Quality and Cost in Health Care Choice in Developing Countries
Item
- Title
-
Quality and Cost in Health Care Choice
in Developing Countries
- extracted text
-
ums
Living Standards
Measurement Study
Working Paper No. 105
Quality and Cost in Health Care Choice
in Developing Countries
Victor Lavy
Jean-Marc Germain
LSMS Working Papers
No. 31
Suarez-Berenguela, Financing the Health Sector in Peru
No. 32
Suarez-Berenguela, Informal Sector, Labor Markets, and Returns to Education in Peru
No. 33
van der Gaag and Vijverberg, Wage Determinants in Cote d'Ivoire
No. 34
Ainsworth and van der Gaag, Guidelines for Adapting the LSMS Living Standards Questionnaires to Local Conditions
No. 35
Dor and van der Gaag, The Demand for Medical Care in Developing Countries: Quantity Rationing in Rural Cote d'Ivoire
No. 36
Newman, Labor Market Activity in Cote d'Ivoire and Peru
No. 37
Gertler, Locay, Sanderson, Dor, and van der Gaag, Health Care Financing and the Demand for Medical Care
No. 38
Stelcner, Arriagada, and Moock, Wage Determinants and School Attainment among Men in Peru
No. 39
Deaton, The Allocation of Goods within the Household: Adults, Children, and Gender
No. 40
Strauss, The Effects of Household and Community Characteristics on the Nutrition of Preschool Children: Evidence from
Rural Cote d'Ivoire
No. 41
Stelcner, van der Gaag, and Vijverberg, Public-Private Sector Wage Differentials in Peru, 1985-86
No. 42
Glewwe, The Distribution of Welfare in Peru in 1985-86
No. 43
Vijverberg, Profits from Self-Employment: A Case Study of Cote d'Ivoire
No. 44
Deaton and Benjamin, The Living Standards Survey and Price Policy Reform: A Study of Cocoa and Coffee Production in
Cote d'Ivoire
No. 45
Gertler and van der Gaag, Measuring the Willingness to Payfor Social Services in Developing Countries
No. 46
Vijverberg, Nonagricultural Family Enterprises in Cote d'Ivoire: A Descriptive Analysis
No. 47
Glewwe and de Tray, The Poor during Adjustment: A Case Study of Cote d'Ivoire
No. 48
Glewwe and van der Gaag, Confronting Poverty in Developing Countries: Definitions, Information, and Policies
No. 49
Scott and Amenuvegbe, Sample Designs for the Living Standards Surveys in Ghana and Mauritania/Plans de sondage
pour les enquetes sur le niveau de vie au Ghana el en Mauritanie
No. 50
Laraki, Food Subsidies: A Case Study of Price Reform in Morocco (also in French, 50F)
No. 51
Strauss and Mehra, Child Anthropometry in Cote d'Ivoire: Estimates from Two Surveys, 1985 and 1986
No. 52
van der Gaag, Stelcner, and Vijverberg, Public-Private Sector Wage Comparisons and Moonlighting in Developing
Countries: Evidencefrom Cdte d'Ivoire and Peru
No. 53
Ainsworth, Socioeconomic Determinants of Fertility in Cote d'Ivoire
No. 54
Gertler and Glewwe, The Willingness to Payfor Education in Developing Countries: Evidence from Rural Peru
No. 55
Levy and Newman, Rigidite des salaires: Donnees microeconomiques et macroeconomiques sur I'ajustement du marche du
travail dans lesecteur moderne (in French only)
No. 56
Glewwe and de Tray, The Poor in Latin America during Adjustment: A Case Study of Peru
No. 57
Aiderman and Gertler, The Substitutability of Public and Private Health Care for the Treatment of Children in Pakistan
No. 58
Rosenhouse, Identifying the Poor: Is "Headship" a Useful Concept?
No. 59
Vijverberg, Labor Market Performance as a Determinant of Migration
No. 60
Jimenez and Cox, The Relative Effectiveness of Private and Public Schools: Evidence from Two Developing Countries
No. 61
Kakwani, Large Sample Distribution of Several Inequality Measures: With Application to Cote d'Ivoire
No. 62
Kakwani, Testingfor Significance of Poverty Differences: With Application to Cote d'Ivoire
No. 63
Kakwani, Poverty and Economic Growth: With Application to Cote d'Ivoire
No. 64
Moock, Musgrove, and Stelcner, Education and Earnings in Peru's Informal Nonfarm Family Enterprises
No. 65
Aiderman and Kozel, Formal and Informal Sector Wage Determination in Urban Low-Income Neighborhoods in Pakistan
No. 66
Vijverberg and van der Gaag, Testingfor Labor Market Duality: The Private Wage Sector in Cote d'Ivoire
No. 67
King, Does Education Pay in the Labor Market? The Labor Force Participation, Occupation, and Earnings of Peruvian
Women
.inues on the inside back cover)
Donated by Dr. C M F rancis in Feb. 2010
Quality and Cost in Health Care Choice
in Developing Countries
The Living Standards Measurement Study
The Living Standards Measurement Study (LSMS) was established by the
World Bank in 1980 to explore ways of improving the type and quality of house
hold data collected by statistical offices in developing countries. Its goal is to foster
increased use of household data as a basis for policy decisionmaking. Specifically,
the LSMS is working to develop new methods to monitor progress in raising levels
of living, to identify the consequences for households of past and proposed gov
ernment policies, and to improve communications between survey statisticians, an
alysts, and policymakers.
The LSMS Working Paper series was started to disseminate intermediate prod
ucts from the LSMS. Publications in the series include critical surveys covering dif
ferent aspects of the LSMS data collection program and reports on improved
methodologies for using Living Standards Survey (lss) data. More recent publica
tions recommend specific survey, questionnaire, and data processing designs and
demonstrate the breadth of policy analysis that can be carried out using LSS data.
LSMS Working Paper
Number 105
Quality and Cost in Health Care Choice
in Developing Countries
Victor Lavy
Jean-Marc Germain
The World Bank
Washington, D.C.
Copyright © 1994
The International Bank for Reconstruction
and Development/THE WORLD BANK
1818 H Street, N.W.
Washington, D.C. 20433, U.S.A.
All rights reserved
Manufactured in the United States of America
First printing July 1994
To present the results of the Living Standards Measurement Study with the least possible delay, the typescript
of this paper has not been prepared in accordance with the procedures appropriate to formal printed texts, and
the World Bank accepts no responsibility for errors. Some sources cited in this paper may be informal
documents that are not readily available.
The findings, interpretations, and conclusions expressed in this paper are entirely those of the author(s) and
should not be attributed in any manner to the World Bank, to its affiliated organizations, or to members of its
Board of Executive Directors or the countries they represent. The World Bank does not guarantee the accuracy
of the data included in this publication and accepts no responsibility whatsoever for any consequence of their
use. The boundaries, colors, denominations, and other information shown on any map in this volume do not
imply on the part of the World Bank Group any judgment on the legal status of any territory or the
endorsement or acceptance of such boundaries.
The material in this publication is copyrighted. Requests for permission to reproduce portions of it should be
sent to the Office of the Publisher at the address shown in the copyright notice above. The World Bank
encourages dissemination of its work and will normally give permission promptly and, when the reproduction
is for noncommercial purposes, without asking a fee. Permission to copy portions for classroom use is granted
through the Copyright Clearance Center, Inc., Suite 910, 222 Rosewood Drive, Danvers, Massachusetts 01923,
U.S.A.
The complete backlist of publications from the World Bank is shown in the annual Index of Publications, which
contains an alphabetical title list (with full ordering information) and indexes of subjects, authors, and countries
and regions. The latest edition is available free of charge from the Distribution Unit, Office of the Publisher, The
World Bank, 1818 H Street, N.W., Washington, D.C. 20433, U.S.A., or from Publications, The World Bank, 66,
avenue d'I6na, 75116 Paris, France.
ISSN: 0253-4517
Victor Lavy is a professor in the Department of Economics at the Hebrew University of Jerusalem, Mt.
Scopus, Jerusalem. Jean-Marc Germain is an economist with the Ministry of Finance in Paris, France.
Library of Congress Cataloging-in-Publication Data
Lavy, Victor.
Quality and cost in health care choice in developing countries /
Victor Lavy, Jean-Marc Germain.
p.
cm. — (LSMS working paper ; no. 105)
Includes bibliographical references.
ISBN 0-8213-2854-9
1. Medical care, Cost of—Developing countries—Econometric
models. 2. Medical care—Utilization—Developing countries—
Econometric models. 3. Health services accessibility—Developing
countries—Econometric models. 4. Medical care—Developing
countries—Finance—Econometric models. 5. Consumer behavior—
Developing countries—Econometric models. 6. Medical care—
Developing countries—Public opinion.
I. Germain, Jean-Marc,
1966n. Title.
HI. Series.
RA410.55.D48L38
1994
362.1'09172'4—dc20
94-12590
CIP
CH -1 0 0
Contents
Foreword............................................................................................................................................ vii
Abstract.....................................................................................................................
ix
Acknowledgments.............................................................................................................................. xi
Introduction.........................................................................................................................................
1
Quality and Health Care Choice: Overview of Model..............................................................
3
Empirical Specification .......................................................................................................
The Health Production Function........................................................................................
The Conditional Utility Function ......................................................................................
Demand Functions and the Distribution of Stochastic Variables ..............................
5
5
6
7
The Demand for Medical Care in Ghana.....................................................................................
9
Ghanaian Health Care......................................................................................................... 9
The Data................................................................................................................................ 10
Definition of the Variables Used in the Estimation ..................................................... 11
Price variables ..................................................................................................................... 12
Demographic and economic variables................................................................................ 12
Estimation Results.............................................................................................................................. 15
Probability of Using Health Care Under Various Policy Assumptions
................................. 18
Improvement in Quality of Care........................................................................................ 18
Improving Access to Public and/or Private Health Care .............................................20
Raising Price to Improve Quality..................................................................................... 23
An Increase in Family Human Capital............................................................................. 23
Willingness to Pay
........................................................................................................................... 24
Conclusions......................................................................................................................................... 29
References
......................................................................................................................................... 31
Appendix Table:
Comparative Characteristics of the Full Ghana 1988 LSMS
Sample and the Sample Used by Lavy and Germain (1993).................. 35
v
Tables
Table 1:
Descriptive Statistics ............................................................................................... 13
Table 2:
Maximum Likelihood Nested Multinomial Estimates of Choice of
Health Care................................................................................................................ 16
Table 3A:
Improved Public Health Care, Predicted Probabilities and Percent
Change in Health Care Use................................................................................... 19
Table 3B:
Improved Access to Health Care and Predicted Probabilities and
Percent Change in Health Care Use.................................................................... 21
Table 3C:
Increasing the Cost of Health Care and Predicted Probabilities and
Percent Change in Health Care Use.................................................................... 22
Table 4:
Willingness to Pay for Improvement of Health Services.................................25
Table 5:
Willingness to Pay for Improvement of Health Services.................................27
vi
Foreword
The effect of quality of health care on demand and choice of treatment is of major
interest to concerned policy makers in many developing countries. The lesson to date from
experiences in cost recovery is that without visible and fairly immediate improvements in the
quality of care provided, increased demand will not support the implementation of user fees.
