According to the United Nations, the world has met the Millennium Development Goal target of halving the proportion of people without access to safe drinking water. However, global figures mask massive disparities between regions and countries, and within countries. For instance, only 64% of the people in sub-Saharan Africa have access to improved water sources. Over 40% of all people globally who lack access to drinking water live in sub-Saharan Africa. Rwanda is used as a case in point in this study. Despite the abundance of water resources in the country, access to improved water sources is limited. Using the Rwandan Demographic and Health Surveys (2000–2010), we examined regional disparities in access to improved water sources. Results from logistic regression models show that overall, access to improved water has declined between 2000 and 2010; except in the western region, where access to water marginally improved. Educated individuals, wealthier and urban dwellers were more likely to have access to improved water sources over time compared to their uneducated, poor and rural counterparts. The persistence of regional disparities in access to improved water over time suggests the need for policy to address insufficient investments in water infrastructure in Rwanda.

## INTRODUCTION

Improved water and human health are inextricably linked (Fink et al. 2011). Unsafe drinking water, and limited accessibility of water for hygiene, among other factors, contributes to about 88% of deaths from diarrheal diseases (Prüss-Üstün et al. 2008). Improvement of drinking water quality will reduce diarrhea episodes by 45% and eventually lead to a 21% reduction in diarrhea morbidity (UN Millennium Project 2005). The proportion of the world's population with access to improved water sources increased remarkably from 76 to 89% globally between 1990 and 2012 (Pullan et al. 2014; WHO & UNICEF 2014). Yet, as of 2012, 748 million people still relied on unimproved water sources for their drinking, cooking, and personal hygiene. Coverage at the global level is approximately 90% in developing regions of the world, but sub-Saharan Africa has coverage of only 64%, and still, there are pervasive disparities between countries and across regions (WHO & UNICEF 2014). Moreover, there are also inequalities between urban and rural areas, where an estimated 96% of the urban population globally used an improved water supply source in 2012, compared to 82% of the rural population (WHO & UNICEF 2014). These disparities are much more remarkable in relation to on-premised piped water, to which 80% of urban dwellers have access, in sharp contrast with just 29% from the rural areas. The disparities are equally obvious in terms of people who remain unserved. Approximately 82% of the global population, which does not have access to an improved source of water, lives in rural areas. Also, 62% of the 2.3 billion people who gained access to improved sources were in urban areas, and more than two-thirds of these gained access to piped water on premises that provide the highest level of health (WHO & UNICEF 2014).

Albeit, these statistics, the drinking water target highlighted by the Millennium Development Goals (MDGs) – which calls for halving the proportion of the population without sustainable access to safe drinking water between 1990 and 2015 – is considered to have been met in 2010. These seemingly noteworthy aggregate statistics at the national and regional levels highlighted above, however, mask huge place-based disparities especially in sub-Saharan Africa where access to improved water already lags behind other geographic areas of the world. Sub-Saharan Africa remains the area of greatest concern because it is a region of the world where access to improved water sources remains low (Hutton 2013; Salami et al. 2014). This stimulates research that seeks to elicit a nuanced understanding of the relationship between place-based disparities in access to safe drinking water and water-related health differentials among populations.

Previous research on disparities in access to improved water sources and its relationship with health differentials among populations have emphasized the importance of either composition or context effect. In terms of composition effect, it is argued that individuals in the same neighborhood tend to be more similar to one another than to those in other neighbourhoods in terms of predisposing factors of limited access to water such as age and socioeconomic status. Within this milieu, extensive research has documented water-related disparities among population subgroups defined by gender, age, education, and relative well-being (Fry et al. 2008; Bambra et al. 2010; Hawkins & Seager 2010; Yang et al. 2013). In terms of context effect, it is suggested that individuals living in the same neighborhood are exposed to similar local factors that have impacts on their water-related health outcomes (Mehta et al. 2007; Bambra et al. 2010; Stephens 2011). In reality, however, both set of factors contribute to varying degrees in explaining water-related health differentials among the population (collective effect). Thus, health disparities occasioned by inequities in access to safe drinking water, represented by differences in the incidence, prevalence, and mortality from disease and other adverse conditions may result from variations in social, cultural, behavioral, biologic, genetic, and environmental factors among population subgroups and geographic locations (Hernandez & Blazer 2006; Adler & Stewart 2010).

Although place-based inequalities of other determinants of health have been widely documented (see Patrick 2011), less effort has been devoted to addressing place-based inequalities in access to improved water sources within or across countries over time. Hitherto, the focus on place-based disparities in drinking water quality, and by extension health differentials, has been restricted to dichotomous conceptualizations of place into rural–urban, and regional north–south. This dualistic framing is problematic as it ignores the heterogeneities within rural and urban areas and within the regional north and south. For instance, both urban and rural areas may occur in the regional north likewise the regional south. Moreover, evaluating water-related health differentials using composition, context, or collective effects alone fail to consider the dynamic nature of people and geographic spaces. For example, urbanization is a continuing phenomenon such that rural areas are constantly changing into urban landscapes with time. This makes it imperative to analyze the role of collective effect in generating water-related health differentials in different places across time. This is a fundamental motivation of this study.

