Water accessibility and child health: Use of the leave-out strategy of instruments
Introduction
Studies in the economic literature have commonly used the leave-out strategy of instruments1 to address the potential endogeneity and possible measurement error of water accessibility on child health (see, for example, Ilahi and Grimard, 2000, Glick et al., 2004, Mangyo, 2007). It is therefore important to examine the conditions that determine the validity of the leave-out strategy of instruments. In this paper, we look at the potential weakness of this instrumentation strategy in examining the effect of water accessibility on child health.
The idea behind this instrument is that community-level access to clean water reflects past infrastructure projects conducted by government or non-governmental organizations (NGOs) rather than individual choice of water access by individual household. Thus, this instrumenting strategy is effective against self-selection of household into better access to clean water but is potentially subject to endogenous project placement. For example, unobserved community wealth may be positively correlated with both community-level access to clean water and child health, leading to an upward bias. For another example, unobserved pollution level of community may be positively correlated with community-level access to clean water (more urbanized areas may have better access to clean water) and negatively associated with child health, leading to a negative bias.
Ilahi and Grimard (2000) address the potential endogeneity of the availability of in-yard water by using the leave-out community ratio of households with in-house access to water in cross-sectional data from Pakistan. The focus of the study was on the relationship between access to water and time allocation of women, who had primary responsibility for water collection. They show that improvement in water-supply infrastructure would lower the total time women spend in all activities including hours in water collection. The examination of hours in water collection using the leave-out strategy in Madagascar and Uganda finds that feasible public investment would not lead to large reduction in water collection times or change the relative burden of overall work in men and women (Glick et al., 2004). They also find that an investment in water does not necessarily have a dramatic effect on female time use. In a panel data study, one of the limited studies in the economic literature, the use of leave-out community ratio as the instrument for change in in-yard water source finds that the access to in-yard water sources improved child health of well-educated mothers in China (Mangyo, 2007).
Behrman and Lavy (1997) have extensively studied the endogeneity of child health, measurement error, and the impact of unobserved fixed and choice inputs on child schooling in Ghana. They suggest that the true impact of child health on child schooling success is not significant and that returns to scale are limited despite the use of typical family and community instruments. Although they have shown the relationship between schooling and child health while addressing endogeneity, their study method can also be applied to a similar study on child health and water access.
Previous studies found a positive relationship between water access and child health. Many of them have examined the relationship between access to piped water and child health, and some of them have also explored the interaction between safe water and other household characteristics (see, for example, Cebu Study Team, 1991, Jalan and Martin, 2003, Esrey, 1996). Galiani et al. (2005), using panel data at the locality level, compare changes in health over time across before and after changes in water accessibility. They control for locality and time fixed effects and exploit within locality variability to identify the effect of water privatization on child health. They find that the privatization of water services and subsequent improvements of water infrastructure only affects water related deaths and not other non-water related deaths in Argentina. A study in Nigeria finds that water supply and sanitation projects have helped to improve weight-for-height but not for height-for-age for children under the age of three years (Huttly et al., 1990). Thomas and Strauss (1992) examine the relationship between parental characteristics, community characteristics, and child height in Brazil. They find that the availability of modern sewerage, piped water, and electricity has significantly affected child height, and the impact of mother's education on child height does not solely reflect resource availability. Esrey et al. (1985) analyzes the impact of water access on child health in five countries. They find no impact of water availability on height or weight in Bangladesh, and water access has a minor impact on weight-for-age in the Philippines. On the contrary, water availability has improved height-for-age and weight-for-age in Colombia. In Nigeria, improved water access has a positive impact on weight but adversely affected height.
The exploration of interaction of safe water and maternal education has found that height-for-age improved with improved water supplies for less well-educated mothers when this variable is interacted with maternal education (Barrera, 1990). Esrey (1996), using multivariate analysis with Demographic and Health Survey data, concluded that improved water quality can improve child height and weight only when sanitation is also improved. Esrey and Habicht (1988), using data from Malaysia, examine the impact of maternal literacy, piped water, and access to toilets on infant mortality. They find a synergistic relationship between maternal literacy and piped water. Butz et al. (1984) measures the impact of improved water and sanitation together with breast-feeding on infant mortality. They find that the presence of both improved water and sanitation has the biggest impact on reduced mortality in Malaysia. Jalan and Martin (2003) estimate the impact of piped water on child health in terms of the incidence and severity of diarrhoea in rural India. They find a lower incidence of diarrhoea among children living in piped water households. Using longitudinal data from household survey conducted in the Philippines, Cebu Study Team (1991) studies the impact of behavioral inputs with other exogenous factors on child health especially on diarrhoeal outcomes. The authors used family specific fixed effects to control for unobserved heterogeneity. The results show that population density, child age, exposure to contaminated water, fecal contamination, and rainfall increased the incidence of diarrhoea. However, none of them have addressed the endogenous nature of child health and water accessibility.
In this paper, we use the data from the Nepal Demographic and Health Survey 2006. In Nepal, the supply of improved drinking water has encouragingly increased since 1990. The water supply coverage reached about 81 percent in 2005, which drastically increased from the 46 percent reported in 1990 (NPC, 2005). However, there is still a problem of physical accessibility to a water source and sanitation. According to WaterAid Nepal (2004), only 42 percent of rural communities of Nepal have access to improved water supplies within return journey times of less than 15 min. Some households in hills have to spend as much as five hours per day for collecting water.
