, 14
received financial support from private and/or donor sources.
'°Clearly, conceptualizing altruism in the health sector with the number of patients treated
is not uncontroversial. See Malani, Philipson, and David (2002) for a review of altruism
models that typically have quantity (and/or quality) of output in the not-for-profit's objective
function.
9



There are S potential services. Thus, a facility can at the most have S workers.
There is a pool of workers who differ with respect to the value placed on working
in a not-for-profit facility. Specifically, a worker j's utility is u(wi,) + 3,NFP,
where u(w) is a concave function, NFP is an indicator variable taking the
value 1 if the worker is employed by an altruistic not-for-profit facility (and
zero otherwise), and 6, is the nonmonetary gains of worker j of working in a
not-for-profit facility. 6j is thus a measure of a worker's altruism or (religious)
beliefs about the importance of working in a not-for-profit facility. We label
3j as the "religious premium". Each worker can get a job in the public sector,
which pays the wage v.11
The manager must also decide what services to provide and prices of these
services. The total cost of producing a given service s e S that x, patients
will be buying is w, + cx8, where w, is the wage cost of the worker assigned
to produce service s, and c is the (constant) marginal cost. We thus assume
that a worker will be paid the same amount irrespective of the number patients
treated. The inverse-demand function for health service s is ps = P(x5) where
pa is the price and PI (xZ) < 0. We let es denote the elasticity of demand with
respect to price for service s. The facility is assumed to be a local monopolist.
3.2   The profit-maximizing not-for-proilt facility
The total cash profits of facility i is 7r = Es, [P(x,)x, - w, - cz,], where S, is
the set of services offered by facility i. We assume that workers do not obtain
any additional noncash gains of working in a profit-maximizing not-for-profit
facility; that is, 6, = 0. A profit-maximizing facility can hire an unlimited
number of workers at the wage iD. Its maximization problem is thus,
SI
max E [P(x,)x, -      C-X] -                  (1)
The first order condition of activity s can be stated as,
P(Xz) 1-    ] -c <0.                        (2)
Equation (2) is a standard condition for profit maximization; the price will
be set to equate the marginal revenue (first term in (2)) with the (constant)
marginal cost. Equation (2) implicitly defines the optimal quantity x* and by
"The assumption of excess demand of workers by the public sector at (the administrative
set) wage iv, is a good approximation of the health market for qualified staff in Uganda given
the economywide shortage of qualified staff.
10



the inverse-demand function the price p* that creates this demand. Since the
marginal cost is the same for each service, the marginal revenue for each service
being provided must also be the same. That is,
P(Xz) [1 -    = P(xt) 1-] = C.                    (3)
Clearly, from (3) it follows that the facility will charge a higher price for
the service with low elasticity of demand.
Equation (2) is a necessary condition for profit maximization. In addition,
each service must yield non-negative profits. That is, the facility will provide
the service s only if
P(Z*8X* )8-  - cx* > 0.                     (4)
3.3 The perquisites-maximizing not-for-profit facility
Following Glaeser and Shleifer (2001), we assume that if the nondistribution
constraint binds, the manager is forced to spend profits on perquisites, denoted
by z. The utility of spending profits on perquisites is v(z) = az, where ca < 1
is a constant.
As with the profit maximizer, we assume that workers do not obtain any
additional noncash gains of working in a perquisites-maximizing not-for-profit
facilities; that is, 53 = 0. Its maximization problem is thus,
SI
max a E [P(xz)xs-w-cxs].                        (5)
Clearly, the first-order condition of activity s, and the non-negative profit
constraint are identical to (2) and (4). Thus, a perquisites-maximizing not-for-
profit facility will set the same prices p* as a for-profit facility. Moreover, it
will pay workers the same wage as a for-profit facility, and it will also choose
to provide the same set of services.
If private-for-profit and private not-for-profit facilities only differ in the
ease in with which a facility can appropriate profits, and if facilities decision
variables are (i) which services to provide, (ii) the prices of these services, and
(iii) wages to their workers, we should not observe any differences between
private-for-profit and private not-for-profit facilities.
3.4 The altruistic not-for-profit facility
The third assumes that private not-for-profit facilities maximize the total health
impact of its activities. Clearly, the total health impact could be defined in a
11



variety of ways. Here we choose to operationalize it as the number of (poor)
patients treated. That is, the private not-for-profit facilities maximize ZSt Xs,
subject to the constraint that Sst [P(X8)X8 -  - cx8] > 0.
Consider first the choice of workers. A manager for an altruistic facility
will try to hire workers biased toward working in a not-for-profit facility. To
simplify the exposition, assume there are two large group of workers, one with
68 = 0 and one with 6, = 6 > 0. The not-for-profit facility will hire workers
with 6, = 6 and pay them the wage w = u-1 [u(wD) - 6]. Note that w < w.
Thus, the not-for-profit facility will exploit the workers' moral gains of working
in a not-for-profit facility by offering a lower wage. The wage is set so that at
the margin, a worker with a positive religious premium is indifferent to working
in a not-for-profit facility or a for-profit facility.
To solve the altruistic not-for-profit manager's maximization problem we
formulate the Lagrange function,
Si                  S,\
L = Exs     + A(    [P(xs)xs - D  -c]) X           (6)
where A is the Lagrange multiplier. Maximizing the Lagrange function yields
the following first order conditions,
+ A[P(X) [      eI]  c] <0 VsESi                 (7)
and
SI
Z  [P(X8)X8- i-cX8] = 0                     (8)
Dividing the first-order conditions (7) for two services s and t, we see that
P(x[) 1 - 6 - c
That is,
P(XZ) [1 -  ]   P(xt) [1   ]                 (10)
Thus, as for the profit- or perquisites-maximizing not-for-profit facility, the
marginal revenue for each service being provided will be the same. Thus, higher
prices will be charged for services with low elasticity of demand. The intuition
is straightforward. Given the zero-profit condition (8), and given that different
patient types are perfect substitutes, if the marginal revenues differ, the facility
12



can provide one less patient with the service with the lowest marginal revenue,
and instead provide more than one extra patient with the service with the
highest revenue. Thus, by shifting types of patients treated, the aggregate
number of patients treated could be raised.
Note that (7) implies that prices will be set such that the marginal return is
strictly lower than the marginal cost. That is, an altruistic not-for-profit facility
will charge lower prices than a profit- or perquisites-maximizing not-for-profit
facility. Thus, an altruistic provider will tend to cross-subsidize services. In
particular, whereas a perquisites-maximizing not-for-profit facility (or a private
for-profit facility) would never provide a service it cannot make a positive profit
from; i.e., for which (4) does not hold, an altruistic provider may do so in order
to increase the total number of patients treated.
3.5 Quality of care
So far we have assumed that quality of care is exogenous. Assume now in-
stead that before (or simultaneously) choosing what services to provide, the
manager/facility also makes an effort choice that influences the quality of the
services being provided. Let the inverse demand function be Pa = P(x8, q),
where q is effort and Pq > 0 and Pxq > 0. We assume that higher quality
services imply both higher financial and nonfinancial (effort) costs. Total cash
profit is now 7r = E  [P(x8,q)x - w- cx, - C(q)], where Cq(q) > 0 and
Cqq(q) > 0. The manager must also bear a noncash cost of exerting effort given
by y (q), where -yq (q) > 0 and Yqq (q) > 0.
Consider first a for-profit provider. The additional first-order condition is
given in (11),
s,
_'Yq (q) + E Pq(x, q)zx - Cq(q) < 0             (11)
The two first-order conditions (11) and (2) define the optimal price and
quality for service s.
The first-order condition for the quality choice for a perquisites-maximizing
provider is
-Yq (q) +   (  Pq(xa, q)X-C,(q)) < 0.              (12)
Totally differentiating the first-order conditions (12) and (2), using the im-
plicit function theorem (see appendix), it is possible to show that the quality of
care of the for-profit facility exceeds that of the perquisites-maximizing facility.
Higher quality of services will also allow the facility to demand a higher price.
13



This is an intuitive result. Providing higher quality services requires nonfinan-
cial costs (effort). Since private for-profit firms are more responsive to profits,
a for-profit provider has stronger incentives to put in high effort.
Consider next the altruistic facility. The first-order conditions of the altru-
istic provider's maximization program are given in (7), (13) and (14).
-7Yq (q) - A  (   Pq(xs, q)x8 - Cq(q)) < 0           (13)
E [P(x3, q)xs - w.- cx] - C(q) < 0                 (14)
Higher quality will increase demand and allow the altruistic provider to
treat more patients.'2 Without further restrictions on the model, however, we
cannot say if the altruistic facility will exert more or less effort than the for-
profit provider. However, what we can say is that only an altruistic provider
will tend to cross-subsidize services, and thus can provide a service it cannot
make a profit from. It will also pay their workers less. Moreover, conditional
on the quality choice being similar, an altruistic provider will charge strictly
lower prices.
3.6   Endogenizing cost
In the baseline model, (marginal) cost is constant and exogenous. However, it
is reasonable to think that facilities can partly influence their cost structure.
For example, a facility could reduce cost by ex post shirking on quality. Below
we consider how such an extension would affect the results.
Glaeser and Shleifer (2001), building on Hansmann (1980), argue that pri-
vate not-for-profit firms face softer incentives which protect consumers from ex
post appropriation. Since private for-profit firms are more responsive to profits,
they will have stronger incentives to pursue cost and nonverifiable quality re-
ductions on the service(s) provided. It is straightforward to incorporate Glaeser
and Shleifer's mechanism in the model
Let the inverse demand function now be given by p5 = P(x,, qe) where qe
is the expected quality of the service being provided, with Pq > 0. Unit cost
is c = C(q), with Cq > 0 and Cq > 0. As in Glaeser and Shleifer (2001), the
manager must bear a noncash cost of 8(qe - q) of shirking on quality.
'2In the standard (reduced form) altruism model, the provider cares about quantity and
quality. Obviously, if quality has its own value for the altruistic provider, this would provide
even stronger incentives to supply high-quality care.
14



