Drinking water and sanitation: progress in 73 countries in relation to socioeconomic indicators

Abstract Objective To assess progress in the provision of drinking water and sanitation in relation to national socioeconomic indicators. Methods We used household survey data for 73 countries – collected between 2000 and 2012 – to calculate linear rates of change in population access to improved drinking water (n = 67) and/or sanitation (n = 61). To enable comparison of progress between countries with different initial levels of access, the calculated rates of change were normalized to fall between –1 and 1. In regression analyses, we investigated associations between the normalized rates of change in population access and national socioeconomic indicators: gross national income per capita, government effectiveness, official development assistance, freshwater resources, education, poverty, Gini coefficient, child mortality and the human development index. Findings The normalized rates of change indicated that most of the investigated countries were making progress towards achieving universal access to improved drinking water and sanitation. However, only about a third showed a level of progress that was at least half the maximum achievable level. The normalized rates of change did not appear to be correlated with any of the national indicators that we investigated. Conclusion In many countries, the progress being made towards universal access to improved drinking water and sanitation is falling well short of the maximum achievable level. Progress does not appear to be correlated with a country’s social and economic characteristics. The between-country variations observed in such progress may be linked to variations in government policies and in the institutional commitment and capacity needed to execute such policies effectively.


Introduction
The United Nations recognizes the basic human right to water and sanitation. 1,2 Accordingly, the international community, through the recent adoption of the Sustainable Development Goals (SDGs), has made a commitment to achieve universal and equitable access to safe drinking water and adequate sanitation by 2030. 3 The SDGs build on the Millennium Development Goal (MDG) target 4 to halve, between 1990 and 2015, the proportion of the population without access to safe water and basic sanitation. During the MDG period, some countries have made substantial progress, while others have stagnated. 5 The national characteristics that may enhance or hinder progress on water and sanitation are poorly understood. For example, external finance should make it easier for governments to improve drinking water and sanitation coverage. While a positive correlation between aid received and improvements in such coverage has been observed in some studies, 6,7 other studies have not detected such a relationship. [8][9][10][11] The differing results may be due to limitations in the methods used 8 and/or the choice of indicator used to measure progress. Progress has been measured as population access to improved drinking water and sanitation -or the change in such access over a specified period. However, changes in population access are not necessarily comparable across different countries because, as a country approaches universal access, it becomes increasingly difficult to reach those who still lack access.
The aim of the present study is to determine whether progress in improving access to improved drinking water and sanitation, achieved by countries between 2000 and 2012, is associated with national socioeconomic characteristics. We used a new indicator of progress -the normalized rate of change in access -to allow countries to be compared, regardless of their initial coverage levels.

Data sources
We obtained estimates of the percentage of national populations with access to improved sanitation and water -for various years between 2000 and 2012 -from the 2013 Country Files of the Joint Monitoring Programme for Water Supply and Sanitation 12 -which were the most up-to-date information available at the time of analyses. This World Health Organization/United Nations Children's Fund programme compiles the results of nationally representative surveys, including Demographic and Health Surveys, Multiple Indicator Cluster Surveys, World Health Surveys and national censuses. We considered only data from 2000 onwards to reflect the progress countries made since the MDGs were set in the year 2000.
We included shared toilet facilities in our improved-sanitation category because data for both shared sanitation and total improved sanitation including shared sanitation -i.e. the two data sets needed to investigate total improved sanitation excluding shared sanitation -were only available for four of our study countries. The Joint Monitoring Programme currently discounts shared sanitation from total improved sanitation by applying a fixed ratio for each country. 13 However, since these ratios are based on data that may have been collected before 2000 and, for some countries, are based on a single data point, we decided not to use them -or any other similar correction factor -in our analyses. We included countries with at least five data points that covered at least three different years. Multiple survey data points from any one year were treated independently.
Objective To assess progress in the provision of drinking water and sanitation in relation to national socioeconomic indicators. Methods We used household survey data for 73 countries -collected between 2000 and 2012 -to calculate linear rates of change in population access to improved drinking water (n = 67) and/or sanitation (n = 61). To enable comparison of progress between countries with different initial levels of access, the calculated rates of change were normalized to fall between -1 and 1. In regression analyses, we investigated associations between the normalized rates of change in population access and national socioeconomic indicators: gross national income per capita, government effectiveness, official development assistance, freshwater resources, education, poverty, Gini coefficient, child mortality and the human development index. Findings The normalized rates of change indicated that most of the investigated countries were making progress towards achieving universal access to improved drinking water and sanitation. However, only about a third showed a level of progress that was at least half the maximum achievable level. The normalized rates of change did not appear to be correlated with any of the national indicators that we investigated. Conclusion In many countries, the progress being made towards universal access to improved drinking water and sanitation is falling well short of the maximum achievable level. Progress does not appear to be correlated with a country's social and economic characteristics. The between-country variations observed in such progress may be linked to variations in government policies and in the institutional commitment and capacity needed to execute such policies effectively.

