Exploring the health impact of economic growth, poverty reduction and public health expenditure
Item
- Title
-
Exploring the
health impact of
economic growth,
poverty reduction and
public health expenditure - extracted text
-
5. PetQ-o T
WH0flC0/MES0.18
Original: English
Oislribirtion: Llmrted
«Macroeconomics,
Health and
Developments Series
Number 18
!
World Health Organization
Geneva, March 1996
Exploring the
health impact of
economic growth,
poverty reduction and
public health expenditure
Other titles in the "Macroeconomic, Health and Development " Series are :
N° 1:
Macroeconomic Evolution and the Health Sector: Guinea, Country Paper - WHO/ICO/MESD.1
N° 2:
Une mdthodologle pour le calcul des coOts des soins de sante et leur recouvrement: Document
technique, Guin4e - WHO/ICO/MESD.2
N° 3:
Debt for Health Swaps: A source of additional finance for the health system: Technical Paper WHO/1CO/MESD.3
N°4:
Macroeconomic Adjustment and Health: A survey: Technical Paper - WHO/ICO/MESD.4
N° 5:
La place de I'aide extdrieure dans le secteur medical au Tchad: Etude de pays, Tchad
WHO/ICO/MESD.5
N° 6:
L’influence de la participation financidre des populations sur la demande de soins de sante: Une
aide d la reflexion pour les pays les plus demunis: Principes directeurs - WHO/ICO/MESD.6
N° 7:
Planning and Implementing Health Insurance in Developing countries: Guidelines and Case Studies:
Guiding Principles - WHO/ICO/MESD.7
N° 8:
macroeconomic Changes in the Health Sector in Guinea-Bissau: Country Paper - WHO/ICO/MESD.8
N° 9:
Macroeconomic Development and the Health Sector in Malawi: Country Paper - WHO/ICO/MESD.9
N° 10:
El ajuste macroecondmico y sus repercusiones en el sector de la salud de Bolivia: Documento de
pais - WHO/ICO/MESD.10
N° 11:
The macroeconomy and Health Sector Financing in Nepal: A medium-term perspective: Nepal,
Country Paper - WHO/ICO/MESD.11
N° 12:
Towards a Framework for Health Insurance Development in Hai Phong, Viet Nam: Technical Paper
WHO/ICO/MESD.12
N° 13:
Guide pour la conduite d'un processus de Table ronde sectorielle sur la sante: Principes directeurs
-WHO/ICO/MESD.13
N° 14:
The public health sector in Mozambique: A post-war strategy for rehabilitation and sustained
development: Country paper - WHO/1CO/MESD.14
N° 15:
La sante dans les pays de la zone franc face d la devaluation du franc CFA - WHO/ICO/MESD.15
(document no longer available)
N° 16:
Poverty and health in developing countries: Technical Paper - WHO/ICO/MESD.16
N° 17:
Gasto nacional y financiamiento del sector salud en Bolivia: Documento de pais - WHO/ICO/MESD.17
Exploring the
health impact of
economic growth,
poverty reduction and
pubflic health expenditure
by
Guy Carrin and Claudio Politi
Division of Intensified Cooperation
with Countries
World Health Organization
Geneva
J
J
J
J
ACKNOWLEDGMENT
The authors are indebted to Michel Jancloes and Jean Perrot of
WHO, and an anonymous referee for useful comments and suggestions. The
views expressed in this article are, however, solely the responsibility of the
authors.
This document is not issued to the general public, and all rights are reserved by the World
Health Organization (WHO). The document may not be reviewed, abstracted, quoted, reproduced
or translated, in part or in whole, without the prior written permission of WHO. No part of this
document may be stored in a retrieval system or transmitted in any form or by any means electronic, mechanical or other- without the prior written permission of WHO.
The views expressed in documents by named authors are solely th responsibility of these
authors
Printed in 1996 by WHO
Printed in Switzerland
HG - toy
U8123
EXPLORING THE HEALTH IMPACT OF ECONOMIC GROWTH,
POVERTY REDUCTION AND PUBLIC HEALTH EXPENDITURE
TABLE OF CONTENTS
Page
INTRODUCTION
I.
........................................................................................
POVERTY AND POOR HEALTH IN THE DEVELOPING
WORLD: A CONTINUING ISSUE.........................................................
II.
IV.
V.
2
ECONOMIC GROWTH: HOW GOOD AN EXPLANATORY
FACTOR OF HEALTH IMPROVEMENT?............................................
III.
2
3
THE ROLE OF POVERTY REDUCTION AND PUBLIC
EXPENDITURE IN HEALTH IMPROVEMENT.................................
7
A.
The Anand and Ravallion model: the basic hypotheses revisited . .
7
B.
The regression estimate........................................................................
9
A REESTIMATION......................................................................................
10
A.
A preliminary look at the expanded data set....................................
10
B.
Regression estimates: loglinear specification....................................
13
C.
Regression estimates: alternative specification..................................
15
CONCLUDING REMARKS........................................................................
16
.........................................................................................
18
FOOTNOTES.................................................................................................
