final project edited(3)
TRANSCRIPT
-
7/30/2019 Final Project Edited(3)
1/52
DETERMINANTS OF INFANT MORTALITY IN KENYA
A HOUSEHOLD LEVEL ANALYSIS
By
SUSAN KABURA NGIGI
X50/64673/2010
Research Paper submitted to the School of Economics, University of Nairobi,
in Partial Fulfillment of the Requirements for the Award of the Degree of
Master of Arts in Economics
November, 2012
i
-
7/30/2019 Final Project Edited(3)
2/52
DECLARATION
I declare this research project is my original work which has not been presented for a degree
award in this University or any other institution of higher learning. Where other peoples work
has been used acknowledgements have been duly made.
SignatureDate.
Susan K. Ngigi
We the undersigned have certified the students work and thus the project has been submitted
with our approval as University supervisors.
SignatureDate.
Prof. Jane Mariara
Signature..Date..
Dr. Mercy Mugo
ii
-
7/30/2019 Final Project Edited(3)
3/52
ACKNOWLEDGEMENT
I am grateful to God for the gift of life and for installing the seed of knowledge in me and more
so for good health and energy to handle my studies well up to the end. To my loving mother for
the sacrifice despite all the challenges, that ensured I attained basic education. To my sisters and
brothers thank you for the support so far. God bless you all.
This work received a lot of support from various people who in one way or the other guided and
motivated me to the end. Special thanks go to my supervisors Prof. Jane Mariara and Dr. Mercy
Mugo whose guidance and comments made this project a reality. I appreciate your patience andmotivation. I also thank Dr. Wawire and Dr. Obere whose lectures on research methods helped
me a lot in writing this paper.
I am greatly indebted to the University of Nairobi for sponsoring my postgraduate studies which
I could not have managed on my own. Special thanks also to the AERC also for facilitating the
Joint Facility for the elective units.
I also appreciate the School of Economics lecturers for their intellectual guidance in the field of
Economics. To my classmates thank you for making the learning process enjoyable and for the
valuable discussions. Special regards to Grace Njeri and Jane Kanina for academic and moral
support and for their valuable input in this project.
Finally my special recognition goes to my husband Geoffrey Muriuki for his valuable support
throughout my entire study period. Thank you for your understanding, your financial and moral
support; you made my learning easier and for this I salute you. God bless you abundantly and all
the best in your studies too.
iii
-
7/30/2019 Final Project Edited(3)
4/52
DEDICATION
I dedicate this research project to my loving husband (Jeff) for your undying love, support and
understanding throughout my Masters study.
To our Angel Shikoh; your kicks in my womb kept me awake in the library and reminded me of
the beauty of life. You greatly motivated my study. I wish you a long healthy life daughter.
iv
-
7/30/2019 Final Project Edited(3)
5/52
TABLE OF CONTENTS
DECLARATION ............................................................................................................................. ii
ACKNOWLEDGEMENT ..............................................................................................................iii
DEDICATION ................................................................................................................................iv
TABLE OF CONTENTS ................................................................................................................ v
LIST OF ACRONYMS .................................................................................................................vii
LIST OF TABLES ...................................................................................................................... viii
LIST OF FIGURES ........................................................................................................................ ix
ABSTRACT .................................................................................................................................... x
CHAPTER ONE: INTRODUCTION ..............................................................................................1
1.1 Project Background ............................................................................................................... 1
1.2 Statement of the problem ....................................................................................................... 5
1.3 Research Objectives ...............................................................................................................7
1.4 Relevance of the Study .......................................................................................................... 7
1.5 Organization of the Study ......................................................................................................7
CHAPTER TWO: LITERATURE REVIEW .................................................................................. 9
2.0 Introduction ...........................................................................................................................9
2.1 Theoretical Literature Review ............................................................................................... 92.1.1 Health care Production and child mortality framework by Mosley and Chen (1984) ....9
2.2 Empirical Literature Review ................................................................................................11
2.2.1 Studies Done in Kenya ..................................................................................................11
2.2.2 Studies outside Kenya ...................................................................................................13
2.3 Literature Summary ............................................................................................................. 16
CHAPTER THREE: METHODOLOGY ...................................................................................... 18
3.1 Introduction ..........................................................................................................................18
3.2 Analytical Framework ........................................................................................................ 18
3.3 Estimation Technique ..........................................................................................................20
3.4 Variable Definitions .............................................................................................................21
v
-
7/30/2019 Final Project Edited(3)
6/52
3.5 Data Source and Analysis tool .............................................................................................24
CHAPTER FOUR: DATA ANALYSIS, RESULTS AND DISCUSSION .................................. 25
4.1 Introduction ..........................................................................................................................25
4.2 Descriptive results ...............................................................................................................25
4.3 Econometric analysis ...........................................................................................................30
4.4 Discussion of Results ...........................................................................................................33
5.1 Introduction ..........................................................................................................................35
5.2 Research Summary ..............................................................................................................35
5.3 Conclusion ...........................................................................................................................35
5.4 Policy Implications based on key Results ........................................................................... 36
5.5 Limitations of the study ....................................................................................................... 37
5.6 Suggestions for further study ...............................................................................................38
REFERENCES
....................................................................................................................................................... 39
APPENDIX ...................................................................................................................................42
vi
-
7/30/2019 Final Project Edited(3)
7/52
LIST OF ACRONYMS
WHO World Health Organization
MDGs Millennium Development Goals
KDHs Kenya Demographic Health Survey
IMCI Integrated Management of Childhood Illnesses
IMR Infant Mortality Rate
DHS Demographic and Health survey
LPM Linear Probability Model
LOGIT Logistic Regression Model
OLS Ordinary Least Squares
HIV Human Immunodeficiency Virus
NFHS National Family Health Survey
SRS Sample Registration System
KSPA Kenya Service Provision Assessment Survey
vii
-
7/30/2019 Final Project Edited(3)
8/52
LIST OF TABLES
Table 1.1: Some Health Indicators in Kenya (1990-2010)..............................................................2
Table 3.2: Variables Acronym Definition.....................................................................................20
Table3.3: Variable Definitions and Apriori expectation...............................................................22
Table 4.4: Infant Mortality Rate as per 2208 KDHs data..............................................................25
Table 4.5: Sample statistic on the Independent variables used in the study..................................26
Table 4.6: Mothers age in relation to Infant death.......................................................................27
Table 4.7: Mothers Education in relation to Infant death.............................................................28
Table 4.8: Other independent variables in relation to infant mortality..........................................29
Table 4.9: Results from Logit model for Infant mortality in Kenya.............................................31
Table 4.10: Marginal effects from logit models............................................................................32
viii
-
7/30/2019 Final Project Edited(3)
9/52
LIST OF FIGURES
Figure 1.1: Infant Mortality and Under-five Mortality trend in Kenya (per 1000 live births)......42
ix
-
7/30/2019 Final Project Edited(3)
10/52
ABSTRACT
The major objective of this study was to establish the determinants of health status in Kenya as
proxied by infant mortality. The trend for infant mortality in Kenya has been on the decline but
the levels are still high as compared to the MDGs target. Infant mortality rate is a good measure
of economic development of a country as it indicates the quality of health services at a basic
level and it is a sensitive indicator since infants depend on the socioeconomic conditions of their
environment for survival.
The study used individual household level data from the 2008 Kenya Demographic and Health
Survey to examine the primary predictors of a child dying before celebrating their first birthday.
The study employed a logit regression model due to the categorical nature of the dependent
variable and also it was mostly used in literature by other researchers. Both standard coefficients
and marginal effects are presented but the study discussion is based on marginal effects.
Analysis of the study shows that in Kenya mothers age, household wealth, infants birth size,
mothers education and tetanus Immunization were the major determinants of infant mortality.
Based on the findings a number of policies are recommended which can help efforts in the
reduction of infant mortality. To begin with the study recommends initiatives to encourage
women to give birth at the middle ages, educating women past primary level, educating women
on healthy lifestyle during pregnancy and the need to attend ante-natal clinic and to receive
recommended immunization and improving households wealth standards would help reduce the
probability of having high infant mortality rate in Kenya.
x
-
7/30/2019 Final Project Edited(3)
11/52
CHAPTER ONE: INTRODUCTION
1.1 Project Background
Health is a state of complete physical, mental and social wellbeing not merely the absence of
disease or infirmity (WHO, 1948). According to the WHO declaration of Alma-Ata (1978),
health is a basic human right and it is fundamental for sustained economic and social well being
of a country. Attainment of a health level that will permit individuals to be socially and
economically productive depends on co-operation between individuals and the government.