Indicator of progress
To compare countries with differing initial levels of population access to improved sanitation and water, we defined the progress of country i as its normalized rate of change in access: normalized rate rate min rate max rate min rate where normalized rate i,j is the normalized rate of change for country i that had a baseline coverage level j in the year 2000; rate i,j is the absolute rate of change for country i at coverage level j; max. rate j is the maximum rate achievable by any country at coverage level j (based on historical data, see below) and the min. rate j is set at zero (no progress). Each country's absolute rate of change was calculated from the earliest available year (2000 in most cases) using linear regression.
We determined values for the maximum rate achievable at each coverage level using the frontier approach. 14,15 Historical absolute rates of change for all countries were plotted as a function of the national coverage level for the year 2000. For countries that had survey data for 2000, we used those values for national coverage level. For countries that did not have surveys for 2000, we used estimates from the Joint Monitoring Programme. 12 The best-performing countries (which we refer to as frontier points) delineate an upper boundary or frontier against which the performance of the other countries can be compared. We used the frontier efficiency analysis package 16 in R software 17 to identify frontier points.
A polynomial curve was fitted through the frontier points to obtain the frontier curve -with the requirement that the curve must pass through the point corresponding to 100% coverage and 0% increase in coverage per year. The frontier curve allowed the maximum achievable rates of improvement in water and sanitation coverage to be estimated for all countries, depending on their initial level of coverage (Table 1). Using the estimated maximum achievable rates and Equation 1, we obtained the normalized rates of change for our study countries. The requirement that the frontier curve must pass through the point corresponding to 100% coverage and 0% increase in coverage per year meant that the frontier curve -which is the fitted polynomial equation -sometimes fell below a frontier point. This resulted in a normalized rate greater than 1 for some frontier countries. We assigned a normalized rate of 1 to all such countries. Similarly, for countries in which we found access to improved drinking water and sanitation to be decreasing, we limited the negative normalized rate to -1. All of the normalized rates we report therefore fall between -1 and 1.

Regression analyses
We used regression analyses to investigate the relationship between progress in water and sanitation and the following national socioeconomic indicators: (i) gross national income per capita -in current United States dollar (US$) values that had been derived using the Atlas method; 18 (ii) government effectiveness; 19 (iii) the per-capita level of official development assistance for sanitation and water -calculated, in constant 2011 values, by dividing the total assistance disbursed from all donors 20 by the total population; 21 (iv) the volume of renewable internal freshwater resources percapita; 18 (v) the percentage of the female population older than 25 years that had completed secondary education; 22 (vi) the percentage of the population with a daily income of less than US$ 1.25; 18 (vii) the Gini coefficient; 18 (viii) the mortality rate among children younger than five years; 18 and (ix) the human development index -a composite index reflecting life expectancy, education and income. 23 For each indicator and country, we used the value for the year 2000 or, if that value was not available, that for the closest available year.
We initially considered data from the World Health Organization's Global Analysis and Assessment of Sanitation and Drinking Water reports, which provide policy and economic indicators such as the per-capita budget for drinking water and sanitation from the year 2010 24 and per-capita expenditure on sanitation and water in the year 2014. 25 However, as these data relate to time periods that are at least 10 years off from our target year of 2000 -and indicators such as expenditures per capita may vary substantially from year to year -we decided not to include them in our analyses.
-- Several of the nine national characteristics we investigated were highly correlated. We therefore used principal components analysis on the nine national indicators to obtain uncorrelated synthetic independent variables (Table 2). However, based on the Kaiser criterion, we only used the three synthetic variables that gave eigenvalues greater than 1 -which together accounted for 76% of the variance in the data observed -in our regression analyses. Backward stepwise regression -with P-values of 0.05 and 0.10 for the addition and deletion of variables, respectivelywas also used to identify a subset of the three synthetic independent variables for the regression analyses.
Univariate and multivariate regression analyses were performed in Stata version 12 (Stata Corp. LP, College Station, United States of America). We ran models using the data from all of our study countries and, separately, using only the data from those study countries that had no armed conflict between 2000 and 2012. 26 While regression results do not necessarily provide information on causality, a predictive empirical model could be useful in estimating the progress towards universal access in countries where sanitation and water data are not available. We analysed the relationship between the normalized rates of change and the nine national indicators that we investigated, as independent variables, using a linear model: and a fractional logistic model: where β 1 to β i are the fitted model coefficient values and x 1 to x i are the independent variables. Countries with negative normalized rates were excluded from the fractional logistic regressions because, for these, the output parameter must lie between 0 and 1. These regressions therefore focused only on countries that Drinking water and sanitation in 73 countries Jeanne Luh & Jamie Bartram had made progress in increasing access to improved sanitation and water. We re-ran the models using the synthetic independent variables.