In this set up, when the manager chooses q, he has already collected revenues
(thus he takes the price and demand as given). The perquisites-maximizing
not2for-profit facility's optimal quality reducing choice is given by,
-aCq(q) (      2S  +, < 0                  (15)
Rational patients will anticipate the manager's ex post incentives. Thus, in
equilibrium q* = qe. The for-profit provider's equilibrium condition is the
same as in (15), with a = 1.
Total differentiating (15) yields,
dq      Cq
-    d = _ C       < 0 .
da     CeCqq
Thus, the nonverifiable quality of the not-for-profit facility exceeds that of
the for-profit facility. Lower quality (which is expected in equilibrium) will lead
to lower costs. Lower quality will also lead to lower demand. Both factors lead
to lower prices. Lower demand will tend to reduce the number of services that
can be provided, although this force is counteracted by lower cost. Without
further restrictions on the model, it is unclear how service provision will be
affected.
The altruistic facility will have no incentives to shirk ex post on quality,
since this will not affect (ex post) the number of patients that could be treated.
4 Implications and specification
The predictions of the baseline and the extended versions of the model are
summarized in Table 1. The baseline model suggests that we could test for the
not-for-profit facilities' objective function by running the following regression
on a sample of facilities with different owners,
Y. = / + MNPNPz + IpFPt + es                   (16)
where the dependent variable y,, is either s,,; a indicator if service s is being
provided or not by facility i, p,,; the price of service s charged by facility S, wu,;
the wage paid to worker of type j in facility i, or q1; the quality of services. NPi
is a dummy indicating if the facility is not-for-profit, and FPz is a dummy in-
dicating if the facility is private for-profit. The ownership category excluded in
(16) is government. The perquisites-maximizing not-for-profit facility hypoth-
esis suggests that 3P = fy p for y., = {sis,pt., w.3}, whereas the altruistic
15



not-for-profit facility hypothesis suggests that /3 p > /5P, OPP < /PFp, and
ONP < FP-
Endogenizing cost and allowing facilities to also choose quality, makes it
more difficult, using observed prices, wages, and service provision, to distin-
guish between the not-for-profit's objectives. In particular, there are parameter
configurations for which we cannot reject either of the two hypotheses. How-
ever, only the altruistic model, under all model specifications, is consistent with
the prediction that /i p > 88 p, /p p < IpFp, wpP < Fwp and I3 p = ,FaP
This implication therefore forms the basis for the empirical analysis.
While the model provides a starting point to assess the behavior of not-
for-profit facilities, it is clearly based on a number of simplifying assumptions.
Thus, the question is whether an association between ownership and outcomes,
from a regression like (16), is a causal relationship. In particular, the different
types may have other characteristics that are also associated with the dependent
variable y. For example, for-profit and not-for profit providers may locate in
different areas and thus face different demand.
We consider two strategies to identify a causal relationship in the data:
controlling for other confounding observables (discussed below) and exploiting
a near natural experiment of financial aid to not-for-profits from government
(discussed in section 7).
In the cross-section analysis, identification is based on the assumption that
we can control for variables that are confounded with ownership and the de-
pendent variable. Thus, we will estimate an equation of the following form,
Yi. = pY +  YNpNPi + I3 pFP, + 13'y XS + ei,       (17)
where Xi, is a vector of other controls. Below we discuss the controls we use.
In the baseline regression, we proxy for the degree of competition by in-
cluding as a control the "number of competitors", i.e., number of dispensaries
and health centers in the facility's catchment area. In the model, each facility
acts like a local monopolist. In reality, patients have some choice about where
to seek health service (although the data suggest that proximity is the most
important factor overall for selecting a given facility), so the market structure
may be important.
Not-for-profit facilities receive (limited) in-kind support (medicine and staff)
that may shift the marginal cost curve and thus influence y. We explicitly con-
trol for this by including a measure of the value of free drugs received and
a variable capturing the full-time equivalent number of staff working in the
facility for free.
Because each facility's location, in principle, is endogenous, determining
whether it is ownership per se or location or some other factor that is correlated
16



with location that drives any observable differences in outcome could present
a difficult identification problem. However, in practice it is less of a problem.
First, as discussed above, most not-for-profit facilities were established many
years ago. Given the large social and economic changes in Uganda during the
last few decades, the local situation may have changed dramatically for many
facilities. Second, empirically, we can (to some extent) control for location
by including controls such as distance to closest subcounty headquarters and
district-specific effects. Thus, we identify the ownership effects from the within-
district variation. Finally, and most importantly, it is possible to reinterpret the
model, letting the choice of services to provide and the prices of these services,
really be a choice of where to locate. If not-for-profit facilities are driven by
altruistic concerns, they would tend to locate in poor areas where they would
not be able to charge high prices. If not-for-profit facilities are not driven by
altruistic concerns, they would instead tend to locate in areas where they, just
like for-profit facilities, could maximize profits. This reduced form approach is
valid as long as we attempt to measure underlying objectives (preferences).
5 Data
A number of tools exist to collect and analyze service provider behavior, in-
cluding facility modules in household surveys and empirical studies to estimate
facility (in particular hospital) cost functions. The approach used here, a quan-
titative service delivery survey (QSDS), is distinct from these other tools in a
number of key respects. First, unlike most other survey-based research tools,
the service provider is the key unit of analysis (as opposed to, say, the firm or
the household). As mentioned above, it is not unusual for household surveys
to include facility modules. The perspective in these surveys, however, is that
of the household rather than the service provider. Consequently, while finding
proxies for quality, they pay little attention to, say, the question of why quality
of and access to services are the way they are. This is reflected in the type
of data collected, which is mainly on simple access indicators and the range
of services offered. In other words, these surveys largely ignore provider be-
havior and the processes and complexities through which public spending is
transformed into services. In most cases, facility information is collected as a
part of community questionnaires, which rely on the knowledge of one or more
informed individuals (Frankenberg 2000). Information supplied by informants
is therefore not only heavily dependent on the perception of a few individuals
but also not detailed enough to form a basis for analysis of service delivery,
such as operational efficiency, utilization, and other performance indicators.
17



To the extent that the information is based on perceptions, there may be addi-
tional problems due to the subjective nature of the data and its sensitivity to
respondents' expectations.
Second, the QSDS does not rely on budgeted costs, but collects detailed
data on actual spending and services provided at the facility level.
Finally, the QSDS explicitly recognizes that agents in the service delivery
system may have strong incentives to misreport (or not report) key data. These
incentives derive from the fact that information provided by, for example, a
health facility partly determine its entitlement of public support or funding.
Also, in case resources (including staff time) are used for other purposes (for
instance in the case of shirking and corruption), the agent involved in the
activity will most likely not report it truthfully. Moreover, certain types of
information, such as official charges, may only partly capture what is intended
to be measured (e.g., the users' costs of the service). The QSDS deals with these
data issues in two ways: (i) by using a multi-angular data collection strategy;
that is, a combination of information from different sources; and (ii) by careful
consideration of which sources and respondents have incentives to misreport,
and identification of data sources that are least influenced by these incentives.
The survey data that we use in this paper consists of 155 randomly selected
primary health care facilities drawn from 10 randomly chosen districts in all
four regions of Uganda. A detailed description of the sample design is provided
in the appendix. The sample is restricted to dispensaries and dispensaries with
maternity units (health center III level facilities) in order to ensure a degree
of homogeneity across facilities. The sample includes facilities from the main
ownership categories: government, private not-for-profit, and private for-profit.
As described earlier, the private not-for-profit health facilities in Uganda are
mostly operated by religious organizations, and in our sample all non-profits
have religious affiliations.'3 The sample was designed so that the proportion of
facilities drawn from different regions and ownership categories broadly mirrors
the population of facilities. However, as noted earlier, no census of private for-
profit health facilities is available in Uganda, and it is hence difficult to assess
the extent to which the sample is representative in this regard.'4 Of the 155
facilities, 81 (52%) are government owned, 44 (29%) are owned by not-for-profit
providers, and 30 (19%) are privately owned.
The survey applied a data triangulation method, that is, data on the health
13Two of the 44 not-for-profit providers did not have a religious affiliation. These facilities,
however, drop out of most regressions due to lack of data.
14A sample of government and private not-for-profit facilities was drawn randomly from the
health facility register kept by the Ministry of Health For-profits were identified on the basis
of information obtained from the sampled government facilities Survey instruments and a
sampling note are available at www.publicspending.org (tools).
18