The purpose of this paper is to underscore the empirical and practical significance of examining the influence of collective effect on water-related health differentials across time using Rwanda, as a case in point. Although Rwanda has done well in the improvement of infrastructure to improve water access, it remains important to investigate the trajectories of potential inequality related to access to improved water sources. The paper therefore investigates changes in access to improved sources of water over time and therefore contributes to the assessment of the country's commitment and progress towards its socioeconomic developmental agenda. The rest of the paper is organized as follows: the next section provides an overview of the data. Then, we describe the theoretical relevance of the outcome variable and covariates. After the presentation of the results, we then discuss the findings in light of the extant literature on access to improved water sources.

## MATERIALS AND METHOD

### Study context

In sub-Saharan Africa, Rwanda is often considered as a success story when it comes to improving access to water across the country. Despite this, urban–rural inequalities and disparities in the underserved still persist. People living in rural areas, the urban poor and marginalized communities often face exclusion from improvements. It is therefore worrying that over the last five years the Government of Rwanda's budget for water has been decreasing, suggesting the lower priority accorded to the sector (WaterAid/Development Finance International 2014). Rwanda is the most densely populated country in Africa, and has approximately 11 million citizens with an annual population growth of 2.6% (National Institute of Statistics of Rwanda 2012; UNDP 2012). About 60% of the population lives below the poverty line. Merely 19.4% of the total inhabitants live in urban areas (UNDP 2012); therefore, it is unsurprising that Rwanda currently has an urbanization rate of 4.5% annually. Although the country remains predominantly rural, there is tremendous transformation taking place as the country undergoes urbanization (Uwimbabazi & Lawrence 2011) and indeed it is one of the countries with the fastest growing urbanization rates in the sub-region; 3.7% annual rate of change of the proportion urban (UNESA 2014).

Although landlocked, Rwanda in sub-Saharan Africa can boast an abundance of water. This abundance of water resources is reflected by the existence of lakes, rivers, and a network of wetlands in various parts of the country covering over 200,000 ha of land (Ministry of Land, Environment and Natural Resources 2004). The Nile covers 67% of the land, which also delivers 90% of the national waters and the Congo covers about 33% of the land (Rwanda Environment Management Authority & UNEP 2009). Yet it is in the midst of this that improved water coverage remains a worry. According to the World Health Organization's (WHO)/UNICEF Joint Monitoring Programme (JMP) 2012 data, progress towards meeting the national MDG target for water is insufficient, with 65% of the population having access to safe drinking water by 2010. The Rwandan Government acknowledged the adequate water supply and sanitation services as drivers for social and economic development, poverty reduction and public health, and has therefore committed to reaching ambitious targets in water supply and sanitation, with a vision to attain 100% service coverage by 2020 (WaterAid/Development Finance International 2014).

To accelerate the move towards the vision 2020 targets of 100% access to water supply and sanitation country-wide, in 2000 Rwanda adopted a seven-year program with sector-based reforms such as decentralization of infrastructure management to accelerate the achievement of 100% access to improved water supply and sanitation facilities by 2017 (Ministry of Land, Environment and Natural Resources 2004). To monitor progress, the Government established a Management Information System (MIS), through the Energy, Water and Sanitation Authority, which has been functional since 2012. The National Water Policy of Rwanda sets its national MDG targets at 84% – a uniform target across urban and rural areas; meanwhile, national data collected through the MIS in 2012 places national water supply coverage at 71%, which would mean the country is on track to meet its MDG target (WaterAid/Development Finance International 2014).

### Data

The study draws on household data from the 2000, 2005, and 2010 Rwanda Demographic and Health Surveys (RDHS-00, RDHS-05 and RDHS-10). The RDHS have measures on ‘access to improved water source’, compositional and contextual variables. RDHS is a large nationally representative data set collected by the National Institute of Statistics of Rwanda (NISR) and Macro, and it is the third, fourth, and fifth of such recent national surveys that form part of the Global Demographic and Health Surveys programme. The study was designed to provide current information about demographic processes in Rwanda. The RDHS-05 and RDHS-10 utilized a two-staged, stratified sample frame in which a systematic sampling with probability proportional to size was used to identify specific enumeration areas to select households in the country. Participants aged 14–49 were randomly selected and interviewed during each of the surveys. A total of 9,696 people were interviewed in 2000; 10,272 people in 2005; and 12,540 people in 2010.

### Dependent variable

The dependent variable for this study is conceptualized as access to an improved source of water, which according to the WHO is critical for global health. Access to an improved source of water is defined as less than 1 kilometer away from its place of use; and an availability of at least 20 liters per member of a household per day. Safe drinking water is defined as water with microbial, chemical, and physical characteristics that meet WHO guidelines or specific national standards on drinking water quality. The outcome variable was therefore created from questions that tapped into respondents' proximity to available and accessible drinking water, principally on whether they have easy access or not. Respondents were asked, ‘what is your source of drinking water?’ and response categories were dichotomized into ‘0’ = unimproved and ‘1’ = improved sources based on the WHO/UNICEF classification (see Table 1). According to the MDG definition, an improved drinking water source is defined as one that by nature of its construction or through active intervention is protected from outside contamination, in particular from contamination with fecal matter (AMCOW 2012). In line with the official indicators for the MDG drinking water target, only users of improved water sources are considered as having access to drinking water.