20
BIBLIOGRAPHY
INTRODUCTION
It is true that, on the whole, developing countries have improved their health status
over the past decades. However, this does not imply that the currently achieved health status
in developing countries, especially that of the poorest, can be denoted as satisfactory. Often
the immediate response to this caveat is that developing countries should continue to grow and
that, eventually, economic growth will trigger sufficient human development. The purpose
of this paper is now to see to what extent variables other than sheer economic development
play a role in health status improvement. In particular, we study the impact of poverty
reduction and public health expenditure in developing economies on health. The latter
variables are often quoted as necessary for the improvement of developing countries' health
status.
In the first section, we give a short overview of some basic indicators of poverty and
health in developing countries, and thus illustrate that poor health in the developing world is
a pressing problem. In section II, we address the question to what extent income growth is
associated with health improvement. Then, the results obtained by Anand and Ravaillon
(1993) regarding the roles of poverty reduction and public expenditure for health improvement
are discussed in section III. New findings from econometric analysis with a new and larger
data set are presented in section IV. A summary of conclusions and indications for further
applied research are presented in the final section.
I. POVERTY AND POOR HEALTH IN THE DEVELOPING WORLD: A CONTINUING
ISSUE
The basic cause of poor health in developing countries is poverty. An immediate and
simple way to indicate the presence of poverty in developing countries (DCs) is to measure the
extent to which income is inadequate for these countries' populations. Informative results are
obtained when Chen et al. (1993) estimate changes in absolute poverty1 incidence between
1985 and 1990 in 40 selected countries2. A poverty line corresponding to average daily
spending of $1 at 1985 purchasing power parity3, is used to obtain these estimates. The
absolute poverty incidence, as an aggregate for the selected sample of countries, declines only
marginally between 1985 and 1990: from 33.3% to 33.1% of the population. The latter
implies that the number of poor people has grown more or less according to overall population
growth, viz. about 2% per year. For all developing countries, it is estimated that out of a total
population of 4,240 million, 1,314 million live in poverty, but especially in rural areas
(UNDP, 1994).
The basic needs approach introduced in the early seventies (ILO, 1976) broadened the
approach to poverty, and advanced that social indicators such as health status and literacy are
as important in revealing poverty as income itself. In fact, an indicator such as the Physical
Quality of Life Index was proposed as a weighted average of income, life expectancy and
literacy (Morris, 1979). The more recent Human Development Indicator proposed by the
United Nations Development Programme (UNDP, 1994) has a similar underpinning, in that
2
it is based on four indicators: life expectancy, literacy, years of schooling and income. The
multidimensional character of poverty is also stressed by Gunatilleke (1995). He writes that4
poverty "is a condition which encompasses deprivation in a variety of forms: inadequate
income, lack of education, knowledge and skill, poor health status, lack of access to safe water
and sanitation, insufficient food and nutrition, lack of control over the reproductive process".
For the purpose of this paper, we concentrate on the evolution of health status measures
in the developing countries, however. When looking at global indicators of health such as life
expectancy at birth or infant mortality, the developing countries as a group made a definite
progress. Whereas average life expectancy was 46.2 in 1960, it rose to 63 in 19925. And
infant mortality decreased from 149 per 1000 in 1960 to 69 per 1000 in 1990. Despite these
relative improvements, it should still be noted that currently 12.2 million children under age
5 die, most of them from causes that could be prevented6. Also the existing gap with
industrialized countries is notable: life expectancy was 69 in 1960 and increased to 74.5 in
1992. The infant mortality rate in industrialized countries was 35 per 1000 in 1960 and 13 per
1000 in 19927.
As we focus on the least developed countries (LDCs)8, it is clear that a special effort
is needed to enhance the health status of their populations and to reduce the gap with respect
to the industrialized world and even other developing countries. In the early 1990s, the
average life expectancy at birth of these countries was only 50.1 years and the average
mortality rate for children under the age of five was 160 per 1000 live births. Average
maternal mortality was 730 per 100 000 live births in 1988. These figures confirm a blatant
inequality with the rest of the world. In fact, average life expectancy in the LDCs is about
67 % of that in industrialized countries. An excessively large gap is observed with regard
to children’s and women's health: the average mortality rates of children under five and of
mothers are at least 10 and 30 times as high, respectively, as the corresponding rates in
industrialized countries. A comparison with other developing countries gives the following
result: life expectancy in LDCs is about 78% of that the other developing countries; average
mortality rates of children under five and of mothers are 2.4 and 2.7 times as high.
II. ECONOMIC GROWTH: HOW GOOD AN EXPLANATORY FACTOR OF HEALTH
IMPROVEMENT ?
The basic question is to what extent overall economic development can reduce poverty
and thus mediate the progress in health. The mechanism is that higher average incomes
increase households' opportunities to achieve better education, health care, nutrition and
improved living standards in general. In turn, this rise in opportunities contributes to a better
health status. The present reasoning can also be referred to as the trickle-down approach to
social development. The potential for economic growth to facilitate health development was
highlighted earlier by Attinger and Ahuja (1980). Ram (1985) is another proponent of the idea
that economic development will positively affect the fulfillment of basic needs. In their study
of human development in poor countries, Anand and Ravallion (1993) also mention economic
3
growth as a first possible determinant of human development as measured by health status and
the literacy rate.
In Figures 1 to 3, we depict the relationship between life expectancy at birth, infant
mortality and underfive mortality vs. Gross Domestic Product (GDP) per capita in $9, for
1990. The data, represented by squares, are for 84 developing countries. These data pertain
to 32 least developed countries10 and 52 other developing countries. The size of the sample
was constrained by data availability on the set of health status indicators and health status
determinants in 1990 that is used in the econometric analysis in section IV11.