facilities were collected both at the district and health facility level, as well as
from patients using an exit poll. At the district level, the district director of
health services was interviewed to obtain data on health infrastructure, staff,
supervision arrangements, and finance. Also at the district level, data were
collected from the district records on each health unit included in the survey.
These data cover staffing, salaries, vaccine supply, immunization, and drug
supply to the facilities for fiscal year 1999/2000. In addition, data on the
number of outpatients, inpatients, and deliveries were collected for 1999/2000
(for a detailed discussion of the data see Lindelow, Reinikka, and Svensson
2003).
At the facility level, a broad range of information relating to the facil-
ity and its activities was collected in the survey, including (i) characteristics
of the facility (location, type, level, ownership, catchment area, organization,
and services); (ii) inputs (staff, drugs, salaries, vaccines, medical and nonmed-
ical consumables, and capital inputs); (iii) outputs (facility utilization, mix of
services, and quality); (iv) financing (user charges, cost of service category,
expenditures, financial and other support); and (v) institutional support (su-
pervision, reporting, performance assessment, and procurement) for fiscal year
1999/2000. Also at the facility level, an exit poll was carried out to interview 10
patients in each facility. The latter interview covered cost of treatment, drugs
received, perceived quality of services, and reasons for selecting for this facility
instead of alternative sources of health care.
The data collected from both the facility and district level reveals fairly poor
consistency. While there is no clear pattern overall across variables, at least for
outpatient numbers, there appear to be a tendency for (government) facilities
to over-report output statistics to the districts relative to the data recorded in
patient registers. In some cases the over-reporting is considerable (see Lindelow,
Reinikka, and Svensson 2003). In the empirical analysis, we therefore use data
obtained directly from the records kept by facilities for their own needs (i.e,
patient registers, medical records) rather than administrative records submitted
to local government. The former, often available in a highly disaggregated
format, was considered to suffer least from any incentive problems in record-
keeping (see Table A.1 for summary statistics).
6    Empirical results
6.1 Staff remuneration
We start by looking at the simple relationship between staff remuneration and
ownership (Table 2). We have data for about 900 employees in a total of 130
19



facilities. We have information on position, skill level, and pay but no other
employee characteristics.
Regression 1 reports a basic wage regression, with dummy variables for
not-for-profit and private for-profit facilities. The dependent variable is the
full-time equivalent salary plus lunch allowances per month.15,'6 As evident,
the religious not-for-profit facilities pay significantly less than both the private
for-profit ones (F-test) and the government operated units (t-test). The pri-
vate for-profit facilities also pay significantly less than government facilities.
On average, religious not-for-profits pay roughly 65,000 Ush per employee less
per month than the government operated facilities and 17,000 Ush less than
for-profit facilities. These are large differences, considering that the average
(unconditional) full-time equivalent salary plus allowances per month is 109,000
Ush. In Regression 1, the district effects are also highly significant (LR-test).
Facilities in more rural areas, i.e., where the distance to the closest subcounty
center is greater, the pay on average is less, but the effect is not significant.
The market proxy also enters insignificantly.
One explanation for the difference in remuneration is that staff composition
differs across ownership. If the average skill (education) level is correlated with
ownership, and better-educated workers are paid more, the average effect cap-
tured in Regression 1 may simply be a composition effect. To control for this,
Regressions 2-6 report the findings from subsamples of the staff. Regression 2
considers only qualified staff."7 The pattern is similar. Government facilities
pay the most, and the religious not for-profits pay the least.
Regression 3 shows wage-setting conditional on ownership for the highest
qualified staff, i.e., staff with a least A-level with subsequent medical training.
For this group of workers, on average the religious not-for-profit providers pay
60,000 Ush per employee less per month than the government-operated facilities
"5In the public sector, lunch allowance was supposed to be paid at a rate of 66,000 shilhngs
per month for health care professionals and 44,000 for support staff in 1999/2000 (a conditional
grant to districts). It was initially intended as an incentive for staff to undertake work in the
community so that they would literally have money to pay for their lunch. In practice it
became a salary supplement and was a mechanism for increasing health workers' pay. Lunch
allowance was not paid to not-for-profit staff which became a point of contention for them
(although some not-for-profits pay their own lunch allowances). In the following year, lunch
allowance was formally rolled up into public sector salaries and is no longer regarded as a
separate item.
'6The qualitative results are similar if we use the more narrow measure for salary excluding
lunch allowances.
7Qualified staff include medical doctor, clinical officer (A level and three years of medical
training), comprehensive nurse (A level and three years of medical training), registered nurse
(A level and two-and-half years of medical training), laboratory assistant (O level and three
years of medical training), and enrolled nurse and midwife (O level and two-and-half years of
medical training).
20



and 56,000 Ush less than for-profit facilities. There is no significant difference
in remuneration between for-profit and government providers. The average
(unconditional) full-time equivalent salary plus allowances per month for the
highest qualified staff is 212,000 Ush. Thus, on average, the highest qualified
staff are paid 28 percent less than both for-profit and government staff in the
same category.
Regression 4 reports the results for the largest group among the qualified
staff, that is, enrolled nurses. While we still observe a large difference between
private and government staff (enrolled nurses employed by private providers
receive 65 percent lower wages than average), there is no significant difference
in remuneration between for-profit and not-for-profit providers.
The same pattern holds for unqualified staff. Regression 5 depicts the re-
lationship between wages and ownership for nursing aides (the largest group
of workers in the unqualified group). Private for-profit dispensaries pay 41,000
Ush less per month (compared with the government facilities), while the not-
for-profit providers on average pay 49,000 less. The coefficient estimates are
not significantly different at the 10-percent level.
The preliminary analysis thus suggests that there exist a religious premium
but only for qualified staff, which makes it possible for religious not-for-profit
facilities to hire qualified workers below market wage. This premium does not
show up in the sample of unqualified or less-qualified staff. One explanation
for this is simply that unqualified staff are paid a very low salary. They may
therefore not be able to accept a lower wage.
One concern with these results is that we are missing not only the usual (in
a wage regression) unobservables, but also some standard observables in deter-
mining wages, such as experience. Unfortunately, information on experience
was not collected in the survey. A priori, it is not clear how this omitted vari-
able bias would influence the results. If health staff in the public sector have
longer tenure and thus are more experienced than their counterparts elsewhere,
we would overestimate the religious (altruistic) wage premium. Conversely, if
the not-for-profit providers' staff is more experienced, the reverse would be
true. Fortunately, what we can do is to quantify how important this experi-
ence bias might be, since we have information on the salary scale for medical
personnel in government health facilities.18 For a qualified nurse; i.e., a nurse
with at least A level with three years of medical training, the maximum re-
turns to experience is 12,000 Ush. Thus, in the extreme where qualified staff in
not-for-profit facilities have little experience and qualified staff in government
and private for-profit units are highly experienced, this would explain roughly
'8Salary schedule B for medical personnel specifies salaries for 10 categories of staff, with a
range of salaries for each category depending on the experience of the respective staff member
21



one-fifth of the difference in the observed wage differential between government
(and for-profit) and not-for-profit providers.
Regression 6 pools the staff similarly to Regression 2, but adds information
on the level of medical training. The variable qualificatton takes the value 0
for enrolled nurses and midwives, 1 for laboratory assistant, 2 for registered
and comprehensive nurses, 3 for clinical officers, and 4 for medical doctors (see
footnote 8 for details of medical training in Uganda). We allow the ownership
effect to be conditional on staff qualifications by interacting qualification with
the ownership dummies. As before, government pay is higher than that of
both types of private providers. Not surprisingly, more qualified staff are gen-
erally better rewarded (positive coefficient of qualificatwon). However, there are
differences in the marginal return to medical training depending on the owner-
ship of the dispensary they work in. More specifically, the marginal return to
medical training is lowest in the government service (i.e., wages are the most
compressed in the government sector) and highest in the private for-profit sec-
tor. Highly qualified not-for-profit staff are paid significantly less than their
for-profit counterparts. Hence, the religious premium in pay falls largely on
the most qualified medical staff. These effects are illustrated in Figure 1.
y
0          1          2          3         4
0
-2.5e+4
-5e+4
-7.5e+4
-le+5
Figure 1: Salaries in relation to govermnent units in for-profit (thin line) and not-for-
profit (thick line) facilities conditional on qualification.
22



