Content uploaded by Wellington Osawe
Author content
All content in this area was uploaded by Wellington Osawe on Apr 08, 2016
Content may be subject to copyright.
Developing Country Studies www.iiste.org
ISSN 2224-607X (Paper) ISSN 2225-0565 (Online)
Vol.4, No.18, 2014
111
Determinant of Infant Mortality Rate:
A Panel Data Analysis of African Countries
OSAWE, Osayanmon Wellington
Department of Agriculture and Rural Development,School of International Development
University of East Anglia, UK
osayanmon.osawe@gmail.com
Abstract
This study examines the factors that affect the rate of mortality among infants under one year of age using panel
data sets from the World Bank’s World Development Indicator database from the year 2000 to 2009 from 53
African countries. Using a random effect model in a 2SLS analytical method, results obtained after correcting for
endogeneity showed that fertility rate significantly affect infant mortality rate in a positive way. Similarly, GDP
per capita as a proxy for income, public health expenditure as a percentage of GDP, Prevalence of HIV and the
participation of adult female in the labour force significantly affect infant mortality rate. Furthermore, of all the
explanatory variables used in the analysis, fertility rate and GDP per capita had the most impact on infant
mortality rate respectively. This study also confirms Chowdhury’s (1988) postulation that there is a dual
causality between infant mortality rate and fertility rate.
Keywords:Infant mortality; Fertility rate; Health; Population; Africa
INTRODUCTION
Infant mortality is the number of deaths of infants under one year of age per 1,000 live births. The infant
mortality rate is an important measure of the well-being of infants, children, and pregnant women. It is related to
a variety of factors, including health status of women, the quality of and access to medical care services, socio-
economic conditions, and public health practices in any given geographical location. High infant mortality
signify the lack of proper childcare owing to poverty, lack of education, and societal preferences (such as the
affinity for a male child), among others (Zakir and Wunnava, 1999). Household decisions shape health, but these
decisions are also inhibited by the income and education of household members.
This study therefore intend to contribute to the quantitative assessment of the interlinking factors that affect the
rate of mortality among infants under one year of age which is one determinant of the health status of any nation
(Nwabu, 1997).
Pritchett and Summers (1996) explore a causal relationship between the the phenomena of health and wealth and
conclude that wealthier nations are healthier nations. These observations are not far-fetched because in principle,
it implies that wealthier nations are in a stronger position to provide better health to the population. Better health
in turn increases labor productivity, thereby enhancing wealth (Nwabu, 1997). Apart from being both a
constituent part of human welfare and a factor of production, good health as measured from infancy through to
progression to adulthood contributes directly to enjoyment of life. In this respect, it is in the same category as
other goods and services such as housing and education that fulfil this function (ibid). High infant mortality rates
could reflect improper childcare. ‘A population with diseased and unhealthy infants who grow up to form sickly
adults prone to disease, dampen economic progress in many ways: it decreases worker productivity; it does not
allow utilization of natural resources that would otherwise be accessible under good health conditions; it harms
the next generation by decreasing enrolment of children in school, and finally, it increases medical-care
expenditure, rendering inefficient allocation of resources (World Bank, 1993).
This study therefore aims at assessing factors from earlier studies on the determinant of infant mortality and to
test the validity of these factors that affect infant mortality using panel data from the World Bank’s World
Development Indicator database from the year 2000 to 2009, covering one decade. Other sections of this study is
organised as follows: a review of some literature and justification for the study is followed by a detailed
methodology for the study and the accompanying empirical analysis. The results and discussion on the findings
of the research are described afterwards. Finally, the conclusion and implication for research and policy is
highlighted.
THEORETICAL FRAMEWORK
Classical demographic transition theory implies that fertility decline is preceded by mortality decline. For this
reason, the interrelationship between mortality and fertility have been a keen concern to demographers as well as
policy-makers (Park et al., 1996). Several studies on health had substantiated a positive relationship between
fertility rate and infant mortality rate. For example, Zakir and Wunnava (1999) observed a strong positive impact
Developing Country Studies www.iiste.org
ISSN 2224-607X (Paper) ISSN 2225-0565 (Online)
Vol.4, No.18, 2014
112
between fertility rate and infant mortality in a cross-sectional data analysis of 117 countries. However, other
studies have put great emphasis on the causal link from high infant mortality to high desired fertility. The reason
being that parents in Africa and other parts of the world do not know whether their children will survive,
therefore they respond by having large families (McCord et al., 2010).The original model of the demographic
transition, indeed, was driven almost solely by infant mortality rates. Exogenous changes to infant mortality (e.g.,
the advent of public health, safe drinking water, immunizations, improved nutrition) were seen as the primary
precursor to reduced fertility rates—albeit with a lag of one or more generations (ibid). In this instance, the
argument has been that at least some of the powerful correlation of high infant mortality and high fertility
represents increased infant mortality due to higher fertility which is due in part to increasing strain on household
resources and decreased parental care and supervision with the addition of more children.