The health status of a country is measured by the level of various indicators. The commonly used
indicators of a society are life expectancy at birth, mortality rate and morbidity rate. Life
expectancy at birth indicates the number of years a newborn infant would live if prevailing
patterns of mortality at the time of its birth were to stay the same throughout its life (World
Bank, 2008). Mortality rates refer to death rate. There are a number of different mortality rates
namely; neonatal mortality rate, infant mortality rate, under five mortality rate, maternal
mortality rate and crude death rate. Morbidity rates on the other hand refer to the rate of disease
prevalence in a population (World Bank, 2008).
Generally, health status indicators have been improving. However these indicators are still poor
in developing countries compared to the similar indicators in developed countries. For example,
life expectancy in Sub Saharan countries in 2007 was 49 years on average compared to 77 years
in developed nations. Similarly infant mortality rates for the more developed nations was on
average of 52 deaths per 1000 while for Sub Saharan countries was on average of 92 deaths per
1000 (Population Reference Bureau, 2007). An overview of Kenyas case is given in table 1.1
below
1
-
7/30/2019 Final Project Edited(3)
12/52
Table 1.1: Some Health Indicators in Kenya (1990-2010)
Years Infant Mortality
Rate (per 1000 livebirths)
Under five Mortality
Rate (per 1000 livebirths)
Life Expectancy
at birth (total)
Fertility rate,
total (births perwoman)
1990 64.3 99.4 59.3 6.0
1991 66.1 102.7 59.0 5.8
1992 67.6 105.4 58.5 5.7
1993 69.1 108.3 57.8 5.5
1994 70.7 111.6 57.0 5.4
1995 71.8 114.2 56.1 5.3
1996 72.6 116.3 55.2 5.21997 72.7 117.2 54.3 5.1
1998 71.8 116.4 53.5 5.1
1999 70.2 114.0 52.8 5.0
2000 68.6 111.1 52.3 5.0
2001 66.9 108.1 52.0 5.0
2002 65.5 105.3 52.0 5.0
2003 64.2 102.7 52.1 5.0
2004 62.8 100.1 52.5 4.9
2005 61.4 97.5 53.0 4.9
2006 60.0 94.6 53.7 4.9
2007 58.8 92.1 54.4 4.8
2008 57.3 89.3 55.1 4.8
2009 56.3 87.0 55.8 4.8
2010 55.1 84.7 56.0 5.0
Source: data.worldbank.org/data-catalog/world-development-indicators
Table 1.1 shows trends in heath indicators in Kenya. There is remarkable improvement in child
mortality since 1997 to 2010. At first infant mortality rate increased from 64 deaths per 1000 in
1990 to 72 deaths per 1000 in 1997. However after 1997 it has been on the decline up to 55
2
-
7/30/2019 Final Project Edited(3)
13/52
deaths per 1000 in 2010. In the same way, under-5 mortality rate increased from 99 deaths per
1000 in 1990 to 117 deaths in 1997 and has also been in a decline since then up to the level of 85
3
-
7/30/2019 Final Project Edited(3)
14/52
deaths per 1000 in 2010. The country is making progress in lowering the rates although 55
infants out of 1000 still die before celebrating their first birthday while 85 children out of 1000
die between one and five years of age. This implies loss in human capital which has negative
impacts on a countrys future economic development.
One of the Millennium Development Goals (MDGs) is the reduction of child mortality with a
target of reducing child and infant mortality rate by two thirds between 1990 and 2015. Kenyas
infant mortality rate was 64.4 deaths per 1,000 live births in 1990 (KDHS, 1993). To achieve the
MDG goal 4 there is need to reduce infant mortality rate to 22 deaths per 1,000 live births by
2015. According to the World Bank data for Kenya, the infant mortality rate in Kenya for the
year 2004-2008 was on average at 60 per 1,000 live births. This shows that despite the declining
trends in infant mortality rates we are still very far from achieving the MDG goal.
Kenyas first health framework; 1994-2010, lays out elements of a sound health care delivery
system, capable of promoting health, preventing disease, promoting life and nurturing well-being
to the highest possible health standards in response to the population needs. To meet its
objectives, the government has implemented various interventions which target various
determinants of health including nutrition, maternal education, safe water provision, adequate
sanitation, proper housing, increased government expenditure on health, intervention on
reproductive health services to control high population growth, and creation of a favorable
environment for increased private sector, and community involvement in health service
provision. Provision of insurance services was also expanded, with increased numbers of
insurance firms and covered persons. In addition introduction of exemptions for user fees for
some specific health services was done, including treatment of children less than 5 years,
maternity services in dispensaries and health centers, TB treatment in public health facilities, and
immunization services.
Other interventions targeting infants included the malezi bora strategy which focused on child
immunization, vitamin A supplementation, deworming of children under five years and pregnant
women, treatment of childhood illnesses, management of HIV, ownership and use of treated
mosquito nets for children and expectant mothers, preventive treatment of malaria during
4
-
7/30/2019 Final Project Edited(3)
15/52
pregnancy, and treatment of childhood fever. There was also the Integrated Management of
Childhood Illnesses (IMCI) approach that looks at improving management and services in health
care facilities as well as improving family and community health practices. Also there has been
exclusive breastfeeding programmes (GOK, 2010).
Despite implementing most of these policies, the health status indicators show a slow response in
performance as shown in table 1.1. For instance according to the KDHS (2008) data, infant
mortality dropped only by 32 percent in the 2008-09 survey as compared to the 2003 survey.
This change is from 77 deaths per 1,000 to 52 deaths per 1,000. Correspondingly, the under-five
mortality rate decreased to 74 deaths per 1,000 live births in 2008-09 from 115 in 2003 giving a
35.7% decline.However, the neonatal mortality rate only reduced by 6.1% from 33 to 31 per
1000 live births. Its effect on under five mortality increased to a contribution of 42% of the underfive mortality compared to its contribution of 29% in 2003 (KDHs).
1.2 Statement of the problem
Kenya is committed to achieving the MDGs targets by 2015. Based on the success made so far,
more progress could even be achieved despite the challenges. This study chose to examine the
determinants of infant mortality in Kenya. The study focus was motivated by the fact that infant
mortality rate indicates the quality of health services at a basic level and it is the most sensitive
of all indicators since infants depend on the socioeconomic conditions of their environment for
survival. Also, infant mortality has negative effects on a countrys future human resource thus
ensuring infants health safeguards their future contribution to the economic well being both at
home and nationally. Thus, the level of infant mortality would present a measure of how well a
society meets the needs of its people (Bicego and Ahmad, 1996). Finally infant mortality rate
illustrates the lasting impact of childhood health into adulthood. Childhood health has an impact
on adult health, education, economic performance and social status which in turn impacts
economic development of a country (Santere and Neun, 2010).
There has been declining trends in infant mortality rates in Kenya since 1997. This decline has
been attributed to the various interventions implemented in the health sector. However, despite
the efforts by the government, these figures are still high as compared to the MDG target of 22
deaths per 1,000 specific for Kenya. Given this, we can hypothesize that there could be many
5
-
7/30/2019 Final Project Edited(3)
16/52
factors behind these low improvements in the health status that cannot be amended through
increasing government health sector spending. There is need therefore, to establish the
underlying factors and to point remedial measures which would work to alleviate the problem.
Similarly there is need to focus on one particular area of concern which could have greater
impacts on the infant mortality rates other than utilizing scarce resources on areas with very low
impacts. For instance neonatal care, need receive more attention since it contributed 42% of the
under five mortality as per KDHs data, (2008). A clear understanding of the determinants of
infant mortality is therefore important which calls for deeper investigation of the determining
factors of Infant mortality in Kenya.
In Kenya the study done by Elmahdi (2008) considered socioeconomic determinants to be more
important in determining infant mortality. His study was more concerned with thesocioeconomic determinants. On the other hand Mutunga (2004) examined infant and child
mortality relationship with households environmental and socioeconomic characteristics and
found both as having significant impact on child mortality. Both studies used KDHS data for
2003. Wamae et al. (2009) assessed the health practices in the management of child illnesses in
health centers and concluded health providers do not conduct full investigation and counseling of
sick children and thus are responsible for the rising trends on child mortality. The study used
KSPA data. This study has used recent demographic data KDHS (2008) which is more
appropriate to consider after the major interventions mentioned above were implemented in the
National Health strategic plan I and II (2000-2010). Also there was need to establish the
important factors in infant mortality since each study reviewed focused on specific determinants
of choice. This study therefore sought to fill this information gap in literature and make
appropriate policies to our policy makers by estimating a model that has demographic,
socioeconomic, biological, environmental, and maternal and child health care factors to establish
which of them greatly determine infant mortality. This was aimed at helping the existing
framework in order to stress the important of various factors in the mediation process for
reducing infant mortality.