6.2 Mix of services
Is there any evidence that the religious not-for-profit facilities are different from
the other providers in their choice of health services? This is what the altruistic
model suggests. Or, are-the private facilities providing roughly the same type
of services, which is what the perquisites-maximizing model predicts. Before
presenting the results, it is worth asking: Why would the offered set of services
differ across ownership types? Our starting point is that the different types
of facilities have different objectives that will show up in different choice of
services. An alternative (complementary) explanation is that demand differs,
and if location is endogenous, this might lead to systematic service differences.
To some extent, we control for this by including proxies of location and market
structure in the regressions. Moreover, even if these variables do not fully
control for differences in demand, this may not be too problematic, since we
can reinterpret the model in such a way that the choice of service mix is in
practice a choice of where to locate. For example, if a facility locates in a poor
area, this implies provision of pro-poor services.
Tables 3a and 3b report a series of regression, where the dependent vari-
able is a 0,1 indicator if a given service is being provided (1), or not (0).19
All facilities are providing general out-patient services (OPD). From Table 3,
it is possible to identify two broad sets of services. The first group includes
in-patients care, medical care, laboratory services, and immunization.20 The
religious not-for-profit and the private for-profit providers are as likely to pro-
vide these services. For medical care, all three facility types are similar, while
government facilities are significantly less likely to provide laboratory services.
Our empirical evidence shows that government units are the most likely
ones to carry out immunizations, followed by the religious not-for-profit facil-
ities. However, this effect is solely driven by differences in vaccine supply.2'
Controlling for the free supply of vaccines in Regression 5, we find no signifi-
cant difference between the three types of facilities. This is also consistent with
the principles of the national (vertical) immunization program.
For the second set of services, the two private sector providers differ. This
19We focus on the most common health services. A handful of facilities also provide mental
care, eye care, and dental care.
20The term "medical care" refers to (non-surgical) curative care
2tImmunization is a special service from the individual facility perspective The national
immunization program (UNEPI) sets countywide standards for immunization services and
manages the program mainly by providing supplies to health facilities. In fact UNEPI is a
monopoly supplier of vaccines in Uganda (both regular supplies and for immunization days).
The program sets its targets for immunization based on population and provides vaccines
to meet those targets directly to the facilities Not-for-profit dispensaries also receive their
vaccines from UNEPI as do some private for-profit providers.
23



set includes outreach, health education, training of nurses and community
health workers, antenatal care, and family planning. All these services, ex-
cept family planning, are more likely to be provided by the not-for-profit than
the for-profit facilities. Not-for-profit facilities are less likely to provide fam-
ily planning. Comparing the not-for-profit and government facilities, the later
is more likely to carry out outreach (almost all did, 77 out of 80), although
the religious not-for-profit and government facilities are similar in number of
staff days per month for outreach (Regression 7). The religious not-for-profit
facilities are more likely to run training programs for community health work-
ers, while government clinics are more likely to provide antenatal care, but the
effect is small.
How do we interpret these results? Clearly, the services depicted in Table
3 differ both in their profit potential, the extent to which they benefit the
poor, and in their public good nature. In general, in-patient care, medical
care, antenatal care, and laboratory testing are services that are demanded by
a broad spectrum of the population and are not typically public goods. Most
of these services are just as likely to be provided by for-profit facilities as the
religious not-for-profit ones, antenatal care being an exception.
It could be argued that outreach is a service that in general has a bias toward
the poor. Fewer not-for-profits provide this service compared to government but
those that do provide more of it. Health education and training of community
health workers have a public good element. Therefore it may be difficult to
make a positive profit from these three services. The perquisites-maximizing
model predicts that such services would not be provided neither by for-profit
nor by not-for-profit facilities. The data does not support this prediction. In
fact, the pro-poor services (outreach) and those with a public good element are
significantly more likely to be provided by the religious not-for-profit facilities
than the for-profit ones.
As can be seen from Table 3, family planning in an exception. A probable
explanation for this is that the not-for-profit facilities have religious motivations
for not providing this type of service, particularly as the Catholic Church is
important as a health care provider.
It is worth noting that one service, laboratory testing, is the only one that
both private providers are significantly more likely to provide. To the extent
that testing is an important quality component, this is in line with findings on
the quality of care reported below.
24



6.3 Price setting
Table 4 reports on the relationship between user-fees and ownership. Again, it
is useful to ask: Why would prices differ across ownership types? In our model,
facilities act as local monopolists, setting prices to maximize their objective
function. If these objectives differ across ownership, we should also observe
differences in user-fees across types.
Another explanation is that what we treat as homogeneous goods (e.g.,
minor surgery) are actually differentiated goods with respect to quality. The
implications of this is analyzed formally in section 3.3.
Another obvious explanation is differences in demand and marginal costs
across units. Wealthier areas may be willing to pay more for health services.
It may therefore be more profitable to locate in such an area, which may have
differential impact on the locality choice of the providers. As discussed above,
we can reinterpret the model so that the choice of prices in practice is a choice
of where to locate. Thus, for evaluating the behavior of not-for-profit facilities,
this may pose less of a problem. In addition, we control for location and degree
of competition.
Table 4 shows that there are large differences across types of facilities. For
general OPD, the government facilities charge almost 2,000 Ush less per first
visit compared to private for-profit providers. The median payment in a gov-
ernment facility is 500 Ush. The religious not-for-profit facilities charge signif-
icantly more than the government facilities, but significantly less (roughly 600
Ush) than the for-profit facilities. A similar pattern matches user fees for the
other services, as can be seen from Regressions 2-5. Private providers charge
more for minor surgery, antenatal care, medical care, and delivery-related ser-
vices. With the exception of antenatal care, the for-profit facilities charge more
than the religious not-for-profit, ranging from around 600 Ush for minor surgery
to 5,000 Ush for delivery.
The baseline model suggests that a perquisites-maximizing not-for-profit
facility will set the same prices as a for-profit facility, while the altruistic model
suggests that the not-for-profit will set prices at which the marginal return is
strictly lower than the marginal cost. In other words, an altruistic not-for-profit
facility will charge strictly lower prices than a for-profit unit. The findings on
user-fee charges are consistent with the altruistic model.
6.4 Quality
In the model, a for-profit facility would choose to exert higher effort (ex ante) to
increase quality than a perquisites maximizing not-for-profit facility. Without
25



further restrictions, we cannot say if an altruistic clinic would choose to exert
more or less effort than a for-profit one. We can, however, draw the conclusion
that if quality of services is the same or higher in the not-for-profit sector than
in the for-profit sector, this is inconsistent with the perquisites-maximizing
model.
Measuring quality is difficult. We provide four complementary measures.
The first measure is based on observed treatment practice. The second measure
is based on observed supply (that is, availability of health infrastructure). The
third captures prescription practices, while the last one is a qualitative indicator
derived from the exit poll data.
One important component in prescribing the correct treatment for malaria
and intestinal worm cases is laboratory testing. We have information on these
two types of test. The number of malaria blood slides carried out (for every
100 suspected malaria patients), and the number of stool tests undertaken (for
every 100 suspected intestinal worm cases). Table 5 reports the findings with
respect to testing. In line with the finding on laboratory services, we find that
the private providers are significantly more likely to test patients for malaria
and intestinal worms. The effect is large. For example, on average, the private
providers test 25 more patients of every 100 suspected malaria patients. It is
interesting to note that this is not due to differences in health equipment and
staff (regressions 2 and 4). Adding these additional controls do not change the
results regarding ownership. Having highly qualified staff and a microscope
increase the frequency of testing, however.
Table 6 reports the result on health infrastructure. Government facilities are
more likely than private for-profit facilities to have sterilization and refrigeration
equipment, and more likely than not-for-profit facilities to have a blood pressure
machine. It is interesting to note that private for-profit facilities are as likely
to have observable health equipment (inputs); that is, equipment that is being
used in the actual treatment process (such as protective clothes, blood pressure
equipment), while they are less likely to have equipment that is more difficult
to observe (like sterilization equipment and refrigerators). One explanation for
this is that the private for-profit clinics are more responsive to profits, and thus
have stronger incentives to cut costs and pursue nonverifiable quality reductions
on the service(s) being provided.
Table 7 depicts the findings on another treatment variable, the prescription
of antibiotics. The extent to which antibiotics are prescribed is generally very
high. In fact, almost half of the patients report receiving an antibiotic. In some
cases, they receive several types at the same time. Government facilities are
significantly more likely to provide antibiotics than private providers. Regres-
sion 2 indicates that the effect is particularly strong in government facilities
26



without qualified (medical) staff. One explanation for this finding, which is
consistent with the altruistic hypothesis, is that antibiotics is a substitute for
effort.22
The last finding on quality comes from exit polls on why the patient had
chosen to visit the facility where she was interviewed. Patients reported that
proximity and good treatment and/or good staff were the most important fac-
tors for selecting the facility. Proximity is the most important factor overall;
this is particularly true for government facilities. In contrast, patients are sig-
nificantly more likely to report good treatment and/or good staff as a reason
for visiting private facilities (Table 7, Regression 3). Not surprisingly, facilities
without qualified staff are less likely to be visited for quality reasons (Regression
4).
The results on quality indicate that for observable outcomes, private providers
appear to provide better quality care. We cannot distinguish between the pri-
vate for-profit and the not-for-profit providers. These findings are inconsistent
with the perquisites-maximizing hypothesis.
7 - The effects of financial aid
A key question in a cross-section framework such as (17) is whether the selection-
on-observables assumption is plausible. Specifically, there might be unobserved
variables that are related to both y and the ownership indicators. Our second
approach avoids this problem by exploiting a near natural experiment of gov-
ernment financial aid for the not-for-profit sector.
As discussed in section 2.3, the financial aid prograim for dispensaries was
initiated by the government of Uganda in the financial year 1999/2000 and
prescribed that each not-for-profit unit was to receive a fixed-amount grant for
the fiscal year. The program was implemented by the local governments (dis-
tricts). Specifically, the Ministry of Finance transferred the funds meant for
lower-level units operated by not-for-profits to the local governments (districts),
who in turn distributed the funds to the units concerned once they had submit-
ted a workplan. In theory, all not-for-profit facilities should have received the
funds in 1999/2000. In practice, however, there was variation in receipts due
to a number of idiosyncratic factors, including not-for-profit dispensaries not
submitting the necessary documentation in time, uncertainty about what the
grants could be used for (it was meant to be an untied grant), and generally
poor communications. As the system of providing financial aid for not-for-profit
22In addition, the cost of prescribing drugs is much lower in government facilities, since
they receive most of their drugs for free.
27