A quick glance at Figures 1 to 3 suggests easily that, on average, economic
development enhances health status: the higher the level of GDP per capita, the higher life
expectancy and the lower infant mortality and underfive mortality. The latter is confirmed by
three simple regression analyses. The dependent variables are the natural logarithms of life
expectancy (LE), the infant mortality rate (IMR) and the underfive mortality rate (U5MR),
respectively. The explanatory variable in the three equations is the natural logarithm of GDP
per capita (GDPC). The results of the regression equations, estimated by ordinary least
squares, are given in Table 1.
Figure 1 Relationship between Life Expectancy and GDP per capita
A Oosorvod (LDCs)
D Oosorvod (DCs)
• Predicted
1000
2000
3000
4000
5000
6000
GDP (PPP$) per capita, 1990
4
7000
8000
9000
10000
Figure 2 Relationship between Infant Mortality Rate and GDP per capita
180 -r
160 -A Observed (LDCs)
140
ln,anl mortol,tyra1e '
” 90
□ Observed (DCs)
• Predicted
120
100
80
60
40
20
—I
0 4-------------------I-------------------r-
0
1000
2000
3000
4000
5000
7000
6000
8000
9000
10000
GDP (PPP$) per capita, 1990
Figure 3 Relationship between Under 5 Mortality Rate and GDP per capita
350
Under 5 mortality rate. 1990
300
& Observed (LDCs)
250
D Observed (DCs)
• Predicted
200
150
100
□
50
•a
□
0
1000
2000
3000
4000
5000
6000
GDP (PPP$) per capita, 1990
5
--------- 1-------------------1—
—t—
□
—I
8000
9000
10000
7000
Table 1
Regression results: Health status explained by Gross Domestic Product per capita
Dependent
variable
In LIFE
In IMR
In U5MR
Explanatory variables
R2
Adjusted R2
intercept
In GDPC
2.8579
(0.0917)
0.1609
(0.0122)
0.6781
0.6742
8.9092
(0.3790)
-0.6321
(0.0506)
0.6554
0.6512
10.2329
(0.4226)
-0.7570
(0.0564)
0.6870
0.6832
Note: Figures in brackets are standard errors
In the life expectancy regression, the GNPC elasticity is quite important, and is
statistically significantly different from zero at the 1% confidence level. This regression
implies that an extra 5 % growth in GNP per capita translates into an increase of life
expectancy of 0.8%'2. Several important outliers vis-a-vis the 'predicted' indicators can be
recognized in Figure 1. The predicted indicators are computed using the results of the
regression equation. The predicted or estimated life expectancy is indicated by the points in
bold in Figure 1. Countries such as China and Myanmar do much better than what would be
predicted for countries in their income group. On the other hand countries like Gabon and
Sierra Leone perform less well than is expected.
From the infant mortality regression, we conclude that that GDPC is significantly
linked to the level of infant mortality. The GDPC elasticity of -0.6321 implies that an increase
in GDPC of 5% is associated with a decline in infant mortality of 3.2%. Finally, GDPC has
a significant effect upon the underfive mortality rate. The GDPC elasticity is -0.7570 which
would imply that an increase in GDPC of 5 % is associated with a reduction in the underfive
mortality rate of 3.8 %. In Figures 2 and 3, several important outliers can be ascertained.
Swaziland and Gabon are performing less well than expected, whereas Zaire and Jamaica have
better indicators than expected.
Although the R2's obtained for these equations are by no means small, they should not
suggest that GDP is only what matters. The fact that there is more to health improvement than
the current level or growth of economic resources, is highlighted by Sen (1993). He has
demonstrated that the countries with a negative growth rate over the period 1960-1985 still
observed an improvement in health indicators. For instance, the country that experienced the
6
largest negative growth rate, namely Kuwait, with an average annual growth of GDP per
capita of-5.82%, saw its under-five mortality decline annually by 6.32%. It is surely granted
that such countries may have built up an important "stock" of health in the earlier part of this
twenty-five period, due to significant improvements in health-related determinants such as
education, and water and sanitation. Severe economic downturns then do not necessarily cause
an immediate reduction in health status achieved. Indeed, achievements in education and other
health-related inputs are apt to act as a buffer against the adverse effects of economic slumps.
In any case, the latter reasoning supports the argument that the interaction between health and
income growth is not instantaneous and quite complex as well. The absence of a unique
relationship between health improvement and economic growth is also demonstrated amply in
the World Bank Development Report of 1993 on "Investing in Health "I3.
III. THE ROLE OF POVERTY REDUCTION AND PUBLIC EXPENDITURE IN HEALTH
IMPROVEMENT
A. The Anand and Ravallion model: the basic hypotheses revisited
It is obvious that quite a number of societal factors impact upon the population health
levels. For instance, the population dynamics, the economy, the educational system, social
infrastructure and the environment have main effects on health. One possible methodology
to analyze these interactions is the construction of a multi-sectoral model. The scope of the
present paper is far more modest, however. We will study and reinterpret a simple model
proposed by Anand and Ravallion (1993). We basically agree with these authors that, in
addition to economic growth, the allocation of resources is also vital for the determination
of health status.