6
-
7/30/2019 Final Project Edited(3)
17/52
1.3 Research Objectives
The general objective of this study was to empirically examine the determinants of infant
mortality in Kenya.
The specific objectives were;
1. To investigate whether demographic, socio-economic, environmental and child/maternal
health care factors have an influence on infant mortality.
2. To analyze the impact of mothers health knowledge on infant mortality.
3. Based on the findings in 1 and 2 above, to make policy recommendations aimed at
reducing infant mortality rate in Kenya.
1.4 Relevance of the Study
This study was relevant as it will help add to the existing literature on child and infant mortality
in Kenya. It used recent demographic data which was more appropriate to consider in assessing
the impacts of various government interventions in child mortality. This is because earlier studies
on infant mortality have used past KDHs data and there was need to use recent KDHs data for
2008 which is more appropriate to consider after the major interventions mentioned above were
implemented in the National Health strategic plan I and II (2000-2010). This study has been done
at the initial years of the implementation of the Kenya health policy plan 2010-2030 and in such;
policy recommendation offered may help the government in its effort of reducing infant
mortality with an aim of achieving the specific MDG target of 22 deaths per 1000 for Kenya by
2015.
1.5 Organization of the Study
In chapter one, the study presented the project background, problem statement, justification of
the study and the study objectives. Chapter two reviewed both the theoretical and empirical
literature relevant to this study. Chapter three outlines the methodological approach that was
employed in this study and it includes the analytical framework, variable definition, estimation
technique together with a description and analysis of data and data sources. Chapter four presents
7
-
7/30/2019 Final Project Edited(3)
18/52
results and interpretation of findings from the study. Finally, Chapter five gives a summary of
the study, conclusions and policy implications.
8
-
7/30/2019 Final Project Edited(3)
19/52
CHAPTER TWO: LITERATURE REVIEW
2.0 Introduction
This chapter presents the theoretical and empirical literature on child health and mortality.
Theoretical literature presents the health production model and child mortality framework by
Mosley and Chen (1984), while the empirical literature discusses studies done on infant mortality
in Kenya, other sub-sahara African countries and outside world.
2.1 Theoretical Literature Review
2.1.1 Health care Production and child mortality framework by Mosley and Chen (1984)
Health care production according to Santere and Neun (2010) is influenced by a variety of inputs
among them nutrients, genetic endowments, exercise, lifestyle and the amount of medical care
consumed given the technological-biological production relationship. In household health
production according to Rosenzweig and Schultz (1983) input preferences made by households
are considered to be determined by the health technology. Health technology is the human
biological mechanisms through which behavior affects health. The household preferences are
also determined by prices and income of the household. They note that in order to conclude on
causal relations between health and inputs there is need to control for heterogeneity and thereforeestimate health technology from models in which inputs affecting health are choice variables by
themselves. E.g mothers education
This production framework was developed further by Mosley and Chen (1984) while trying to
provide an approach to the investigation of determinants of child survival in developing
countries. In their framework all economic, environmental and social determinants of child
mortality operate through a common set of biological mechanisms (proximate determinants) to
impact on mortality. For instance educated mothers may have lower infant mortality because the
skills learnt will help them in making choices regarding contraception, preventive care, treatment
and child nutrition. This model was first introduced by Davis and Blake (1956) in a study on
fertility. The proximate determinants framework is very similar to the health production function
for households.
9
-
7/30/2019 Final Project Edited(3)
20/52
The model identifies a set of proximate determinants that determine mortality directly. They are
grouped into five categories. The first four categories affect the rate of a child moving from a
state of healthiness to that of sickness while the last category; personal illness control, influence
the rate of getting sick as well as the rate of recovery. This is through preventive measures and
treatment seeking behavior. The categories include: Maternal factors (age, parity and birth
interval); environmental factors (air, skin/soil, insect vectors, inanimate objects and food/water);
nutrients deficiency (calories, proteins and the micronutrients); injury (intentional or accident)
and Personal illness control (preventive measures and medical treatment). A change in any of the
determinants will influence a childs health either to the better or to the worse.
The Mosley and Chen (1984) framework was integrated by Shultz (1984) into an economic
choice model. This model explores the biasness of the direct association between health inputsand an individual's health outcome as employed in epidemiological research. The argument is
that individuals initial health endowment differs and their health inputs are determined by their
knowledge of their endowment. The framework is based on the microeconomic model of the
family. Individuals allocate their time and resources in response to the value of their time,
nonhuman capital endowments and the relative prices for their inputs and outputs. The market
determines the value of the time of persons working since the market sets the wage
rate.
Thus individuals allocate their resources to yield optimal quantities for all their choice variables
for instance contraceptive use, nutrients intake, medical care, whether to breastfeed or not, food
purchase and child care time for a given set of constraints; wages and prices and for a given
production technology. Mothers education is used as a close correlate of women wages. This
model gives two sets of equations and suggests methods of estimation for each. The first is the
demand equation which predicts the proximate determinants as a function of cultural, social and
economic factors. It can be estimated using OLS but for discrete choice variables, Logit or probit
functions can be used. The second equation is the health production equation and it predicts the
health outcome as a function of proximate determinants. It can be estimated using Logit or
structural equation methods.
10
-
7/30/2019 Final Project Edited(3)
21/52
Millard (1994) also proposed an approach to the investigation of determinants of child survival
in developing countries. The model by Millard is similar to that by Mosley and Chen in terms of
levels of causes of child mortality. However it differs by the introduction of three tiers of causes
of child mortality in trying to emphasize parental behavior and child care traditions in the
economic, social, cultural, political and environmental context. The three layers are proximate
(immediate biomedical causes of death e.g. malnutrition), intermediate (behavioral patterns that
increase exposure to the proximate causes e.g. breastfeeding patterns) and ultimate causes
(economic, political and socio-cultural factors)
According to Hill (2003), other modeling strategies have been in use after the Mosley and Chen
model. For instance there are models that link background factors and proximate factors with
mortality; such as factors related to immunization. The models are implemented step by stepstarting with background factors and then the proximate factors. A reduced-form model of net
association of background variables and child mortality is then established. For example
Ajakaiye and Mwabu (2007) have employed such a modeling strategy for the determination of
birth weight while considering tetanus vaccination as a health input factor while behaviors like
prenatal care and behavioral changes during pregnancy to be the intermediaries.
2.2 Empirical Literature Review
There are various studies regarding the determinants of infant mortality which show significant
association between demographic, socioeconomic factors, environmental factors and infant
mortality.
2.2.1 Studies Done in Kenya
In Kenya there are several studies on infant and child mortality using KDHs and KSPA data.
Kamau (1998) for instance used KDHs data for 1993 in his unpublished research paper focusing
on child survival determinants in Machakos, Kilifi and Taita Taveta districts. The study showed
significant relationship between marital status, mothers level of education, place of delivery, age
at first birth, religion and availability of toilets with incidence of child mortality.
Mutunga (2004) in his unpublished Masters research paper sought to show that socio-economic
and environmental characteristics (mothers education, sources of drinking water, electricity,
11
-
7/30/2019 Final Project Edited(3)
22/52
-
7/30/2019 Final Project Edited(3)
23/52
The study suggests that improved skills for health workers are needed for better services in child
health management.
2.2.2 Studies outside Kenya
Studies done in regard to determinants of infant and child mortality globally show evidence of
significant association between socioeconomic, demographic, environmental and infant-child
characteristics (Hosseinpoor, 2005; and Caldwell, 1979). Caldwell (1979) for instance gave input
in regard to effect of mothers education on infant mortality with reference to Nigeria. The study
concluded that higher education lowered the rate of infant mortality through factors like hospital
delivery, increased ante natal care for pregnant mothers and changing traditional family
relationships. Supporting Caldwells explanation, Hobcraft (1993) explained that education can
contribute to child survival by making women more likely to marry and enter motherhood later,
have fewer children, utilize prenatal care and immunize their children.
Maternal education is usually used as a proxy for other household characteristics. Medrano et al.
(2000) used mothers education as a measure of health knowledge and the socioeconomic status.
Their study used cross sectional data for South Africa Integrated Household Survey for 1993.
They used nave regression to assess the relationship between mothers education and child
health and the findings show positive relation with increased knowledge leading to childs good
health. Similarly Kovsted et al. (2003) using mothers religion as a measure of health knowledge
showed mothers knowledge was important in determining child health.