This study contributes to our understanding of the effect of quality improvement on demand,
and explores the implications for the use of user fees in the public health sector.
This paper is part of broader research effort in the Policy Research Department (PRD)
that examines the effect of the quality of social services on human capital investment
outcomes. This work is located in the Poverty and Human Resources Division. The data
used are from the Ghana Living Standards Survey, which is one of the Living Standards
Measurement Study (LSMS) household surveys which the World Bank has implemented in
many developing countries.
Lyn Squire
Director
Policy Research Department
vii
Abstract
The definition of health care quality and the impact of improved quality on the
demand for health care have not been the subject of rigorous econometric studies. This
study models theoretically and empirically the quality of health care in household decision
making with respect to demand for health care and presents empirical evidence concerning
the impact of various policy options on these decisions. Besides modeling quality explicitly,
our model relaxes some of the restrictive assumptions that were common in recent studies of
the demand for health care.
Acknowledgments
This study was supported by a grant from the World Bank Research Committee. The
views presented here are those of the authors and should not be interpreted as reflecting the
views of the World Bank. We greatly appreciate comments on earlier drafts made by M.
Ainsworth, H. Aiderman, P. Gertler, M. Over, M. Jimenez, and participants in seminars at
the World Bank and the Hebrew University.
Introduction
The effect of quality of health care on demand and choice of treatment is of major
interest to concerned policy makers in many developing countries. The lesson to date from
experiences in cost recovery is that without visible and fairly immediate improvements in the
quality of care provided, increase demand will not support the implementation of user fees.1
However most of the studies that examined the impact of user fees on demand are of limited
use in understanding the effect of quality improvement on demand, since they did not control
effectively for variation in quality of care. Not controlling for quality limit actually even the
interpretation of the coefficients on prices as price effects: higher priced options, such as
treatment in the private sector, are likely to provide higher quality health care. If quality is
not controlled for, the effect of cost on choice will be a combination of negative price effects
and positive quality effects
An important focus of the present paper is the simultaneous influence of price and
quality on decision making.23A unique feature of the present paper is that it controls for the
quality of health care when simulating the effect of cost (travel costs and user fees) on the
choice between health care options. The definition of health care and the impact of user fees
on the use of health care facilities have been the subject of numerous recent econometric
studies (Denton et al., 1990; Aiderman and Gertler, 1989, Gertler and Van der Gaag, 1990;
and Lavy and Quigley, 1991). However, all these studies treat quality of care as an
unobservable and discuss the importance of quality-adjusted measures of price effects by
designating different options of health care as different levels of quality of care (for example,
visiting a doctor versus a nurse, or being treated in a health clinic versus a hospital) without
measuring quality directly. The assumption was that a doctor provides higher quality care
than a nurse and that a hospital provides better treatment than a clinic. This approach,
however, does not allow measuring the sensitivity of consumer demand to various quality
characteristics of services, and therefore is not useful in addressing the main relevant policy
issue, namely identifying which quality improvements can "pay for themselves" with
increased user fees.
Our study, which aims at answering this question, departs from the above literature
by directly measuring the quality of the various options available to the consumer? The
1. The problem of low utilization of modem health care facilities and poor health status in many developing
countries was recently aggravated by budget constraints faced by many governments in developing countries. The
cut in resources available to the social sectors adds a dimension of urgency to the problem. Cost recovery, a
mechanism whereby patients pay part or all of the cost of care in a public facility, is being considered as a means
to generate additional resources for the public health sector (see Jimenez 1987).
2. There has been extensive research on the influence of household characteristics on choice of health care. Some
studies include cost in the form of the value or opportunity cost of travel time (Akin, Griffin and Popkin, 1986;
Gertler and van der Gaag, 1990).
3. Two other recent studies that looked at this issue are Mwabu, Ainsworth and Nyamete (1993) and Alrin and
Guilkey (May 1993). The first of these two studies have used a very small sample of sick people (251) and facilities
(15) from a single district in Kenya. There data on quality was very minimal and their results with regard to quality
2
objective is to empirically model the quality of health care in household decision making with
respect to demand for health care, and to gather and present empirical evidence on the
impact of various policy options on these decisions. In order to do so, we first develop and
evaluate a model of the underlying preference structure for health care. We then examine
how decisions might be affected by changing three factors that affect the choice of each
option: access to health care (distance to the nearest health facility), the price of health care,
and its quality (all are variable in the short run). We also examine an option for the longer
run, that of increasing the level of schooling of heads of households. The decision to choose
between forms of health care involves evaluating the cost and quality of care in each of the
modes available, given household resources and preferences. Utilizing our estimated model,
we simulate the impact on health care choices and demand of improving various dimensions
of quality (drugs and services availability, qualified personnel, and adequate equipment),
improving accessibility (reducing distance) and increasing or decreasing user fees in the
public sector. We also calculate the amount families are willing to pay for improvements in
accessibility and quality of public health services, where a household’s willingness to pay is
measured as a compensating variation. The discrete choice model we specify and estimate
allows willingness to pay to vary with income. As a result we are able to consider the
distributional effects of improving access and quality in the health care system.
Besides modeling quality of care explicitly in the theoretical and empirical
framework, our model extend in two other important aspects the theoretical and empirical
framework used previously to study the demand for health care. First, we depart from the
convention of specifying the conditional utility function as linear in health and quadratic in
consumption in order to allow for differential price effect by income. This specification was
criticized to be restrictive in many ways and we instead use a more general function that is
still consistent with stable utility maximization. Second, the cost of time involved in medical
consultation, both travel and queuing in line for treatment, which have been included in past
studies in the cost of consultation, are treated in our model as facility specific fixed effects,
and are parameterized in the health production technology.
The data used in this study are drawn from the second round (carried out in 1988/89)
of the Ghana Living Standard Survey, a household survey conducted by the government of
Ghana in collaboration with the Living Standards Unit of the World Bank (see Ainsworth and
Munoz, 1986, for a description of the survey methodology). In 1989, a very detailed health
facility survey was specially conducted (immediately following the completion of the
household survey) in exactly the same communities as those covered by the household
survey. We match these facility-level data with the household and individual information to
create a very rich multi-level database for our empirical analysis of the impact of quality of
health services on health care choice.
were not statistically significant and very sensitive to specification. The second study have used a richer data set
from Nigeria.
Quality and Health Care Choice: Overview of Model
The initial structure of our framework resembles the model used in previous studies
of the demand for health care, but the details of the model developed below depart from it in
some important aspects. Individuals who take ill must first decide whether or not to seek
medical care.4 Conditioned on this decision, individuals also have to choose the preferred
type of medical care. The cost of medical care is a reduction in the consumption of other
goods. Decisions are based on maximizing utility which is a function of the individual’s
health and the consumption of other goods. Individuals have to choose between a finite
number of alternatives, including self-treatment and treatment by various care providers.
Each provider offers an expected improvement in health for a certain price. Therefore, in a
sense, there exists a household health production function, conditional on the quality of the
provider and on the characteristics of the household, with health expenditures as inputs.
We define the quality of provider j as the improvement of health status offered at a
cost of XH. This price includes the cost of travel to the health facility, the opportunity cost of
travel time and consultation time and the consultation fee.5 The type of care and provider
chosen implies the health technology and quality of care, which in turn determines the full
cost of treatment. The improvement in health status therefore depends on the provider
chosen. For example, households can expect better treatment in a facility with a modem
infrastructure where they will be attended by doctors and nurses and where needed drugs will
be available. The expected improvement in health status also depends on the ability of the
household to implement the recommended treatment (Gertler and Van der Gaag, 1990). For
example, the expected improvement in health from professional care relative to self-care may
increase with education, since an individual with better education may be better equipped to
implement a recommended treatment.
Formally, the health production function is written as:
=
(1)
where XH are expenditures on health services,
represents the characteristics of provider j
as observed by household n and
denotes some of household n’s is characteristics. From
now on, we will drop the superscript n for simplicity. Expected utility conditional on
receiving care from provider j is written as:
u;^ xc) = ufax^^.c),
(2)
4. We focus on the economic determinants of the choice of the type of medical care when the sample is constraint
to individuals who are ill. We do not discuss the determinants of illness itself because of lack in the data (both
individual and facility characteristics) set of information on factors that affect morbidity but not the selection of
health care provider.
5. The cost of medicine, however, is excluded from the decision process since the demand for medicine is an ex
post decision and individuals do not know ex-ante the type and cost of medicines that they will need and therefore
cannot estimate their cost.
4
where C is the aggregate consumption good and Xc are the expenditures associated with that
consumption. We replace consumption C by total consumption expenditures Xc since we are
not interested in consumption price effects.
The conditional maximization of the consumer is:
Max U^XH,XC) ,
Xu + XcsY
W
where Y is household income. Letting Uj denote the maximum of (3), and using the
convention that j =0 denotes self-treatment, the maximum utility reached using provider j is:
Uj =
.
(4)
The unconditional maximization of the household is
Max{Uj / j = 0 to j],
(5)
where j is the number of alternatives available to the household. The solution of the
optimization problem yields the provider who maximizes the utility function and the ex-ante
amount of resources to be spent on health. This level of expenditures can be written as:
XB = X(Y^r,t)
(6)
where j* denotes the provider which maximizes (5).
An important issue here is whether households have discretion over the total resources
spent on treatment by a given provider. For example, can an individual choose the type of
treatment, medicine (brand and quantity) or number of consultations? If so, then it may be
useful to specify a choice model which takes into account both the choice of health provider
and health expenditures in a composite discrete-continuous choice model as in King (1980).6
However, in order to estimate a discrete-continuous choice model, we would require data on
expected health expenditures which are not available to us. We could use the ex-post
expenditures as a measure of expectations, assuming that they are equal to the ex-ante
expenditures. Such an assumption may be appropriate in the case of housing (King, 1980) or
education (Gertler and Glewwe, 1990), but in the case of health care this assumption is
irrelevant since the information set often changes dramatically when the care provider reveals
the type of illness and the recommended treatment, following which the individual may have
to modify his choices. Therefore, even if a household behaves ex-ante and ex-post as if they
can choose the level of expenditures, we cannot take advantage of this fact in a composite
model. The analysis must therefore be divided into two parts: a discrete-choice model for the
choice of provider (ex-ante) and an ex-post analysis of expenditures conditioned on choice of
provider. In this paper we focus on the discrete choice model.