Two types of allocation mechanisms are relevant. First, the intersectoral allocation,
or the allocation of economic resources between the private and public sector, and more
specifically, the degree of provision of social services. Secondly, the interhousehold allocation
of economic resources, or the income distribution, also matters.
Intersectoral allocation
Regarding the intersectoral allocation, it is evident that a country can make use of
economic growth in different ways: enhancing public expenditure, e.g. health or military
expenditure, or increasing the private consumption, e.g. housing or cars, or investing for
ensuring further economic growth. The direct and indirect effects of economic improvement
on the health status of the population vary according to the ways a higher level of resources
is allocated. This is the reason why countries with a similar level of economic development
are associated with different health outcomes, as shown in Figures 1 to 3.
We argue that it is the distribution of GDP or GNP14 in favour of public resources for
basic health services, including family planning and nutrition, as well as elementary education,
that matters significantly for health improvement of all socio-economic population categories.
7
Governments can make a good case for public expenditure for health and education. First,
governments can invoke that these development activities entail considerable positive
externalities. A consumption externality exists as soon as the consumption of a commodity
or service by a consumer has an effect on other consumers' welfare. The externality is
positive (negative) when the consumption by that consumer increases (reduces) the welfare of
other consumers. In the health sector, vaccination activities can be said to result in positive
externalities, when the vaccination of one patient or group of patients benefits other patients
as well. To use this example further, letting the provision of vaccines depend strictly on
private decision-making, would be sub-optimal: private citizens would not take account, or
only insufficiently, of the importance of their decisions on other citizens. The government is
an agency that could rectify this sub-optimality by duly taking account of the externalities and,
hence, by assuming responsibility for the provision of the vaccines. In other words,
government can correct this so-called market failure by seeing to it that these development
activities are provided in satisfactory amounts. It is safe to say that, generally, social and
economic development requires a proper input from government.
Note that the
complementary between publicly financed goods and services and private sector activities has
already been emphasized by Barten (1970). A second reason for government expenditure in
health and education is simply the concern for equity, whereby governments want to stimulate
an equitable access to health and education. For instance, governments may want to allocate
grants across regions in such a way that they advance especially the poorer regions.
Governments may also wish to finance as well as provide health and education services
themselves, so as to ensure that they reach the poor as well.
Income distribution
Interhousehold distribution of resources matters in that the spreading of wealth among
the socio-economic groups plays a relevant role in translating economic growth in health
benefits. A fairer distribution of private income leads to a better distribution of consumption
of health services, food, shelter, sanitation and education across households. In turn, the latter
results in a more equal distribution of health status. Still, we think the most urgent concern
is not so much about the income distribution on the whole, but about the extent of poverty15.
Indeed, especially poverty and ill health are very much intertwined. Poverty implies lack of
food, safe water and sanitation, and basic health services. This lack is the overriding cause
of insufficient health. In turn, ill health reduces people's ability to work or to secure income
increases, thus sustaining poverty. In view of the synergistic relationship between poverty and
ill health, poverty reduction is seen as a major instrument of health improvement. As the
recent 1995 World Health Report states
The link between health and poverty should be
obvious, as poverty has continued and will continue to be a major obstacle to health
development"16.
8
B. The regression estimate
The considerations above, based on Anand and Ravallion (1993), can best be reflected
in the following simple equation:
HS = HS (GNPC, PESS, POV)
where
HS
=
GNPC =
PESS =
POV
=
(1)
health status indicator
gross national product per capita (PPP)
public expenditure on social services
poverty indicator
Anand and Ravallion subjected equation (1) to a regression analysis. The data used
pertain to the years 1985-1990 and cover 22 selected developing countries. The health status
indicator used is "the difference between desirable life expectancy (80 years) and actual life
expectancy (LE)". To represent government's effort for social services, they restricted
themselves to the variable "public health expenditure per capita" (PHEC). Poverty is
measured via the "proportion of the population consuming less than $117" (POP< 1).
A loglinear specification is used for the regression. Also note that the natural
logarithm of the health status variable is multiplied by -1. In this way, one measures the
impact of the explanatory variables on a reduction in the shortfall of observed from desirable
life expectancy. The coefficients associated with GNPC and PHEC are expected to be
positive, whereas the coefficient associated with POP< 1 is expected to be negative.
The results obtained are presented in Table 2. The GNPC elasticity has the wrong sign
and is not statistically significant. The coefficients associated with the determinants POP< 1
and PHEC have the expected sign, and are statistically significant. Anand and Ravallion do
remark that the correlation between GNP per capita and public health spending per capita is
0.91. Yet, they do not discuss the possibility that multicollinearity might have caused the
coefficient of GNPC to be negative and not significantly different from zero. For Anand and
Ravallion (1993, p. 142), the result obtained does not indicate that GNPC is unimportant,
however. They notice that the result "...says that the importance of growth lies in the way that
its benefits are distributed between people, and the extent to which growth supports public
health services".
Several doubts have also to be expressed regarding the validity of the PHEC variable.