We assess the extent to which the community-level access to clean water is subject to endogenous project placement by utilising data on disease symptoms. If the correlation between child health and community-level access to clean water is at least partially due to endogenous project placement, such unobserved differences across communities not only affect water-relevant symptoms but also affect water-irrelevant symptoms. Here, water-relevant symptoms include diarrhoea, while water-irrelevant symptoms include fever and cough. For example, there is no reason to expect that unobserved community differences in wealth affect only the incidence of diarrhoea but not the incidence of fever and cough.
Our econometric results show that the leave-out community-level access to clean water appears to be correlated with unobserved differences across communities, implying that our instrument is not valid. Our econometric results find that community-level access to clean water is negatively correlated with the incidence of both water-relevant and irrelevant symptoms. One plausible story behind this finding would be unobserved community wealth is positively correlated with community-level access to clean water and negatively correlated with the incidence of both water-relevant and irrelevant symptoms. Thus, our econometric estimates of the impact of fetching water time on child health fails to pin down the true effect. However, our study presents a potential weakness of the leave-out instrumenting strategy which has been popular in the economic literature. Our study clearly shows that the leave-out community-level instruments are subject to endogenous project placement and could do more harm than good.
The remainder of this paper is organized as follows. In Section 2, we describe the dataset, descriptive statistics, and variables description. Section 3 presents the empirical specification. The results and discussion are present in Section 4 and we close with conclusions and recommendations in Section 5.
Section snippets
Data description
Analysis of data was based on 5783 children aged 0–59 months included in the Nepal Demographic and Health Survey (DHS), which was carried out from February to August 2006. The DHS collected demographic, socioeconomic, and health data and formed a nationally representative sample of 10,793 women aged 15–49 years in 8707 households included in the survey. The DHS sampling design2
Empirical specifications
The theoretical framework linking child health to access to clean water is conceptually similar to that of Thomas and Strauss (1992). The model is based on Becker (1981) in which a household maximizes a utility function. In this case, a household may be assumed to choose child health H, leisure L, and consumption of goods and services C.
The problem is:where Xh is a vector of household characteristics including educational level; μ unobserved
General results for OLS, community fixed effects, and IV estimates
Table 4, Table 5 report the results of regressing child health on the main variable of our interest, hours in fetching water time and other control variables. The z-scores of height-for-age and weight-for-age were used as health measures. Four regression models: (i) weighted ordinary least square (OLS), (ii) weighted community fixed effects, (iii) weighted two-stage least square (2SLS), and (iv) unweighted two-stage least square (2SLS) were computed for each health measure. The standard errors
Conclusions
This paper examined the effect of water accessibility in terms of fetching water time on child health, addressing the endogeneity of fetching water time. First, self-selection of wealthy households into convenient access to clean water would be a source of bias if no measures were taken. To deal with this problem, we used community-level access to clean water as instruments for household-level access to clean water. The idea behind this instrumental variable strategy was that community-level
Acknowledgements
We are grateful to Prof. Donghun Kim, Christopher Murphy, Sabina Shrestha, and two anonymous referees for their helpful comments. We would also like to express our thanks to Demographic and Health Survey data archive for providing us data and Noureddine Abderrahim for providing responses to our data inquiries.
References (27)
The role of maternal schooling and its interaction with public health programs in child health production
Journal of Development Economics
(1990)- et al.
The Imo state (Nigeria) drinking water supply and sanitation project, 2. Impact of drcunculiasis, diarrhoea and nutritional status
Transactions of the Royal Society of Tropical Medicine
(1990) - et al.
Does piped water reduce diarrhoea for children in rural India?
Journal of Econometrics
(2003) - et al.
Determinants of malnutrition in Senegal: individual, household, and community variables, and their interaction
Economics and Human Biology
(2008) - et al.
Model specification and endogeneity
Journal of Econometrics
(1998) - et al.
Price, infrastructure, household characteristics, and child height
Journal of Development Economics
(1992) A Treatise on the Family
(1981)- Behrman, J.R., Lavy, V., 1997. Child health and schooling achievement: association, causality and household...
- et al.
Problems with instrumental variables estimation when the correlation between the instruments and the endogenous explanatory variable is weak
Journal of the American Statistical Association
(1995) - et al.
Environmental factors in the relationship between breast-feeding and infant mortality: the role of water and sanitation in Malaysia
American Journal of Epidemiology
(1984)
Underlying and proximate determinants of child health: the Cebu longitudinal health and nutrition study
American Journal of Epidemiology
Water, waste, and well-being: a multicountry study
American Journal of Epidemiology
Interventions for the control of diarrhoeal diseases among young children: improving water supplies and excreta disposal facilities
Bulletin of the World Health Organization
Cited by (11)
Drinking water facilities and inclusive development: Evidence from Rural China
2024, World DevelopmentHousehold welfare effects of ROSCAs
2023, World DevelopmentEditorial Cost-Benefit Analysis to Inform Water Policy Debate
2023, Water Economics and PolicyDoes piped water improve adolescent health? Empirical evidence from rural China
2021, Review of Development Economics