6. King (1980) modeled and estimated the choice of type of housing and the level of expenditure on housing in a
simultaneous framework.
5
Empirical Specification
In order to translate this theoretical framework into an empirical model, we specify
functional forms for the conditional utility function, the health production function and the
distributions of the stochastic variables. We then derive the probability of choosing a given
alternative which is equal to the probability that the utility of this choice is the highest among
all alternatives. These probabilities will provide the likelihood function which is then used to
estimate the utility function through a maximum likelihood estimation procedure.
The Health Production Function
We assume that consumers, given resource level XH, expect to improve their health
status as follows:
(7)
where Qj is the quality factor of provider j expected by household n (equal to the marginal
product of health expenditures) and where m- is the "fixed cost" of being treated by provider
j. It includes the cost of travel and the opportunity cost of time, both of which have a
negative effect on m“. However, it could also include elements with a positive effect on m“,
such as the free provision of health care to government employees in public clinics or
hospitals, which would affect the choice of health provider for these consumers.
We can compute the "exact” cost of transportation using the price of transportation
and the distance to the facility; likewise, a proxy for the opportunity cost of time can be
derived from hourly wages and the time spent in consultation with a doctor (as in Gertler and
Van der Gaag, 1990). However, this approach may yield a cost variable measured with
error since the opportunity cost of time is not always nonzero. For example, illness may
impair the ability of an individual to work and thus reduce his alternative or opportunity cost
of time, perhaps even to zero. Moreover, we are not interested in the actual cost of
transportation (the ex-post price) but rather in the price as evaluated by the decision maker.
These two prices are different as a result of different sets of information. Therefore, instead
of calculating these costs, we specify them as parameters in equation (7) and estimate them.
Formally, we write the expected fixed cost as:
= y'nJ
where -x-j is a vector of fixed cost characteristics (including a constant term as explained
above) and y is the vector of the expected "prices" of these factors, which is to be estimated.
The quality factor depends on the characteristics of the provider and those of the
household. We assume that:
Q" = exp(a'i|r" +
,
where a and & are vectors of parameters. Note that this functional form leads to a quality
factor which is always positive.
(9)
6
The Conditional Utility Function
We assume that the utility function is Cobb-Douglas:
+
.
(10)
The maximization of (5) subject to the budget constraint XH + Xc < Y gives us
log Uj = alogQj + log(Y+mj). Assuming that nij is lower than Y, we can replace the
second element of the right-hand side of the equation by its first-order approximation to
obtain the following linear form:
log(Y + mj) = log(Y) + log(l + my/y) # log(Y) + m^Y .
(11)
With Qj determined by (9), we finally derive the following:
Uj = a 6 'i|ry + a p'/ +
+ log(Y) .
(12)
In this last equation, we have replaced logUj by Uj. We see from this equation that
the parameters are unidentified. However, it provides us with the form of Uj to be estimated.
We will later replace the products ad and a& by 6 and 0, respectively.
In fact, the right-hand side of equation (12) is only an approximation of the utility
function. This is due to two distinct factors: an error in the utility function (including the
first-order approximation error introduced above) and an error in the quality factor.
Therefore, we introduce two random terms:
UJ = VJ + ZJ +
(13)
where Vj is equal to the right-hand side of equation (12). The term e, is a zero-mean random
taste disturbance with finite variance. It is uncorrelated across individuals and alternatives.
The term »jj is also a zero-mean random term with finite variance, uncorrelated with Et.
arises from the fact that we cannot completely explain the quality factor. For example, we do
not control for the expected severity of illness. If the illness is expected to be severe, the
individual has a strong incentive to seek professional care (whoever the provider), rather than
to choose self-treatment. This demonstrates how the tjj terms might be correlated across
alternatives. For self-treatment, the expression for utility reduces to
Uo = 0O + jq/Y 4- log(Y) + e0 since by normalization, the household attributes no quality
to self-treatment (all the coefficients are zero as is the random term j/q) and there are no fixed
costs involved in seeking medical care.
The individual chooses the alternative with the highest utility. If a variable is constant
between alternatives, it will not influence his choice. Taking the difference between the
utility functions of any two alternatives reveals the factors that influence the choice of health
care:
7
Uj-U, =
+
•
<14)
Consequently, we can assume that all the coefficients of the provider-specific variables are
equal, i.e. we do not need provider-specific parameters y. If the coefficients of the
demographic variables are to be identified, they cannot be constrained to be equal across
providers; therefore, we will allow them to vary across alternatives, i.e. 0 will depend on j.
By the same token, the price factor coefficients which do not affect provider-specific
parameters must be provider-specific in order to be identified. Therefore, some of the y,
such as the constant term, will be provider-specific.
The form in which income enters the utility function is important. If we had assumed
an indirect utility function which is linear with respect to income, the income terms would
have canceled out in (14), implying that income has no effect on the choice of provider.
This is inconsistent with the idea that health is a normal good. To avoid this drawback,
some authors have allowed the parameters to vary across alternatives, as we have done for
the individual characteristics. However, this is inconsistent with the hypothesis of the
maximization of a stable utility function such as (10). An alternative is to introduce the
square of income in the utility function, as in Gertler-Van der Gaag (1990).
We see from the above calculation that the Cobb-Douglas specification of the utility
function leads to an indirect utility function in which income does influence the choice of
facility. A more general parameterization would have been to start from a second-order
translog indirect utility function (which includes a log squared term in (12)), as suggested in
Gertler, Locay and Sanderson (1987). The advantage of this function lies in not imposing
second-order restrictions on the marginal rate of substitution as we have done implicitly by
choosing a Cobb-Douglas utility function. However, rather than introducing a second-order
term, we chose to let the data determine the price effect as explained above. Although
neither are incompatible, our data did not allow us to estimate a first and second-order
income term and price coefficients simultaneously.
Demand Functions and the Distribution of Stochastic Variables
The demand function for a provider is the probability that the utility from that
alternative will be higher than that of any other. In our case, since the random terms may be
correlated, we use a nested multinomial logit (NMLG) model rather than the simple
multinomial logit. The latter assumes that the conditional utilities are uncorrelated and that
the cross-price elasticities are the same across alternatives.
In this paper, we examine the choice between self-treatment (j=0), treatment in the
private sector (j = l) and treatment in the public sector (j=2). Allowing for correlation
between
and £2, the error terms of Uj and U2, and assuming that £0 is independent from Uj
8
and U2, the NMLG determines the following bivariate cumulative distribution function of Ei
and E2'
F^ve^) = exp{-[exp(-c1/ o) + exp(-e2/a)]°}
(15)
and the following c.d.f of £q:7
(16)
GCeq) = exp[-exp(-e0)] .
The probability that self-care is chosen (i.e., the self-care demand function) is then:
expCpy
(17)
exp(K0) + {expCV, / a) + exp(K2/ a)}’
and the probability of choosing a public or private health facility is:
exp(KJo)
= (1 - nJ ----------------- --------------J
v exp(K1 / a) + exp(K2 / a)
J=(l,2) .
(18)
These demand functions can be used to compute the willingness to pay for improved
quality of health care (in a public or private facility) or for the reduction of user fees or
travel time or for improved access).
7. The correlation coefficient between C| and £2 can be shown to be 1-<F. McFadden (1981) shows that £ must be
between 0 and 1 to be consistent with utility maximization. Note that when <r=l, the NMNL reduces to the MNL.
The Demand for Medical Care in Ghana
Ghanaian Health Care
In the late 1980s the health system in Ghana was comprised of some 1,220 service
facilities. The public sector accounted for 46 hospitals and 250 rural health centers; missions
— 35 hospitals and a similar number of clinics; the private sector — for 400 clinics and
nearly 300 maternity homes. Little expansion has taken place in the public sector. Ghana has
trained an impressive number of medical personnel, but the years of economic decline have
led to a mass exodus of qualified people both from the public sector and from Ghana in
general.
A 1989 World Bank health sector review pointed to two main problems with regard to
coverage of health services: (i) poor access: the present health system effectively reaches
only 65 % of the population; and (ii) inequality of access, both between urban and rural areas
(almost 100% of the urban population being covered and only about 50% of the rural
population) and even more so between regions (rural coverage varies by region from 11 % to
100%). The report raises major concerns about the quality of health care in the public sector:
inadequately trained staff, uneven geographical distribution of clinical personnel, widespread
shortages of drugs and inadequate and improperly used equipment. With regard to the
deterioration in quality in the public health care system, the report cites the dramatic decline
in utilization of health services. The number of outpatients fell from 10-11 million in 1973 to
approximately 5 million in 1987.
The costs of consultation and medical care are borne by the individual and there is no
system of health insurance in Ghana. The only exception is public-sector employees and their
families, whose medical expenses are reimbursed by the government if incurred at public
health facilities.
The government of Ghana has adopted major reforms since 1989 in an effort to
improve the quality of health services and the coverage of the population. These reforms
included in-service training of all medical staff, reallocation of personnel to underserved
regions, maintenance of drug supplies and replacement of inadequate medical equipment. The
World Bank study suggests that in the medium term, most service charges should either be
increased or indexed to inflation. Drug prices should be determined by a cost-based formula.
The study also recommended setting up a national health insurance scheme to be managed by
a new and separate institution which would be independent of the Ministry of Health.
This study is designed to shed light on the efficacy of some of the proposed policy
changes. Specifically, it delves into the interrelationship between health care quality and the
goals of increased utilization and the possibility of financing quality improvements through
increased user fees.
10
The Data
We use individual, household- and community-level data from the second year (1988)
of the Ghana Living Standard Survey (GLSS), which was also conducted in 1987. This was
an integrated survey of 15,000 individuals in 3,200 households comprising 200 clusters,
randomly drawn from the population (see Scott and Amenuvegbe, 1989, for details). The
GLSS collected socio-economic information such as household composition, demographic
characteristics, time use, income and consumption, education, and health status. The health
statistics provide a detailed description of health care and the incidence of morbidity during
the thirty days prior to the survey date including information on the length of illness, choice
of treatment (self, nurse, doctor, traditional healer), type of health facility visited (public or
private, clinic or hospital), expenditure on consultation and drugs, travel time and cost.
Over 5,000 individuals, approximately one third of the sample, experienced a period of
illness or were injured during the relevant four weeks; forty percent of them sought some
form of modem health care treatment.