units was new, this pattern was not surprising. The outcome for the fiscal year
1999/2000 was that some units did not receive their entitlement. Instead their
first grant reached them the following financial year. Thus, de facto the grant
program was phased in. It is this variation in receipt over time that we exploit.
A possible objection to this approach is that the de facto phasing-in was not
random, or more specifically that the incidence of receipts could be correlated
with the error term in equation (17). In that case, correlation between transfers
and outcomes may be spurious. Although we cannot empirically fully reject this
alternative hypothesis (since there might be unobserved factors influencing both
transfer and y), we can check if the groups of grant recipients and nonrecipients
differ on observables.
Tables 8a and 8b report a set of regressions using observable facility charac-
teristics as dependent variables. The regressor is a dummy variable taking the
value 1 if the facility received the entitled grant, and 0 otherwise. Regressions
1 and 2 show that grant recipients and nonrecipients do not differ significantly
in age, measured either as the year the facility was established (Regression 1),
when the facility has been renovated last (year), or whether the facility had
been renovated (Regression 2).23 The recipients and nonrecipients do not differ
in access to communication infrastructure (Regression 3), that is, a nonrecipi-
ent is as likely as a grant recipient to have access to telephone, newspapers, and
radio at the facility. We also do not find any difference in distance to district
or subcounty headquarters (Regression 4), size of the facility (Regression 5),
whether or not the facility was staffed with at least one qualified nurse or a
doctor (Regression 6), or in access to health infrastructure (Regression 7-10).
Thus, there is no (observable) evidence suggesting that the grant recipients
and nonrecipients differ on observable characteristics (apart from receiving the
transfer or not).24
The reason we use the variation in grant receipt to identify the effects of
ownership is that a profit or a perquisites-maximizing not-for-profit provider's
behavior would not be affected by the inflow of aid. Untied aid does not affect
the marginal cost or revenue schedules. Thus, it would set the same prices and
provide the same services as without aid.
The altruistic not-for-profit facility's maximization program would however
23The variable established/renovated depicts for each facility the year the facility was last
renovated, or if no renovation has occurred, the year the facility was established.
24Although the groups of facilities do not differ in observables, they may still differ in some
unobserved dimension. However, this unobserved dimension must then be uncorrelated with
the set of observable characteristics reported in Table 7.
28



be affected. Formally, with aid, the facility maximizes,
SI            5.
L = Z Ex + A   a + L  [P(x8, q)x,-w -cx. - C(q))
where a is financial support. As shown in the appendix, for an altruistic
provider, aid will lead to lower prices (and possibly more services) and to higher
quality care. These results are intuitive. The altruistic provider cares about
the number of (poor) people treated and this number can be increased by ei-
ther lowering prices or increasing the quality of care. Both strategies are costly.
Aid relaxes the provider's budget constraint and at the margin it is optimal to
increase the the number of people treated using both strategies.
It is worth noting that government units also receive financial aid, primarily
from public sources, although not from the same financial aid program. The
allocation criteria and disbursement mechanism for aid to the government units
are not clear. On the one hand, some local governments may use financial aid
to improve service delivery. On the other hand, as most government programs
executed at the local (district and subcountry) level, it is likely that well-
organized facilities in more affluent (and politically powerful) areas are likely
to capture a larger share of this funding (for capture of education grants see
Reinikka and Svensson 2002).
While most government and religious not-for-profit facilities receive finan-
cial aid from public sources, no for-profit facility did.25 Conditional on receiv-
ing financial assistance, the median receipt for not-for-profit dispensaries (with
maternity unit) was 3.2 million Ugandan shillings (Ush), which is close to the
amount allocated and disbursed by the central government (3.4 million Ush).
Roughly 25 percent of the not-for-profit facilities did not receive financial aid.
Sixty-seven percent of the government facilities received public financial sup-
port. Conditional on receiving financial aid, the mean receipt in the government
sector was 0.6 million Ush.
When evaluating the effects of financial aid, it is important to identify which
potential variables might be affected by the inflow in a short time interval. We
look at three sets of variables that facilities can easily adjust in the short run:
testing procedures, prices, and staff remuneration.
In Table 9, Regressions 1 and 2, report the correlation between financial aid
and laboratory testing. Financial aid is positively correlated with testing for
malaria and intestinal worms. The estimated effects are large. A not-for-profit
provider with the median grant receipt test on average 24 more patients out
25Dropping two suspected misrecorded observations, we have data from 152 of the 155
sample facilities.
29



of every 100 suspected malaria case. The relationship, however, only holds for
religious not-for-profit facilities.
When we test the relationship between user fees and financial aid for the
specific services we have information on (minor surgery, antenatal care, medical
care, and delivery services), we find no impact of financial aid. However, as
depicted in Regression 3, financial aid is negatively correlated with OPD user
charges. This conditional finding again only holds for religious not-for-profit
facilities. A not-for-profit provider with the median grant receipt charge, on
average, 900 Ush less for general OPD. As the median number of outpatients
treated for this subsample of facilities was 230 patients per month, the upper
bound on the foregone revenues of this price cut is 2.4 million Ush, or three-
quarters of the total grant.26
Finally, we analyzed the relationship between salaries and financial aid.
There is no robust relationship in any staff category (we report the results for
qualified staff and nursing aides only).
To summarize, we find evidence that financiai aid leads to more testing of
suspected malaria and intestinal worm cases and lower prices for OPD services,
but only in religious not-for-profit facilities. Since the variation in financial
assistance is, we argue, to a large extent exogenous, these findings provide
strong evidence in support of the altruistic hypothesis.
$ (Concluding remarks
In this paper we exploit a unique micro-level data set on primary health care
facilities in Uganda to explore the motivation of religious not-for-profit health
care providers. We develop a simple model to guide the empirical work. To
identify whether working for a (religious) not-for-profit has an effect, we use two
strategies. The first builds on the assumption that we can identify the behavior
of the not-for-profit providers by comparing their performance in various di-
mensions with government and for-profit providers. The second approach relies
on a near natural policy experiment of public financial aid for the not-for-profit
sector.
In the cross section, we find that religious not-for-profit facilities hire quali-
fied medical staff below the market wage. They pay significantly less than both
the private for-profit sector and the government. The finding that the govern-
26That is 900*230*12 = 2.4 million Ush. This is the upper bound since the price cut most
likely increased the number of patients treated In addition, we do not have information on
when during the year the price cut occurred. For example, if the price cut occurred in the
middle of the year, the foregone revenues should be half as large.
30



ment pays the highest salaries is quite surprising, as the commonly held view
in Uganda is that public sector pay is well below the private sector, including
the health care profession. For example, a major pay comparator study, which
underpins the government pay reform, puts the pay of a clinical officer and an
enrolled nurse employed by government at 20-40 percent of that in the private
sector (Republic of Uganda 1999). Our findings are in stark contrast to this
commonly held view, at least in the case of lower-level health care units.
Moreover, religious not-for-profit facilities are more likely to provide pro-
poor services and services with a public good element, and charge lower prices
for services than for-profit units. Private not-for-profit and for-profit facilities
both provide better quality care than their government counterparts, although
government facilities have better equipment.
Finally, the quasi-experiment reveals that financial aid leads to more testing
of suspected malaria and intestinal worm cases, and hence to better quality
of care, and lower prices in religious-not-for-profit facilities. Moreover, the
estimated effects are substantial.
These findings are consistent with the existence of a premium in working
in a religious not-for-profit facility and that religious not-for-profits are driven
(partly) by altruistic concerns.
It is worth pointing out what we have not measured and other possible ex-
planations for the pattern we observe. First, we interpret the evidence above
in favor of the altruistic model. However, as argued by Glaeser (2002), it
may still be the case that the not-for-profit providers are captured by their
workers/managers, but that their preferences themselves are also altruistic and
therefore they (partly) internalize the stated goals of the provider. There is
some qualitative evidence to support this interpretation, as many practitioners
in the field report that the working environment in religious not-for-profit fa-
cilities are considerably better (i.e., revenues are spent on perks that improve
the working environment for the staff). This in turn could also help explain
why salaries in not-for-profits are lower (i.e., compensated by perks). On the
other hand, there are reports that labor practices in religious not-for-profits
are not always ideal- dismissing single pregnant workers, compulsory religious
activities-and that these policies can be resented by the workers.
Second, since all the not-for-profits in our sample have religious affiliations,
it is possible that the objective is not pure altruism, but to convert people.
The provision of services and the service delivery choices may therefore be
guided by this goal. Distinguishing between these two objectives would require
data on nonreligious not-for-profit providers and a model of a health provider
maximizing the number of people converted (for example by maximizing public
relations). Clearly, this is an important area for future research.
31



Third, it is possible to think of alternative explanations for each individ-
ual finding reported above. For example, the religious wage premium may be
due to rigidities in the labor market combined with recent increases in the
pay for government employees. On the other hand, paid training (where a
per diem typically makes up a significant part of the monthly wage) is more
prevalent in the public sector, which suggest that what we pick up is a lower
bound. Also, this effect cannot explain the wage differential between private
for-profit and not-for-profit providers. If the type of workers (within a category
of workers, say, nurses) differs across ownership types, this may also explain the
wage differential. In particular, if workers in the not-for-profit sector are less
competent health care providers, this may explain why they are paid signifi-
cantly less. However, this interpretation is difficult to reconcile with the fact
that not-for-profit facilities provide better quality care than their government
counterparts.
An alternative explanation for the financial aid findings is that better-run,
well-organized, not-for-profit providers managed to get the financial aid sooner
than poorly functioning ones. But if these well-organized units also pay higher
wages, we should observe a positive relationship between monthly salaries and
aid. We do not. In addition, on observables, the early and late aid recipients
look similar. We believe the strength of the argument put forward in the paper
lies in the fact that we find consistent evidence across different aspects of service
delivery (price and wage setting, service mix, and quality choices) and across
empirical techniques. We cannot think of one particular alternative explanation
that would explain all these facts.
32