Table 1

Categorization of sources of water

Unimproved sourcesImproved sources
Unprotected dug well Piped water into dwelling, yard, or plot
Unprotected spring Public tap or standpipe
Cart with small tank or drum Tubewell or borehole
Tanker truck Protected dug well
Surface water (river, dam, lake, pond, stream, canal, irrigation channel) Protected spring
Boiled water Rainwater collection
Unimproved sourcesImproved sources
Unprotected dug well Piped water into dwelling, yard, or plot
Unprotected spring Public tap or standpipe
Cart with small tank or drum Tubewell or borehole
Tanker truck Protected dug well
Surface water (river, dam, lake, pond, stream, canal, irrigation channel) Protected spring
Boiled water Rainwater collection

Source: WHO/UNICEF JMP for Water Supply and Sanitation.

### Key independent variable

The key independent variable is place, which reflects contextual factors. It was operationalized using region, place of residence (rural–urban), and time to get to source of water (proximity). The five regions were coded as ‘1’ = Kigali, ‘2’ = South, ‘3’ = West, ‘4’ = North, and ‘5’ = East. Place of residence was coded as ‘1’ = Rural and ‘2’ = Urban.

### Control variables

Using the criteria set out in Pol & Thomas (2000), theoretically relevant variables such as biosocial (age and sex of household head) and sociocultural factors (level of education and wealth index) were controlled. According to Pol & Thomas (2000), biosocial characteristics are those that have an underlying biological and physical component while sociocultural factors reflect the position of members within the social structure; these are not traits one is born with. Variables were therefore carefully selected based on their established relationship with access to water in the literature (e.g., Fry et al. 2008; Bambra et al. 2010; Yang et al. 2013). Although ethnicity is an important biosocial factor, it was omitted in this study due to historical antecedents in Rwanda. In 2006, Rwanda's 12 provinces were abolished and replaced with five provinces hence ‘Region’ in 2000 and 2005 data were recoded to match the current divisions – Kigali City, South, West, North, and East. The 2010 data already had this classification, so no recoding was necessary. Age was also re-categorized into four groups of ‘1’ = 13–30, ‘2’ = 31–50, ‘3’ = 51–65, and ‘4’ = 66 or older. The original categorizations of all other independent variables in the data were maintained. Sex/gender was operationalized as dummy variable ‘0’ = female and ‘1’ = male. Educational attainment of respondents was coded as ‘0’ = no education, ‘1’ = primary, ‘2’ = secondary, and ‘3’ = tertiary. Wealth quintile of the household was re-categorized into quartiles due to the limited number of responses on the poor category. Wealth quartile was then coded as ‘1’ = poorest and poor, ‘2’ = middle, ‘3’ = richer, and ‘4’ = richest. Other control variables included number of individuals in the household and distance to nearest improved water source.

### Data analysis

The data were processed in and analyzed in Stata version 13. Logistic regression analysis was performed to determine whether or not any of the independent variables influenced the dependent variable (access to improved source of water). A logistic regression model is frequently used when the dependent variable is dichotomous (Gliner & Morgan 2000; Agresti & Finlay 2009). Let Y be the dependent variable, which takes on values 1 (event) and 0 (nonevent). In this context, the event represents access to improved source of water and nonevent represents the lack of it, Further, let p denote the probability that an observation is an event, that is, P = P(Y = 1). The logistic regression models the log-odds of an event as a function of a linear combination of the intercept and slope parameters:
With the obtained estimates, it can be shown that:

p = exp{α + β1x1 + β2x2 + … βkxk}/1 + α + β1x1 + β2x2 + …βkxk

which gives the estimated probability that an observation is an event.

Usually, when this probability is greater than 0.5, the observation is classified as event, otherwise, it is classified as nonevent (Agresti & Finlay 2009). In the present study, apart from theory, parsimony, and previous research, independent variable selection was based on a backward-elimination method. The advantage of this method is that it can include a variable that does not have a strong association with the dependent variable by itself but has some contribution in the model with the presence of other variables. Such a variable will not be detected when a forward-selection method is used. To check the model fit, the correct classification rate was considered and the Wald Chi-Square test was used. The Wald Chi-Square test statistic measures the correspondence of the actual and predicted values of the dependent variable. A better model fit was indicated by a smaller difference in the observed and predicted classification (Hair et al. 1998). The significance of the test was assessed by a chi-square distribution. A good model fit was indicated by a significant test result (Hair et al. 1998).

Modeling building followed a nested pattern in which place-based variables were estimated in Model 1. An interaction term was introduced in Model 1 to test if place of residence and time to reach drinking water source were concurrently associated with source of drinking water. In Model 2, biosocial factors were incrementally introduced to the variables in Model 1. In Model 3, sociocultural factors were incrementally added to Model 2 variables. In each of the models, exponentiated coefficients (odds ratios (ORs)) are reported in addition with robust standard errors. An OR of 1.00 implies that the two groups are equally likely to access improved source of water. An OR higher than 1 implies that the first group is more likely to experience the event (access water) than the second group. An OR of less than 1 implies that the first group is less likely to experience the event (access water). The OR is a measure of effect size and therefore provides information on the strength of relationship between two variables.

## RESULTS

### Univariate

Table 2 shows that more than three-quarters of respondents surveyed had access to improved sources of water. More than half of the respondents had primary education. More than three-quarters of the respondents lived in rural areas. A greater percentage of the respondents reside in the West, followed by Kigali City, South, East, and North.