The 1988 GLSS was supplemented with a health facility survey. Responses were
obtained from 231 facilities over a six week period.89 Facilities were selected for
interviewing on the basis of proximity to a household cluster. The nearest health facility to
each cluster of the GLSS was surveyed first; if the nearest facility was private, the nearest
public facility was also surveyed.10 The health facility survey collected information about
infrastructure (beds, vehicles, laboratory, operating room, etc.); personnel (number of
doctors, nurses, medical assistants, etc); availability of health services and drugs (number of
hours open per week, type of services provided, stocks of 16 types of drugs) as well as fees.
The empirical analysis in this paper is based on a sample of 6,000 individuals from 88
clusters for which both the nearest public and private health facilities were surveyed.11
Approximately 2,150 of the individuals in the sample reported an incidence of illness or
injury. We use their revealed choices between self-treatment, a visit to a private clinic or to a
public health facility in order to study the demand for health care and the willingness to pay
for improved quality. The survey methodology suggests that in each of the clusters the
8. A cluster is a geographic area such as a village or neighborhoods of a city. Approximately 16 households were
interviewed in each cluster.
9. A detailed description of the health facility survey and an analysis of the data is given in Reed (1990)
10. This procedure does not necessarily lead to a representative sample of health facilities. However as noted in
Reed (1990), the distribution of the population and the sample of facilities across Ghana’s ten administrative reoinnc
do not differ significantly from one another.
11. This sample included 68 profit^riented health facilities and 20 private mission facilities The profit-orient^
sample included 6 hospitals and 62 clinics. The mission facilities accounted for 10 hospitals and 10 clinics in the
private facility sample. The sample of public health facilities included 14 hospitals and 74 clinics.
“
11
neares
ac ty is a private one and the data confirm this. However, our sample is
S1, ..ar ° . e one 0 . ^^ividuals who live in communities where the nearest health facility is
public. This conclusion is based on the comparison of the distributions of various
characteristics, such as age, gender, expenditures, income, educational attainment,
emp oyment, school enrollment, morbidity and utilization of health care services (see table in
e appendix). The geographical distribution of the sample clusters resembles the dispersion
o the GLSS sample of clusters as a whole. We therefore assume that there is no sample bias
in our results.
Definition of the Variables Used in the Estimation
Provider quality Q/ The facility survey provided us with a long list of characteristics.
Due to the strong multicolinearity among the variables, we have decided to group them
according to five measures of quality.12 The availability of essential drugs is an obvious
category.13 The second measure is the number of medical staff as an indicator of the level
of human resources available at the facility which may reflect the sophistication and range of
health services provided. The third is the provision of basic adult and child health services
measured by the availability of a functioning laboratory, the ability to vaccinate children and
the ability to provide prenatal, postnatal and child growth monitoring services (grouped
together as ‘mother and baby care’). The availability of electricity and running water are
good indicators of the quality of infrastructure, since electricity is essential for the
refrigerated storage of vaccines14 and running water is required in order to offer obstetric
care.
The variables used are defined as follows:
Drugs - mean availability of ampicillin, chloroquine and paracetemol.15
12. Very few studies in the health economics and bio-medical literature provide useful guidelines for building or
constructing health quality indices from facility level data. Garner, Thomason and Donaldson (1990) and Peabody
et al. (1993) attempted to deal with this problem.
13. Low availability of drugs may actually indicate high demand and intensive utilization of a health facility,
signaling higher rather than lower quality of health care. We have no way to deal with this endogeneity of measured
stocks of drugs, or, for that matter, any other consumable measure of quality. We expect that the sign of the
coefficient in the estimation will indicate the net effect of drugs on the demand and choice of health care. This point
is also discussed in the conclusion of Mwabu et al (1993).
14. This statement should be qualified since many developing countries use kerosene-fueled refrigerators when
electricity is unavailable.
15. The questionnaire focused on 11 essential drugs (excluding vitamins). Chloroquine in the form of tablets, syrup,
or injection and any other anti-malarial drug constituted four of those listed and all are included in the variable
drugs.
12
Personnel - number of doctors and nurses.16
Infrastructure - equals 1 if the facility has running water and electricity and 0 otherwise.17
Services - mean availability of immunization, laboratory and mother and baby services.18
Oprm - equals 1 if the facility includes an operating room and 0 otherwise.19
Price variables
Distance - distance of facility from the cluster in kilometers.
Price of consultation - price of a regular consultation.20
Government employee - equals 1 if the head of the household is a government employee and
0 otherwise. This status entitles the family to free health care in public health facilities.
Demographic and economic variables
Income - monthly per capita expenditures, deflated by a monthly price index (divided by
10000/12 in the estimations).21
Schooling - Own years of schooling.
Male - equals 1 if individual is a male, 0 if female.
Head of household’s schooling - years of schooling of head of household.
Table 1 presents the means and standard deviations of the variables. Sixty six percent
of the sample chose self-treatment, 20 percent chose to visit a private facility and 14 percent
chose a public facility. The sample included roughly the same number of males and females;
16. Information is also available on the number of administrators and non-medical staff. We decided to focus on
medical staff only, since this measure probably has a high correlation with the quality of health care at the facility.
The data refer to the actual number of working staff rather than the book value. Recent studies on the impact of
health care quality on health outcomes indicate the importance of distinguishing between actual and book value of
personnel (Thomas, Lavy and Strauss, 1992; Lavy, Strauss, Thomas and de Vreyer, 1992).
17. Detailed information on the number of refrigerators, fans and air-conditioners are also given, but are highly
collinear with the availability of electricity.
18. Availability of mother and baby services, prenatal and postnatal care and programs for the malnourished child
are highly correlated and we therefore decided to include only the former in ‘services’.
19. All the hospitals in the sample had an operating room as compared to only 28 percent of the clinics. Similar
proportions were found for the presence of a laboratory.
20. In cedis (divided by 100). The official exchange rate in 1988 was 200 cedis per U.S dollar. The ’free’ market
rate was close to 300 cedis per dollar.
21. Per capita expenditure includes all expenditure except that on health care. It also includes imputed rent for home
owners. Since the survey was conducted over 12 months, we computed real values by deflating with the monthly
price index.
13
Table 1: Descriptive Statistics
Sample Size 2,126
Mean
Standard
deviation
0.88
0.92
0.59
0.37
0.36
0.70
1.90
0.50
0.36
0.48
Price of consultation
3.42
1.06
5.64
1.43
Characteristics of public facilities
Drugs
Personnel
Infrastructure
Services
Operating room
0.66
2.00
0.50
0.76
0.35
0.28
3.10
0.50
0.24
0.48
8.33
0.44
8.12
0.35
5853.00
2.38
3.26
0.48
4.42
0.07
0.20
0.14
0.66
4949.00
2.10
4.41
0.50
4.93
0.25
0.39
0.35
0.47
Variable
Characteristics ofprivate facilities
Drugs
Personnel
Infrastructure
Services
Operating room
Price factors
Distance to facility
Price factors
Distance of facility
Price of consultation
Household characteristics
Monthly per capita expenditure
Age (Divided by 10)
Schooling
Male
Head of household education
Government employee
Private facility chosen
Public facility chosen
Self-treatment chosen
14
the average age in the sample was 24 years with the mean years of schooling at just over 3.
The private health sector is, on average, better stocked with drugs and possesses better
infrastructure and equipment. The public health facilities have, however, a greater number of
medical personnel and provide more diversified services. As discussed above, the average
distance to the nearest public health facility is more than twice the average to the nearest
private facility. However, the mean consultation fee in the private sector is more than twice
that in the public sector.
Estimation Results
The NMNL model was estimated with full information maximum likelihood using two
different specifications. The first is the model described above, which includes the various
characteristics of the care provider; the second is similar to the first except that it excludes
the quality variables. The results are presented in Table 2. Sigma is equal to 0.50 in the first
model and 0.48 in the second model. Both are significantly different from zero and from
one. Therefore, the model is consistent with utility maximization (0<a< 1) and the
correlation between the private and public alternative is different from zero (a< l).22
Analyzing the individual coefficients, we are able to conclude that the individual’s sex
and own education level do not affect the choice of health care (both variables have
coefficients equal to zero).23 Age has a significant negative effect both in the private
alternative (-0.19) and the public (-0.21), suggesting that the probability (conditional on
illness) of seeking medical care decreases with age. Better educated heads of households tend
to favor health care at private facilities.
In discussing "price factors", recall that they are divided by household per capita
expenditure. The effect of distance is negative, large, and significantly different than zero,
suggesting that the probability of seeking professional care would significantly increase if
health care were more accessible. This result is similar to the one reported in Lavy and
Quigley (1991) who used the 1987 (first year) GLSS data. Similarly, the price of consultation
has a significant negative effect. An increase in a facility’s user fees will lower both the
probability of that facility being chosen and that of choosing modem health care. The
"constant" term (constant/income) is negative and quite large for the private alternative and is
significant. Recall that this term arises from the fixed cost of health production. It could be
an indication that households consider the government subsidization of treatment in a public
facility as better ’quality’ (getting more ’service’ for the actual fee paid), which, ceteris
paribus, enhances the probability of choosing public health care. Finally, the government
employee dummy variable has no effect on the probability of the private alternative being
chosen, but has a significantly large positive effect on the public alternative being chosen.
This could, of course, be a reflection of the policy which entitles government employees and
their families to free health care at public facilities which significantly increases the
probability of their choosing a public facility. This suggests that the dummy variable should
be interpreted as a price effect rather than a quality effect.
22. Note that a is slightly lower in the specification which excludes the quality variables. This could be due to the
fact that the random term in this model includes the random term of model one plus the quality terms, and
consequently, the residual terms should be less correlated. However, this difference is very small, indicating that
the correlation between the private and public alternative is not due primarily to the quality of the provider. This
result is not surprising since the correlation was interpreted as capturing the severity of illness and therefore, should
not be influenced by the quality of the facility.
23. Similar results, of no effect of gender and own education on choice of health care, are reported in Mwabu,
Ainsworth and Nyamete (1993), who used 1981 household and facility level data from Kenya.