9 Appendix
9.1 Sample design
The sample design was governed by three principles. First, attention was re-
stricted to dispensaries and dispensaries with maternity units (i.e., health center
III) to ensure a degree of homogeneity across sampled facilities. Second, subject
to security constraints, the sample was meant to capture regional differences.
Finally, the sample had to include facilities from the main ownership cate-
gories: government, private nonprofit (churches, NGOs), and private for-profit
providers. These three considerations lead us to choose a stratified random
sample. The sample was based on the Ministry of Health (MOH) facility regis-
ter for 1999. The register includes government, private non-profit, and private
for-profit facilities, but is known to be inaccurate with respect to the latter two.
A total of 155 health facilities were surveyed. On the basis of existing infor-
mation, it was decided that the sample would include 81 government facilities,
44 private non-for-profit facilities, and 30 private for-profit facilities. The exit
poll of clients covered 1,617 individuals. The field work was carried out during
October to December 2000. For summary statistics see Table A.1.
As a first step in the sampling process, 8 districts (out of 45) had to be
dropped from the sample frame due to security concerns.27 From the remaining
districts, 10 districts, stratified according to geographical location with the
size distribution determined by population shares, were randomly sampled in
proportion to district population size. Thus, three districts were chosen from
the Eastern and Central regions, and two from the Western and Northern
regions.28
From the selected districts, a sample of government and private non-profit
facilities was drawn randomly from the MOH register. A reserve list of replace-
ment facilities was also drawn from the sample frame. Due to the unreliability
of the register for private for-profit facilities, it was decided that for-profit facil-
ities would be identified on the basis of information from the government facil-
ities sampled.29 The administrative records for facilities in the original sample
were reviewed first at the district headquarters, where some facilities that did
not meet selection criteria and data collection requirements were dropped from
27The eight districts were Bundibugyo, Gulu, Kabaxole, Kasese, Kibaale, Kitgum, Kotido,
and Moroto
28The study districts were Mpigi, Mukono and Masaka in the Central region; Mbale, Iganga
and Soroti in the East; Arua and Apac in the North; and Mbarara and Bushenyi in the West
29Specifically, the x private facilities in region y would be determined by the in-charge in
the first x randomly drawn government facilities in region y, where each in-charge would be
asked to identify the closest private dispensary or dispensary with maternity unit.
33



the sample. These were replaced by facilities from the reserve list. Overall 30
facilities were replaced.
9.2   Proof of claim in section 3.5
The claim in section 3.5 is that a for-profit facility provides higher quality care
and charges higher prices than a perquisites-maximizing facility. Without loss
of generality, consider the case of one service. We then want to show that
P F=dp/dce > 0 and Qa =_ dq/dce > 0. It is analytically more convenient to
let the facility choose quantity and quality instead of price and quality (both
approaches are equivalent). Thus, demand is x = X(p, q). To prove the claim
let F(p, q; ca) and G(p, q; a) denote the first-order conditions for optimal price
and quality (corresponding to (12) and (2)),
F(q, x; a) = o (x + (p-c) Xp(p, q)) = 0        (18)
G(q, x; ce) = a (pXq(p, q) - cXq(p, q) - CO) - 'Yq  0  (19)
Total differentiate F and G yields,
FpPoa + FqQa + F, =0                    (20)
GpPa.+GqQo,+G.=0                        (21)
where Fx = dF/dx. Thus,
P, =FqG, > 0                          (22)
Qa       FpG, > 0                     (23)
since Fq = a (Xq + (p - c)Xpq) > 0, G:x = -yq/O > 0 (from (19)) and, by the
second order conditions, A > 0, Fp < 0.
9.3   Proof of clarim in section 7
The claim in section 7 is that aid to an altruistic not-for-profit provider leads
to higher-quality care and lower prices. Without loss of generality, consider the
case of one service and assume y(q) = 0 Vq. We want to show that Pa < 0 and
Qa > 0. The facility's problem can be restated as maximizing the Lagrange
function,
L = X(p, q) + A (a +pX(p, q) -w-cX(p, q) -C(q)).
34



Let F(A, p, q; a), G(A, p, q; a) and H(A, p, q; a) denote the first-order condi-
tions for A, p and q, respectively.
F(A,p, q; a) = a + pX(p, q) -w-cX(p, q) -C(q) = 0     (24)
G(A, p, q; a) = Xp(p, q) + A (X(p, q) + (p - c)Xp(p, q)) = 0  (25)
H(A, p, q; a) = Xq(p, q) + A ((p -c)Xq(p, q) - Cq) = 0.  (26)
The second-order condition for a constrained optimum is
A _ Fq [FpGq - FqGpJ + Fp [GqFq - FpHq] > 0.       (27)
Since Fq = -Xq/A < 0 (from (26)) and Fp = -Xp/A > 0 (from (25)), a
sufficient condition for an optimum is that the first term in brackets in (27) is
negative and the second term is positive. By the implicit function theorem we
have,
P.      [GqFq-FpHq] < 0                    (28)
Qa.      [FpGq-FqGp] > 0                   (29)
where it follows from (27) that the term in brackets in (28) is positive while
the term in brackets in (29) is negative.
35



IRefeiren¶ces
[1] Appleton, Simon. 2001. "Poverty in Uganda, 1999/2000: Preliminary Estimates
from the UNHS." School of Economics, University of Nottingham, United King-
dom.
[2] Duggan, M.G. 2000. "Hospital Ownership and Public Medical Spending." Quar-
terly Journal of Economics 115(4): 1343-73.
[3] Frankenberg, Elizabeth. 2000. "Community and Price Data." In M. E. Grosh
and P. Glewwe, eds., Destgning Household Survey Questionnaires for Developing
Countries: Lessons from 15 Years of the Living Standards Measurement Study.
Washington, D.C.: World Bank.
[4] Glaeser, Edward L. 2002. "The Governance of Not-for-Profit Firms." Discussion
Paper 1954. Harvard Institute of Economic Research, Harvard University. Cam-
bridge, Mass. April.
[5] Glaeser, Edward L., and Andrei Shleifer. 2001. "Not-for-profit entrepreneurs."
Journal of Public Economtcs 81: 99-115.
[61 Hutchinson, Paul. 2001. "Combating Illness." In R. Reinikka and P. Collier, eds.,
Uganda's Recovery: The Role of Farms, Firms, and Government. World Bank
Regional and Sectoral Studies. Washington, D.C.: World Bank.
[7] Hansmann, Henry. 1980. "The Role of the Nonprofit Enterprise." Yale Law Jour-
nal 89: 835-901.
[8] Jeppson, Anders. 2001. "Financial Priorities under Decentralization in Uganda."
Health Policy and Planning 16(2): 187-92.
[9] Lakdawalla, Darius, and Tomas Philipson. 2001."The Nonprofit Sector and In-
dustry Performance." University of Chicago. Processed.
[10] Lindelow, Magnus, Ritva Reinikka, and Jakob Svensson. 2003. "Health Care on
the Frontline Survey Evidence on Public and Private Providers in Uganda." Africa
Region Human Development Working Paper. World Bank, Washington, D.C.
Forthcoming.
[11] Malani, Anup, Tomas Philipson, and Guy David. 2002. "Theories of Firm Behav-
ior in the Non-Profit Sector: A Synthesis and Empirical Evaluation." University
of Chicago. Processed.
[12] McClellan, Mark, and Douglas Staiger. 2000. "Comparing Hospital Quality at
For-Profit and Not-For Profit Hospitals." Working Paper 7324. National Bureau
of Economic Research, Cambridge, Massachusetts.
36