First, public health expenditure contains both current and capital expenditure. All of this
expenditure is not necessarily geared to basic health services. For instance, investment in the
construction of a university hospital, and the subsequent health services provided, is likely to
be less important for basic health than basic primary health care services and prevention. In
fact, in many developing countries half of public health expenditures are allocated to hospital
services that are not accessible to the larger part of the population18. Hence, by construction,
PHEC is not a very valid indicator, as it does not completely reflect what we intend to
measure. In addition, public health expenditure incorporates government expenditure at the
central level only. Hence, PHEC may be biased downwards for countries with important
9
>
provincial or municipal health expenditure. Secondly, it is very difficult for the current PHEC
variable to take account of the efficiency of the public health expenditure. Some government
budgets may be sufficiently important in terms of GDP, but their effectiveness is not
adequately revealed. Public health expenditure is effective as long as there is a proper mix of
inputs. A public health budget that allocates most of its budget to salaries at the detriment of,
say, vaccines or pharmaceuticals, is likely to entail inefficiencies. Thirdly, public health
expenditure needs to be properly complemented by other expenditure essential for health
improvement. For instance, sanitation infrastructure and water development positively
influence health. Also private health expenditure may be needed as a complement to public
health expenditure in many countries. Currently the importance of these complementary
factors is captured, but in an indirect way only, via the GDPC variable.
Table 2
Regression result: life expectancy regression from Anand and Ravallion (1993)
Dependent variable
- In LE (80-LE)
Explanatory variables
R2
intercept
In GNPC
In
POP<1
In PEHC
-1.08
(0.46)
-0.28
(0.21)
-0.21
(0.04)
0.30
(0.10)
0.71
Note: Standard errors are in brackets
IV. A REESTIMATION
A. A preliminary look at the expanded data set
We reestimate equation (1), but utilizing the larger data set for 1990 referred to
earlier in section II. We also modify two of the three explanatory variables. First, in our
analysis, public effort for public health is measured via the share of public health
expenditure in GDP (PHE%). We are not disregarding here our own reservations,
expressed earlier, about the validity of public health expenditure as an indicator. Still, we
are interested to see whether similar econometric results are obtained when government's
commitment to health is measured differently. Secondly, poverty is measured either by the
incidence of total absolute poverty (TPOV) or by absolute rural poverty (RPOV)19. In
addition to life expectancy, we retain two other health status variables for further study,
namely infant mortality and underfive mortality.
As an initial exercise, we rank the countries according to the sign of the residuals
estimated in the life expectancy equation presented in Table I20. A negative residual means
that the country does less well than expected, whereas a positive residual is associated with
a country that performs better than expected. Next, we select the first 10 countries with
10
the highest positive deviation from estimated life expectancy (we call these the "high"
performers). We also select the first 10 countries with the highest negative deviation from
estimated life expectancy (we call these the "low" performers). We refer to Table 3 for the
comparison between the high and low performers.
From section II, we already know that GNPC is highly correlated with health
status. Therefore, we focus on possible differences in public health expenditure and the
poverty indicator between the high and low performers. Public expenditure as a share of
GNP varies widely both in the group of high and low performers, respectively. In
particular, the first group has a range from 0.8% of GNP (Myanmar) to 5.6% (Costa Rica)
with an average of 3% of GNP spent on health. The countries with observations below the
predicted life expectancy in Figure 1 have a similar varying pattern, from 1.3% (Guinea
Bissau) to 5.5% (Mauritania), with an average of 2.5% of GNP. It is striking that there is
no clearcut difference in the importance of public health expenditure, as a % of GDP,
between the two groups.
Further analysis investigating the role of poverty indicates that the level of rural
poverty is substantially higher in the low performers as compared with the high
performers. The 10 selected countries below the expected line in Figure 1 have an average
of 63% of the rural population below the absolute poverty line. However, the 10 selected
countries who performed better, that is with higher life expectancy than predicted, have an
average of 49% of absolute rural poverty. This is already some indication that the level of
poverty matters in the explanation of differences in health status across countries.
Unfortunately, it proves to be difficult to compare total absolute poverty between the high
and low performers, due essentially to lack of data.
A similar analysis based on the residuals estimated in the infant mortality and
underfive mortality equations was performed. In general, similar conclusions as above are
obtained. These analyses are therefore not reported separately.
Table 3
Comparison between high and low performers
Countries
Deviation
from
expected life
expectancy
(%)
GDP per
capita
Absolute Poverty
(%)
Total
Rural
Public Health
Expenditure
(% of GDP)
HIGH PERFORMERS
Myanmar
23.8
659
35
40
0.8
China
18.5
1,990
9
13
2.1
Zaire
17.6
367
70
90
0.8
Sri Lanka
16.2
2,405
39
46
1.8
Jamaica
15.8
2,979
na
80
2.9
Honduras
15.2
1,470
37
55
2.9
Nicaragua
14.7
1,497
20
19
6.7
Kenya
11.7
1,058
52
55
2.7
Tanzania
11.6
572
58
60
3.2
Costa Rica
10.9
4,542
29
34
5.6
average
15.6
1,754
39
49
3.0
Congo
-11.7
2,362
na
80
3.0
Benin
-11.9
1,043
na
65
2.8
Senegal
-12.0
1,248
na
70
2.3
Mauritania
-12.1
1,057
na
80
5.5
Oman
-14.1
9,972
na
6
2.1
Afghanistan
-15.3
714
53
60
1.6
Gambia
-15.7
913
na
85
1.6
Guinea-Bissau
-17.5
841
na
75
1.3
Gabon
-21.2
4,147
na
41
3.2
Sierra Leone
-21.7
1,086
na
65
1.7
average
-15.3
2,338
53
63
2.5
LOW PERFORMERS
12
B. Regression estimates: loglinear specification
The econometric analysis focuses on the roles played by GDP per capita (GDPC),
public health expenditure as a % of GDP (PHE%) and total absolute poverty (TPOV) in the
determination of the three health status variables. The data sample consists of 57 countries.