16
Table 2: Maximum Likelihood Nested Multinomial Estimates of Choice of Health Care
Without quality
c oefficient
With quality
Variable
(1)
T-Values
(2)
Coefficient
(3)
T-Values
(4)
a
0.50
5.47
0.48
5.05
Quality offacility
Drugs available
Personnel
Infrastructure
Services
Operating room
0.64
0.10
0.34
0.55
-0.19
2.39
3.03
2.30
2.55
1.05
-0.07
-0.11
8.15
4.36
-0.07
-0.08
-8.52
-3.50
-1.68
-0.18
-0.00
0.03
0.07
-4.57
-2.94
-0.07
0.16
2.48
-0.68
-0.19
-0.00
0.06
0.07
-2.29
-2.97
-0.08
0.27
2.57
-0.22
-0.00
-0.13
-0.20
-0.26
-0.00
-1.75
-0.01
-2.14
-0.21
0.04
0.05
0.02
-5.65
-3.1
1.36
0.26
0.69
-0.75
-0.20
0.04
0.11
0.02
-2.31
-2.84
1.24
0.49
0.57
-0.07
0.55
-0.53
2.53
-0.20
0.57
-1.41
2.54
Price factors
Distance/Income
Price of consult./Income3
Private alternative
Constant
Age
Schooling
Male
Head of household education
Specific price factor
Cons tant/Income
Member of govemment?/Income
Public Alternative
Constant
Age
Schooling
Male
Head of household education
Specific price factor
Constant/In come
Member of govemment?/Income
a Income was multiplied by 12/100000 for the estimation.
17
All the quality factors have significant and large positive effects. The individual and
joint statistical significance of these variables indicates that households take into account the
various dimensions of the quality of the health provider in making their choice. Households
prefer facilities with better infrastructure and are more likely to visit a facility where drugs
and diversified services are available. They also attach importance to the probability of being
treated by a doctor or nurse. The only quality variable not correlated with choice is the
dummy variable ’operating room’ which is negative, although not significantly different
from zero at the 5 % level. These results emphasize the importance of including quality
variables in a health demand choice model.24
In order to clearly demonstrate this fact, we re-estimated the model without the
quality factors (columns 3-4 in Table 2). In comparing our two models, we see that the
coefficient of the variables included in both models are very similar. However, note that the
price coefficient fell from -0.11 to -0.08, suggesting that omitting quality from the demand
equation leads to a downward bias in the price elasticity. The distance coefficient, on the
other hand, does not change following the exclusion of the quality variables, perhaps
indicating low correlation between distance and quality.25
The distance variable was interpreted as a price factor arising from the cost of travel
and opportunity cost of travel time. This variable is therefore divided by income as is the
other price factor. As a result, the distance price effect declines in importance as income
rises. However, it can also be argued that the opportunity cost of time is the more important
of the two costs incurred in travelling to the facility. This cost is equal to the time lost during
travel (proportional to distance) multiplied by the hourly wage of the individual (which, on
average, is proportional to income). Therefore, although distance is a price factor, it could
have been introduced without being divided by income. When we estimated such a model,
the value of a jumped to 0.65, indicating that the private and public alternatives are less
correlated in the latter model. This correlation might be due to the distance factor, which was
underestimated for wealthier households.
24. In a recent study, Mwabu, Ainsworth and Nyamete (1993), the authors also analyzed the effect of quality of
medical care on choice of medical treatment. With somewhat limited data (only fifteen facilities with only two
quality indicators in the analysis: the number of different types of drugs available in a health facility and number
of health workers in the facility) they report no significant effect of any of the quality indicators on the choice of
health care in Kenya. The number of staff at the facility was included as a control for the size of the facility. The
coefficient on the total number of drugs at the facility was positive and insignificant. In another specification the
authors replaced this variable was replaced by variables on the availability of two specific drugs- aspirin and
antimalarials. the coefficient on absence of aspirin was negative at a 0.1 level of significance and on absence of
antimalarials was positive and significant at a 0.02 level of significance. In an attempt to explain these
counterintuitive results, the authors discuss the possible endogeneity of the quality variables they have used.
25. Some of the private health facilities are ’missions’ facilities. We estimated our two models once more including
a dummy for ’mission’ health facilities. This dummy was not statistically significant in the specification that included
the quality variables but became significant when these variables were dropped. This is an indication that mission
health centers provide high quality health care which is measured adequately by the quality indices.
Probability of Using Health Care Under
Various Policy Assumptions
Using the multinomial logit model, we simulated the impact of several hypothetical
policy decisions on health care demand and the choice between private and public care based
on the average characteristics of the individuals and facilities in the sample. Tables 3A-3C
present several simulations involving improvements in quality: easier access (reducing the
distance to the nearest public and/or private facility) and changes in user fees for the public
and/or private facilities. In each scenario we calculated the probability of each choice and the
percentage change relative to a referenced baseline, which in most cases was the choice at
the mean of the sample characteristics.
Improvement in Quality of Care
In the first scenario the public health care system emphasizes the quality of health
care. The first three rows of the top panel in Table 3A present the outcomes of single
complete improvements in each of the three dimensions of quality in public facilities.2627
Relative to the baseline rates, these improvements reflect an increase of 51, 100 and 31
percent, respectively, in drug availability, infrastructure and services. The corresponding
changes in the predicted probabilities of using a public facility are quite large: e.g., the
improvement in drug availability leads to a 44 percent increase in utilization. A similar
response (elasticity of 0.8) is obtained from an increase in the variety of services available at
the public facility. Improving infrastructure by 100 percent leads to an increase of only 25
percent in the predicted probability of choosing treatment at a public facility. The
implementation of all these improvements simultaneously (row 5 in the table) increased the
probability of an individual choosing a public facility to 0.1 (from a baseline of 0.04).
Similarly, a total collapse of these three dimensions of quality (from their mean value to
zero) leads to a 75 percent decline in the predicted probability of using public health care
(row 6 in Table 3A). This result could explain part of the fall in utilization rates of public
health facilities in Ghana during the early 1980s. Ministry of Health data suggest that public
health facilities experienced a 40 percent decline in utilization from 1979 to 1983. During
the same period the quality of public health services deteriorated dramatically due to
inadequate staffing, shortages of medication, cancellation or lapses in immunization programs
and an overall breakdown in health facility physical infrastructure.77
Increasing the number of doctors and nurses in public facilities is also effective in
inducing an increase in the utilization of public health care. Increasing the number of doctors
and nurses to three (a 50 percent increase from the sample mean) leads to a 20 percent
26. By complete improvements we mean a rise in the quality indicator from its sample mean value to a value of
1.0.
27. The years of economic decline have led to a mass exodus of qualified people both from the public health sector
and from Ghana in general. The depletion and expiry of drug stocks became. an all-pervasive problem in the public
sector. See Ghana (1989) for more detail.
19
Table 3A: Improved Public Health Care, Predicted Probabilities and
Percent Change in Health Care Use
Self-care
Public
Facility
Private
Facility
0.82
0.04
0.14
Drug =1.00
0.81
(1.2)
0.06
(43.7)
0.13
(-7-0)
Infrastructure =1.00
0.81
(-0-9)
0.06
(33.0)
0.13
(-5.3)
Service = 1.00
0.81
(-0.6)
0.06
(24.9)
0.13
(-4.0)
Personnel = 3.00
0.81
(-0.5)
0.05
(18.9)
0.14
(-3.1)
Drug = Infr.=Serv. = 1
0.79
(-3.5)
0.10
(127.6)
0.11
(-19.5)
Drug=Infr.=Serv.=0
(a total collapse of public
health care)
0.83
(1.2)
0.01
(-75.0)
0.16
(14.3)
0.70
(-14.0)
0.07
(61.0)
0.23
(60.0)
Probabilities at the mean
Improve Quality of Public Facilities
from Sample Mean to 1.00
Improving AU Quality Factors Simultaneously
in the Private and Public Sectors
Drug = Infr. = serv. = 1
Note: The percent changes are given in parentheses.
20
increase in the predicted probability of choosing public health care.
o
Another interesting simulation (last row of table 3A) involves a complete
improvement of all the quality factors in the public and the private sector simultaneously.
Such changes lead to a 0.30 predicted probability of consultation — 0.07 in the public sector
and 0.23 in the private sector. The relative change in the predicted probabilities is almost
identical for the two sectors, 61 percent in the public sector and 60 percent in the private
sector.
Improving Access to Public and/or Private Health Care
A second scenario involves the construction of additional public health facilities in
order to reduce the distance to the nearest clinic in each community. The top panel in Table
3B suggests that such a government program would have an immediate payoff by
dramatically increasing the utilization rate of health facilities. Halving the mean distance to
the nearest public clinic would increase the probability of utilization by 250 percent, from
0.04 to 0.15, and would increase the overall probability of seeking professional care from
0.19 to 0.24. The results are almost identical if the distance to the nearest private facility is
halved (middle panel in Table 3B). Obviously, if the distance to both the nearest public and
private health facility were halved simultaneously, the impact on the probability of seeking
health care would be much larger. The bottom panel in Table 3B simulates the predicted
probabilities following such a simultaneous reduction in distance. The predicted probability
of consultation is increased to 0.28 (from 0.18), but most of the change is in the demand for
the public health care, which rises from 0.04 to 0.13 (the private probability is almost
unchanged, increasing from 0.14 to 0.15). This result reflects the unique feature of our data
— that the nearest facility is always private; thus, halving the distance to the nearest public
facility is always a larger absolute change. At the mean, for example, the change implies a
4.2 km reduction in distance to a public facility, compared to only 1.7 km in the case of a
private facility.
Table 3C suggests that the impact of increasing user fees in the public sector is fairly
modest: doubling fees would reduce demand for public health care by only 11.3 percent,
while tripling them would result in a 21.5 percent decline in predicted use (implying an arc
own price elasticity of 0.12). Note that in this simulation we allow individuals to substitute
private care and self-care for public care, which became relatively more expensive.
However, the cross-price effect on the demand for private and self-care is very modest,
leaving their predicted demand probabilities almost unchanged. On the other hand, changes
in the price of private health care lead to larger own and cross effects: the own-price
elasticity is -0.17 and the cross-price elasticity is 0.15 (see middle panel in Table 3C).
Increasing user fees simultaneously for public and private health care leads to a relatively
large decline in demand for private care (an elasticity of about 0.1) and a very modest
increase in the demand for public care: when the set of all modem health care opportunities
becomes more expensive, consumers tend to resort more to public care.
21
Table 3B: Improved Access to Health Care and Predicted Probabilities
and Percent Change in Health Care Use
•
Private
Public
Facility
Facility
Self-care
Improving Access to
Public Health Facilities
Reduce distance by 25%
0.81
(-1.1)
0.06
(41.5)
0.13
(-6.7)
Reduce distance by 50%
0.79
(-2.6)
0.09
(95.9)
0.12
(-14.9)
Reduce distance by 100%
0.76
(-7.2)
0.15
(244.9)
0.09
(-34.8)
Reduce distance by 25%
0.81
(-1.1)
0.04
(-8.0)
0.15
(10.0)
Reduce distance by 50%
0.79
(-3.0)
0.03
(-15.0)
0.17
(20.0)
Reduce distance by 100%
0.77
(-6.0)
0.03
(-28.0)
0J0
(-41.0)
Reduce distance by 25%
0.79
(-2.0)
0.06
(32.0)
0.15
(3-0)
Reduce distance by 50%
0.77
(-5.0)
0.08
(72.0)
0.15
(5.0)
Reduce distance by 100%
0.72
(-U.0)
0.13
(182.0)
0.15
(6.0)
Improving Access to
Private Health Facilities
Improving Access to Public
and Private Health Facilities
Note: The percent changes are given in parentheses.