Table 2

Sample characteristics of selected dependent and independent variables

Percentages
Source of water
Unimproved 21.75
Improved 78.25
Region
Kigali 21.50
South 20.15
West 23.24
North 16.69
East 18.41
Place of residence
Rural 81.03
Urban 18.97
Sex of household head
Female 33.88
Male 66.12
Level of education
No education 35.18
Primary 53.44
Secondary 9.64
Higher 1.74
Wealth
Poor 29.09
Middle 28.90
Richer 20.23
Richest 21.78
Age of household member
13–30 24.22
31–50 46.28
51–65 19.01
> 65 10.49
Distance
Near 10.06
Far 89.94
Percentages
Source of water
Unimproved 21.75
Improved 78.25
Region
Kigali 21.50
South 20.15
West 23.24
North 16.69
East 18.41
Place of residence
Rural 81.03
Urban 18.97
Sex of household head
Female 33.88
Male 66.12
Level of education
No education 35.18
Primary 53.44
Secondary 9.64
Higher 1.74
Wealth
Poor 29.09
Middle 28.90
Richer 20.23
Richest 21.78
Age of household member
13–30 24.22
31–50 46.28
51–65 19.01
> 65 10.49
Distance
Near 10.06
Far 89.94

Figure 1 shows that a greater percentage of respondents in the North (89.8%) and West (89.2%) had access to improved water sources, followed by those in the South (86.4%) and Kigali City (84.9%), then those in the East (70.2%) in year 2000. With the exception of the South, where access to improved water sources declined minimally, the other regions experienced a sharp decline in access to improved water sources in 2005. Between 2005 and 2010, access to improved water sources continued to decline in the North, West, and South while access began to improve in Kigali and the East.
Figure 1

Graphical representation of access to improved source of water over time by region.

Figure 1

Graphical representation of access to improved source of water over time by region.

Four multivariate models are fitted as shown in Table 3. In the first model, we examine the effects of the two spatial variables, place of residence and region of residence, on access to improved water source. In the second model, we control for psychosocial and sociocultural variables. In Models 3 and 4, we examine how time moderates the relationship between place of residence and region of residence, and access to improved water source, respectively. In doing so, we capture how access to improved water has changed over time in these spatial contexts while controlling for other theoretically relevant factors.

Table 3

OR (standard error) from complementary log-log regression analysis of access to improved sources of water

Model 1Model 2Model 3Model 4
Independent variablesORSEORSEORSEORSE
Region (Kigali)
South 0.93 (0.05) 1.20 (0.07)*** 1.29 (0.07)*** 1.41 (0.15)***
West 0.88 (0.04)** 1.05 (0.06) 1.12 (0.06)* 1.55 (0.15)***
North 0.99 (0.05) 1.20 (0.07)*** 1.28 (0.07)*** 1.77 (0.19)***
East 0.45 (0.02)*** 0.51 (0.03)*** 0.55 (0.03)*** 0.41 (0.04)***
Place of residence (Rural)
Urban 3.20 (0.17)*** 2.00 (0.12)*** 2.16 (0.26)*** 2.03 (0.12)***
Age of household head (13–30)
31–50   1.13 (0.05)*** 1.13 (0.05)*** 1.14 (0.05)***
51–65   1.13 (0.05)** 1.16 (0.06)*** 1.17 (0.06)***
> 65   1.23 (0.07)*** 1.26 (0.07)*** 1.28 (0.07)***
Sex of household head (Female)
Male   0.92 (0.03)** 0.93 (0.03)* 0.92 (0.03)*
Level of education (No education)
Primary   1.06 (0.04) 1.10 (0.04)*** 1.11 (0.04)***
Secondary   1.29 (0.09)*** 1.37 (0.10)*** 1.36 (0.10)***
Higher   1.85 (0.44)** 2.14 (0.51)*** 2.00 (0.48)***
Wealth (Poor)
Middle   1.90 (0.07)*** 1.58 (0.06)*** 1.53 (0.06)***
Richer   2.11 (0.09)*** 1.87 (0.08)*** 1.83 (0.08)***
Richest   2.85 (0.16)*** 2.57 (0.15)*** 2.49 (0.14)***
Distance to source of water (Near)
Far   0.68 (0.04)*** 0.65 (0.04)*** 0.67 (0.04)***
Number of household members   0.98 (0.01)*** 0.98 (0.01)*** 0.98 (0.01)***
Place of residence # time
Rural#2005     0.69 (0.03)***
Rural#2010     0.67 (0.03)***
Urban#2005     0.79 (0.12)
Urban#2010     0.54 (0.07)***
Region # time
Kigali 2005       0.72 (0.06)***
Kigali 2010       0.82 (0.1)
South 2005       0.93 (0.11)
South 2010       0.56 (0.05)***
West 2005       0.45 (0.04)***
West 2010       0.50 (0.04)***
North 2005       0.60 (0.07)***
North 2010       0.41 (0.04)***
East 2005       0.91 (0.08)
East 2010       1.10 (0.08)
Model 1Model 2Model 3Model 4
Independent variablesORSEORSEORSEORSE
Region (Kigali)
South 0.93 (0.05) 1.20 (0.07)*** 1.29 (0.07)*** 1.41 (0.15)***
West 0.88 (0.04)** 1.05 (0.06) 1.12 (0.06)* 1.55 (0.15)***
North 0.99 (0.05) 1.20 (0.07)*** 1.28 (0.07)*** 1.77 (0.19)***
East 0.45 (0.02)*** 0.51 (0.03)*** 0.55 (0.03)*** 0.41 (0.04)***
Place of residence (Rural)
Urban 3.20 (0.17)*** 2.00 (0.12)*** 2.16 (0.26)*** 2.03 (0.12)***
Age of household head (13–30)
31–50   1.13 (0.05)*** 1.13 (0.05)*** 1.14 (0.05)***
51–65   1.13 (0.05)** 1.16 (0.06)*** 1.17 (0.06)***
> 65   1.23 (0.07)*** 1.26 (0.07)*** 1.28 (0.07)***
Sex of household head (Female)
Male   0.92 (0.03)** 0.93 (0.03)* 0.92 (0.03)*
Level of education (No education)
Primary   1.06 (0.04) 1.10 (0.04)*** 1.11 (0.04)***
Secondary   1.29 (0.09)*** 1.37 (0.10)*** 1.36 (0.10)***
Higher   1.85 (0.44)** 2.14 (0.51)*** 2.00 (0.48)***
Wealth (Poor)
Middle   1.90 (0.07)*** 1.58 (0.06)*** 1.53 (0.06)***
Richer   2.11 (0.09)*** 1.87 (0.08)*** 1.83 (0.08)***
Richest   2.85 (0.16)*** 2.57 (0.15)*** 2.49 (0.14)***
Distance to source of water (Near)
Far   0.68 (0.04)*** 0.65 (0.04)*** 0.67 (0.04)***
Number of household members   0.98 (0.01)*** 0.98 (0.01)*** 0.98 (0.01)***
Place of residence # time
Rural#2005     0.69 (0.03)***
Rural#2010     0.67 (0.03)***
Urban#2005     0.79 (0.12)
Urban#2010     0.54 (0.07)***
Region # time
Kigali 2005       0.72 (0.06)***
Kigali 2010       0.82 (0.1)
South 2005       0.93 (0.11)
South 2010       0.56 (0.05)***
West 2005       0.45 (0.04)***
West 2010       0.50 (0.04)***
North 2005       0.60 (0.07)***
North 2010       0.41 (0.04)***
East 2005       0.91 (0.08)
East 2010       1.10 (0.08)