Country pairings
We selected countries where, despite similar initial coverage, we observed marked differences in progress. To understand possible reasons for these differences in progress, we chose discordant pairs of countries within the same geographic region and with similar characteristicsas defined by the country clusters of Onda et al. 27 -and compared their national socioeconomic indicators.

Results
National access to improved sanitation and water in the year 2000 and historical absolute rates of change are shown in Table 1. Relatively few relevant data were available from high-income countries that are approaching or have already achieved universal access. High-income countries were therefore not well represented in our analyses. The absolute rates of change in access to improved drinking water and sanitation ranged from -0.9% to 3.5% per year (67 countries) and from -1.4% to 3.2% per year (61 countries), respectively. The frontier curves used to calculate the maximum rates of change in Equation 1 -shown as solid lines in Fig. 1 -were constructed using five frontier points for water -based on data from Armenia, Egypt, Ethiopia, Ghana and Samoa -and eight frontier points for sanitation -based on data from Benin, Egypt, Estonia, Ethiopia, Honduras, Lao People's Democratic Republic, Rwanda and Sri Lanka. For water, Mali was identified as an outlier 28,29 and not used to construct the frontier curve. The frontier curves for both sanitation and water indicate decreases in the maximum achievable rate of change as countries approach 100% coverage.
While positive and negative absolute rates indicate countries with increasing and decreasing coverage, respectively, only the normalized rates in Table 1 should be used to compare the performances of the study countries. These normalized rates indicate that, over our study period and for both water and sanitation, only about one in every three of our study countries progressed at a rate that was at least half of their maximum

Research
Drinking water and sanitation in 73 countries Jeanne Luh & Jamie Bartram achievable rate -i.e. they had normalized rates that were greater than 0.5. Among the countries with relevant data, 20 (30%) of 67 had normalized rates for water that fell below 0.25 and 21 (34%) of 61 had the same low normalized rates for sanitation.
Using the normalized rate as our indicator of progress, only two univariate regression models for access to drinking water -and no models for sanitation -were statistically significant overall (P ≤ 0.05; Table 3). However, the model fit was poor (adjusted R 2 < 0.2) and Fig. 2 and Fig. 3 show the poor agreement between the observed and modelled estimates.
Multivariate regression with the three synthetic independent variables resulted in two models -i.e. one for water and one for sanitation -that were statistically significant (Table 3). Again, however, there was poor agreement between the observed and the modelled estimates ( Fig. 3 and Fig. 4).
Overall, our results show no correlation between the normalized rates of change in the improvement of access to drinking water or sanitation and any of the nine national indicators that we investigated or any of the principal components obtained from these indicators. A similar lack of correlation was observed when the analyses were performed using the most recent data available for each of the nine national indicators (available from the corresponding author).
An analysis of the illustrative pairs of countries with differing progress indicate that no single indicator was consistently associated with progress in coverage for water or sanitation (Table 4 or Table 5, respectively, available at: http://www.who. int/bulletin/volumes/94/2/15-162974).

Discussion
The historical absolute rates of change in access to sanitation and water varied greatly at all coverage levels. Over our study period, most countries increased their sanitation and water coverage. Ethiopia and the Lao People's Democratic Republic, for example, showed absolute rates of change -in access to both drinking water and sanitation -in excess of 2.2% per year. Although several countries were found to have decreasing sanitation or water coverage, only one of the countries we investigated -Zimbabwe -showed decreasing coverage for both sanitation and water. We determined normalized rates of change to compare progress between countries. For example, while both Kenya and South Africa had an absolute rate of change of 0.70% per year for water, the corresponding normalized rate for Kenya (0.28) was markedly lower than that for South Africa (0.49)indicating that South Africa was making greater progress than Kenya.
National socioeconomic characteristics may not be primary determinants of progress in access to water and sanitation. For example, from the illustrative country pairings, Peru might be expected to make better progress than Paraguay -since, per capita, Peru has the greater gross national income, external financial assistance and renewable freshwater resources. However, the normalized rates that we calculated indicate that, over our study period, Paraguay was making good progress whereas Peru was making no progress. Factors