22
Table 3C: Increasing the Cost of Health Care and Predicted Probabilities
and Percent Change in Health Care Use
Self-care
Public
Facility
Private
Facility
Increasing user fees by 50%
0.82
(0.1)
0.04
(-5.8)
0.14
(1.0)
Increasing user fees by 100%
0.82
(0.3)
0.04
(-11.3)
0.14
(1.9)
Increasing user fees by 200%
0.82
(0.5)
0.03
(-21.5)
0.15
(3.6)
Increasing user fees by 50%
0.81
(1.0)
0.05
(7.0)
0.13
(-9.0)
Increasing user fees by 100%
0.83
(2.0)
0.05
(15.0)
0.12
(-17.0)
Increasing user fees by 200%
0.85
(5.0)
0.10
(31.0)
0.06
(-33.0)
Increasing user fees by 50%
0.83
(1.0)
0.04
(2.0)
0.14
(-8.0)
Increasing user fees by 100%
0.84
(2.0)
0.05
(3.0)
0.12
(-15.0)
Increasing user fees by 200%
0.85
(5.0)
0.05
(5.0)
0.10
(-29.0)
Reducing the Subsidy on Public
Health Facilities
Increasing User Fees in Private
Health Facilities
Increasing User Fees in Public and
Private Health Facilities
Note: The percent change* are given in parentheses.
23
Raising Price to Improve Quality
What is the tradeoff between the price and quality of health care? What would be the
result of public care providers raising their fees in order to improve the quality of the care
they provide? The price simulations clearly indicate that given the low price responsiveness
of demand, prices would have to increase drastically in order to offset the effect of
improving quality. For example, if the availability of drugs, services and infrastructure in
public health care were improved by 100 percent, user fees would have to increase by more
than 1,200 percent to offset the increased predicted probability of choosing a public facility.
Given that many individuals in the sample report being treated at public facilities for no fee,
this result should be interpreted with some caution. If the distance demand elasticity is
interpreted as a price effect, then the tradeoff between distance and quality is a relevant
simulation as well. The government could build fewer facilities, increasing the mean distance
to the nearest public facility, but offset this effect by improving quality. Improving quality by
100 percent is equivalent to doubling the mean distance to the nearest pubic health provider
in terms of the impact on the predicted choice probability. These simulations clearly indicate
that there is large scope for quality improvements to be financed, at least partially, by an
increase in user fees.
An Increase In Family Human Capital
The level of household human capital, as measured by the years of schooling of the
head of household, has a strong positive effect on the preference for modem health care,
especially in the private health sector. It is interesting to look at the elasticity of demand for
modem health care in response to an increase in the level of completed schooling in the
economy. The last row of Table 3B suggests that if the mean years of schooling were to
double from 4 to 8, utilization of modem health care would increase by 30 percent. Such a
magnitude suggests large cross-sectoral benefits from investments in the education sector.
Willingness to Pay
In this section we calculate the welfare-neutral price changes under various
scenarios.28 The compensating variation can be computed using equations (19) and (20).29
We calculated the welfare-neutral price changes under several different scenarios. The first
set of scenarios focuses on the distance effect: how much is an individual willing to pay to
have his travel costs reduced by one km (we will refer to this scenario as "Minus 1km"), by
two km (scenario "Minus2km"), halved ("Dist/2") and reduced to zero ("Dist-zero"). The
second set of scenarios looks at the welfare-neutral price change from two different angles:
the price being halved ("Price/2") and the price equal to zero ("Px=Zero"). Finally, we
analyze the willingness to pay for improved quality in existing public facilities: starting from
a state in which the public facility has the lowest quality rating, we successively increase
each of the quality factors. More precisely, let us denote the personnel, drug, service and
infrastructure quality variables by P, D, S and I, respectively. We assume for initial values
P=2 and D=S=I=0, and then successively compute the willingness-to-pay for D = l, S = l,
1=1, P=4, D=S = 1, D=I = 1, S=I=1, D=S=I=1. The results are presented in Table 4.
Each simulation is computed using the mean individual characteristics for three different
values of income: the sample mean income (12), 12 minus the sample standard deviation (II)
28. Willingness to pay is measured by the compensating variation of income necessary to keep the individual at least
as well off after the improvement as before. Small and Rosen (1981) have shown that in discrete models, this
compensating variation can easily be expressed using the demand function. In the case of the NMNL, the amount
of resources that an individual must be given to be as well off after a change as before the change is:
cv =
^exp^ + (expCK, / a) + exp(K2/ a))°]
(19)
- ln[exp(P,0)) + (exp(K/1 / a) + exp(l"2/ a))’],
where Vj and V, are the utilities before and after the change and X is the marginal utility of income. With the
functional form of the utility given by (10), we find that:
BU =
1
=
BY
Y+mj
Y '
(20)
29. Note that combining these two equations gives us the compensating variation as a percentage of income:
CV
— = ln^exp(F0) + (exp(P1/a) + exp(KJ/a))°)
-ln(exp(r/0)) + (exp^/a) + exp(V,Ja))a'))),
where Vj is the initial level of utility and vj is the level of utility following the improvement in social services.
25
Table 4: Willingness to Pay for Improvement of Health Services
(Prices in cedis; 175 cedis = 1$)
Distance
Price
Quality
II
12
13
I2WQ
Scenario
40000
70242
100000
70242
Minus 1
14
26
34
34
Minus2
34
58
72
74
Dist/2
375
422
436
526
Dist-Zero
375
422
436
526
Px/2
4
9
12
8
Px=zero
9
18
23
16
D=1
13
53
117
S=1
10
42
68
1=1
3
20
33
P=4
3
11
17
D=1 &S = 1
47
186
305
D=1&I=1
30
120
197
S = 1 & 1=1
23
98
161
S = 1 &I=1
23
98
161
D&S&I =1
91
353
580
-
26
and 12 plus the sample standard deviation (13). For comparison purposes, we also computed
willingness to pay in the model without quality variables, again using the mean individual
characteristics. 111656 results are presented in the last column of Table 4 (12WQ).
Reading the table across rows shows the change in willingness to pay as a percentage
of income as income increases. Reading down a column shows, first, the change in welfare
neutral price effects as the distance to the nearest public facility decreases; then, as the price
of consultation decreases; and finally, as quality increases.
First, we note that the welfare-neutral price change increases with income, no matter
what the scenario. However, the income effect is more important for the quality variables
(infrastructure, drug availability, services and personnel) than for distance or price. In fact,
examining willingness to pay as a percentage of income, we note that for the price of
consultation, this percentage decreases as income increases, while for the quality factors it
increases as income increases. Not surprisingly, the richest households are more concerned
with quality of health services than with its cost (price or distance).
Next, we compare the willingness-to-pay for each of the simulated changes. The most
important factor for households seems to be distance to the health facility: the representative
household is willing to pay 7.2 percent (i.e., 422 cedis) of their monthly income to have
their travel distance (or travel time) to the nearest public facility reduced to zero. The second
most important factor is drug availability: households are willing to pay 117 cedis to have a
facility stocked with the basic drugs. The importance of access and drug availability to
consumers can also be seen from a simulation in which public and private facilities were both
the same distance from the cluster and had the same level of drug availability. In this case,
the probability of consulting would be slightly higher in the public sector (11.6 percent in the
public sector and 10.6 percent in the private sector). The willingness to pay for such a
change was no less than 4.75 percent of monthly income.
Households are also concerned with the services provided by the health facility.
Households in the higher income category of our sample are willing to pay 68 cedis (0.8
percent of monthly income) to insure the availability of child services, immunization services
and a laboratory in the nearest public health facility. The lowest level of compensating
variation is the one required following an increase in quality of infrastructure and number of
qualified personnel: households are willing to pay a third of a percent of their monthly
income to insure infrastructure quality in the nearest public health facility or to pay 17 cedis
(0.2 percent of income) to double the number of doctors and nurses.
Are the welfare-neutral prices consistent with the level of user fees in public
facilities? To answer this question, we examined the ’willingness to pay’ for free public
health care in Table 5. The welfare-neutral price is about 20 cedis which is not far from the
average price of consultation (40 cedis), especially since about 50% of those who choose a
public facility do not pay for consultation. Therefore, our estimate of the ’welfare-neutral
price’ for free consultation is approximately equal to the sample mean price of consultation.
27
Table 5: Willingness to Pay for Improvement of Health Services
(Percent of monthly income)
Distance
Price
Quality
II
12
13
12WQ
Minus 1
0,43
0,45
0,40
0,58
Minus2
1,01
0,99
0,86
1,27
Dist/2
3,19
2,65
2,14
3,37
Dist-zero
11,25
7,21
5,23
8,89
Px/2
0,13
0,15
0,14
0,14
Px=zero
0,28
0,31
0,27
0,28
D= 1
0,39
0,91
1,04
-
S= 1
0,31
0,71
0,81
-
1=1
0,09
0,35
0,40
-
P=4
0,08
0,18
0,20
-
D=1 & S = 1
1,40
3,18
3,66
-
S=1 &I=1
0,90
2,05
2,36
-
S=1 &I=1
0,69
1,68
1,93
-
D&S&I=1
2,72
6,03
6,95
-
28
We did not compare the amounts households are willing to pay for quality
improvements to the cost of these improvements which would have provided a complete cost
benefit evaluation of quality improvements. However, the comparison of public health care
with the private sector does provide an opportunity to simulate a "complete" cost-benefit
analysis, assuming that the private sector maximizes profit and that user fees are set at levels
appropriate to recover both fixed and marginal costs. The relevant simulation set the public
sector quality factors equal to those in the private sector. Note that since average values for
the personnel and service variables are higher in the public sector than in the private sector,
we did not adjust them down to the private mean values. Thus, the simulation consisted of
adjusting ’Drugs’ and ’Infrastructure’ in the public sector to their mean values in the private
sector (0.87 and 0.56, respectively). In the first simulation, the price of consultation was
kept constant; in the second, user fees in the public sector were set to equal the mean private
sector fees. Willingness to pay in the first simulation amounted to almost one percent of
monthly income (60 cedis), which is almost two thirds of the mean private sector user fee.