OR, odds ratio; SE, standard errors.

* P < 0.05; ** P < 0.01; *** P < 0.001.

Variables categories in parentheses () are the reference.

Model 1 shows that people who lived in the West (OR: 0.88 P < 0.01) and East (OR: 0.45 P < 0.00) were less likely to have access to improved water sources compared to those in Kigali. Rwandans who resided in urban areas (OR: 3.20 P < 0.00) were, however, more likely to have access to improved water sources.

After controlling for psychosocial and sociocultural variables in Model 2, we found that people who resided in the South (OR: 1.20 P < 0.00) and North (OR: 1.20 P < 0.00) were more likely to have access to improved water sources relative to those in Kigali. However, those in the East (OR: 0.51 P < 0.00) were still less likely to have access to improved water sources. Although marginally attenuated, the positive effect of place of residence on access to improved water was retained. Household heads with secondary (OR: 1.29 P < 0.00) and primary (OR: 1.84 P < 0.01) education were more likely to have access to improved water compared to their uneducated counterparts. Respondents from middle (OR: 1.90 P < 0.00), richer (OR: 2.11 P < 0.00), and richest (OR: 2.85 P < 0.00) households were also more likely to have access to an improved source of water. The age of household heads was positively associated with access to improved water source. Male-headed households (OR: 0.91 P < 0.01) were less likely to have access to improved water compared to female-headed households. Every additional household member decreased the likelihood (OR: 0.98 P < 0.01) of having access to improved water. Likewise, distance to a water source was negatively associated with the likelihood of having access to improved source of water.

In Model 3 we controlled for time and found that over time respondents from the South (OR: 1.29 P < 0.00), West (OR: 1.12 P < 0.05), and North (OR: 1.28 P < 0.00) were more likely to have access to improved water compared to those in Kigali. Those in the East (OR: 0.55 P < 0.00) were, however, less likely to have access to improved water relative to their counterparts in Kigali. In year 2000, respondents who resided in urban areas (OR: 2.16 P < 0.00) were more likely to have access to improved water compared to their counterparts in rural areas. Respondents who lived in rural areas in 2005 were 31% less likely to have access to improved water compared to their counterparts in 2000. Rural populations in 2010 were 33% less likely to have access to improved water relative to their counterparts in 2000. This implies that access to improved water in rural areas declined over time. The urban population in 2010 were 46% less likely to have access to improved water relative to their counterparts in 2000. Although urban populations in 2005 were also less likely to have access to improved water than their counterparts in 2000, the observed difference was not statistically significant.

Model 4 shows that, in year 2000, respondents who lived in the South (OR: 1.41 P < 0.00), West (OR: 1.55 P < 0.00), and North (OR: 1.77 P < 0.00) were more likely to have access to improved water, but those who resided in the East (OR: 0.41 P < 0.00) were less likely to have access to improved water compared to their counterparts in Kigali. Respondents who lived in Kigali in 2005 were 28% less likely to have access to improved water compared to their counterparts in 2000. Respondents who lived in the South in 2010 were 44% less likely to have access to improved water relative to their counterparts in 2000. Compared to respondents who lived in the West in 2000, those who lived in the West in 2005 and 2010 were 55% and 50% less likely to have access to improved water, respectively. This suggests that access to improved water in the West improved over time. However, in the North, access to improved water declined over time. Respondents who lived in the North in 2005 were 40% less likely to have access to improved water compared to their counterparts in 2000; however, in 2010, respondents who resided in the North were 59% less likely to have access to improved water relative to their counterparts in 2000.