[13] McPake, Barbara, Delius Asiimwe, Francis Mwesigye, Mathias Ofumbi, Lisbeth
Ortenblad, Pieter Streefland, and Asaph Turinde. 1999. "The Economic Behavior
of Health Workers in Uganda: Implications for Quality and Accessibility of Public
Health Services." Social Science and Medicine 49 (7, October): 849-65.
[14] Moeller, Lars Christian. 2002. "Uganda and the Millennium Development Goals."
Human Development Network. World Bank, Washington, D.C. Processed.
[15] Okello, D.O., R. Lubanga, D. Guwatudde, and A. Sebin&-Zziwa. 1998. "The Chal-
lenge to Restoring Basic Health Care in Uganda." Social Sctence and Mediczne
46(1):13-21.
[161 Pauly, M., and M. Redisch. 1973. "The Not-for-Profit Hospitals as a Physicians'
Cooperative." Amencan Economzc Revzew 64(1): 87-99.
[17] Philipson, Tomas. 2000. "Asymmetric Information and the Not-for-Profit Sector:
Does Its Output Sell at a Premium?" In David M. Cutler, ed., The Changing
Hospital Industry: Comparzng Not-for-Profit and For-Profit Institutions. Chicago:
University of Chicago Press.
[18] Reinikka, Ritva, and Jakob Svensson. 2002. "Local Capture and the Political
Economy of School Financing." CEPR Discussion Paper No. 3227.
[19] Republic of Uganda. 1999. "Public Service Pay Reform Project." A report by
KPMG to the Ministry of Public Service. Kampala. June.
[20] Republic of Uganda. 2000. "National Health Policy and Health Sector Strategic
Plan 2000/01-2004/05." Ministry of Health, Kampala, Uganda.
[21] Republic of Uganda. 2001a. "Policy for Partnership with Facility-Based Private
Not-for-Profit Health Providers." Ministry of Health, Kampala.
[22] Republic of Uganda. 2001b. "Facility-Based Private Not-For-Profit Health
Providers: A Quantitative Survey." Ministry of Health in collaboration with
Uganda Catholic, Muslim, and Protestant Medical Bureaux, National Health
Consumers Organization, and lTopical Business Research, Kampala, November.
[23] Republic of Uganda. 2002. "Infant Mortality in Uganda 1995-2000: Why the Non-
Improvement?" Discussion Paper 6. Ministry of Finance, Planning, and Economic
Development. Kampala.
[24] Rose-Ackerman, Susan. 1996. "Altruism, Nonprofits, and Economic Theory."
Journal of Economic Literature XXXIV(June): 701-28.
[25] Sloan, Frank A., Gabriel A. Picone, Donald H. Taylor, Jr., Shin-Yi Chou. 1998.
"Hospital Ownership and Cost and Quality of Care: Is There a Dime's Worth
of Difference?" Working Paper 6706. National Bureau of Economic Research,
Cambridge, Massachusetts.
37



TWabRle AA1 Summary statistics
Sample variable           All           Government       Not-for-profit     For-profit
No. staff                 7.0               7.8               7.6              4.0
7.0               7.0              6.0               3.0
3.8               3.2              4.6               2.7
No. doctors, clinical     0.6               0.7               0.3               0.7
officer                   0                 1.0               0                0
0.69              0.55              0.6               1.0
No. nurses                1.7               2.0               1.7               1.1
1.0               2.0              1.0               1.0
1.6               1.5              1.9               1.0
Established or last      1993              1993              1992              1993
renovated                1996              1995              1998              1998
12               8.9               10.7             20.2
Distance to District      32                36                28               26.0
H[Q (kcm)
29                35                22               22.0
22.5              21.6              25.5              18.6
Distance to
subcounty HQ (Im)         3.9               3.8              4.5               3.5
3.0               2.0               3.0              3.0
Competition               4.1               4.4               3.9              3.8
(number of other          1.7               1.2               1.8              3.2
health-care
providers)                1.0               1.0               1.0              2.0
No. outpatients per       2.0               1.3               1.7              3.0
month                     419               500              346               270
358               474               252              204
Sample size               296              284               283               279
155               81                44                30
Note: Mean, median, and standard deviation reported in subsequent rows for each variable. Sample size is maximum
number of observations. Because of missing data, not al variables have maximum number of observations.
38



Table 1. Testable implications
Model specification              Perquisites maximizing               Altruistic
AVkP = l.FP                   fl/P >FP
Basic model                            /,P = flFP                     AVP < ,FP
fliNP = flFP                   AV < flFP
f13kP > flFP
Endogenous costs                      /3kP > /3FP                    /lNP </p3,
flNP > 8lFP                   flP < fiFP
1kNP < fiFP                    /k       P > /P
Endogenous quality                    /3P </;P
3f9F                     flZP < pFP
iNP < flFP
39



1121 bie 2. Remuneration
Regression           (1)          (2)          (3)          (4)           (5)          (6)
High
Qualified    qualified     Enrolled     Nursing      Qualified
Sample               Full         staff       staff        nurses       aides         staff
Dep. variable                 Full-time equiv. Salary + lunch allowances per staff and month
NFP               -65,046°°    -95,493°°°   -59,914°°°  -102,213"'     -48,792"' -109,647..
(5,709)      (8,467)     (24,326)      (6,178)       (3,780)      (8,700)
FP                -47,949°°    -60,191°°°     -3,471    -105,919"'     -40,883     -120,762"'
(9,182)     (12,943)     (28,457)     (10,749)       (6,580)     (15,001)
distance             -808         1183         2,021        1,506"      -1,788$       1,677"
(563)        (831)      (2,190)        (621)       (1,063)        (737)
competition          1,495        -653          512        -2,582         -219         -964
(1,727)      (2,879)      (6,347)      (2,301)         (394)      (2,527)
qualification                                                                        13,334
(4,054)
qualification*                                                                       16,248$
NFP                                                                                   (6,295)
qualification*                                                                       34,785
FP                                                                                   (6,666)
F (NFP=FP)          3.16*        6.81         3.190        0.11          1.28
[.076]       (.009)       (.078)       (.740)        (.259)
District effects     Yes          Yes          Yes          Yes           Yes          Yes
LR                  98.2         31.2         13.5         32.6         128.9         40.1
(.000)       [.000]       (.142)       (.000)        (.000)       (.000)
Facilities           138          116           70           100          117          116
Observations         848          288           84           204          259          288
Adj. R2              0.74         0.90         0.88         0.95          0.93         0.93
Notes: OLS regressions with standard errors clustered by facility in parenthesis.
(**) [00°] denotes significance at the 1 (5) [10] percent level.
F is F-test statistic for testing the null hypothesis that NFP=FP.
LR is likelihood ratio test statistic for testing the null hypothesis that all district effects equal.
40



Table 3a. Service provision
Regression         (1)         (2)        (3)         (4)         (5)         (6)         (7)
Staff days
per month
In-patients  Medical       Lab.     Immuni-     Inmuni-                    for
Dep. variable     care        care      services     zation     zation     Outreach    outreach
NFP              0.262***     0.068     0.387*     -0.167*      -0.001      -0.126**    -1.111
(.081)      (.044)      (.085)      (.057)      (.033)     (.062)      (1.260)
PP                0.104       0.018     0.378***    -0.687***    -0.029    -0.823***   -5.830"'
(.111)      (.065)      (.101)      (.091)      (.028)      (.066)      (.936)
Distance         -0.005       0.001     -0.015*      0.005       -0.003      -0.001      0.035
(.009)      (.004)      (.009)      (.006)      (.002)      (.006)     (0.178)
competition       0.001      0.010       -0.005      0.005       0.025       0.006       0.043
(0.021)      (.008)     (.017)      (.020)      (.017)      (.020)      (.199)
free supply of                                                   0.968"                   16.5
vaccinations                                                     (.033)                  (.000)
F (NFP=PP)        1.86        0.76        0.01        26.6        1.15        67.5
(.175)      (.386)      (.938)      (.000)      (.285)      (.000)
District          Yes         Yes         Yes         Yes         Yes         Yes         Yes
effects
LR                29.1        11.8        23.4        34.4        11.0        23.0        18.3
(.001)      (.225)      (.005)      (.000)      (.273)      (.006)      (.031)
Facilities         152         154        153         151         151         153         151
Adj. R2           0.71        0.94        0.52        0.90        0.99        0.90        0.41
Notes: OLS regressions with robust standard errors in parentheses.
* (**) [***] denotes significance at the 10 (5) [1] percent level.
F is F-test statistic for testing the null hypothesis that NFP=PP.
LR is likelihood ratio test statistic for testing null hypothesis that all district effects equal.
41



TabRe 3b. Service provision
Regression             (8)           (9)            (10)          (11)          (12)
Dep. variable         Health      Training of    Training of  Antenatal care   Family
education       nurses     health workers                 planning
NFP                   -0.037        -0.002         0.116         -0.096       -0.360*
(.024)        (.080)         (.083)        (.054)         (.079)
PP                   -0.256*       -0.238c*Q     -0.196*        -0.2840        -0.049
(.078)        (.090)         (.059)        (.073)         (.074)
Distance              0.007         -0.006        -0.003          0.003         0.007
(.003)        (.008)         (.007)        (.004)         (.012)
competition           0.019          0.030        -0.009         0.016          0.001
(.009)        (.021)         (.018)        (.008)         (.012)
F (NFP=PP)             7.9            5.1           12.5          4.9            9.6
(.006)        (.024)         (.000)        (.028)         (.002)
District effects       Yes           Yes            Yes           Yes            Yes
LR                     20.0          48.8           24.3          28.9           13.9
(.018)        (.000)         (.004)        (.001)         (.126)
Facilities             153            154           152           152            152
j. R2                  0.96          0.47           0.31          0.93           0.88
See notes to table 3a.
42