We first use a loglinear specification to estimate the various effects. This has the known
advantage that the coefficients immediately measure the elasticities. The latter elasticities are
constant, however. The results are reported in Table 4.
Impact of GDP per capita
A first conclusion is that the coefficients indicating the GDPC elasticities are all
statistically significantly different from zero at the 2.5% significance level. Also notice that
the absolute values of all coefficients are less, though not in a very important way, than those
obtained from the simple regressions of health status on GDPC (see Table 1). In any case,
these results suggest that the overall level of economic resources remains important for the
determination of health status. Indeed, health is not only dependent upon medical care per se.
The state of health is heavily intertwined with the overall process of resource generation, as
the latter creates more opportunities to achieve better health.
Using the econometric results, we can compute that an increase of 10 % in the level
of GDP per capita would result in an increase of life expectancy by about 1.4 %
(= 0.1360 x 10%). And a 10 % increase in GDP per capita decreases infant mortality and
underfive mortality by about 4.8% (= 0.4848 x 10%) and 5.8% (0.5812 x 10%),
respectively21.
Impact of poverty
The coefficient of the poverty variable (TPOV) has the expected sign: a negative sign
in the life expectancy equation (higher levels of poverty are associated with lower life
expectancy) and a positive one in the mortality equations (higher poverty levels are correlated
positively with the mortality rates). The coefficients in the life expectancy and mortality
equations are statistically significantly different from zero at the 5% and the 2.5% significance
level, respectively.
The level of the estimated poverty elasticities are clearly more important in the
mortality equations than in the life expectancy equation. This suggests that poverty hits the
youngest population groups harder than other population categories. To appreciate further the
impact of poverty, let us calculate the effect of a halving of the average poverty level (from
43.49 to 21.75) on the health status variables. Let us calculate this impact for countries whose
variables, initially, take on average sample values (59.76 years, 77.40/1000 and 119.32/1000
for life expectancy, infant mortality and underfive mortality, respectively). Using the
coefficient estimates, we then calculate that a 50% drop in poverty means an extra 1.7 years
to life expectancy, and leads to a reduction of 14.94 per thousand and 27.47 per thousand in
infant mortality and underfive mortality, respectively. Notice, however, that the magnitude
13
of these impacts decreases with higher levels of health status. The latter is a specific feature
of the logarithmic specification.
Impact of public health expenditure
The PHE% variable does not have the impact that could be anticipated on the basis of
the regression presented in section III. The coefficients do not have the correct sign and, in
addition, are not statistically different from zero either.
We cannot really blame
multicollinearity22 between GDPC and PHE% for this result, as the correlation coefficient
between those variables is a mere 0.038. It seems that our doubts regarding the PHE%
variable are supported by this result. We reiterate that government's effort for basic health
services is reflected inadequately by public health expenditure.
Two examples are given to illustrate that public health expenditure as a share of GDP
is insufficiently linked to health status. For instance, Madagascar and Myanmar have a similar
GDP per capita. Myanmar allocates less of its GDP to health than Madagascar: 0.8% vs.
1.3%. Yet, Myanmar's life expectancy exceeds that of Madagascar: 61.3 years vs. 54.5
years. Another example is that of Somalia and Angola that both have a similar GDP per
capita. Angola spends double the amount on public health as compared to Somalia, yet they
arrive at a similar life expectancy. The latter illustrates that other factors are at work that are
not adequately captured via our simple regression model.
Do the impacts change when absolute rural poverty is used as the poverty indicator ?
All equations were reestimated, using absolute rural poverty (RPOV) instead of total
absolute poverty (TPOV). The data sample is larger in this case, viz. 84 countries. The
results obtained are similar to those reported in Table 4, in two respects23. First, the values
of the GDP elasticities are quite alike. Secondly, the PHE% coefficients are not statistically
different from zero. We found that the poverty elasticities remain statistically different from
zero; yet, they assume values that are approximately half of the values of the elasticities
reported in Table 4. The latter is not unexpected given that rural poverty is an imperfect
indicator of a country's poverty. Still, it is interesting to note that the poverty effects are
confirmed, subsequent to using another data set and a different poverty indicator.
14
Table 4
Regression results: Loglinear specification
Dependent
variable
In LIFE
In 1MR
In U5MR
R2
Explanatory variables
Adjusted R2
intercept
In GDPC
In TPOV
In PHE%
3.2215
(0.1777)
0.1360
(0.0162)
-0.0410
(0.0208)
-0.0077
(0.0190)
0.7414
0.7268
6.8150
(0.7888)
-0.4848
(0.0721)
0.2548
(0.0921)
0.0947
(0.0844)
0.6966
0.6794
7.7535
(0.8629)
-0.5812
(0.0789)
0.2990
(0.1008)
0.0935
(0.0923)
0.7317
0.7165
Note: The figures in brackets are standard errors.