Second component from principal components analysis used as independent variable
Notes: The plots show estimates from linear regressions, with either the Gini coefficient or the second component from principal components analysis used as the independent variable. The solid lines indicate a perfect match between the observed rates and the modelled estimates.
Drinking water and sanitation in 73 countries Jeanne Luh & Jamie Bartram other than the nine national indicators we investigated are probably more important than those indicators in determining progress towards universal access. For example, government policies -and variation in the provision of the institutional commitment and capacity needed to execute such policies effectively -may be important determinants of such progress. The lack of association we observed between progress and per-capita level of official development assistance is consistent with previous studies 8-11 -although these earlier investigations used different measures of progress and varied in their scale, from global to city level. Our study has several limitations. We calculated absolute rates of change in coverage of water and sanitation using a linear fit to the data points -even though progress may have been nonlinear during our study period. This may affect the estimated rates of change, the identification of frontier countries and consequently, the frontier curve, the corresponding maximum rates and the normalized rates. Household surveys used as our data sources did not include extra-household settings -e.g. educational institutions, workplaces and health-care settings -and therefore did not represent sanitation and water access for all dimensions of society. Neither did the surveys distinguish between the different levels of improved sanitation or water services -e.g. between a household tap and a community hand pump or between a pit latrine and a sewer connection. Furthermore, inequalities in access often exist. Coverage and service levels tend to be relatively poor among marginalized and vulnerable groups and this may not be captured by national surveys. Identification of the disadvantaged groups in each country is needed so that progress among these groups can be compared with that in the general population.
With respect to our regression analyses, we recognize that the variables we used as national economic indicators may not accurately reflect the levels of investment in sanitation and water. For example, such indicators exclude the many household investments, particularly in sanitation, that occur in developing countries. In addition, the data for the nine national indicators that we investigated were for a single year and did not cover all of our 2000-2012 study period. Alternatives to linear and logistic regression, such as generalized additive models, need to be tested in future studies.
Use of normalized rates allowed countries to be compared regardless of their coverage level, aligns with the human rights principle of progressive realization and could be extended to measure progress in other health sectors -e.g. to measure rates of improvement in the maternal mortality ratio. Use of such quantitative measures of progress allow policy-makers to make evidence-based decisions and provide the human rights community and others with an objective method for country comparison. Our results indicate that, in many countries, the progress being made towards universal access to improved drinking water and sanitation is far from the maximum achievable. The lack of relationship between the normalized rates of change and the nine national indicators that we investigated is important -particularly with respect to the economic variables. The finding that official development assistance is not correlated to our indicator of progress suggests that investment alone is not sufficient to ensure progress. In future studies, the effect on progress of additional variables that assess the enabling environment and governance should be investigated. ■

Agua potable y saneamiento: progreso en 73 países en relación con los indicadores socioeconómicos
Objetivo Evaluar el progreso del suministro de agua potable y saneamiento en relación con indicadores socioeconómicos nacionales. Métodos Se utilizaron los datos de una encuesta domiciliaria realizada en 73 países (recogidos entre los años 2000 y 2012) para calcular la tasa de variación lineal del acceso mejorado de la población al agua potable (n = 67) y/o saneamiento (n = 61). Para poder comparar el progreso entre países con distintos niveles iniciales de acceso, las tasas de variación calculadas se normalizaron para abarcar entre -1 y 1. Se realizaron análisis de regresión en los que se investigó la relación entre las tasas de variación normalizadas del acceso de la población y los indicadores socioeconómicos nacionales: el producto interior bruto per cápita, la eficiencia del gobierno, la ayuda oficial al desarrollo, los recursos de agua dulce, la educación, la pobreza, el coeficiente de Gini, la mortalidad infantil y el índice de desarrollo humano. Resultados Las tasas de variación normalizadas indicaron que la mayoría de los países investigados estaban progresando hacia un acceso universal mejorado al agua potable y saneamiento. No obstante, únicamente un tercio de ellos mostró un nivel de progreso equivalente a, al menos, la mitad del máximo del nivel alcanzable. Las tasas de variación normalizadas no mostraban signos de estar relacionadas con ninguno de los indicadores nacionales que se investigaron. Conclusión En numerosos países, el progreso conseguido hacia un acceso universal mejorado al agua potable y saneamiento no logra alcanzar el nivel máximo alcanzable. El progreso no parece estar relacionado con las características sociales y económicas de un país. Las variaciones entre países observadas en dicho progreso pueden estar relacionadas con las modificaciones de las políticas gubernamentales y el compromiso y la capacidad necesarios de las instituciones para ejecutar tales políticas de forma eficaz.  27 b The lower the Gini coefficient, the greater the equality. c As defined by the World Bank. 19 The higher the value, the stronger the performance of governance. The lower the Gini coefficient, the greater the equality. c As defined by the World Bank. 19 The higher the value, the stronger the performance of governance.