The relative income-health hypothesis postulates that income distribution is one of the key determinants of
population health (Torre and Myrskyla, 2011). One may not expect a direct linear relationship between infant
mortality and income per head to exist, because national income is simply a measure of the rate of production of
new goods and services. However, income is likely to influence the level of mortality indirectly through its
effects on the rate of consumption of items affecting health, such as food, housing, sanitation, medical care and
education (Flegg, 1982). Similarly, other studies have strongly suggested the income inequality-health
hypothesis (Lynch et al 2004)
1
.
There are many other possible determinants of infant mortality besides fertility rate and income. Unfortunately,
due to the limitation of access to adequate data, only few additional explanatory variables will be considered
here. These variables could include maternal education (proxy by literacy rate of adult female as a percentage of
females ages 15 and above), expenditure on health care as a percentage of GDP, percentage of women in the
labour force, prevalence of HIV, measure of income inequality (Gini index) etc. A case for the importance of
maternal education as a determinant of child mortality was made by Caldwell (1979). He concluded that
maternal education was the most important single determinant of mortality in childhood in the area of study.
Similarly, Zakir and Wunnava (1999) in their analysis of infant mortality suggest a significant coefficient for
both maternal labour force participation and literacy rate in their study of 117 countries. Similarly, the World
Development Report of 1993 argued that a limited package of public health measures and essential clinical
interventions is a top priority for government finance; some governments may wish, after covering the minimum
for everyone, to define their national essential package more broadly. Government expenditure on health care
services as it relates to infant and maternal health will be a crucial determinant of infant mortality rate. This
study is therefore aimed at examining the effect of some of these factors on infant mortality rate among Africa
countries.
METHOD AND EMPIRICAL MODEL SPECIFICATION
This research report uses a panel data set for the empirical analysis. The observations are 53 African countries
covering ten-year period from 2000 to 2009. The dependent variable is Infant Mortality rate (IMR) while the
explanatory variables are fertility rate (births per woman), GDP per capita (PPP, constant 2005 international $),
public health expenditure (percentage of GDP), prevalence of HIV (percentage of population ages 15-49), and
female labour force participation (percentage of total labour force). Undoubtedly, a substantial body of
knowledge have shown that other variables such as the measure of income inequality (measured by Gini index),
maternal literacy rate etc has strong influence on infant mortality (Deaton 2003; Lynch et al, 2004; Gonzalez and
Tseng, 2009; Torre and Myrskyla, 2011). However, these variables were not included in this analysis due to
unavailability of adequate data from the World Bank’s World Development Indicator database for the case study.
In choosing an appropriate specification model that produces efficient and consistent estimates, a test of
simultaneity is important. Gujarati (2004) noted that the OLS estimators produce consistent and efficient
estimators if there is no simultaneity problem. On the other hand, if there is evidence for simultaneity, the OLS
estimators are not even consistent. Similarly, Abrevaya et al (2010) opined that failure to correct for endogeneity
can result in incorrect inference and that with the availability of appropriate instruments, two-stage least squares
(2SLS) yields consistent estimates in linear models without the need for making parametric assumptions on the
error disturbances. In view of this, Chowdhury (1988) has suggested that there is a dual causality between infant
mortality rates and fertility rates. He argued that when a woman has multiple pregnancies, the chances of her
child’s survival are significantly reduced. A woman may therefore, decide to bear more children in the hope that
some will survive. Other studies have affirmed the endogeneity of infant mortality and fertility rate (Benefo and
Schultz, 1996; Schultz, 1997; Galloway et al, 1998; Dreze and Murthi, 2001). From these studies therefore, a test
for endogeneity between infant mortality and fertility is expedient for this study and this was subsequently tested
for.