C. Regression estimates: alternative specification
A further question is whether the use of another specification would produce similar
conclusions. We therefore tested one alternative specification, which also takes account of the
natural bounds of the dependent variables. In the case of infant mortality and underfive
mortality, a logistic curve is estimated. This specification ensures that estimated values from
the regression will be positioned between 0 and 1. The estimation of a logistic curve for life
expectancy proves to be somewhat cumbersome. Therefore, a logarithmic reciprocal-type
model is estimated; the latter ensures that the estimated life expectancy is between 0 and a
maximum value.
The results are presented in Table 5 and 6. In general, the conclusions from section
4.2 apply. PIIE% is not statistically significantly different from zero. Both the coefficients
of GDPC and TPOV remain significantly different from zero. Note, however, that the
explanatory power of the current life expectancy regression is weaker than in the case of the
logarithmic specification.
In contrast to the previous estimates, the elasticities implied in the current regression
equations are no longer constant, however. For instance, at average levels for the dependent
variables and poverty indicator, the poverty elasticities of infant mortality and underfive
mortality are 0.2568 and 0.3317, respectively. And in the case of the life expectancy
equation, the poverty elasticity of life expectancy is -0.0225. However, at maximum levels
for the dependent variables and the poverty indicator, the implied poverty elasticities of infant
mortality and underfive mortality are 0.4500 and 0.5175, respectively. And the poverty
elasticity of life expectancy becomes -0.0115.
These alternative specifications were also used to undertake reestimations with absolute
rural poverty as the poverty indicator24. The GDPC coefficients are quite similar to those
reported in Tables 5 and 6. Again, the PHE% variable does not exert a statistically
significiant effect on health status. The poverty coefficients are all statistically significantly
15
different from zero at the 2.5% significance level. It is to be noted that the poverty
coefficients in the mortality equations are about half as large as those reported in Table 6,
however. Still, the results concerning the impact of poverty are generally coherent with those
that were obtained when TPOV was used as the poverty indicator.
Table 5
Regression result: logarithmic reciprocal specification for the life expectancy regression
Explanatory variables
Dependent variable
In LE
intercept
1/GDPC
1/TPOV
1/PHE%
4.1649
(0.0346)
- 166.363
(20.4516)
0.9793
(0.3957)
0.0225
(0.0387)
R2
Adjusted
R-
0.6616
0.6425
Note: The figures in brackets are standard errors.
Table 6
Regression results: logistic specification for the mortality regressions
Explanatory variables
Dependent variable
In IMR/l-IMR
In U5MR/1-U5MR
R2
Adjusted
R2
intercept
GDPC
TPOV
PHE%
-2.4206
(0.2182)
-0.00025
(0.00003)
0.0064
(0.0030)
0.0347
(0.0371)
0.6858
0.6680
-1.9752
(0.2525)
-0.00030
(0.00004)
0.0087
(0.0035)
0.0364
(0.0429)
0.7133
0.6972
Note: The figures in brackets are standard errors.
V. CONCLUDING REMARKS
Among the most important findings from this analysis is the support for the hypothesis
that poverty reduction is important in the determination of health status indicators. The
analysis shows that lower rates of total or rural absolute poverty are associated with better
health status. Whereas this may seem intuitively obvious, econometric analyses that highlight
the role of poverty are relatively scarce. Furthermore, in our analysis, GDP per capita keeps
its role in determining health status, unlike the finding by Anand and Ravallion (1993). It can
be understood that the level of income and its growth provide an important support for health
development in developing countries. In fact, the growth in the volume of economic resources
is apt to facilitate the financing of inputs, like water and sanitation infrastructure, that are often
as important for health as medical services.
16
We do not obtain, however, the result reported by Anand and Ravallion (1993), namely
that public health expenditure has a statistical significant effect on health status. In view of
the current analysis, one can question whether their result can be extrapolated to a larger set
of countries. One of the main reasons we submit for our finding is that the government effort
to finance basic health services is not captured well by the PHE% variable. Moreover, the
PHE% does not take account of the extent to which public health service systems are efficient.
Indeed, a significant level of public health expenditure may hide large inefficiencies, whereas
a modest level of public health expenditure may result in a quite performing system. It is
precisely this difference in efficiency that is of capital importance for health status. In
addition, PHE% does not capture efforts to finance the complementary determinants of health
such as sanitation and water infrastructure and basic education.
Which recommendations could be made for future empirical research in this area ?
First, it is clear that indicators should be used that better reflect the true public effort for basic
health services. Secondly, explanatory variables reflecting expenditure for infrastructure and
basic education may have to be considered as well. In addition, the level of private health
expenditure can be introduced among the set of explanatory variables. Thirdly, the impact of
poverty can be reassessed via the use of the other poverty indicators or inequality indicators
such as the GINI coefficient. Of course, the application of these recommendations risks to be
hampered by the lack of international data. Fourthly, an important question is to what extent
the level and distribution of economic resources is not, in turn, influenced positively by the
state of health. A single regression approach then no longer suffices. In other words, we may
well have to consider the study of poverty and health via a simultaneous modeling framework.
Finally, although the health policy maker may appreciate that part of his intuition is
confirmed, he may want to express greater expectations towards the economist. It is evident
that most of the health policy decisions are being taken at country level. Hence, a fruitful line
of research would be the analysis of the relationship between health status, on the one hand,
and socio-economic variables, on the other hand, using data on a country's regions or districts.