Figure 2 shows that access to improved sources of water was higher in urban areas compared to rural areas at any given point in time. Access to improved water sources, however, declined in both rural and urban areas over time, albeit the rate of decline was higher in rural areas compared to urban areas.
Figure 2

Graphical representation of access to improved source of water over time by place of residence.

Figure 2

Graphical representation of access to improved source of water over time by place of residence.

## DISCUSSION AND CONCLUSION

The 2013 report from the Rwandan Ministry of Finance and Economic Planning shows a consistent increase in the GDP per capita income from $206 in 2002 to$644 in 2012. The 2013 World Bank report also shows a decrease in the poverty headcount ratio at national poverty line from 56.7% in 2006 to 44.9% in 2011. As to whether this translates into countrywide improvement in infrastructural developments over time begs the question. Place-based disparities still exist with regards to access to improved water sources in Rwanda. The study found significant regional disparities exist in Rwanda over the ten-year period with the East region experiencing consistent lack in access to improved water source. More significantly, we found deterioration in access to improved water source across all regions over the period. This key finding is consistent with the 2006 report of UNICEF, which reported that urban–rural disparities in access to improved drinking water source are higher in Eastern and Southern Africa than in any other region.

In Rwanda, while urban areas had coverage of 90%, the rural areas had a lesser coverage of 40% (UNICEF 2006). These findings are also consistent with the National Water Supply and Sanitation Policy of Rwanda (2010), which indicated a general decline in access to improved drinking water since 2005 with 81% of urban populations having access in 2012. Our study shows consistently that this gap has not changed as of 2010. In Rwanda, population growth has contributed to reduction in access to improved water source over the ten-year period. Although measures have been put in place to increase access, population growth rate seems to be much higher than the economic and infrastructural growth in the country. For instance, between 1990 and 2004 the population growth rate was 5.6% on average, while the GDI per capita income had an average increase of 0.9% (UNICEF 2005). In reference to this study, the Eastern Province has been found to have the highest population share of 24.7% (2,600,814) people with an average population growth rate of 4.3% relative to all the other regions while Kigali was found to have just over one million people (National Institute of Statistics of Rwanda 2012). Hence, the finding that those in the Eastern region were less likely to have access to improved water source relative to those from Kigali could be attributed to the difference in population growth within the regions.

Some other reasons have been suggested in explaining the worsening of access to improved water source. One such reason is the increase in water-intensive agriculture in Rwanda. As a land-locked country, with most of the farmers engaged in subsistence agriculture based on rain-fed production systems and with a recent increase in irrigation farming (IFAD 2014), water used for farming is drawn from the available surface water in the country. This limits what would have been channeled into improved sources of drinking water. Also, the decrease in access to improved water sources has been associated with an increase in pollution (Sorenson et al. 2011). The 2011 Rwandan Water quality monitoring report indicated that there has been an increase in the pollution of surface water by anthropogenic activities. Most farmers have been found to use fertilizers and pesticides to increase their yield productivity which finally results in the pollution of rivers during run offs. Farming activities are predominant in valleys near rivers and streams, leading to their contamination. Further, pollutants such as heavy metals and other chemicals resulting from industrial processing have been found to increase the levels of pollution, limiting what could have been sources of improved drinking water.

In terms of socioeconomic status, wealthier households were more likely to have access to improved sources of drinking water relative to poorer households over time, which is an indication of inequality in water quality access in Rwanda. This finding is consistent with quite a number of studies on WASH (Garriga & Foguet 2013; Yang et al. 2013; Joshi et al. 2013). Most significantly, the report of WHO and UNICEF JMP for Water Supply and Sanitation suggested that as of 2010, for 35 countries representing 84% of the population in sub-Saharan Africa, wealthier households were more likely to have access to improved water source than poorer households (UNICEF & World Bank 2012; Luh et al. 2013). The inequality in access to water quality has also been associated with the high cost of operating water systems in both urban and rural environments due to the poor quality of the raw water and the mountainous terrain that increases the cost of treatment and pumping (Fitzpatrick 2014). Consequently, the cost of access creates a disadvantage to poorer households in accessing improved water sources. Further, the quest to privatize water supply as advocated in the policy arena, with the hope of increasing accessibility, have rather been found to exacerbate the gap between the wealthier and the poor (Budds & McGranahan 2003; Galiani et al. 2005). Consequently, policies in lieu of increasing water quality access need to consider pro-poor approaches.

Consistent with past studies (Larson et al. 2006; Keshavarzi et al. 2006; Abebaw et al. 2010; Nayak 2013; Adams et al. 2015), this study found a significant positive association between level of education of household heads and access to improved water source. Households with highly educated heads (tertiary education) relative to those with no education were more likely to have access to improved source of water over time. Education is a robust predictor of water quality access over time. Education has been found to be highly correlated with health seeking behaviors (Cutler & Lleras-Muney 2012). Osabuohien et al. (2012) has argued that with increase in education, households increase their ability to seek for ways to mitigate water challenges. Consequently, to avoid water-borne diseases, household heads that are well educated and understand the consequences associated with the use of untreated or unimproved water sources are more likely to make decisions on accessing improved water sources.