1
For a detailed review of this literature on income inequality and health, see Deaton (2003) and Lynch et al. (2004).
Developing Country Studies www.iiste.org
ISSN 2224-607X (Paper) ISSN 2225-0565 (Online)
Vol.4, No.18, 2014
113
After performing a simultaneity test using Hausman (1978) Specification Tests, fertility rate was found to be
endogenous. The predicted value of fertility rate from the OLS regression of fertility rate on the instrumental
variable (population density) from the reduced form equation (2) was added as an extra regressor to the model in
equation (3) in order to correct for the problem of endogeneity. The empirical specification therefore for the
2SLS regresion analysis, is given below:
I
it
= β
0i
+ β
1
E
it
+ β
2
G
it
+ β
3
H
it
+ β
4
HI
it
+ β
5
L
it
+ µ
it
---------------(1)
Instrument: Population density (P
it
) for Fertility rate, E
it
STAGE 1
E
it
= α
0
+ α
1
P
it
+ ν
it
---------------------------------------------------(2)
STAGE 2
I
it
= β
0i
+ β
1
Ê
it
+ β
2
G
it
+ β
3
H
it
+ β
4
HI
it
+ β
5
L
it
+ ω
it
-------------(3)
β
0i
= β
0
+ e
i
i = 1, 2, ..., 53
e
i
= random error term
ω
it
= e
i
+ µ
it
Where i = 1, 2, 3, . . ., 53; t = 2000, 2006, . . , 2009;
I = Infant Mortality Rate (per 1,000 live births)
E = Fertility Rate (births per woman)
G = Natural logarithm of GDP per capita (PPP, constant 2005 international $)
H = Public Health Expenditure (percentage of GDP)
HI = Prevalence of HIV (percentage of population ages 15-49)
P = Population density (people per sq. km of land area)
L = Female labour force participation (percentage of total labour force)
µ, ν and ω= error term
Hausman specification test was conducted to select between fixed effect and random effect regression model.
According to the result from the test, the null hypothesis for selection of fixed effect was rejected at the 5 percent
level of significance. Therefore, random effect model was used for this analysis in the second stage of regression.
RESULTS AND DISCUSSION
Before delving into the econometric analysis, a brief overview of the description of the data used will suffice.
Table 1 therefore, shows a descriptive statistics of the data that was used for the analysis. The table revealed that
for all the observations (Africa countries), the mean infant mortality in Africa between the year 2000 to 2009 is
71.9779 and the maximum is 141.9. The average fertility rate for the sampled African countries is 4.7507.
As highlighted above, the a priori expectation is that fertility rate affects infant mortality rate in a positive way;
the variable parameter β
1
is therefore expected to be positive. The variable G (the natural logarithm of GDP per
capita) is used to capture a diminishing effect of GDP on infant mortality. The coefficient β
2
therefore, is
expected to be negative implying that an increase in income decreases infant mortality at a decreasing rate. β
3
is
expected to be negative, implying that an increase in government expenditure in the health sector, will ultimately
lead to better and improve access to health care services, which will therefore reduce infant mortality rate. The
prevalence of HIV was chosen because a country where HIV is prevalent especially among reproductive females
could exacerbate the rate of infant of mortality. Therefore, β
4
is expected to carry a positive sign. Similarly, the a
priori expectation is that female participation in the labour force could increase infant mortality (Zakir and
Wunnava, 1999). Accordingly, β
5
should have a positive sign. Table 2 shows the results of selected ordinary
least square (OLS) regression estimation with infant mortality as the dependent variable while ignoring the
possibility of simultaneity between infant mortality and fertility.
The value of R
2
for the first regression of Table 2 indicates that the explanatory variable, fertility rate, contribute
substantially to the explanation of inter-country differences in infant mortality rate in Africa; moreover, this
variable is also significant at the 1 per cent level of significance. Similarly, it is worth noting that a comparison
of the results in the first and last regression shows that fertility rate has a strong impact on infant mortality in
Africa with the R
2
value drastically reduced after exclusion of the effect of fertility rate in the last regression
model. It is equally worth noting that most of the explanatory variables are significant at the one per cent level in
the regression results.