The second simulation, which increased public-sector user fees, suggests that households are
willing to pay non-zero values for improved quality which is accompanied by increased fees
(about 20 cedis per month). These results indicate that quality improvements in the public
sector could be financed by increased user fees.
Conclusions
The design of policy reforms in the public health care sector requires reliable
estimates of the effects of improvement in quality and access and the extent to which these
improvements can be financed by raising user fees for health services. We have utilized
household data from Ghana in order to estimate the effect of these policy changes.
The use of distance to health care facilities and user fees as measures of cost has
appeared in previous studies and the results suggest a strong negative effect of distance and
travel cost on health- care utilization. But the direct price effect of user fees is of lower
magnitude. Our results confirm this pattern: distance is an important factor in deterring
individuals from seeking modem health care while prices are less important in the demand
for and choice of the care provider.
This paper’s main contribution is in its attempt to rigorously evaluate the effect of
quality of service and assess the likely outcome of various policy scenarios involving
improvements in the quality of public health services. Improving basic services such as
vaccinations, child care and the availability of drugs is likely to have a significant effect on
demand for health care. The estimated effects of improvements in facility infrastructure and
staff are positive, but have lower elasticities. The tradeoff between improvements in quality
and an increase in cost, measured either by user fees or by distance, suggests that there is a
wide scope for financing quality improvements in the public health sector through raising
fees or by increasing distance (building fewer facilities).
The results of the policy simulations are supported by the outcome of the willingnessto-pay computations. Households are willing to pay almost one percent of their monthly
income for improvements in drug and service availability. They are willing to pay even more
for improving their access to the public facility: the willingness to pay for a reduction in
distance of two kilometers is as high as that for having drugs available on demand. To
gauge the welfare implications of increasing user fees to finance improvements in the quality
of health care or, alternatively, to finance the improvements through increased travel time
(i.e., trading quality for access or for travel time and cost requires comparing willingness-topay estimates with the capital and operating costs of the various investment scenarios. This
difficult task is left for future work.
However, as an alternative, we compare the willingness to pay for equalizing the
quality of public and private health care to the level of user fees in the private sector.
Presuming that fees in the private sector are set to cover costs, we find that quality
improvements in the health sector in Ghana can be financed by increasing user fees.
References
Alderman, H. and P. Gertler (1989). "The Substitutability of Public and Private Health Care
for the Treatment of Children in Pakistan." Living Standard Measurement Study
Working Paper 57. Washington, D.C.: World Bank.
Akin, J., and N. Birdsall (1987). "Financing of Health Services in LDCs," Finance and
Development, June, 40-43.
Akin, J., C. Griffin, D. Guilkey and B. Popkin (1986). "The Demand for Primary Health
Care Services in the Bicol Region of the Philippines", Economic Development and
Cultural Change, 344: 755-782.
Akin, J., D. Guilkey and E. Hazel (1993). "Multinomial Probit Estimation of Health Care
Demand: Ogun State, Nigeria", Draft, Department of Economics, University of North
Carolina, Chapel Hill, N.C.
de Ferranti, D. (1984). "Paying for Health Services in Developing Countries: An Overview,"
Staff Working Paper, No. 721, The World Bank, Washington, D.C. Processed.
Denton, H. et al., (1990). Federal Republic of Nigeria: Health Care Cost, Financing and
Utilization, Western Africa Department, Population and Human Resources Operations
Division, The World Bank, Washington, D.C. Processed.
Gamer, P., J. Thomason, and D. Donaldson (1990). "Quality Assessment of Health
Facilities in Rural Papua New Guinea." Oxford University Press.
Gertler, P., and P. Glewwe (1989). "The Willingness to Pay for Education in Developing
Countries: Evidence from Rural Peru," Living Standards Measurement Study,
Working Paper, No. 54, The World Bank, Washington, D.C. Processed.
Gertler, P., L. Locay and W. Sanderson (1987). "Are User Fees Regressive? The Welfare
Implications of Health Care Financing Proposals in Peru,", Journal of Econometrics,
36: 67-80.
Gertler, P., and J. Van der Gaag (1990). The 'Willingness to pay for Medical Care: Evidence
from two Developing Countries. Baltimore: Johns Hopkins University Press.
Ghana: Population, Health and Nutrition Sector Review, 1989, Washington D.C.: World
Bank.
Hensher, D. (1986). "Sequential and Full Information Maximum Likelihood Estimation in
Nested Multinomial Logit Models." The Review of Economics and Statistics, 56: 652667.
32
Jimenez, Emmanuel (1987). Pricing Policy in the Social Sectors. Baltimore: Johns Hopkins
University Press.
King, Mervyn (1980). "An Econometric Model of Tenure Choice and Demand for Housing
as a Joint Decision," Journal of Public Economics, 14: 137-159.
Lavy, V., and J.M. Quigley (1991). "On the Economics of Medical Care: The Choice of
Quality and Intensity of Medical Care in Ghana," The World Bank, Washington, D.C.
Processed.
Lavy, V., J. Strauss, D. Thomas, and P. de Vreyer (1992). "Quality Health Care and
Survival and Health Outcomes in Ghana," The World Bank, Human Resources
Department, Poverty Analysis Division. Mimeo: August.
Mwabu, G., M. Ainsworth, and A. Nyamete (1993). "Quality of Medical Care and Choice
of Medical Treatment in Kenya: An Empirical Analysis," Africa Technical
Department, Working Paper No.9, April, The World Bank.
Manski, C., and D. McFadden (ed.) (1981). Structural Analysis of Discrete Data: with
Econometric Application. Cambridge, MA.: MIT Press.
McFadden, Daniel (1981). "Econometric Models of Probabilistic Choice". In C. Manski and
D. MacFadden, (eds.), Structural Analysis of Discrete Data with Econometric
Applications. Cambridge, MA: MIT Press.
Peabody, J. W., O. Rahman, K. Fox, and P. Gertler, "Measuring Quality in Primary Health
Facilities in Jamaica," RAND, March 1993.
Pollak, R., and M. Wachter (1975). "The Relevance of the Household Production Function
and its Implications for the Allocation of Time," Journal of Political Economy,
83: 255-299.
Reed, Renya, (1990). "Descriptive Analyses of Ghana Facility Data", August 1990, The
World Bank, Africa Technical Department, Population and Nutrition Division.
Scot, Christopher, and Ben Amenuvegbe. (1989). "Sample Designs for the Living Standards
Surveys in Ghana and Mauritania." LSMS Working Paper No. 49. The World Bank,
Washington D.C.
Small, Kenneth, A., and Harvey S. Rosen (1981). "Applied Welfare Economics with
Discrete Classic Models." Econometrica, 49 (1) 105-130.
33
Thomas, Duncan, Victor Lavy and John Strauss (1992). "Public Policy and Anthropometric
Outcomes in C6te d’Ivoire," LSMS Working Paper No. 89. Washington, D.C.: The
World Bank.
Waddington, C.J., and K.A. Enyimayew (1989). "A Price to Pay: The Impact of User
Charges in Ashanti-Akim District, Ghana," International Journal of Health Planning
and Management, 417-447.
World Bank (1989). "Ghana: Population, Health and Nutrition Sector Review," 1989.
Appendix Table
Comparative Characteristics of the Full Ghana 1988 LSMS
Sample and the Sample Used by Lavy and Germain (1993)
All Law and Germain
Sample
Sample
Age
Proportion of age <14
Rural
Sex
Head of household schooling
Own schooling
Size of household
Log per capita expenditure
Government employee
Proportion who chose self treatment
Number of days ill
Number of days inactive
Cost of consultations
Cost of medicine
Cost of transportation
Total cost of treatment
Cost of preventive care
24.1
0.43
0.52
0.48
0.42
3.26
5.89
9.83
0.07
0.62
7.56
3.15
201.7
608.5
117.8
981.1
494.8
24.1
0.43
0.51
0.48
4.49
3.28
6.06
9.76
0.07
0.58
7.59
2.93
209.3
594.1
94.6
936.2
551.8
Number of observations
5965
2291
The two samples include all the people who reported being ill in the last 30 days. The list of variables is just a
select group of variables that suggest that the two samples are drawn from the same population. In other words, it
implies that the sub-sample used in Lavy and Germain (1993) is not a selective sample in any way. The means of
other economic and demographic characteristics support the same conclusion.
Distributors of World Bank Publications
ARGENTINA
Carlos HIrsch, SRL
Galena Guemes
Florida 165,4th Floor-Ofc. 453/465
1333 Buenos Aires
AUSTRALIA, PAPUA NEW GUINEA,
F1JL SOLOMON ISLANDS,
VANUATU, AND WESTERN SAMOA
DA Information Services
648 Whitehorse Road
Mitcham 3132
Victoria
The Middle East Observer
41, Sherif Street
Cairo
FINLAND
Aka teem inen KIrjakauppa
P.O. Box 128
SF-OOIOI Helsinki 10
FRANCE
World Bank Publications
66, avenue dTina
75116 Paris
AUSTRIA
Ceroid and Co.
Graben 31
A-1011 Wien
GERMANY
UNO-Verlag
Poppolsdorfer Allee 55
D-5300 Bonn 1
BANGLADESH
Micro Industries Development
Assistance Society (MIDAS)
House 5, Road 16
Dhan mondi R/Area
Dhaka 1209
HONG KONG, MACAO
Asia 2000 Ltd.
46-43 Wyndham Street
Winning Centre
2nd Floor
Central Hong Kong
BnrncA offices:
Pine View, 1st Floor
100 Agrabad Commercial Area
Chittagong 4100
BELGIUM
Jean De Lannoy
Av. du Roi 202
1060 Brussels
CANADA
Le Diffuseur
151A Boul. de Mortagne
Boucherville, Quebec
J4BSE6
Renouf Publishing Co.
1294 Algoma Road
Ottawa, Ontario
K1B3W8
CHILE
In vertec I GT S.A.
Av. Santa Maria 6400
Edifido INTEC Of. 201
Santiago
CHINA
China Financial 4c Economic
Publishing House
8, Da Fo Si Dong Jia
Beijing
COLOMBIA
Infoenlace Ltd a.
Apartado Aereo 34270
Bogota D.E.
COTE D'IVOIRE
Centre d'Edition et de Diffusion
Africaines (CEDA)
04 B.P. 541
Abidjan 04 Plateau
CYPRUS
Center of Applied Research
Cyprus College
6, Diogenes Street, Engomi
P.O. Box 2006
Nicosia
DENMARK
SamfundsLitteratur
Rosenoems Alli 11
DK-1970 Fred enksberg C
DOMINICAN REPUBLIC
Editors Taller, C por A.