The study also showed that distance is negatively associated with access to improved water source. An increase in the distance to water source decreases the likelihood of having access to an improved source of drinking water. Previous research has recorded that the time burden of fetching water influences the volume of water fetched as well as time spent on other income generating activities and child care, especially by women (Moriarty et al. 2004). Likewise, there are potential benefits of reducing distance to improved water sources. After analyzing about 26 countries in sub-Saharan Africa, Pickering & Davis (2012) report a 15-minute reduction in one-way distance to improved water source could result in a concomitant reduction in diarrhea prevalence and under-five mortalities.

Findings regarding the relationship between sex and access to an improved water source are mixed. While a number of researchers, on one hand, have argued that female-headed households are more likely to have access to improved water sources (Crow & Sultana 2002), others argue that male-headed households tend to have more access to improved water sources (Bhorat et al. 2009). For instance, Abebaw et al. (2010) found that in Ethiopia, female-headed households have a greater likelihood of using an improved water source than male-headed households. Our results here extend this finding whereby female-headed households reported having more access to improved water sources over time relative to males.

Furthermore, older household heads were more likely to have access to an improved water source than those in the younger cohort over time. This has been associated with the level of experience that comes with age, which positively affects the ability of these household heads to mitigate the challenges that comes with either the scarcity of water or weather fluctuations (Osabuohien et al. 2012).

In sum, this study evaluated access to improved water sources in Rwanda over time. The overarching finding in this study showed a worsening in access to improved water source. Regional disparities with regards to access to water quality were found to have widened, with the Eastern region having the lesser likelihood of access to improved water source. Also, rural–urban disparities in access to improved water sources were evident, with urban areas having a greater access to water than the rural. Wealth, education, and age were found to be significant and robust predictors of water quality access. Any policy directed at dealing with the problem of water access should consider primarily densely populated areas while employing pro-poor approaches.

## ACKNOWLEDGEMENTS

We are most grateful to Frederick Ato Armah and Hanson Nyantakyi-Frimpong, Department of Geography, Western University, Canada for reading through an initial draft of this article.