In another level, mothers education has been shown to have positive and inverse impact on child
health. Desai and Alva (1998) using DHS data from 22 countries found that infant mortality was
lower among educated women, and that although this effect intensify with the inclusion of other
socioeconomic factors in their models, maternal education remained significant. In contrast, Hill
et al. (2001) observed an inverse relationship between mothers educational level and economic
status and child mortality in the late 1980s and 1990s using the 1993 and 1998 Kenya DHS
data. This study was done using multivariate analysis. The variables examined were mothers
education, wealth status, residence, maternal age, preceding birth interval and birth order. The
authors made a conclusion that HIV epidemic was the most probable cause of increased child
mortality not the socioeconomic or demographic factors.
13
-
7/30/2019 Final Project Edited(3)
24/52
Beenstock and Sturdy (1990) also found important role for female literacy in determining infant
mortality in their study among several Indian states. However the study concluded that maternal
literacy had a weaker effect on child survival in Sub-Saharan Africa. This could be due to
environmental factors, family wealth and other demographic factors unique to this part of the
continent. Similarly in India Claeson et al. (2000) analyzed the countrys SRS and NFHS data
and observed that non income factors (maternal and child health interventions) play a significant
role in reducing infant and child mortality in India. The study also observed that girls are 30
percent more likely than boys to die before their fifth birthday. This was attributed to son
preference in India, which is manifested in lower spending on health for girls and higher
prevalence of immunization among boys. Conversely the United Nations secretariat (1988)
carried out a study on sex differentials on life expectancy and mortality in less developed
countries and their results showed that male infant have higher probability of dying than female
infants.
According to Jones et al. (2006) the main causes of child mortality presented in India are disease
related (diarrhoea, pneumonia for under-five children and tetanus for new born), pre-term
delivery and sepsis. However the major cause of death is under nutrition. Due to inadequate data
the study employed an analytical overview through study groups.
While Uganda DHS for 3 surveys and employing a reduced form probit model, Ssewanyana and
Younger (2005) investigated the infant mortality determinants in Uganda in order to assess the
likelihood of Uganda meeting MDG goal 4. The results show that improvement in mothers
primary education and vaccination service greatly impacts infant mortality. However household
income and infant mortality were correlated negatively to a small extent. This implied that even
if Ugandas rapid growth continued the impact on infant mortality would be small.
A study by Hosseinpoor et al. (2005)on the socioeconomic inequality in infant mortality in Iran
using ordered probit model concluded that other factors beyond the health care delivery system
significantly contribute to high infant mortality rate in Iran. The study recommended additional
interventions to be done in the area of environmental and sanitation in order to reduce infant
mortality. This would work best if implemented in rural areas so as to curb the socioeconomic
inequality in mortality among urban and rural areas in Iran. The study used DHS data done in
14
-
7/30/2019 Final Project Edited(3)
25/52
Iran in 2000. This data covered a long period of time (early 1980s to early
2000s).
Alves and Belluzzo (2005) investigated the determinants of infant mortality rates in Brazil using
panel data for the years 1970, 1980, 1990 and 2000. Pooled OLS was used to estimate the model
while considering variables like sewage services, illiteracy rates and dummy variable for
economic development. Their findings confirm that poor child health (in terms of mortality rates)
in Brazil could be explained by the levels of education, sanitation and poverty. Moreover, the
paper showed that mothers education was the most important variable given that for every
additional year of schooling, average mortality rates declined by more than 7%.
Pandey, et al. (1998) estimated the socioeconomic determinants of infant mortality and observed
that mothers literacy, household heads religion, place of delivery, cooking fuel, mothers
exposure to mass media and wealth were significant variables. However religion was found to
have a modest impact in some region of Iran. Hazard models were used to analyze the
relationship using NFHS data. Mothers tetanus immunization during pregnancy was also found
to reduce neonatal mortality. In another level, the study also found a U-shaped relationship
between birth order and mothers age at birth with infant mortality. Thus when age increases
from teenage to matured mother mortality falls and it rises as one move to elderly mother. Other
studies have also documented evidence of a reverse pattern in the association between maternal
age at birth and infant mortality, with teenage and older mothers having elevated risks of child
loss (Koenig, 1992; Pebley, 1991).
A study by Da vanzo et al. (2004) supports these findings. They used high-quality longitudinal
dataset in Bangladesh to establish the relationship between birth interval and child mortality. The
findings using Cox proportional hazard model show that preceding inter-birth intervals of less
than 24 months in duration are associated with significantly higher risks of early neonatal
mortality whereas, intervals of less than 36 months are associated with higher risks of late child
mortality. Effects of short intervals are stronger the younger the child.
Madise, (2003) did a study which aimed at comparing the socioeconomic and demographic
determinants of child mortality in Zambia and found that at neonatal stage demographic factors
15
-
7/30/2019 Final Project Edited(3)
26/52
(e.g. birth orders, mothers age ) were more important determinants of infant mortality while at
the post neonatal stage both demographic and socioeconomic e.g. place of residence, were
important determinants.
Several studies have considered households socioeconomic status in terms of their source of
drinking water, sanitation, source of cooking oil and income level. Where data on income levels
is not collected it is proxied by wealth. According to Mosley and Chen (1984) the effect of
socioeconomic factors on mortality is through environmental hazards, maternal factors, injury
and nutritional status. Kabubo-Mariara et al. (2012) shows that water and sanitation variables
reduce the hazard ratio when other factors affecting mortality are not controlled but when
controlled the variables gave a positive relationship with mortality. Fayehun (2010),
hypothesized that variations in household environments among sub-Saharan countries couldaffect childrens survival chances. The study found that there are significant relationships
between the household environment and child survival. Some of the differences in childhood
mortality could be explained by levels of household environmental health hazards and by
maternal socioeconomic status.
Source of cooking fuel is another considered variable that affect child mortality. Mutunga (2004)
found clean sources of cooking fuel to be significant in reducing death in households. This could
be due to a reduction in indoor air pollution.
2.3 Literature Summary
In conclusion, major health determinants that have been put forward in literature are
socioeconomic, demographic and environmental factors. Also health services and behavior that
promote and increase health stock have been associated with improved health status. For instance
tetanus injection for pregnant mother, higher education/health knowledge for mothers, access to
clean water and sanitation were found to have high impacts on child survival. In another level,
high mortality rates were observed in girls than boys due to some households preference of
boys.
Therefore there are many factors that have been argued to be the determinants of infant
mortality. The choice of variables used in the present study was determined by data availability.
16
-
7/30/2019 Final Project Edited(3)
27/52
Nevertheless most of the determinants of infant mortality presented in literature, like mothers
age at the time of birth, mothers education, place of residence, religion, household wealth, birth
interval, access to sanitation, source of drinking water and tetanus injection of mothers were
analyzed.
Majority of the studies reviewed in Kenya were mainly concerned with some determinants not
all. For instance, Mutunga (2004) was concerned with impacts of environmental and socio-
economic factors on child mortality, Kabubo-Mariara et al. (2012) focused on effect of physical
environment on child survival while Elmahdi (2008) was more concerned on the socio-economic
determinants of infant mortality. On another level Kamau (1998) focused on child survival
determinants in arid and semi arid lands of Kenya thus did not consider the whole country. This
study will add to the existing knowledge on child mortality in Kenya while employing a modelthat incorporates most of the variables available in the socio-economic, demographic,
environmental and biological categories as well as maternal and child health care factors to
establish which of them greatly determine infant mortality.
17
-
7/30/2019 Final Project Edited(3)
28/52
CHAPTER THREE: METHODOLOGY
3.1 Introduction
This chapter presents the tools of analysis and describes the data as well as methods of analysis
that will be used. The methodology of the study was motivated by the reviewed literature both
theoretical and empirical in chapter 2 and the type of data that is available.
3.2 Analytical Framework
From reviewed literature child/infant survival was found to be determined by socioeconomic
factors which affect the intermediate/proximate determinants of health. These proximate
determinants are the intermediate variables between the socioeconomic determinants and themortality risk (Mosley and Chen, 1984). The variables include; (maternal and environmental
factors, nutrient deficiency, injuries and disease control). In this regard the household production
framework given by Schultz (1984) will be modified and used to analyze the impacts of various
covariates on child survival. As Hill (2003) puts it; this framework has stood the test of time and
still provides the conceptual basis for many studies on child survival. For instance this model
was modified and used by Mwabu (2008) in his study of child health production in Kenya and
Foloko (2009) Masters thesis on determinants of child mortality in Lesotho.