ResuuracuSn e Isabel la Catblica 309
Apartado de Correos 2190 Z-l
Santo Domingo
EGYPT, ARAB REPUBLIC OF
Al Ahram
Al Galaa Street
Cairo
HUNGARY
Foundation for Market Economy
Dombovari Ut 17-19
H-1117 Budapest
INDIA
Allied Publishers Private Ltd.
751 Mount Road
Madras-600 002
Branch offices:
15 JN. Heredia Marg
Ballard Estate
Bombay - 400 038
13/14 Asaf All Road
New Delhi-110 002
17 Chittaranjan Avenue
Calcutta - 700 072
Jayadeva Hostel Building
5th Main Road, Gandhinagar
Bangalore - 560 009
5-1129 Kachiguda
3Cross Road
Hyderabad - 500 027
Prarthana Rats, 2nd Floor
Near Thakore Baug, Navnmgpura
Ahmedabad - 380 009
Patiala House
16-A Ashok Marg
Lucknow - 226 001
Central Bazaar Road
60 Bajaj Nagar
Nagpur 440 010
INDONESIA
Pt Indira Limited
Jalan Borobudur20
P.O. Box 181
Jakarta 10320
IRAN
Kowkab Publishers
P.O. Box 19575-511
Tehran
IRELAND
Government Supplies Agency
5 Harcourt Road
4Dublin 2
ISRAEL
Yozmot Literature Ltd.
P.O. Box 56055
Tel Aviv 61560
ITALY
Licosa Commissionaria Sansoni SPA
Via Duca Di Calabria, 1/1
Casella P os tale 552
50125 Firenze
JAPAN
Eastern Book Service
Kongo 3-Chome. Bunkyo-ku 113
Tokyo
KENYA
Africa Book Service (E_A.) Ltd.
Quaran House, Mfangano Street
P.O. Box 45245
Nairobi
KOREA, REPUBLIC OF
Pan Korea Book Corporation
P.O. Box 101, Kwangwhamun
Seoul
Korean Stock Book Centre
P.O. Box 34
Yeoeido
Seoul
MALAYSIA
University of Malaya Cooperative
Bookshop, Limited
P.O. Box 1127, Jalan Pantai Baru
59700 Kuala Lumpur
MEXICO
INFOTEC
Apartado Postal 22-860
14060 Tlalpan, Mexico D.F.
NETHERLANDS
De Lindeboom/InOr-Publikatdes
P.O. Box 202
7480 AE Haaksbergen
NEW ZEALAND
EBSCONZLtd.
Private Mail Bag 99914
New Market
Auckland
NIGERIA
University Press Limited
Three Crowns Building Jericho
Private Mail Bag 5095
Ibadan
NORWAY
Narvesen Information Center
Book Department
P.O. Box 6125 Etterstad
N-0602 Oslo 6
PAKISTAN
Mirza Book Agency
65, Shahrah-e-Quaid-e-Azam
P.O. Box No. 729
Lahore 54000
PERU
Editorial Desarrollo SA
Apartado 3824
Lima 1
PHILIPPINES
International Book Center
Suite 1703, Cityland 10
Condominium Tower 1
Ayala Avenue. H.V. dela
Costa Extension
Makati, Metro Manila
POLAND
International Publishing Service
Ul. Piekna 31/37
00-677 Warza wa
For subscription orders:
IPS Journals
UI. Okrezna 3
02-916 Warszawa
PORTUGAL
Livrar.a Portugal
Rua Do Carmo 70-74
1200 Lisbon
SAUDI ARABIA. QATAR
Jarir Book Store
P.O. Box 3196
Riyadh 11471
SINGAPORE. TAIWAN,
MYANMAR.BRUNEI
Gower Asia Pacific Pte Ltd.
Golden Wheel Budding
41, Kallang Pudding, #04-03
Singapore 1334
SOUTH AFRICA. BOTSWANA
For single titles:
Oxford University Press
Southern Africa
P.O. Box 1141
Cape Town 8000
For subscription orders:
International Subscription Service
P.O. Box 41095
Craighall
Johannesburg 2024
SPAIN
Mundi-Prensa Libros, S_A.
Castcllo 37
28001 Madrid
Libreria Intcmadonal AEDC6
Consell de Cent 391
08009 Barcelona
SRI LANKA AND THE MALDIVES
Lake House Bookshop
P.O. Box 244
100, Sir Chittampalam A
Gardiner Mawatha
Colombo 2
SWEDEN
Far single titles:
Fritzes Fackboksforetaget
Regeringsgatm 12, Box 16356
S-103 27 Stockholm
For s ubs cription orders *
Wennergren-Williams AB
P. O. Box 1305
5-171 25 Soina
SWITZERLAND
For single titles:
Librairie Payot
Case postale 3212
CH 1002 Lausanne
Far subscription orders:
Librairie Payot
Service des Abonnements
Case postale 3312
CH 1002 Lausanne
THAILAND
Central Department Store
306 Silom Road
Bangkok
TRINIDAD TOBAGO, ANTIGUA
BARBUDA, BARBADOS,
DOMINICA. GRENADA, GUYANA,
JAMAICA. MONTSERRAT, ST.
KITTS 4c NEVIS, ST. LUCIA,
ST. VINCENT 4c GRENADINES
Systematics Studies Unit
•9 Watts Street
Curepe
Trinidad, West Indies
UNITED KINGDOM
Microinfo Ltd.
P.O. Box 3
Alton. Hampshire GU34 2PG
England
LSMS Working Paper (continued)
No. 68
Kozel, The Composition and Distribution of Income in Cote d'Ivoire
No. 69
Deaton, Price Elasticities from Survey Data: Extensions and Indonesian Results
No. 70
Glewwe, Efficient Allocation of Transfers to the Poor: The Problem of Unobserved Household Income
No. 71
Glewwe, Investigating the Determinants of Household Welfare in Cote d’Ivoire
No. 72
Pitt and Rosenzweig, The Selectivity of Fertility and the Determinants of Human Capital Investments: Parametric and
Semiparametric Estimates
No. 73
Jacoby, Shadow Wages and Peasant Family Labor Supply: An Econometric Application to the Peruvian Sierra
No. 74
Behrman, The Action of Human Resources and Poverty on One Another: What We Have Yet to Learn
No. 75
Glewwe and Twum-Baah, The Distribution of Welfare in Ghana, 1987-88
No. 76
Glewwe, Schooling, Skills, and the Returns to Government Investment in Education: An Exploration Using Data from
Ghana
No. 77
Newman, Jorgensen, and Pradhan, Workers' Benefits from Bolivia's Emergency Social Fund
No. 78
Vijverberg, Dual Selection Criteria with Multiple Alternatives: Migration, Work Status, and Wages
Nd. 79
Thomas, Gender Differences in Household Resource Allocations
No. 80
Grosh, The Household Survey as a Tool for Policy Change: Lessons from the Jamaican Survey of Living Conditions
No. 81
Deaton and Paxson, Patterns of Aging in Thailand and Cote d'Ivoire
No. 82
Ravaillon, Does Undernutrition Respond to Incomes and Prices? Dominance Tests for Indonesia
No. 83
Ravaillon and Datt, Growth and Redistribution Components of Changes in Poverty Measure: A Decomposition with
Applications to Brazil and India in the 1980s
No. 84
Vijverberg, Measuring Income from Family Enterprises with Household Surveys
No. 85
Deaton and Grimard, Demand Analysis and Tax Reform in Pakistan
No. 86
Glewwe and Hall, Poverty and Inequality during Unorthodox Adjustment: The Case of Peru, 1985-90
No. 87
Newman and Gertler, Family Productivity, Labor Supply, and Welfare in a Low-Income Country
No. 88
Ravaillon, Poverty Comparisons: A Guide to Concepts and Methods
No. 89
Thomas, Lavy, and Strauss, Public Policy and Anthropometric Outcomes in Cdte d'Ivoire
No. 90
Ainsworth and others, Measuring the Impact of Fatal Adult Illness in Sub-Saharan Africa: An Annotated Household
Questionnaire
No. 91
Glewwe and Jacoby, Estimating the Determinants of Cognitive Achievement in Low-Income Countries: The Case of Ghana
No. 92
Ainsworth, Economic Aspects of Child Fostering in Cote d’Ivoire
No. 93
Lavy, Investment in Human Capital: Schooling Supply Constraints in Rural Ghana
No. 94
Lavy and Quigley, Willingness to Pay for the Quality and Intensity ofMedical Care: Low-Income Households in Ghana
No. 95
Schultz and Tansel, Measurement of Returns to Adult Health: Morbidity Effects on Wage Rates in Cote d'Ivoire and Ghana
No. 96
Louat, Grosh, and van der Gaag, Welfare Implications of Female Headship in Jamaican Households
No. 97
Coulombe and Demery, Household Size in Cote d’Ivoire: Sampling Bias in the CILSS
No. 98
Glewwe and Jacoby, Delayed Primary School Enrollment and Childhood Malnutrition in Ghana: An Economic Analysis
No. 99
Baker and Grosh, Poverty Reduction through Geographic Targeting: How Well Does It Work?
No. 100
Datt and Ravaillon, Income Gains for the Poor from Public Works Employment: Evidence from Two Indian Villages
No. 101
Kostermans, Assessing the Quality of Anthropometric Data: Background and Illustrated Guidelines for Survey Managers
No. 102
van de Walle, Ravaillon, and Gautam, How Well Does the Social Safety Net Work?: The Incidence ofCash Benefits in
Hungary, 1987-89
No. 103
Benefo and Schultz, Determinants of Fertility and Child Mortality in Cdte d’Ivoire and Ghana
No. 104
Behrman and Lavy, Children's Health and Achievement in School
Headquarters
1818 H Street, N.W.
Washington, D.C. 20433, U.S.A.
European Office
66, avenue d'lena
75116 Paris, France
Telephone: (202) 477-1234
Facsimile: (202) 477-6391
Telex: wui 64145 worldbank
Telephone: (1) 40.69.30.00
Facsimile: (1) 40.69.30.66
Telex: 640651
RCA 248423 WORLDBK
L av y /G erm ain
The World Bank
Tokyo Office
Kokusai Building
1-1, Marunouchi 3-chome
Chiyoda-ku, Tokyo 100, Japan
Telephone: (3) 3214-5001
Facsimile: (3) 3214-3657
Telex: 26838
Cable Address: intbafrad
WASHINCTONDC
Q u ality an d C ost in H ealth C are C hoice in D ev elo p in g C o u n tries
ISBN 0-8213-2854-9
The W orld Bank
12854
POP 100
0-8213-2854-9
QUALITY & COST HEALTH CA
- Media
12206.pdf
Position: 1184 (7 views)