## REFERENCES

Abebaw
D.
Fentie
Y.
Kassa
B.
2010
.
Food Policy
35
(
4
),
286
293
.
Adams
E. A.
Boateng
G. O.
Amoyaw
J. A.
2015
Socioeconomic and demographic predictors of potable water and sanitation access in Ghana
.
Social Indicators Research 1–15
.
Adler
N. E.
Stewart
J.
2010
The Biology of Disadvantage: Socioeconomic Status and Health (Vol. 1186)
.
The New York Academy of Sciences
,
New York
,
USA
.
Agresti
A.
Finlay
B.
2009
Statistical Methods for the Social Sciences
, 4th edn.
Pearson Prentice Hall
,
New York
,
USA
, pp.
590
597
.
AMCOW
2012
A Snapshot of Drinking Water and Sanitation in Africa – 2012 Update. A regional perspective based on new data from the WHO/UNICEF Joint Monitoring Programme for Water Supply and Sanitation, Fourth Africa Water Week, Cairo, Egypt, 14–15 May, 2012
.
(accessed 12 December 2014)
.
Bambra
C.
Gibson
M.
Amanda
S.
Wright
K.
Whitehead
M.
Petticrew
M.
2010
.
J. Epidemiol. Community Health
64
(
4
),
284
291
.
Bhorat
H.
van der Westhuizen
C.
Jacobs
T.
2009
Income and Non-Income Inequality in Post-Apartheid South Africa: What are the Drivers and Possible Policy Interventions? Development Policy Research Unit, University of Cape Town, Working Paper 09/138, August 2009
.
Budds
J.
McGranahan
G.
2003
.
Environ. Urban.
15
(
2
),
87
114
.
Crow
B.
Sultana
F.
2002
.
Society Natural Resour.
15
(
8
),
709
724
.
Cutler
D. M.
Lleras-Muney
A.
2012
Education and Health: Insights from International Comparisons (No. w17738)
.
National Bureau of Economic Research
,
Cambridge, MA
,
USA
.
Fink
G.
Günther
I.
Hill
K.
2011
.
Int. J. Epidemiol.
40
(
5
),
1196
1204
.
Fitzpatrick
C.
2014
AFRICA: African countries make progress in water and sanitation investment. IWA Publishing, London, UK
.
(accessed 17 December 2014)
.
Fry
L. M.
Mihelcic
J. R.
Watkins
D. W.
2008
.
Environ. Sci. Technol.
42
(
12
),
4298
4304
.
Galiani
S.
Gertler
P.
Schargrodsky
E.
2005
.
J. Political Econ.
113
(
1
),
83
120
.
Garriga
R. G.
Foguet
A. P.
2013
.
Water Resour. Manage.
27
(
5
),
1501
1515
.
Gliner
J. A.
Morgan
G. A.
2000
Research Methods in Applied Settings: An Integrated Approach to Design and Analysis
.
Lawrence Erlbaum Associates
,
Mahwah, NJ
,
USA
.
Hair
J. F.
Jr
Anderson
R. E.
Tatham
R. L.
Black
W. C.
1998
Multivariate Data Analysis, 5 ed. Institute of Medicine (US) Committee on Assessing Interactions Among Social, Behavioral, and Genetic Factors in Health (L. M. Hernandez & D. G. Blazer, eds). Prentice Hall, Upper Saddle River, NJ, USA
.
Hawkins
R.
Seager
J.
2010
.
Professional Geographer
62
(
1
),
16
31
.
Hernandez
L. M.
Blazer
D. G.
2006
Institute of Medicine (US) Committee on Assessing Interactions Among Social, Behavioral, and Genetic Factors in Health.
Genes, Behavior, and the Social Environment: Moving Beyond the Nature/Nurture Debate.
National Academy of Sciences
,
US
.
Hutton
G.
2013
.
J. Water Health
11
(
1
),
1
12
.
IFAD
2014
Investing in rural people in Rwanda
.
Joshi
A.
Prasad
S.
Kasav
J. B.
Segan
M.
Singh
A. K.
2013
.
Global J. Health Sci.
6
(
2
),
23
34
.
Keshavarzi
A. R.
Sharifzadeh
M.
Kamgar
A. A.
Amin
S.
Keshtkar
S.
Bamdad
A.
2006
.
Water Res.
40
(
6
),
1173
1178
.
Larson
B.
Minten
B.
Razafindralambo
R.
2006
.
J. Develop. Stud.
42
,
22
40
.
Luh
J.
Baum
R.
Bartram
J.
2013
.
Int. J. Hyg. Environ. Health
216
(
6
),
662
671
.
Mehta
L.
Marshall
F.
Stirling
A.
Shah
E.
Smith
A.
Thompson
J.
Movik
S.
2007
Liquid Dynamics: challenges for sustainability in water and sanitation. STEPS Working Paper 6, STEPS Centre, Brighton, UK
.
(accessed 12 December 2014)
.
Ministry of Land, Environment, Natural Resources
2004
Sectorial policy on water and sanitation
.
(accessed 18 January 2015
)
.
Moriarty
P.
Butterworth
J.
van Koppen
B.
Soussan
J.
2004
Beyond Domestic: Water, Poverty and Productive Uses of Water at the Household Level
.
IRC International Water and Sanitation Centre
,
Delft
,
the Netherlands
.
National Institute of Statistics of Rwanda
2012
The fourth population and housing census
.
(accessed 25 January 2015)
.
Nayak
D.
2013
The pitfalls of privatisation of higher education
.
Econ. Political Week.
48
(
5
),
15
17
.
Osabuohien
E. S.
Efobi
U. R.
Gitau
C. M.
2012
Environmental Challenge and Water Access in Africa: Empirical Evidences based on Nigeria's Households Survey. Paper presented at the 2012 Berlin Conference of the Human Dimensions of Global Environmental Change on ‘Evidence for Sustainable Development’, 5–6 October 2012, Berlin, Germany
.
Patrick
R. J.
2011
.
Health Place
17
(
1
),
386
389
.
Pickering
A. J.
Davis
J.
2012
.
Environ. Sci. Technol.
46
(
4
),
2391
2397
.
Pol
L. G.
Thomas
R. K.
2000
The Demography of Health and Health Care
.
Springer Science & Business Media
,
New York
,
USA
.
Prüss-Üstün
A.
Bos
R.
Gore
F.
Bartram
J.
2008
Safer Water, Better Health: Costs, Benefits and Sustainability of Interventions to Protect and Promote Health
.
World Health Organization
,
Geneva
,
Switzerland
.
Pullan
R. L.
Freeman
M. C.
Gething
P. W.
Brooker
S. J.
2014
.
PLoS Med
11
(
4
),
e1001626
.
Rwanda Environment Management Authority & UNEP
2009
Rwanda State of Environment Outlook. Water and wetland resources
.
Salami
A. O.
Stampini
M.
Kamara
A. B.
Sullivan
C. A.
Namara
R.
2014
.
Water Int.
39
(
3
),
294
314
.
Sorenson
S. B.
Morssink
C.
Campos
P. A.
2011
.
Social Sci. Med.
72
(
9
),
1522
1526
.
Stephens
C.
2011
.
Environ. Urban.
23
(
1
),
29
40
.
UNDP
2012
About Rwanda. UNDP in Rwanda
.
(accessed 20 January 2015)
.
UNESA
2014
World Urbanization Prospects: The 2014 Revision, Highlights. Department of Economic and Social Affairs, Population Division. United Nations
. .
UNICEF
2005
The State of the World's Children 2006: Excluded and Invisible
.
UNICEF
,
Washington DC
,
USA
.
UNICEF
2006
Progress for Children: A Report Card on Water and Sanitation (No. 5)
.
UNICEF
,
Washington DC
,
USA
.
UNICEF, World Bank
2012
Progress on Drinking Water and Sanitation. 2012 Update
.
(accessed 12 January 2015)
.
UN Millennium Project
2005
Health, Dignity, and Development: What Will it Take? Task Force on Water and Sanitation
.
(accessed 12 January 2015)
.
Uwimbabazi
P.
Lawrence
R.
2011
Compelling factors of urbanization and rural-urban migration in Rwanda
.
Rwanda J.
22
(
1
),
9
26
.
WaterAid/Development Finance International
2014
Financing of the water, sanitation and hygiene sector in Rwanda. Case Study
.
(accessed 20 December 2014)
.
WHO, UNICEF
.
2014
Progress on Drinking Water and Sanitation: 2014 Update
.
World Health Organization
,
Geneva
,
Switzerland
.
Yang
H.
Bain
R.
Bartram
J.
Gundry
S.
Pedley
S.
Wright
J.
2013
.
Environ. Sci. Technol.
47
(
3
),
1222
1223
.