Abstract
This chapter analyses multidimensional aspects of child poverty in Kenya. We carry out poverty and inequality comparisons for child survival and also use the parametric survival model to explain childhood mortality using DHS data. The results of poverty comparisons show that: children with the lowest probability of survival are from households with the lowest level of assets; and poverty orderings for child survival by assets are robust to the choice of the poverty line and to the measure of well-being. Inequality analysis suggests that there is less mortality inequality among children facing mortality than children who are better off. The survival model results show that child and maternal characteristics, and household assets are important correlates of childhood mortality. The results further show that health-care services are crucial for child survival. Policy simulations suggest that there is potential for making some progress in reducing mortality, but the ERS and MDG targets cannot be achieved.
JEL Classification J13, I12, I32, I38, D63
This is a preview of subscription content, log in via an institution to check access.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
Notes
- 1.
Functionings are the “beings and doings” of a person, whereas capabilities are the various combinations of functionings that a person can achieve. Capability is thus a set of vectors of functionings reflecting the person’s freedom to lead one type of life or another (Sen, 1985). In terms of mortality, capabilities would embrace the ability to live through to mature age whereas the equivalent measure of functioning would be the mortality rates.
- 2.
- 3.
See Duclos et al. (2006a) for important features and conditions of bi-dimensional dominance surface.
- 4.
Schultz (1984) has shown that the best approach would be to estimate both the demand equations for health inputs and the production function linking health inputs to child survival by simultaneous structural equation methods. This however requires that there are accurate data on prices, wages, programs, and environmental conditions, which can be used as exogenous instruments. In the absence of appropriate instruments, the reduced-form equation for child survival may be estimated without imposing a great deal of structure on the problem, as is commonly done in the literature.
- 5.
The survey interviews women between the ages of 15 and 49 today, so the information it has on births and deaths (say 15 years ago) is for women who were 0–34 years old at that time, and the estimated mortality rates for the earlier period would be biased upward because the sample of mothers is younger. At the same time, there are some mothers who were less than 15 but no information was collected on them. So an estimate of current mortality will be biased downward. Another possible bias arises from the fact that some women would have died in the years between the date for which a mortality rate is to be estimated and the survey date, as there is no information about them in the survey. If these women’s infants were more likely to die than other infants then the mortality rates estimated for years prior to the survey will be biased downward (see Mosley, 1984; Ssewanyana, 2005). One problem of analyzing retrospective birth histories is the quality of information: misplacement of dates of birth is always possible, and so is misreporting of death as well as omissions of birth reporting of children who die very early in life. These problems are however more likely to bias neonatal and infant mortality, but not under-five mortality (Hobcraft et al., 1984). We investigate for possible biases in the data using education and mother’s heights by women’s birth cohort, but uncover no evidence of any of these problems (this is presented in a separate appendix and is available from the authors or from PEP).
- 6.
Survival analysis is a technique for analyzing time to event or failure data. It helps to model the risk of failure or the probability of experiencing failure (hazard) at time t +1 given that the subject is at risk at time t. The higher the hazard, the shorter the survival. In the case of mortality, the survival time of a child is a continuous non-negative random variable T with a cumulative distribution function \(F(t)\), and probability density function \(f(t)\). The survivor function is defined as \(S(t) \equiv 1 - F(t)\), the probability of being alive at time t.
If we let the cumulative density function for the failure function to be \(\Pr (T \leq t) = F(t)\), then the survivor function can be estimated as:
$$S(t) \equiv \Pr (T > t) = 1 - F(t).$$((17))The Weibull model is parameterized as both a proportional hazard and accelerated failure-time model. It is suitable for modeling data with monotone hazard rates that either increase or decrease exponentially with time. The proportional hazard model for Weibull regression is specified as:
$$\theta(t\!:\!X) = \alpha t^{\alpha - 1}\exp (\beta 'X) = \alpha t^{\alpha - 1}\lambda,$$((18))where \(\lambda \equiv \exp (\beta' X), \alpha \) is the shape parameter to be estimated from the data and exp(.) is the exponential function. The hazard rate either rises monotonically with time (\(\alpha > 1\)), falls monotonically with time (\(\alpha < 1\)), or is constant (\(\alpha = 1\)). The last case is the special case of the Weibull model known as the exponential model. For a given value of α, larger values of λ imply a larger hazard rate at each survival time. Like other probabilities, the survivor function lies between zero and one, and is a strictly decreasing function of t. The survivor function is equal to one at the start of the spell (\(t = 0\)) and is zero at infinity: \(0 \leq S(t) \leq 1\). The density function is non-negative but may be greater than 1 in value, i.e., \(F(t) \geq 0\).