Table 4. User fee charges
Regression             (1)           (2)            (3)           (4)            (5)
Dep. varnable          OPD       Mmor surgery  Antenatal care  Medical care    Delivery
NFP                   1,309*         950           388          2,106**        3,587***
(184)         (290)          (87)          (355)          (734)
FP                    1,930*        1,598*         471           3,103*        8,566*
(223)         (351)         (123)          (460)         (983)
distance               0.25         -61.1           9.02         -7.05          -44.3
(43.6)        (29.2)         (8.66)        (37.1)         (74.9)
competition           -12.8          109.0          10.0         -58.5         -345*
(19.1)        (97.5)         (20.2)        (76.0)         (155)
F (NFP=FP)            6.73w          2.78*          0.38          4.28           20.3
(.011)        (.100)         (.539)        (.042)         (.000)
District effects       Yes           Yes            Yes           Yes            Yes
LR                     39.8          32.6           22.9          24.6           42.6
(.000)        (.000)         (.006)        (.003)         (.000)
Facilities             130            80            99             94             87
Adj. R2                0.79          0.70           0.83          0.70           0.83
Notes: See notes to table 3.
43



aIblle . Qality - testing
Regression                  (1)               (2)               (3)              (4)
Dep. variable           Blood slides      Blood slides    Stool tests made  Stool tests made
NFP                        25.5             27.7              19.0              20.9 
(5.71)            (5.30)           (4.74)            (4.73)
FP                         25.2             28.4              15.9              18.9 
(6.50)            (6.21)           (5.30)            (5.32)
distance                                     0.13                               -0.18
(.540)                              (0.45)
competition                                  -0.20                              -0.40
(1.23)                              (1.04)
microscope                                  22.8                                16.4'
(5.13)                              (4.50)
high qualified                              20.1                                13.7 
staff                                       (5.25)                              (4.59)
F (NFP=FP)                  0.00             0.01              0.27              0.13
(.958)            (.917)            (.605)           (.723)
District effects            Yes              Yes               Yes               Yes
LR                         28.5              10.5              37.1              29.5
(.000)            (.314)            (.000)           (.000)
Facilities                  155               153               149               149
Adj. R2                     0.42             0.57              0.39              0.51
Notes: See notes to table 3.
44



Table 6. Quality - access to health infiastructure provision
Regression             (1)           (2)           (3)            (4)            (5)
Sterilization               Blood pressure                 Protective
Dep. variable       equipment    Refrigerator    equipment     Microscope      clothes
NFP                   0.004        -0.241***      -0.044         0.147          0.019
(.012)        (.080)         (.065)         (.089)        (.086)
FP                   -0.170*       -0.719*9       0.112*         0.011         -0.043
(.069)        (.084)         (.045)         (.111)        (.108)
distance             -0.004         0.008         -0.005         -0.016         0.004
(.004)        (.006)         (.013)         (.010)        (.009)
competition           0.005        -0.026*        -0.008         0.019          0.026
(.007)        (.015)         (.013)         (.021)        (.029)
F (NFP=FP)            6.25           21.8           6.4           1.41           0.32
(.013)        (.000)         (.012)         (.236)        (.571)
District effects      Yes            Yes           Yes            Yes            Yes
LR                    13.0           7.7            6.7           35.8           42.1
(.162)        (.564)         (.666)         (.000)        (.000)
Facilities             154           154            154           153            154
Adj. R2               0.97           0.80          0.91           0.58           0.54
Notes: See notes to table 3.
45



Tahble 7. Exit poll
Regression                   (1)              (2)               (3)              (4)
Dep. variable             Antibiotics      Antibiotics     Good treatment   Good treatment
GOV                         0.065"           0.021           -O.122"           -0.144 
(.037)           (.046)            (.047)           (.050)
FP                          0.012            -0.007            0.079            0.073
(.043)           (.057)            (.058)           (.059)
GOV*                                         0.109
unqualified                                  (.079)
FP*                                          0.035
unqualified                                  (.092)
unqualified                                  -0.041                            -0.083
(.057)                             (.042)
District effects             Yes              Yes               Yes              Yes
LR                           20.8             19.7             128.8            133.1
(.013)           (.020)            (.000)           (.000)
LR1                                            8.2
(.017)
Facilities                   155              155               155              155
Patients                     1516             1516             1156              1156
Adj. R2                      0.47             0.47             0.51              0.52
Notes: See notes to table 2. LR1 is likelihood ratio test statistic for testing the nul hypothesis that (GOV
+ GOV*unqualified) is zero.
46



Table 8a. NFP financial aid - recipients and nonrecipients
Regression          (1)          (2)        (3)        (4)        (5)         (6)
Dep. variable                                Receipt of aid
year facility     0.002
established       (.006)
year establ.                   -.000
renovated                       (.003)
access to                                  0.129
telephone                                 (.381)
access to                                  -.017
newspaper                                  (.192)
access to radio                            -.121
(.164)
distance to                                           -.015
subcounty HQ                                          (.020)
distance to                                           0.002
subcounty HQ                                          (.003)
number of staff                                                   0.009
(.017)
qualified staff                                                              -.019
(.154)
Observations          42         44         44          43         44          44
R2                   0.00       0.03        0.02       0.03       0.01        0.00
Notes: OLS regressions with standard errors in parenthesis, constant not reported.
47



TalRe Ob. NFP financial aid - recipients and nonrecipients
Regression                   (7)              (8)             (9)            (10)
Dep. variable                                   Receipt of aid
refrigerator                 0.048
(.161)
blood pressure equipment                     -.079
(.222)
microscope                                                  0.255
(.152)
protective clothes                                                          0.000
(.034)
Observations                  44              44              44              44
R2                            0.00           0.00             0.06           0.00
Notes: OLS regressions with standard efrors in parenthesis, constant not reported.
48



Table 9. Financial aid
Regression            (1)            (2)           (3)           (4)            (5)
Dep. variable      Blood slides   Stool tests  OPD charges   Qualified staff  Nursing aides
NFP                  -17.8         -12.7           -93         -32,316*        -1,602
(10.8)         (9.8)         (528)        (18,523)       (10,378)
GOV                 -25.0          -15.5*        -1,972.       61,256**       42,713*
(7.10)         (6.0)         (262)        (13,573)       (6,800)
NFP*AID             7.5E-6*        6.5E-6'       -3.1E-4*       -0.003         -0.002
(3.4E-6)      (3.2E-6)       (1.5E-4)       (.005)         (.003)
GOV*AID              1.3E-6        -5.0E-6        1.6E-4         0.002        -0.005
(6.7E-6)      (5.6E-6)       (9.2E-5)       (.008)         (.004)
District effects      Yes            Yes           Yes           Yes            Yes
LR                    31.2          39.6           37.2          24.8           113.3
(.000)        (.000)         (.000)        (.003)         (.000)
Facilities            150            144           127            113           114
Observations          150            144           127           278            253
Adj. R2               0.43          0.40           0.82          0.90           0.93
Notes: See notes to table 3.
49






Policy Research Working Paper Series
Contact
Title                           Author                  Date             for paper
WPS3043 The Incentive-Compatible Design  Thorsten Beck           May 2003          K. Labrie
of Deposit Insurance and Bank                                             31001
Failure Resolution: Concepts and
Country Studies
WPS3044 Impregnated Nets Cannot Fully     Mead over              May 2003          H Sladovich
Substitute for DDT Field        Bernard Bakote'e                          37698
Effectiveness of Malaria Prevention  Raman Velayudhan
in Solomon Islands              Peter Wilikai
Patricia M. Graves
WPS3045 Causes and Consequences of Civil  Klaus Deininger        May 2003          M. Fernandez
Strife: Micro-Level Evidence from                                         33766
Uganda
WPS3046 Migration, Spillovers, and Trade  Ross Levine            May 2003          E Khine
Diversion' The Impact of        Sergio L. Schmukler                       37471
Internationalization on Stock Market
Liquidity
WPS3047 Comparing Mortgage Credit Risk    Robert Buckley         May 2003          0. Himid
Policies An Options-Based Approach Gulmira Karaguishiyeva                 80225
Robert Van Order
Laura Vecvagare
WPS3048 Targeted Transfers in Poor Countries: Martin Ravallion   May 2003          P. Sader
Revisiting the Tradeoffs and Policy                                       33902
Options
WPS3049 Hidden Impact? Ex-Post Evaluation  Shaohua Chen          May 2003          P. Sader
of an Anti-Poverty Program      Martin Ravallion                          33902
WPS3050 The Changing Financial Landscape:  Wafik Grais           May 2003          R. Vo
Opportunities and Challenges for  Zeynep Kantur                           33722
the Middle East and North Africa
WPS3051 Regional, Multilateral, and Unilateral  Glenn W. Harrison  May 2003        P. Fiewitt
Trade Policies of MERCOSUR for  Thomas F. Rutherford                      32724
Growth and Poverty Reduction in  David G. Tarr
Brazil                          Angelo Gurgel
WPS3052 Long-Run Impacts of China's WTO   Kym Anderson           May 2003          S Lipscomb
Accession on Farm-Nonfarm Income  Jikun Huang                             87266
Inequality and Rural Poverty    Elena lanchovichina
WPS3053 Economic Impacts of China's       Elena lanchovichina    May 2003          S. Lipscomb
Accession to the World Trade    William Martin                            87266
Organization
WPS3054 The New Comparative Economics     Simeon Djankov         May 2003          G. Sorensen
Edward Glaeser                            37088
Rafael La Porta
Florencio Lopez-de-Silanes
Andrei Shleifer
WPS3055 The Global Growth of Mutual Funds  Deepthi Fernando      May 2003          A. Yaptenco
Leora Klapper                             31823
Viktor Sulla
Dimitri Vittas



Policy Research Working Paper Series
Contact
Title                           Author                  Date             for paper
WPS3056 Avoiding the Pitfalls in Taxing   Patrick Honohan        May 2003          A. Yaptenco
Financial Intermediation                                                  31823
WPS3057 Teaching Adults to Read Better and  Helen Abadzi         May 2003          H. Abadzi
Faster: Results from an Experiment                                        80375
in Burkina Faso