Correcting for the problem of endogeneity, the second stage regression estimation results using the Random
Effect Model are shown in table 3 below:
An analysis of the results in the study shows that all the explanatory variables are significant at the one per cent
Developing Country Studies www.iiste.org
ISSN 2224-607X (Paper) ISSN 2225-0565 (Online)
Vol.4, No.18, 2014
114
level. However, the coefficient estimate of fertility rate after correcting for the problem of endogeneity was
found to be very high as shown in table 3. This may suggest that fertility rate has a strong impact on the rate of
death of infants within Africa countries. Similarly, GDP per capita and government expenditure in health also
have strong impact on infant mortality rate as shown in the estimates of -14.291 and -5.536 respectively. This
seems to imply that as the economy grows and the government place more priority in the provision of health care
services, African countries could achieve a rapid fall in their infant mortality rate. Similarly, the result in table 3
also shows that the participation of adult female in the labour force and the prevalence of HIV both have positive
impact on infant mortality. However, the estimates show that the impact of these variables is rather low
compared to other variables.
CONCLUSION
This empirical research examined the determinants of infant mortality in Africa using a panel data from the
World Bank’s World Development Indicator database covering ten years from 2000 to 2009. The results from 53
African countries after correcting for endogeneity showed that fertility rate significantly affect infant mortality
rate in a positive way. Similarly, GDP per capita as a proxy for income, public health expenditure as a
percentage of GDP, Prevalence of HIV and the participation of adult female in the labour force significantly
affect infant mortality rate in the area of study. Furthermore, of all the explanatory variables used in the analysis,
fertility rate and GDP per capita had the most impact on infant mortality rate respectively. This study also
confirms Chowdhury’s (1988) postulation that there is a dual causality between infant mortality rate and fertility
rate.
However, it is pertinent also to note certain limitations of the analysis. First, there are the challenges inherent in
drawing inferences from results, regarding the likely changes over time in the infant mortality rates of particular
countries or sets of countries (Flegg, 1982). Secondly, finding an appropriate Instrumental Variable for fertility
rate to correct the problem of simultaneity was a daunting challenge in arriving at an appropriate model
specification. Thirdly, limitations of data meant that many potentially important explanatory variables had to be
neglected. For instance, the measure of income inequality measured by Gini indext and the literacy rate of adult
female that are both important indicators for infant mortality rate were not included in the analysis.
Despite the limitations highlighted above, the results presented here in this study seems to suggest that African
countries whose government place low priority on individual family health (e.g. appropriate family planning
measures), achieving economic growth, and provision of adequate health care services are unlikely to
accomplish a rather rapid fall in their infant mortality rates.
REFERENCES
Abrevaya, J., Hausman, J.A., Khan, S. (2010) Testing for Causal Effect in a Generalised Model with
Endogenous Regressors. Econometrica, Vol. 78 (6), pp. 2043-2061
Benefo, K. and Schultz, T.P. (1996) Fertility and Child Mortality in Cote d’Ivoire and Ghana. The World Bank
Economic Review, Vol. 10(1), pp. 123-158.
Caldwell, J.C. (1979) Education as a Factor in Mortality Decline: An Examination of Nigerian Data. Journal of
Population Studies, 33 (3), pp. 395-413
Chowdhury, A. (1988) The Infant Mortality-Fertility Debate: Some International Evidence. Southern Economic
Journal. Vol. 54(3), pp. 666-674.
Deaton, A. (2003). Health, Inequality, and Economic Development. Journal of Economic Literature, 41(1), pp.
113-158
Dreze, J. and Murthi, M. (2001) Fertility, Education, and Development: Evidence from India. Population and
Development Review, Vol. 27(1), pp. 33-63.
Flegg, A.T. (1982) Inequality of Income, Illiteracy and Medical Care as Determinants of Infant Mortality in
Underdeveloped Countries. Journal of Population Studies, Vol. 36 (3), pp. 441-458
Galloway,P.R., Lee, R. and Hammel, E.A. (1998). Infant Mortality and the Fertility Transition: Macro Evidence
from Europe and new Findings from Prussia. in From Death to Birth: Mortality Decline and Reproductive
Change. Montgomery, Mark R. and Barney Cohen, eds. Washington, D.C., National Academy Press
Gonzalez, R.L., Tseng, F. (2009) Socio-economic Determinants of Mortality in Taiwan: Combining Individual
Data and Aggregate Data. GRIPS Policy Information Centre. Discussion paper, 09-01.