- 7.
To save on space, this analysis is based on under-five mortality only. The advantage of using under-five mortality is that duration models allow us to explore the determinants of probability of a child dying before the fifth birthday without losing any information. Yet, because most deaths occur before the first birthday (neonatal and infant mortality), this analysis allows us to make inferences about the correlates of infant mortality as well. For this reason, we use the term childhood mortality or child survival in the discussion of all results.
- 8.
In this chapter, the asset index is used as a proxy for household resources or income. These two are measures of household welfare and may therefore be seen as indicators of household wealth and well-being. For this reason, some of these terms may be used interchangeably in the chapter depending on the context.
- 9.
The asset index can be defined as: \(A_i = \sum\limits_k {\tau_k \alpha_{ik}}\), where A i is the asset index for household i, the \(\alpha_{ik}\)’s are the k individual assets recorded in the survey for that household, and the τ k ’s are the weights. Most studies use the standardized first principal component of the variance–covariance matrix of the observed household assets as weights, allowing the data to determine the relative importance of each asset based on its correlation with the other assets, following Filmer (2000). This study uses the factor analysis approach to derive the index, following Sahn (2000, 2003). This approach is similar to principal components but has certain statistical advantages and assumes that the one common factor that best explains the variance in the ownership of a set of assets is the measure of economic well-being (Sahn, 2000). The assets that are included in the analysis are ownership of a radio, TV, refrigerator, bicycle, a motorcycle, a car, the household’s source of drinking water (piped or surface water relative to well water); the household’s toilet facilities (flush or no facilities relative to latrine facilities); the household’s floor material (low quality relative to higher quality); and the years of education of the household head (and of respondent if not the head) to account for household’s stock of human capital. The scoring coefficients from the factor analysis are applied to each household to estimate its asset index and will rank the households on a –1 to 1 scale. To avoid arbitrary assignment of weights to the variables, we rely on the factor loadings results for weights (see Table 1 for weights so generated).
- 10.
Cluster- and district-level shares are used instead of the individual responses to adjust for design effects in the survey and also to control for endogeneity of individual level data on service use. Endogeneity of these variables spring from the fact that they depend on household characteristics among other factors and may also be jointly determined with other factors that affect mortality. For some variables, it is convenient to use cluster averages because of decentralization of service delivery (e.g., family planning clinics) but for other variables, delivery is less decentralized (e.g., medical professionals) and so these variables are measured at the district level.
- 11.
The probability of child survival is defined as 1 minus the number of children that faced mortality weighted by the total number of children aged 0 to 60 months born to a household. We could as well use the predicted probabilities of survival from a childhood mortality regression, with the only difference being that the latter smoothens out the expected probability of survival. Both approaches gave us comparable results. Recall from the methodology that childhood mortality is measured through under-five mortality.
- 12.
Recall that dominance in poverty refers to the distribution that is better in well-being. There are two steps in testing for the statistical robustness of stochastic dominance (see Araar, 2007). The first step is to draw the difference between FGT curves. This is a necessary but not a sufficient test because it can only tell us whether there is stochastic dominance or not. The second step is to test whether the upper bound of the difference between poverty orderings is anywhere below zero at a desired confidence interval. If both conditions/tests are satisfied, the stochastic dominance is statistically robust. If the first is satisfied and the second is not, there is stochastic dominance but the difference is not statistically significant. Our results show that the stochastic dominance of urban over rural areas is statistically robust. There is stochastic dominance of Central over all other provinces except Nairobi and Coast provinces, but this dominance is not statistically robust.
- 13.
Inequality here is measured by the absolute Gini index. One can recall here that in the presence of negative values in the index of well-being, one cannot use the usual Gini index (See Araar 2006).
- 14.
Overlap refers to the cases where childhood mortality may be the same but using the asset index ranks the children into different wealth groups. For instance, some less poor children may be from households that experienced more deaths than poorer children. When decomposing inequality by area of residence, the overlap would refer to children ranked as less poor in one area but would be ranked poorer if the mean level of the wealth measure in another area is considered.