Child health production function is given as a linear function;
H = F (Y, I,k, )..(1)
Where,
Y is the proximate biological inputs to child health (immunization, cooking fuel, water andsanitation environment)
I is a child health input such as curative and preventive medical care
K is the health knowledge possessed by the household (e.g. fertility and child spacing
techniques) and
is the child health endowment due to genetic or environmental conditions not influenced by the
parents.
18
-
7/30/2019 Final Project Edited(3)
29/52
The proximate inputs Y are chosen by the household in a manner to reduce the health outcome
(mortality). They depend on the child health endowment (), maternal/household preferences
(PR), prevailing prices in the market and specific constraints posed by the households physical
environment (P) and household wealth (W). Thus our reduced-form input demand function will
be;
Y=F (, PR, P, W).. (2)
The reduced form demand functions were given as the objects of study in the Shultz (1984)
model.
Equation (1) and (2) shows that child health can be explained by proximate biological inputs to
child health (Y), child health input (I), Household knowledge (K), household wealth (W), and achilds health endowment .
Based on equation (1) and (2) this study estimated a model which included variables from the
categories in the equations.The choice of variables was motivated by the Mosley
and Chen (1984) framework.
These factors consisted of;
Xi,j which is a vector of maternal characteristics (maternal age, education, religion, number of
children ever had, marital status and family/residential factors). This was to help capture the
households choice of proximate inputs and its impact on mortality.
Xk,j which is a vector of household characteristics (household wealth, access to water and
sanitation, source of fuel). This helped capture the variables impact on child health.
Xl,j which is a vector of biological and child endowment factors (birth order, birth size, gender
and birth spacing). This helped capture the childs health endowment and its impact on mortality.
Xm,j which is a vector of health services variables (place of delivery and tetanus injection). This
helped capture the impact of health intervention in reducing mortality risk
ej is the error term.
19
-
7/30/2019 Final Project Edited(3)
30/52
The study estimated the following model;
Lj = 0 + iXi,j + kXk,j + Xl,jl + mXm,j + ej.(3)
Where Li is the dependent variable based on the probability of a child dying before the first
birthday. It took only two values; 1 if the child was reported alive and 0 if the child was reported
dead.
The final model while including all variables was as given in equation (4)
Lj = 0 + 1 MAGE + 2 MAGESQ + 3TNO + 4MRTS + 5MEDU1 + 6MAREL1 + 7 TORE1 +
8 HW1 + 9 ATW1 + 10 ATS1 + 11 BORD1 + 12 BSIZE11 + 13 G + 14 BSPA1 + 14
PLOD1 + 15 TTI1 + ei.. (4)
Table 3.2: Variables Acronym Definition
Acronym Definition
MAGE mothers age
MAGESQ Mothers age squared
TNO Total number of children the mother has
MRTS Marital status of the mother
MEDU mothers highest level of education
MAREL mothers religion
TORE households type of residence
HW households wealthATW households access to water
ATS households access to sanitation
SCF households source of cooking fuel
BORD birth order of the infant
BSIZE birth size of the infant
G Gender of the infant
BSPA birth spacing
PLOD place of delivery of the infant
TTI tetanus toxoid injection for the infants mother
3.3 Estimation Technique
One problem the estimation encountered was on prevailing prices for health input goods given
that DHS data did not collect information on prices. We followed, Kovsted et al. (2003) where
20
-
7/30/2019 Final Project Edited(3)
31/52
we assumed identical prices for all the households and thus we estimated child health as a
function of proximate biological inputs to child health (Y), child health input (I), household
health knowledge (K), household wealth (W), and a childs health endowment .Similarly since
health knowledge is assumed to be endogenous a bias could arise when parents are aware of a
given health condition not known to the researcher. In this regard, mothers education was used
to assess health knowledge (Medrano et al.2000).
The estimation was done using the binary choice models since mortality rate is a discrete
variable. Thus the dependant variable took only two values; 1 if the child was reported dead and
0 if the child was reported alive. The objective was to find the probability of the child dying or
not dying. Three approaches could have been used; the linear probability model (LPM), the
logistic regression model (logit) and the probit model (Gujarati, 2007).
The LPM is the simplest to use but has some serious limitations. The error term violates the
assumption of normality, the model suffers from heteroskedasticity and more seriously, there is
always the possibility of the estimated probability lying outside the 0-1 bounds (Gujarati, 2007).
Other limitations are limited usefulness of R2 and marginal effects are constant. Furthermore, the
LPM is logically not a very attractive model even if the above limitations are solved in that it
assumes that the conditional probabilities increase linearly with the values of the explanatory
variables. A probability model that has an S-shaped feature of the cumulative distributionfunction is preferred. In practice the logistic and the normal cumulative distribution functions are
chosen, giving rise to the logit and the probit models respectively. The study presented estimates
using the logit model although the study estimated both logit and probit models with no
significant differences in coefficients. Logit model is often used in literature (see Sewanyana and
Younger, 2005; Da vanzo, 2004 and Elmahdi, 2008). This was noted to be as a result of the ease
in interpreting results for logit model in terms of marginal effects which is impossible with probit
model.
3.4 Variable Definitions
This section defines the variables which were analyzed in the study.
21
-
7/30/2019 Final Project Edited(3)
32/52
Table3.3: Variable Definitions and Apriori expectation
Variables Definitions Apriori expectation
Maternal characteristics
Mothers age (MAGE) Very young and very old
women are expected to
increase child mortalityrisk. (Pandeyet al. 1998)
Mothers education (MEDU) Mothers highest level of
education;
None-(0)
Primary -(1)
Secondary -(2)
Higher -(3)
Higher education is
expected to improve child
health and survival,(Alves and Belluzzo, 2005)
Type of residence (TORE) Urban (1),
Rural (0)
Rural residents are
expected to have high
child mortality risks thanurban residents, (Kabubo-
Mariara et al., 2012)
Mothers Religion(MAREL) Has Religion (1),
No religion (0)
Religion impactsknowledge which is
expected to improve
health. (Pandey, et al.,
1998)
Total number of children (TNO)
Marital status of the mother
(MRTS)
Married (1)
Not married(0)
Children Characteristics
Gender (G) Male (1),
Female (0)
Boys are expected to havehigh mortality risks than
girls, (Claeson et al.,2000)
22
-
7/30/2019 Final Project Edited(3)
33/52
Birth Order (BORD) The birth order of the child in
the family;
First order -(1)
2-3 birth order-(0)
Above 3 birth order -(2)
First order and above 3
birth order are expected tohave negative impacts on
mortality, (Da vanzo et al.
2004)
Birth spacing (BSPA) Less than 24 months (1)
More than 24 months (0)
Better spacing is expected
to have positive impacts
on mortality, (Da vanzo etal. 2004)
Birth size (BSIZE) small/ very small -(1)
average size-(0)
Large/ very large -(2)
Small size and very large
infants are expected to
have high mortality risks,(Elmahdi, 2008)
Household characteristics
Households wealth (HW) Poor -(1)
Middle -(0)
Rich -(2)
Household wealth is
expected to impact child
health positively, (Pandey,
et.al., 1998)
Source of cooking fuel (SCF) Electricity-(1)
LPG-(2)
Kerosene (3)
wood -(4)
coal/charcoal -(5)
dung -(6)
grass- (7)
Availability of clean
cooking fuel free from airpollution is expected to
improve child health,
(Mutunga, 2004)
Access to sanitation (ATS) No facility-(0)
Flush toilet-(1)
Ventilated improved pit (VIP) -
(2)
Clean human waste
disposal facilities are
expected to improvehealth, (Mutunga, 2004)
23
-
7/30/2019 Final Project Edited(3)
34/52
Pit Toilet-(3)
Other-(4)
Access to water (ATW) Piped water -(1)
public tap -(2)
open well -(3)
Rainwater -(4)
Other -(5)
River/Natural reserves- (6)
Bottled water- (7)
Availability of clean
drinking water is expectedto improve child health
and survival, (Mutunga,
2004)
Health service variables
Place of delivery (PLOD) Hospital (1)(0) Otherwise
Availability of basichealth amenities improveshealth, (Mosley and Chen,
1984)
Tetanus Toxoid injection (TTI) Immunized (1)
(0) Otherwise
Injection with tetanustoxoid for pregnant
mothers improves child
health and survival,
(Ssewanyana andYounger, 2005)
3.5 Data Source and Analysis tool
The study used KDHS data for 2008/9 collected in Kenya. The data was downloaded from
Measure DHS website after obtaining permission to download. The sample size for all
interviewed women was 8,444 women aged between 15 to 49 years selected from 400 samples
throughout Kenya. The survey utilized a two-stage sample based on the 1999 Population and
Housing Census and was designed to produce separate estimates for key indicators for each
province in Kenya.The study population for this analysis includes a sample of women whoreported information on infants born one year before the study period. The observations at this
level were 2425 representing mothers who had infants born for the period.