Gujarati, D.N. (2004) Basic Econometrics (4
th
Edition). New Delhi: Tata McGraw-Hill Publishing Company
Limited
Hausman, J.A. (1978) Specification Tests in Econometrics. Econometrica, Vol. 46, pp. 1251- 1271
Lynch, J., Smith, G.D., Harper, S. and Hillemeier, M. (2004.) Is Income Inequality A Determinant of Population
Health? Part 1. A Systematic Review. MilbankQuarterly, 82(1), pp. 5-99
McCord, G.C., Conley, D., Sachs, J.D. (2010) Improving Empirical Estimation of Demographic Drivers:
Developing Country Studies www.iiste.org
ISSN 2224-607X (Paper) ISSN 2225-0565 (Online)
Vol.4, No.18, 2014
115
Fertility, Child Mortality & Malaria Ecology. International Relations and Pacific Studies (IRPS). Available at
http://irps.ucsd.edu/assets/001/502663.pdf
Nwabu, G. (1997) Human Development in Africa. Economic Research Papers, No 38. African Development
Bank.
Park, C.B., Islam, M.A., Chakraborty, N., Kantner, A. (1996) Partitioning the Effect of Infant and Child Death
on Subsequent Fertility: An Exploration in Bangladesh
Pritchett, Lant and Lawrence H. Summers (1996), 'Wealthier is Healthier', Journal of Human Resources, Vol. 30
(4), pp. 841-68.
Schultz, T.P. (1997) Demand for Children in Low Income Countries. in Rosenzweig, M. R. and Stark, O. (eds.)
Handbook of Population and Family Economics
Torre, R. and Myrskyla, M. (2011) Income Inequality and Population Health: A Panel Data Analysis on 21
Developed Countries. Max Planck Institute for Demographic Research, MPIDR Working Paper, WP 2011-006
World Bank, (1993) World Development Report: Investing in Health. New York: Oxford University Press
Zakir, M. and Wunnava, P.V. (1999) Factors Affecting Infant Mortality Rates: Evidence From Cross-Sectional
Data. Applied Economics Letters. Vol. 1999 (6), pp. 271–273
TABLE 1: Mean and Standard Deviation of Variables in the Sample
Variable Obs. Mean Std. Dev. Min. Max.
Year 530 2004.5 2.8749 2000 2009
Infant Mortality Rate 530 71.9779 28.3535 11.5 141.9
Fertility Rate 530 4.7507 1.3523 1.5 7.503
GDP 509 3440.829 4835.954 247.9451 31738.23
Public Health Exp. 512 2.5518 1.2107 0.0031 8.1991
Prevalence of HIV 470 5.4657 6.9301 0.1 26.3
Female Labour Force
Participation
520 41.9691 7.7312 21.7928 53.1525
Population Density 530 79.0402 108.1542 2.3027 628.2379
TABLE 2: Selected Ordinary Least Squares (OLS) Regression Estimation Result with Infant Mortality as
the Dependent Variable.
E G H HI L R
2
Ȓ
2
17.041***
(32.06)
-
- - - 0.6607 0.6600
14.742***
(19.07)
-3.983***
(-4.10)
- - - 0.6631 0.6618
14.381***
(18.17)
-4.162***
(-4.28)
-1.259**
(-2.01)
- - 0.6658 0.6638
13.775***
(17.38)
-3.826***
(-3.73)
-3.533***
(-4.85)
0.922***
(7.02)
- 0.6267 0.6233
13.578***
(17.04)
-2.848**
(-2.50)
-3.691***
(-5.05)
0.810***
(5.67)
0.253*
(1.94)
0.6298 0.6256
- -13.045***
(-10.48)
-5.815***
(-6.28)
0.661***
(3.61)
0.537***
(3.23)
0.3876 0.3821
Note: t-statistics in parentheses. The level of statistical significance of the estimates is indicated by ***, ** and *
for the one, five and ten per cent level of significance.
TABLE 3: Random Effect Model Regression Estimation Result with Infant Mortality as the Dependent
Variable.
Explanatory
Variables
Estimated Coefficient
Standard
Error
p-value
R
2
= 0.4431
Ê 50.070*** 7.527 0.000
G -14.291*** 1.203 0.000
H -5.536*** 0.885 0.000
HI 0.476*** 0.177 0.007
L 0.568*** 0.158 0.000
Note: The level of statistical significance of the estimated coefficients is indicated
by ***, ** and * for the one, five and ten per cent level of significance.