- 15.
Studies on fertility argue that multiple births may not be purely exogenous because women with high fertility are more likely to experience a twin birth. For this reason, such studies recommend the use of first-born twins or ratio of twins to total births as experiments to instrument fertility. There is no evidence that twins may be endogenous to mortality. Either way, we try to use twins at first birth and the results are robust with the use of any multiple birth, except for urban areas where the number of observed first born twins is too small.
- 16.
Examination of possible collinearity between the environmental variables and other explanatory variables suggests however that there is no serious correlation problem as all the correlation values are less than 0.5 (see separate appendix accompanying this report).
- 17.
If some women who died between a given year and the survey date, and were thus not surveyed, also had children more likely to die (most frail children could have died before the survey), then their exclusion from the sample would cause a downward bias in the estimated mortality rate for years before the sample, and that bias should be greater with longer lags (Strauss, 1995; Ssewanyana, 2005).
- 18.
The Economic Recovery Strategy Paper (ERS) was designed in 2003 to implement the Poverty Reduction Strategy Policies and the Government’s development agenda to restore economic growth and reduce poverty through employment and wealth creation.
- 19.
A separate appendix of this paper reviews the conceptualization of child poverty and recent policy interventions for child welfare and survival in Kenya.
- 20.
Since the estimated model is nonlinear, we carry out simulations for each observation in the estimation sample and then find the average impact on the mortality risk. If we define the mortality hazard as \(Y = \alpha + \beta X\), the new mortality hazard resulting from a change in X is given by \(Y^{1} = \alpha + \beta X^{1}\), where a Y 1 is the new probability of death resulting from a change in the policy variable from X to X 1.
References
Araar, A. 2006. The absolute Gini coefficient: Decomposability and Stochastic Dominance. Mimeo PEP and CIRPÉE.
Araar, A., and J.-Y. Duclos. 2007. DASP: Distributive analysis stata package. Laval University, World Bank, PEP and CIRPÉE.
Araar, A. 2007. Poverty, inequality and stochastic dominance, theory and practice: The case for Burkina Faso. PMMA Working Paper 2007-08. PEP.
Atkinson, A.B. (1987). On the measurement of poverty, Econometrica, 55:749–64.
Becker, G. 1965. A theory of the allocation of time. Economic Journal 75:493–517.
Duclos, J.-Y., and A. Araar. 2006. Poverty and equity: Measurement, policy and estimation with DAD. Boston/Dordrecht/London: Spinger/Kluwer.
Duclos, J.-Y., D.E. Sahn, and S.D. Younger. 2006a. Robust multidimensional poverty comparisons. The Economic Journal 116:943–968.
Duclos, J.-Y., D.E. Sahn, and S.D. Younger. 2006b. Robust multidimensional spatial poverty comparisons in Ghana, Madagascar, and Uganda. World Bank Economic Review 20(1):91–113.
Duclos, J.-Y., D.E. Sahn, and S.D. Younger. 2006c. Robust multidimensional spatial poverty comparisons with discrete indicators of well-being. Working Paper 06-28 CIRPÉE.
Filmer, D., and L. Pritchett. 2000. Estimating wealth effects without expenditure data- or tears: An application of educational enrollment in states of India. Mimeo. The World Bank, Washington, DC.
Filmer D., and L. Pritchett. 1997. Child mortality and public spending on health: How much does money matter? Policy research working paper No. 1864. The World Bank Washington, DC.
Foster, J.E., and A.F. Shorrocks. 1988. Poverty orderings and welfare dominance. Social Choice Welfare 5:179–198.
Foster, J.E., J. Greer, and E. Thorbecke. 1984. A class of decomposable poverty measures. Econometrica 52(3):761–776.
Glick, P., S.D. Younger, and D.E. Sahn. 2006. An assessment of changes in infant and under-five mortality in demographic and health survey data for Madagascar. CFNPP working paper 207.
Hill, K., G. Bicego, and M. Mahy. 2000. Childhood mortality in Kenya: An examination of trends and determinants in the late 1980s to mid 1990s. HPC working papers WP01_01. John Hopkins University, Baltimore.
Hobcraft, J.N., J.W. Mcdonald, and S.O. Rustein. 1984. Socio-economic factors in infant and child mortality: A cross-national comparison. Population Studies 38(2):193–223.