It is hoped that the use of variables on economic activity, on poverty and income distribution,
and on the use of public and private health expenditure at a more micro-level would result in
firmer and more useful conclusions for policy making.
17
BIBLIOGRAPHY
Anand S. & Ravallion M., 1993, Human Development in Poor Countries: On the Role of
Private Incomes and Public Services, Journal of Economic Perspectives, vol.7, no.l. 133-150.
Attinger E.O. & Ahuja D.R., 1980, Health and Socioeconomic Change, IEEE Transactions
on Systems, Man and Cybernetics, vol SMC-10, no. 12, 781-796.
Barten A.P., (1970), Some Reflections on the Relation between Private Consumption and
Collective Expenditure in an Expanding Mature Economy, Programming for Europe's
Collective Needs, ASEPELT, 4, 82-98.
Barten A.P. e.a. (1989), Econometrische Lessen, (Academic Service, Schoonhoven).
Chen S., Datt G. and Ravallion M. (1993), Is Poverty Increasing in the Developing World,
World Bank working paper, WPS 1146 (World Bank, Policy Research Development).
Gunatilleke G. (1995), Poverty and its
WHO/ICO/19/NAM/95.4 (WHO, Geneva).
Effects
on
Health,
WHO
document,
ILO (1976), Employment, Growth and Basic Needs: A One-World Problem, (International
Labour Organization, Geneva).
Morris M.D. (1979), Measuring the Condition of the World's Poor: The Physical Quality of
Life Index, (Frank Cass, London).
Ram R. (1985), The Role of Real Income Level and Income Distribution in Fulfillment of
Basic Needs, World Development, vol. 13, no.5, 589-594.
Sen A. (1993), Economic Regress - Concepts and Features, Proceedings of the World Bank
Annual Conference on Development Economics 1993. Supplement to the World Bank
Economic Review and the World Bank Research Observer, 1994.
UNDP (1992), Human Development Report 1992, (Oxford University Press, Oxford).
UNDP (1993), Human Development Report 1993, (Oxford University Press, Oxford).
UNDP (1994), Human Development Report 1994, (Oxford University Press, Oxford).
Waldman R.J. (1992), Income Distribution and Infant Mortality, Quarterly Journal of
Economics, vol. CVII, no.4, 1283-1302.
WHO (1994), Progress towards health for all. Statistics of member states 1994, WHO
document, WHO/HST/GSP/94.1, (WHO, Geneva).
WHO (1995), World Health Report. Bridging the gaps, (WHO, Geneva).
18
World Bank (1993). World Development Report. Investing in Health, (Oxford University
Press, Oxford).
19
FOOTNOTES
1. Chen et al. (1993) define absolute poverty as a situation whereby one can no longer satisfy
the basic needs of food, clothing and shelter.
2. These are developing countries belonging to East Asia. Latin America, Middle-East and
North Africa, South Asia and Sub-Saharan Africa and countries belonging to Eastern Europe.
3. Note that countries' dollar exchange rates, expressed in purchasing power parity (PPP),
reflect these countries' true purchasing power. Thus, PPP exchange rates allow one to
compare countries’ real consumption baskets.
4. Gunatilleke (1995, p.2).
5. UNDP (1994).
6. WHO (1995).
7 . We also wanted to analyze the evolution of the underfive mortality rate. However, data
for 1960 were not readily available to us.
8. Least developed countries are part of "A group of developing countries that was established
by the United Nations General Assembly. Most of these countries suffer from one or more
of the following constraints: a GNP per capita of around $300 or less, land-locked location,
remote insularity, desertification and exposure to natural disasters" (UNDP, 1994).
9. GDP per capita is measured at purchasing power parity (PPP) exchange rates.
10. The UNDP classifies Botswana as a LDC. However, given its level of GDP per capita,
we chose to rank it in the group of other developing countries.
11. The data collected are all from UNDP (1992), UNDP (1993) and UNDP (1994).
12. These calculations are approximations, as the computation of the impact (coefficient times
the change in the explanatory variable) is only truly correct for small changes in the
explanatory variables.
13. World Bank (1993, ch.3).
14. Note that Anand and Ravallion (1993) use GNP for their empirical analysis.
15. Some authors would continue to stress the role of the income distribution, however. For
instance, Waldham (1992) arrives at the anomolous result that a country, with the same real
income as another country, but in which the rich are wealthier, has a higher infant mortality.
16. WHO (1995, p.40).
17. At PPP values.
18. For data on the allocation of public health expenditure, see WHO (1994).
20
19. The poverty line is defined by the UNDP (1994, p.221) as "that income level below which
a nutritionally adequate diet plus essential non-food requirements are not available". Also note
that when TPOV is used to measure poverty, the data set consists of 57 countries. Due to
better data availability on rural poverty, the data set consists of 84 countries when RPOV is
used.
20. Recall that for the estimation of this equation, the data set of 84 countries was used.
21. See endnote 12.
22. See Barten (1989, p.194) on the multicollinearity problem.
23. For reasons of space, they are not reported here. The detailed results can be obtained
from the authors, however.
24. The results can be obtained upon request.
21
O 81
- Media
8123.pdf
Position: 1781 (5 views)