The effect of water accessibility on child health in China

Abstract

Using a dynamic panel model of child anthropometrics from China, the effect of an in-yard water source on child health was measured. Changes in within-community averages of household access to in-yard water were used as the instrument for changes in access to in-yard water sources. Further, to address the concern of non-random placements of water projects, correlations between changes in disease symptoms and community-level changes in access to in-yard water sources were examined. It was found that access to in-yard water sources improved child health only when mothers were relatively well educated.

Introduction

This paper is concerned with physical proximity of a drinking water source to the dwelling and measures its effect on child health. Specifically, the effect of an in-yard water source on child health was examined. One plausible mechanism through which access to an in-yard water source affects child health is the improved quality of water, and thus, the reduced exposure to germs. If a source of clean water is physically proximate to the dwelling, people may obtain fresh water more frequently, reducing the duration of water storage, which, in turn, decreases the chance of bacterial growth. Alternatively, water sources within the yard of one's house (typically, wells and processed water from the tap) may be better in quality in the first place, compared to water obtained outside of one's house (such as rivers, ponds, and public wells). It is also possible that the quantity, as well as the quality, of water matter for child health. Access to an in-yard water source would help people obtain the amount of clean water they need whenever they need it. For example, Gilman et al. (1993) argues that hygiene education is of limited value, unless clean water is available in sufficient quantity at critical times, such as before meals and after defecation. Finally, access to an in-yard water source may release mothers from fetching water. If they use the saved time for childcare or for more general home production, child health may increase as a result.

China has been strongly committed to providing clean drinking water for its residents. The Chinese government announced that by the end of 2020, every rural family in China would have clean drinking water. Nowadays, more than 300 million rural residents throughout China still lack clean drinking water. Over the past five years, more than 14 million rural families throughout 27 provinces have gained access to drinking water, with more than 800,000 new water processing facilities going into operation. China earmarked a record 18 billion yuan (US$ 2.1 billion) for rural drinking water supplies during the 10th Five-Year Plan (2001–2005). The funds mostly come from local revenues and national debt (China Daily, November 29, 2004).

Measuring the effect of water accessibility on child health is an important policy question. Water-supply projects have been popular in China, as well as in many developing countries. Although much human labor and monetary resources have been devoted to water projects, the effect of clean water on child heath (a principal outcome of interest) and its mechanism is far from conclusive.

Previous research found a positive relationship between access to piped water and child health (Merrick, 1985, Cebu Study Team, 1991, Thomas and Strauss, 1992, Lee et al., 1997, and Jalan and Ravallion, 2003). None of these studies, however, exploited changes in health within individual children, before and after changes in water accessibility, by using panel data. The only exception is a study by Galiani et al. (2005) which compared municipality-level child mortality before and after the privatization of water services in both treatment and control municipalities. In this case treatment municipalities did experience the privatization of water services while control municipalities did not and it was found that the privatization of water services and subsequent improvements of water-provision infrastructure reduced child mortality in Argentina. In this paper, child-specific fixed effects available using panel data controlled for time-invariant characteristics that were likely to be correlated with both water accessibility and child health. Such characteristics included parents’ knowledge about health and also community characteristics affecting child health.

Nonetheless, even with child fixed effects, measuring the effect of water accessibility on child health using observational data is potentially problematic. First, unobserved changes in household wealth may increase child health by allowing purchase of not only water access, but also better nutrition and health care, creating upward bias. Alternatively, households could react to bad child health by investing in better access to drinking water, creating downward bias. More generally, as long as households have control over convenient access to water, many unobserved factors affecting the household could be correlated with both changes in water access and changes in child health. In contrast, if access to clean water is mostly determined by government or NGO investments in water projects, access to water would be largely exogenous to a household's own demand.

However, even if changes in access to clean water are mostly determined by governments or NGOs, the placements of water projects may not be random across villages (Pitt et al., 1999, Molyneaux and Gertler, 2000, Frankenberg and Thomas, 2001). For example, local governments or NGOs could give priority to areas where other health-related infrastructure is deteriorating quickly, leading to downward bias. Alternatively, investments in water access could be either positively or negatively correlated with other health-related investments. On the one hand, investments in water access may crowd out other health-related projects due to budget constraints, creating downward bias. On the other hand, if communities can afford investments in water access, they may also be able to afford other health-related projects, leading to upward bias. Note that with child fixed effects, project placements endogenous to time-invariant characteristics do not bias the estimated effect of water accessibility on child health. Rather, the identification issue is whether the dynamic characteristics of project sites are correlated with both changes in water access and changes in child health.

This paper addresses these problems by using the following two methods, in addition to controlling for child fixed effects. First, to avoid a potential household-level dynamic correlation between changes in water access and unobserved changes in a household's demand for water access, we used community-level changes in water access as the instrument for household-level changes in water access. Specifically, we calculated the ratio of sample households with water access for each sample community and used changes in the calculated ratio as the instrument for household-level changes in water access (Ilahi and Grimard, 2000, Glick et al., 2004). This enabled us to utilize changes in water access that were not endogenous to household-level demand. The rationale behind this identification strategy was that community-level changes in water access would reflect government or NGO projects in water access rather than individual demand by each household. Second, to address a potential community-level dynamic correlation between the placements of water projects and other community circumstances relating to health, we made use of data on disease symptoms, such as bone fracture, muscular injury, skin rash, burns, fever, headaches, diarrhea, stomach aches, and coughs. While it is likely that water projects would mostly affect the incidence of diarrhea and stomach ache, other health-related circumstances would influence not only the incidence of diarrhea and stomach ache but also other symptoms as well. If a dynamic correlation between the placement of water projects and other health-related circumstances does exist, changes in community-level water access would be correlated not only with changes in reports of diarrhea and stomach ache but with changes in reports of other symptoms. Our results show that changes in community-level water access and changes in the water-irrelevant symptoms are positively correlated for educated households. This occurs if local governments selectively placed water projects in communities where the health-related circumstances were worsening. This implies that our estimated impact of water accessibility on child health could be underestimated if the selective placement of water projects biased our results, suggesting that the true impact could be larger.

The effect of maternal schooling on child health or, more generally, home production has been discussed in the literature (Leibowitz, 1974, Behrman and Wolfe, 1987, Datcher-Loury, 1988, Behrman and Wolfe, 1989, Thomas et al., 1991). Convenient access to water and maternal education could be either complements or substitutes in producing child health (Thomas and Strauss, 1992, Jalan and Ravallion, 2003). On the one hand, well-educated mothers may make better use of convenient access to clean water than less educated mothers by, for example, encouraging children to have better hygiene, which is easier with access to clean water. On the other hand, educated mothers may be more skillful than less educated mothers in obtaining safe water (e.g. they would know to boil water) when easy access to clean water is unavailable. In this case, convenient access to water could benefit children of uneducated mothers more than children of educated mothers. Whether parental education and convenient access to water are complements or substitutes has important policy implications. It might be that increasing parental education would make water projects more effective in promoting child health. Alternatively, water projects could decrease gaps in child health between educated and less educated households.

The rest of the paper is organized as follows. Section 2 describes the data and also provides summary statistics, partly to make sure that there is enough within-child variation in the variables used in the econometric analysis. Econometric models and identification strategies are discussed in Section 3, results are presented in Section 4, and a conclusion is presented in Section 5.