24
-
7/30/2019 Final Project Edited(3)
35/52
CHAPTER FOUR: DATA ANALYSIS, RESULTS AND DISCUSSION
4.1 Introduction
This chapter presents the research findings for this study. Descriptive statistics are presented first
then the findings on infant mortality rate in Kenya as well as the determining factors. The results
are presented in tables followed by a discussion on the same. Logit results are presented and
discussed.
4.2 Descriptive results
This study utilized data collected for women between the ages of 15-49years. Before the analysiswas done various variables were regrouped to make the analysis more meaningful. For instance
mothers age was grouped into seven categories to allow for analysis of very young and very old
womens effect on infant mortality. Birth order was recoded to 2-3 birth order, first order and
above 3 birth order following the study by Da Vanzo et al. (2004). Similarly as per the studys
finding this study adopted their categorization on birth spacing into two; less than 24 months and
more than 24 months. Other grouping was done on access to sanitation, place of delivery and
access to water depending on the number of observations recorded and the appropriateness to the
analysis.
Table 4.4: Infant Mortality Rate as per 2208 KDHs data
Number of infant dead 127
Number of infants alive 2,298
Total 2425
infant deaths per 1000 live births (IMR) 55.27
As shown in table 4.4 infant mortality rate as per the KDHs data was 55.27 deaths per 1000 live
births. This was as reported in the KDHs report. As earlier alluded in the introduction of thisresearch Kenya still suffers from high infant mortality rate as compared to the MDGs target of
22 deaths per 1000 live births specific for Kenya. Our findings indicate that 5% of infants born
are likely to die before celebrating their first birthday.
25
-
7/30/2019 Final Project Edited(3)
36/52
Table 4.5: Sample statistic on the Independent variables used in the study
Variable Value Names Frequency Percent
MAGE 14-19 years 260 10.72
20-24 years 774 31.92
25-29 years 616 25.430-34 years 445 18.35
35-39 years 231 9.53
40-44 years 80 3.3
45-49 years 19 0.78
MEDU No education 478 20
Primary Education 1,395 58
Secondary Education 428 18
Higher Education 124 5
MAREL No religion 86 3.55
Has religion 2,339 96.45
TORE Rural 1,812 74.72
Urban 613 25.28
HW Middle 380 15.67
Poor 1,117 46.06
Rich 928 38.27
ATW Piped Water 404 16.66
Public Tap 304 12.54
Open well 947 39.05
Rainwater 35 1.44
Other 53 2.19
River/Natural reserves 682 28.12
ATS No facility 589 24.29
Flush toilet 246 10.14
VIP 348 14.35
Pit toilet 1,221 50.35
Other 21 0.87
G Female 1,145 47.22
Male 1,280 52.78
PLOD Home 1,313 54.14
Hospital 1,107 45.65
TTI Not Immunized 330 14.35Immunized 1,951 84.86
The study utilized various variables as motivated by the Mosley and Chen (1984)
framework. Table 4.5 gives a summary of the variables included in the model
for infant mortality.
26
-
7/30/2019 Final Project Edited(3)
37/52
From the findings in regard to maternal characteristics, majority of the
women interviewed were between 20 and 24 years (32%) while only 0.78%
of the interviewed women were between 45-49 years. In total all women
interviewed were between 14 and 49 years. 58% had primary education, 18% had
secondary education, 20% had no education while only 5% had higher education. In regard to
religion, 96% of the respondents had some religion; Christianity or Muslim while 4% did not
have any religion. 75% resided in rural areas with only 25% residing in urban areas.
Households reported various characteristics which helped in analysis. Majority of the households
(46%) were reported to be poor, 38% were reported to be rich while 16% were in the middle
category. They reported their sources of water to be piped water (17%), public tap (13%), open
well (39%), rainwater (2%), other sources (3%) and river/natural reserves (28%). Consequently50% used pit toilets, 14% used ventilated improved toilets (vip), 10% used flush toilet, while
25% reported they did not have any facility.
In regard to child characteristics, 53% of reported infants were male while 43% were female.
In this study, place of delivery and tetanus injection were used as health facility related variables.
Out of the respondent women 54% were reported to have delivered their children at home while
46% had delivered in a hospital either private or public. 84% had received tetanus injection while
pregnant while only 16% did not receive the injection.
Other important statistics are reported as below.
Table 4.6: Mothers age in relation to Infant death
age groups in
years
Number of infants
alive
Number of infant
dead
infant deaths per 1000 live births
(IMR)
15-19yrs 246 14 57
20-24yrs 741 33 45
25-29yrs 585 31 53
30-34yrs 418 27 6535-39yrs 218 13 60
40-44yrs 73 7 100
45-49yrs 17 2 117.6
27
-
7/30/2019 Final Project Edited(3)
38/52
The mean age of interviewed women was 26years with the minimum age in child bearing
reported as 15 years. According to Pandey et al. (1998) study very young and very old women
had high risks of child and infant deaths. This study supports these findings as shown in table 4.3
where women aged15-19 years, 40-44 years and 45-49 years had high infant deaths of 57, 100
and 117 per 1000 live births respectively as compared to 45 and 53 infant deaths per 1000 live
births among women aged 20-24years and 25-29 years respectively.
Table 4.7: Mothers Education in relation to Infant death
Mothers Education
(MEDU)
Number of infants
alive
Number of infant
dead
infant deaths per 1000 live
births (IMR)
No education 456 22 48.2
Primary Education 1,317 78 59.2
Secondary Education 407 21 51.6
Higher Education 118 6 50.8
This study used mothers education to assess health knowledge. In contrast to literature findings
women with no education had the lowest infant mortality rate (48%) followed by women with
higher education (50.8%), women with secondary education (51.6%) and the highest infant
mortality rate was reported among women with primary education.
28
-
7/30/2019 Final Project Edited(3)
39/52
Table 4.8: Other independent variables in relation to infant mortality
Variable
Value
names
Number of
infants alive
Number of
infant dead
infant deaths per 1000 live
births (IMR)
MAREL No religion 82 4 48.8
Has religion 2216 123 55.5
TORE Rural 1723 89 51.7
Urban 575 38 66.1
HW Middle 355 25 70.4
Poor 1067 50 46.9
Rich 876 52 59.4
ATW Piped Water 377 27 71.6
Public Tap 287 17 59.2
Open well 907 40 44.1
Rainwater 34 1 29.4Other 48 5 104.2
River/Natura
l reserves 645 37 57.4
ATS No facility 566 23 40.6
Flush toilet 230 16 69.6
VIP 332 16 48.2
Pit toilet 1150 71 61.7
Other 20 1 50
G Female 1089 56 51.4
Male 1209 71 58.7PLOD Home 1255 58 46.2
Hospital 1042 65 62.4
BSIZE Average 1216 50 41.1
Small/Very
small 380 30 78.9
Large/Very
large 684 42 61.4
From table 4.8 infant mortality rate was reported to be high among women with religion (55%),
those women who resided in urban areas (66%), women from household reported to be in the
middle wealth quantile (70%), those who used piped water (71%), those who used flush toilets
(69%) and those women who delivered in hospital. In regard to child characteristics infant
mortality rate was reported to be high among male infants and infants born with small or very
small birth size.
29
-
7/30/2019 Final Project Edited(3)
40/52
On the other hand infant mortality rate was reported to be low among women with no religion,
women residing in rural areas, those who delivered at home, those from poor households and
those from households that used rainwater. In regard to child characteristics infant mortality rate
was reported to be low among female infants and large infant birth size.
4.3 Econometric analysis
The study estimated the relationship between infant mortality and various determinants; maternal
characteristics, household characteristics, infants biological and health endowment as well as
health service variables.
The study examined how well the model fitted the data in order to obtain meaningful results. In
terms of likelihood ratio statistics, all the models reported passed the goodness of fit. This
implies that for every one of these models, there is at least one variable that is not equal to zero.
This means that the dependent variable (infant mortality) is better explained by at least some of
the independent variables than the constant alone.
The study estimated the model using LPM and Logit. However only logit results are presented
and discussed. In order to assess effect of health knowledge on infant mortality a second model
was estimated while omitting mothers education so as to check the change on the significance,
sign and effect of the remaining variables on infant mortality. This followed Medrano et al.,
(2000) study.
