The effect of water quality on demand for safe drinking water in rural and peri-urban Cambodia

Abstract

Executive Summary

Policy Question

How does perceived drinking water quality affect household willingness to pay (WTP) for clean water in rural Cambodia?

Background

Inadequate access to improved water and sanitation in the developing world continues to be a major public health and development challenge. UNICEF estimates that nearly 2 million child deaths are attributable to diarrhea (UNICEF, 2008). The excessive burden of diarrhea-related morbidity and mortality is partly attributable to inadequate access to safe drinking water. Poor access to high-quality and convenient water sources may partly be a problem of low demand (Whittington et al., 2009). In fact, there is relatively limited evidence in the literature on demand for water and sanitation services.

In this paper, we consider the demand for improved water quality only (rather than changes in both quality and convenience) in two communities in Kandal province, Cambodia. Importantly, many households in these communities already have access to convenient sources of water, either in the form of private connections to piped water networks, or via rainwater harvesting and storage where they live. These sources, however, are of variable quality, and water treatment, both at the system or household-level, is inconsistent.

We hypothesize that household WTP for improved water quality depends in part on individual perceptions of the safety of existing drinking water sources. Investigating the relationship between perceptions of and demand for water quality is important for two reasons. First, households may not accurately perceive the risks associated with their existing water, which would suggest the need for education on such risks, or the provision of information related to their water quality. Second, promoters of improved access to water supply need to understand what features of such services are most important to households to achieve better targeting to consumer preferences. Understanding what role perceptions play as a determinant of WTP for improved water quality will allow for targeted interventions to affect demand for safe water.

Methodology

The contingent valuation methodology (CVM) is a commonly used way to measure demand for goods or services where markets are incomplete, imperfect, or inexistent (Dutta et al. 2005). Since markets for private water services tend to be imperfect, the CVM can be useful for capturing more accurate household demand values for improved water supply.

A number of CVM studies have been conducted in the developing world to measure the WTP for water services and treatment options. In addition, most CVM studies focusing on the water sector in developing countries have attempted to measure consumer WTP for improved water supply rather than improved water quality. Studies that have measured WTP for improved water quality in developing countries have not attempted to measure household perceptions of existing sources as a determinant of WTP for clean water.

Research in the United States that has included perceptions in WTP estimates concludes that as the perception of water quality increases, WTP for improvements in drinking water decreases. It seems probable that many of the same unobserved characteristics that influence WTP may also affect subjective quality perceptions (Whitehead 2006). Therefore, past studies that have attempted to include perceptions as an indicator of WTP for quality improvements may have had a biased coefficient on the quality perception variable in the WTP regression, because of the correlation in the error terms in the WTP and quality perception models.

In this paper, I estimate WTP for improved water quality using a basic logit model, investigating the effects of perceptions of water quality and actual water quality as covariates. I also employ a two-stage least squares model (2SLS) to address the potential endogeneity of perceptions in predicting WTP for improved water quality. I utilize the coefficients from these models to estimate the mean WTP for improved water quality.

Data

The data for this study come from a household survey of 915 households in two communities in Kandal province, Cambodia. The survey included questions on household demographics, diarrhea incidence, water storage, handling, and treatment practices, water quality perceptions, and WTP for improved water quality. Water samples were collected for all households with piped connections and tested in laboratory settings.

Perceptions of water quality were measured on a scale of 1-10 using subjective probability games. Respondents were asked both about perceptions of source water quality and water quality after in-house handling, storage, and treatment. Respondents also participated in a WTP scenario in which they tasted three samples of treated water and were offered a binary price choice for 200 liters of their favorite sample.

Results

Simple correlation tests showed that there is little correlation between actual water quality and perceptions of water quality. Households were generally very confident that their in-house water was 100% safe to drink. Results from the WTP scenario illustrate a downward shaping demand curve for improvements to improved water quality.

Estimates from the basic logit model do not show that perceptions of water quality are statistically significant in affecting WTP for improved water quality. Under certain specifications of the basic model, actual water quality is statistically significant in affecting WTP for improved water quality. However, the marginal probabilities associated with actual water quality in these models are very small.

Estimates from the 2SLS model show that perceptions of both source and in-house water quality are negative and statistically significant in affecting WTP for clean water. Both of these perception measures have very large marginal probabilities. This result implies that as households believe that their water quality is better, the likelihood of rejecting the price offer for improved water quality decreases significantly.

From the regression output, I estimate mean household WTP for improved water quality of approximately $2.50 per month for households in this sample. I also find that beliefs that existing levels of water quality are high have a significantly negative impact on WTP for improved water quality. Households that believe their in-house water quality is perfectly unsafe are willing to pay approximately $1.50 more per month than households that believe their in-house water quality is perfectly safe.

Discussion

The data gathered concerning perceptions about water quality and WTP in this study may be used to inform the design of future interventions in similar sites that have already made partial investments to improve access to water supplies. While there was not a very large sample size of actual water quality, the data suggests low correlation between perceived and actual water quality. The 2SLS models provide evidence that perceptions of water quality are greatly influencing demand, while the basic models imply that actual water quality does not have a large effect on demand. Since perceptions of in-house water quality in particular seemed to be overstated, one way to increase demand for improved water quality may be through educational campaigns to help align perceptions with actual water quality.