Kabubo-Mariara, J., G.K. Nd’enge, and D.M. Kirii. 2006. Evolution and determinants of non-monetary indicators of poverty in Kenya: Children’s nutritional status, 1998–2003. Mimeo, African Economic Research Consortium, Nairobi.
Macro International, Demographic and Health Surveys, Accessed on 20 July 2006. http://www.measuredhs.com/countries/country.cfm? ctry_id=20.
Matteson, D.W., J.A. Burr, and J. R. Marshall. 1998. Infant mortality: A multi-level analysis of individual and community risk factors. Social Science and Medicine 47(11):1841–1854.
Ministry of Health. 2005. The second national health sector strategic plan of Kenya (NHSSP II 2005-10): Reversing the trends. Mimeo, health sector reform secretariat, Nairobi.
Mosley, W.H., and L.C. Chen. 1984. An analytical framework for the study of child survival in developing countries. Population and Development Review 10(Supplement):25–45.
Muhuri, P.K., and S.H. Preston. 1991. Effects of family composition on mortality differentials by sex among children in Matlab, Bangladesh. Population and Development Review, 17(3):415–434.
Paxton, C., and N. Schady. 2004. Child health and the 1988–1992 economic crisis in Peru. policy research working paper No.3260. The World Bank Washington, DC.
Pitt, M., M. Rosenzweig, and D. Gibbons. 1995. The determinants and consequences of the placement of government programs in Indonesia. In Public spending and the poor: Theory and evidence, eds. Dominque van de Walle and Kimberly Nead. Johns Hopkins University Press.
Rutstein, S.O. 2000. Factors associated with trends in infant and child mortality in developing countries during the 1990s. Bulletin of the World Health Organization 78(10):1256–1270.
Republic of Kenya. 2004. Investment programme for the economic recovery strategy for wealth and employment creation, 2003–2007.
Sahn, D., and D. Stifel. 2000. Poverty comparisons over time and across countries in Africa. World Development 28(12):2123–2155.
Sahn, D., and D. Stifel. 2003. Exploring alternative measures of welfare in the absence of expenditure data. Review of Income and Wealth 49(4):463–489.
Schultz, T.P. 1984. Studying the impact of household economic and community variables on child mortality. Population and Development Review 10(Supplement):215–235.
Sen, A. 1985. Commodities and Capabilities. Amsterdam: North Holland.
Ssewanyana, S., and S. Younger. 2005. Infant mortality in Uganda: Determinants, trends and the millennium development goals. CFNPP working paper 186.
Strauss, J., and D. Thomas. 1995. Human resources: Empirical modeling of household and family decisions. In Handbook of Development Economics, Vol. 3, eds. J. Behrman and T.N. Srinivasan. Amsterdam; North-Holland.
Trussell, J., and C. Hammerslough. 1983. A hazards-model analysis of the covariates of infant and child mortality in Sri Lanka. Demography 20(1):1–26.
UNDP, Government of Kenya, and Government of Finland. 2005. MDGs status report for Kenya. Ministry of planning and national development, Nairobi.
Wang, L. 2002. Health outcomes in poor countries and policy options: Empirical findings from demographic and health surveys. Policy research working paper No. 2831. The World Bank Washington, DC.
Ware, H. 1984. Effects of maternal education, women’s roles, and child care on child mortality. Population and Development Review 10(Supplement):191–214.
Younger, S. 2001. Cross-country determinants of declines in infant mortality: A growth regression approach. CFNPP working paper 130.
Author information
Authors and Affiliations
Corresponding author
Editor information
Editors and Affiliations
Rights and permissions
Copyright information
© 2010 Poverty and Economic Policy (PEP) Research Network
About this chapter
Cite this chapter
Kabubo-Mariara, J., Karienyeh, M.M., Mwangi, F.K. (2010). Multidimensional Poverty, Survival and Inequality Among Kenyan Children. In: Cockburn, J., Kabubo-Mariara, J. (eds) Child Welfare in Developing Countries. Springer, New York, NY. https://doi.org/10.1007/978-1-4419-6275-1_2
Download citation
DOI: https://doi.org/10.1007/978-1-4419-6275-1_2
Publisher Name: Springer, New York, NY
Print ISBN: 978-1-4419-6337-6
Online ISBN: 978-1-4419-6275-1
eBook Packages: Business and EconomicsEconomics and Finance (R0)