30
-
7/30/2019 Final Project Edited(3)
41/52
Table 4.9: Results from Logit model for Infant mortality in Kenya.
Model One Model Two
Coef. Std Error Coef. Std Error
Mage 0.168 0.129 Mage 0.176 0.127
Magesq-0.004* 0.002 Magesq
-0.004* 0.002
Tno 0.149 0.102 Tno 0.147 0.101
Married 0.17 0.302 Married 0.175 0.302
Primary Education -0.22 0.358
Secondary Education -0.026 0.447
Higher Education -0.11 0.604
Has religion 0.134 0.639 Has religion 0.065 0.63
Urban 0.201 0.334 Urban 0.201 0.333
Poor 0.598* 0.311 Poor 0.617* 0.309
Rich 0.25 0.336 Rich 0.266 0.333
Public Tap 0.063 0.411 Public Tap 0.054 0.411
Open well 0.509 0.363 Open well 0.489 0.361
Rainwater 0.671 1.068 Rainwater 0.697 1.066
Other -0.832 0.546 Other -0.821 0.542
River/Natural reserves 0.015 0.384 River/Natural reserves -0.002 0.383
Flush toilet -0.003 0.545 Flush toilet -0.032 0.531
Vip 0.082 0.434 Vip 0.054 0.429
Pit toilet -0.219 0.341 Pit toilet -0.266 0.328
Small/Very small size
-
0.651* 0.287 Small/Very small size
-
0.641* 0.287
Large/Very large size
-
0.449* 0.242 Large/Very large size
-
0.456* 0.242
First order 0.536 0.386 First order 0.548 0.383
Above 3 birth order -0.334 0.377 Above 3 birth order -0.354 0.374
Male -0.081 0.214 Male -0.081 0.214
Less than 24 months -0.292 0.276 Less than 24 months -0.291 0.276
Hospital -0.237 0.253 Hospital -0.229 0.251
Immunized
-
0.135* 0.332 Immunized
-
0.155* 0.33
_cons 1.116 2.126 _cons 0.939 2.087
Number of observations 2265
Number of
observations 2265
Prob > chi2 0.0636 Prob > chi2 0.0337
* signifies parameter was statistically significant at the 10% level
31
-
7/30/2019 Final Project Edited(3)
42/52
In model one, the independent variables used include, mothers age, mothers age squared, total
number of children, marital status, religion, education, households type of residence, wealth,
access to sanitation, sources of water, tetanus injection, place of delivery, birth size, birth order
and birth spacing of the infant. The second model omits education.
Table 4.10: Marginal effects from logit models
Marginal effects after
Logit
y = Pr(L) (predict)= 0.96416021
Marginal effects after
logit
y = Pr(L) (predict)= .96401291
Variable dy/dx Std Error dy/dx Std Error
MAGE 0.006 0.004 MAGE 0.006 0.004
MAGESQ 0.000* 0.000 MAGESQ 0.000* 0.000
TNO 0.005 0.004 TNO 0.005 0.003
Married 0.006 0.012 Married 0.006 0.012
Primary Education -0.007 0.012 Primary Education
Secondary Education -0.001 0.016 Secondary Education
Higher Education -0.004 0.023 Higher Education
Has religion 0.005 0.025 Has religion 0.002 0.023
Urban 0.007 0.011 Urban 0.007 0.011
Poor 0.020* 0.011 Poor 0.021* 0.011
Rich 0.008 0.011 Rich 0.009 0.011
Public Tap 0.002 0.014 Public Tap 0.002 0.014
Open well 0.017 0.011 Open well 0.016 0.012Rainwater 0.017 0.020 Rainwater 0.018 0.020
Other -0.042 0.038 Other -0.041 0.038
River/Natural reserves 0.001 0.013 River/Natural reserves 0.000 0.013
Flush toilet 0.000 0.019 Flush toilet -0.001 0.019
VIP 0.003 0.014 VIP 0.002 0.014
Pit toilet -0.008 0.012 Pit toilet -0.009 0.011
Small/Very small size
-
0.028* 0.015 Small/Very small size
-
0.028* 0.015
Large/Very large size
-
0.017* 0.010 Large/Very large size
-
0.017* 0.010
First order 0.016 0.010 First order 0.017 0.010Above 3 birth order -0.012 0.014 Above 3 birth order -0.013 0.014
Male -0.003 0.007 Male -0.003 0.007
Less than 24 months -0.011 0.011 Less than 24 months -0.011 0.011
Hospital -0.008 0.009 Hospital -0.008 0.009
Immunized
-
0.004*
0.011 Immunized
-
0.005* 0.010
32
-
7/30/2019 Final Project Edited(3)
43/52
* signifies parameter was statistically significant at the 10% level
4.4 Discussion of Results
Mothers age squared
An increase in the age of the mother by one year increases the probability of infant mortality by
0.014 percent. The impact is the same when controlling for health knowledge using mothers
education. These findings concur with earlier expectations on the effect of mothers age on infant
mortality. It is expected that very young mothers may experience difficult pregnancies and
deliveries because of their physical immaturity. They are also likely to have limited knowledge
and confidence in caring for infants. Similarly, women who are very old may also experience age
related problems during pregnancy and delivery. Similar results were also found by Madise et al
(2003) in Zambia. Our findings support literature findings but contradicts Medrano et al. (2000)
findings where educated mothers were expected to be more knowledgeable and thus to have low
infant mortality.
Poverty
An increase in the proportion of women from households classified as poor relative to
households classified as middle class by 1 percent would increase the probability of infant
mortality by 2.05 percent. The impact of poverty increases to 2.12 percent when the model is
controlled for health knowledge. This could imply that even though households could be poor,with knowledge they may come up with ideas to help them improve their living condition and
thus reduce infant mortality. The findings support our apriori expectations where wealthy
households are expected to have low infant mortality.
Birth size
The size of the baby at birth, either small or large, as reported by the mother is used in the study
as a proxy for the childs birth weight since most of the children did not have birth weights.
Infants reported born of small/very small size and those reported born of large/very large size
had a 2.79 percent and 1.69 percent lower probability of infant deaths as compared to infants
reported born of medium size. These findings contradict the expected results as infants born with
small and large sizes were expected to have high mortality risk as discussed by Elhamadi (2003).
Nevertheless other reserchers like Pandey, (1998) had reported similar results in India.
33
-
7/30/2019 Final Project Edited(3)
44/52
Immunization
Mothers immunized against tetanus while they were pregnant had a 0.45 percent reduced risk of
having their children die before celebrating their first birthday. The impact was higher (0.51
percent) where health knowledge was not considered.
34
-
7/30/2019 Final Project Edited(3)
45/52
CHAPTER FIVE: SUMMARY, CONCLUSIONS AND POLICY
IMPLICATIONS
5.1 Introduction
This chapter outlines the summary on research findings, conclusion of the study as well as some
policy recommendations with limitations of the study and suggestions for future research.
5.2 Research Summary
The major objective of this study was to establish empirically the determinants of infant
mortality in Kenya. This was motivated by the fact that despite the reported decline in infant
mortality the level was still high as per the MDG standard. Similarly infant mortality rate
illustrates the lasting impacts of childhood health into adulthood and will thus impact economic
performance of a country.
The objectives of the study were to investigate whether demographic, socioeconomic,
environmental and child health characteristics were important in explaining infant mortality in
Kenya. This study was relevant as it would add to the existing knowledge in literature on infant
mortality in Kenya.
The study used individual household level data from the 2008 Kenya demographic and healthsurvey. The logistic regression model was used due to the categorical nature of the dependent
variable and he marginal effects were obtained using STATA statistical package. The results are
given in chapter four and marginal effects were discussed.
5.3 Conclusion
From the analysis the study found out that mothers age was a key determinant to infant health. It
did not matter whether the mother was knowledgeable or not as the impact was the same. As a
woman ages their chances of having their children not celebrating their first birthday were higher
by 0.014 percent. Therefore its important to encourage women to give birth at the middle ages
where infant mortality rate risk was lower.
35
-
7/30/2019 Final Project Edited(3)
46/52
The study found that households classified as poor had high risks of infant deaths as compared to
the middle household category. Household wealth explains a lot in regard to households living
standard in terms of education attained, ability to pay for hospital visits and use of better
sanitation. Therefore there is need to improve households wealth so as to improve child survival.
