Since 2015, the number of people without safely managed drinking water in sub-Saharan Africa has increased from 703 to 766 million (United Nations, 2021). The wealth gradient in health—indicating poorer health among less affluent individuals—has long intrigued health policy and social science discussions (Kawachi et al., 1997; Kennedy et al., 1996). Wealthier people on average appear to have better health (Haan et al., 1987; Krieger et al., 1997; Marmot 2001). Theory suggests that the same might apply to drinking water quality and that wealthier people, on average, might have better drinking water. Yet, the wealth gradient in water quality has scarcely been explored. The literature that has examined drinking water quality has mainly focused on the following research strands: fecal contamination of various drinking water sources (Bain, 2014), examined risk factors for Escherichia coli (E. coli) contamination (Bain et al., 2021), household- and community-level risk factors for E. coli contamination (Cronin et al., 2017; Harris et al., 2017; Kandel et al., 2017; Kirby et al., 2016; Kumpel et al., 2016; Pickering et al., 2010; Wang et al., 2017; Wardrop et al., 2018; Yang et al., 2013). This paper looks at the wealth gradient in drinking water quality at the point of use, through improved sources or water treatment.
In order to analyze the wealth gradient in drinking water quality, the primary variable of interest I use is a measure of having clean drinking water. I define that in the following way: I construct a binary variable, “Enhanced water” that takes the value of “1” if a household has better water at the point of use than point of collection. I use this variable as my measure of clean drinking water because studies typically do not take into account the degradation in drinking water quality that happens from the source to the point of consumption as Jessoe states.13 A primary contribution of this paper is to use this degradation in water quality as a measure of having clean drinking water.
It is important to note that this study is purely descriptive and does not attempt to make causal claims. Rather, it seeks to identify and describe the wealth gradient in drinking water quality and its seasonal variations across different countries in sub-Saharan Africa. TI analyze data from 27,408 households from 14 African countries, drawn from the Multiple Indicator Cluster Survey (MICS) datasets from the sixth round (MICS6). I observe a wealth gradient in the probability of having enhanced drinking water in the dry season. Being in the poorest wealth quintile is associated with a 22.1 percentage point decrease in the probability of having enhanced water at point of use compared to households in the richest wealth quintile in the dry season.
A strand of literature has found the adverse effect rainy seasons have on water quality. Systematic reviews suggest fecal contamination in drinking water sources follow a seasonal trend of greater contamination during the rainy season. Kostyla (2015) identified 22 studies in developing countries through a systematic review and analyzed seasonal variation in fecal contamination. The greater contamination disproportionately affects poorer households (WHO, 2017). The report only focuses on how the poor do not have access to clean water infrastructure and claims the lack of infrastructure might affect them negatively in the rainy season. In contrast, this paper quantitatively examines the wealth gradient in drinking water quality in the rainy season and finds a differential wealth gradient in the pooled data. However, I find that there is heterogeneity at the country level where some countries experience the wealth gradient getting steeper and others experience the wealth gradient getting flatter in the rainy season. In Benin, Central African Republic, and DR Congo the rainy season makes the wealth gradient steeper. For these countries, among those households surveyed in the rainy season, a household being in the richest quintile is associated with a 6.3 to 24.8 percentage point increase in the probability of having enhanced drinking water compared to the poorest households. However, in Guinea, Lesotho, and Madagascar the wealth gradient changes in a different way. For these countries, among households surveyed in the rainy season, being in the richest quintile is associated with a 17.8 to 27.1 percentage point decrease in the probability of having enhanced drinking water compared to the poorest households.
To address the issue of water quality crisis, studies have evaluated the effectiveness of a wide variety of interventions across a variety of contexts. Kremer et al. (2021) conducted a meta-analysis of 15 RCTs on water treatment and Clasen et al. (2007) conducted a meta-analysis of 32 studies on interventions to improve water quality. Systematic reviews suggest that households do one of two things to attempt to improve water quality at point of use (Clasen et al., 2007). According to the review, some households use better quality sources to have better water at point of use. Whereas others decide to treat their water prior to consumption (Kremer et al., 2021). I observe both the decision to treat water prior to consumption and the selection of water source in my dataset. To study why there is a wealth gradient in drinking water quality, I explore a household’s choice of water source and the decision to treat their water as potential mechanisms.
I find that being in the richer wealth quintiles is associated with an increase in the probability of water treatment on average in the dry season. In the dry season, being in the fourth and fifth wealth quintile is associated with a 3.0 and 3.8 percentage point increase in the probability of water treatment when compared to households in the poorest wealth quintile. However, the rainy season does not appear to be associated with a household’s water treatment decision in the pooled data.
Households also endogenously choose their source of water to have better quality drinking water.5 In the pooled data, richer households on average use better quality water sources and in the rainy season, decrease their usage of surface water sources to shift to using basic water sources even more. A household being in the richest quintile is also associated with a 6.1 percentage point drop in the probability of using surface water compared to households in the poorest wealth quintile in the dry season. A household being in the richest wealth quintile is also associated with a 13.1 percentage point increase in the probability of using basic water sources. Among those surveyed in the rainy season, being in the second, third and fourth quintile is associated with a 4.2, 6.5, and 7.0 percentage point decrease in surface water usage respectively compared to households in the poorest quintile.
These patterns suggest that there is a wealth gradient in drinking water quality in African countries. The endogenous decision to treat water prior to consumption and selection of water sources are mechanisms that contribute to the wealth gradient. I find no differential wealth gradient in my pooled data. However, I find that there is a differential wealth gradient at the country level; but because there is heterogeneity in how the wealth gradient changes at the country level, the estimate is close to zero in the pooled data.
My paper compliments, and adds to, the literature on the determinants of water quality in households. Jessoe (2013) for example, focuses on the substitutability between home treatment and choice of water source. I also contribute to the literature on unequal allocation of drinking water quality across wealth quintiles. Yang et al. (2013) report inequality in access to safe drinking water indicated by type of water source. In contrast, my contribution is to show that there is a wealth gradient in drinking water quality in Africa and that there is a differential wealth gradient in the rainy season. I show that there is heterogeneity in the wealth gradient at the country level. Finally, I contribute to the literature on unintended degradation of water quality from the source to the point of use. Bain et al. (2021) assess the differential presence of E. coli at point of consumption and at point of use. However, to the best of my knowledge, my paper is the first to use the difference in risk levels between the source and point of use as a measure of water quality and show how that varies in the rainy season.
This paper has several potential limitations. The findings in this paper are subject to unobservable variables at both the household and at the country level that might be confounded with drinking water quality. Additionally, the findings in this study might not generalize to countries that are in different stages of development and growth compared to countries in my African sample. I also use just one dimension of having clean drinking water: the presence of E. coli. The results may not be robust to other definitions of having clean water.
The remainder of the paper proceeds as follows: Section two presents the background and describes the data for the sample. Section three presents the evidence on the wealth gradient in the African household water quality in dry and rainy seasons, and Section four tests decisions of water treatment and water source as mechanisms to explain the wealth gradient in drinking water quality. Section five concludes.