We found that education is very important in explaining infant survival. In this study education
was used to assess health knowledge of mothers. Mothers with higher education were less likely
to have their infant die before celebrating their first birthday as compared to mothers with no
education. Similarly when education was included in the model the impact of the other variables
on infant mortality was lower as compared to their impacts when education was omitted. Thus
education is key to ensuring women are knowledgeable in terms of infants health during
pregnancy and afterwards. This is important as it determines infants survival.
Finally the study established lower infant mortality rate among women who had been immunized
against tetanus while they were pregnant. Similarly large and small sized infants were found to
have lower risk of infant death as compared to the middle infants. These findings contradict
literature findings and could be as a result of the sample used. Majority of the interviewed
women were from rural areas where belief on having a large sized infant signifies health could
be rampant. Also the categorization was based on a womans judgment and thus the results
would have been biased as size could be subjective.
5.4 Policy Implications based on key Results
The findings of the present study indicate that infant mortality risk is high as a woman ages.
Therefore this study recommends that women be encouraged to bear children between 20 and 34
years of age and to have fewer children so as they get enough time to take care of the children.
This can be done through making family planning services accessible to all especially in rural
areas.
Poverty has been found to be a great determinant of infant mortality in Kenya. The study found
that majority of the respondent women were poor and lived in rural areas. There is need therefore
to come up with programmes to assist women get financial empowerment. For instance the
government can encourage commercial farming by promoting infrastructure development,
opening up markets for various farm products by farmers, offering credit facilities at lower rate
36
-
7/30/2019 Final Project Edited(3)
47/52
to farmers, selling farm inputs at subsidized prices and also educating farmers on new farming
techniques that will yield more output. Similarly women need be encouraged to join hands and
mobilize resources for development.
It was also noted from the results that the birth size of the infant is important in determininginfant mortality in Kenya. The focus here could be directed on promoting healthy lifestyle for
pregnant woman to take care of factors that could contribute to birth weight. This could be in the
area of nutrition, exercise and avoidance of cigarette smoking.
Similarly the results have shown that women with education are more knowledgeable and
experience lower infant mortality than those with no primary education. With the inception of
free primary education many children are acquiring primary education but are unable to progress
to secondary due to factors like lack of fees despite the government subsidizing the fees in
secondary school. In this regard the government could introduce a policy of providing free
secondary education so that more females are motivated to go further with their studies which
may in the long run result in reduction of infant mortality.
Finally the study observed that women who had been immunized when pregnant had lower risk
of encountering infant death. Thus its important to encourage women to attend ante natal clinic
regularly for checkup and for general education.
5.5 Limitations of the study
This study aimed to inform policy regarding the determinants of infant mortality in Kenya. It was
limited to the data already collected which was collected in the year 2008. This is the most recent
available data and thus may not give up to date information. This was as a result of limited time
to carry out the study and the amount of resources one may require to collect such data.
Similarly the data had missing values which affected the quality of output. This was as a result of
recall problem among interviewed women and biasness in reporting especially in the area ofbirth size. This could have been as a result of misplaced birth records, misreporting of deaths and
omissions of birth reporting for the children who die very young.
37
-
7/30/2019 Final Project Edited(3)
48/52
5.6 Suggestions for further study
In the light of the above shortcomings, a more comprehensive study would be required that
would investigate the determinants of infant mortality by estimating two different models for
neonatal, infant and child mortality capturing all the determinants of infant mortality including
the ones outlined in this section . Recent collected data would also be appropriate. This would
give to date information and would help capture the impacts of various variables in the children
categories as what could affect infants would be different from what affects under five children.
This will help inform policy effectively.
38
-
7/30/2019 Final Project Edited(3)
49/52
REFERENCES
Ajakaiye, O. and Mwabu, G. (2007), The demand for reproductive health services: An
application of control function approachCollaborative Project on Reproductive Health,Economic Growth and Poverty Reduction in Africa, AERC, Nairobi.
Alves, D. and Belluzzo, W. (2005), Child health and infant mortality in Brazil. Paper NO.1
Seminar on Child Health, Poverty and the Role of Public Policies. InteramericanDevelopment Bank, Washington DC.
Beenstock, M. and Sturdy, P. (1990), The determinants of infant mortality in Regional InIndia, World Development; Vol. 18 No. 3:443-453.
Bicego, G., and Ahmad, B. (1996). Infant and child mortality. Demographic and Health
Surveys, Comparative Studies No. 20. Macro International Inc., Calverton, Maryland.
Caldwell, J. (1979), Education as a Factor in Mortality Decline: An Examination of Nigerian
Data.Population Studies; 33(3):395-413
Claeson, M., Mawji, T. and Pathmanathan, I (2000), Reducing child mortality in India in the
new millennium. Bulletin of the World Health Organization 78:1192- 1199.
Curtis, S. (1995), Assessment of the quality of data used for direct estimation of infant and child
mortality in DHS-II surveys. DHS Occasional Papers No. 3. Macro International Inc.,Calverton, Maryland.
Da vanzo, J. Razzaque, A. Rahman, M. and Hale, L. (2004), The effects of birth spacing oninfant and child mortality, pregnancy outcomes, and maternal morbidity and mortality in
Matlab, Bangladesh working paper No.198, RAND Labor and Population workingpaper series.
Desai, S. and Aava, S. (1998), Maternal education and child health: Is there a strong causalrelationship?Demography 35:71-81.
Elmahdi, H. (2008), Socioeconomic Determinants of infant mortality in Kenya: Analysis of
Kenya DHS 2003.Journal of humanities & social sciences Volume 2, Issue 2, 2008
Fayehun, A. (2010), Household Environmental Health Hazards and Child Survival in Sub-
Saharan Africa. DHS Working Papers No. 74. Calverton, Maryland, USA: ICF Macro.
Foloko, N. (2009), Determinants of Child Mortality in Lesotho Unpublished Masters of Arts
Research Paper .University of Nairobi
Government of Kenya, (2010), Kenya Health Policy Framework (1994 2010): Analysis of
Performance. Analytical Review of Health Progress, and Systems performance Ministryof Medical services and Public health publications
39
-
7/30/2019 Final Project Edited(3)
50/52
Government of Kenya, (2010), Draft progress in attainment of MDGs and way forward towards
achieving MDGs. UNDP report.
Gujarati, D. (2007), Basic Econometrics. Fourth Edition. Published by Tata McGraw Hill
education private limited.New Delhi
Hill, K. (2003), Frameworks for studying the determinants of child survival bulletin of the
World Health Organization 2003, 81 (2)
Hill, K., Bicego, G. and Mahy, M. (2001), Childhood Mortality in Kenya: An Examination oftrends and determinants in the late 1980s to mid 1990s.
Hobcraft, J. (1993), Womens education, child welfare and child survival: a review of the
evidence. Health Transition Review; 3(2):159-173.
Hosseinpoor, A., Mohammad, R., Majdzadeh, M., Naghavi, F., Abolhassani, A., Sousa, N.,
Speybroeck, H., Jamshidi, R. and Vega J. (2005), Socioeconomic inequality in infant
mortality in Iran and across its provinces. Bulletin of the World Health Organization;83(11):837-844.
Jones, G., Schultink, W. and Babille, M. (2006), Child Survival in India Indian Journal ofPediatrics, Volume 73June, 2006 pg 160-325
Kabubo-Mariara, J. Karienyeh, M and Kabubo, F. (2012), child survival, poverty and inequality
in Kenya: does physical environment matter?African Journal of Social Sciences 2(1):
65-84
Kamau, D. (1998), Child Survival determinants in the arid and semi arid lands. A study of
Machakos, Kilifi and Taita Taveta district. Unpublished masters of Arts researchpaper .University of Nairobi
Kenya National Bureau of Statistics (KNBS) and ICF Macro (2010),Kenya Demographic and
Health Survey 2008-09. Calverton, Maryland: KNBS and ICF Macro
Koenig, A. and Gillion, H. (1992), Patriarchy, womans status and Reproductive behaviour in
rural North India.Demography India 21:145-66.
Kovsted, J. Portner, C. and Tarp, F. (2003), Child Health and Mortality: Does Health
Knowledge Matter?Journal of African Economies, Vol. 11, No. 4, pp. 542-560.
Madise, N. (2003), Infant Mortality in Zambia: Socioeconomic and demographic correlates.Social Biology 50(1-2): 148-166.
Medrano, P., Rodriquez, C. and Villa, E. (2000), Does Mothers Education Matter in ChildsHealth? Evidence from South Africa.
Millard, A. (1994), A causal model of high rates of child mortality. Social Science andMedicinevol38:253-268.
40
-
7/30/2019 Final Project Edited(3)
51/52
Mosley, W. and Chen, L. (1984), An analytical framework for the study of child survival in
devel