Researchers have long speculated that the demand for water supply improvements is related to subjective perceptions of water quality (Mitchell and Carson, 1989; Whitehead, 2006). Yet surprisingly little is known about the role such perceptions play in whether or not development projects in the water sector actually succeed. This is at least partly because of the fact that isolating the effects of such perceptions has proven difficult in practice, due to both measurement and endogeneity concerns. Challenges in measurement stem from the difficulty of obtaining meaningful responses about water quality beliefs from typical survey methods; endogeneity concerns arise because the same unobserved characteristics of respondents that explain willingness-to-pay (WTP) are also likely related to subjective perceptions.

This study aims to contribute to bridging this knowledge gap in two ways. First, we incorporate new methods for eliciting water quality beliefs from respondents, borrowed from the literature on perceptions of climate change (Delavande et al., 2010). Through these exercises, we obtain household-specific beliefs about drinking water safety immediately following collection from the source, as well as just prior to drinking (following transport to the house, storage and in-house handling). Second, in order to address concerns over the endogeneity of perceptions with WTP, we produce an exogenous information shock to water quality beliefs in a random subset of our sample. This exogenous shock is created by providing water quality testing and results to half of households in our sample, for both source and drinking water quality, prior to elicitation of WTP. This design is developed to test the hypothesis that information on water quality shifts perceptions of the safety of consuming it, which in turn affects WTP for water treatment.

 Prior to the surveys carried out during this experiment, we interviewed 915 in 16 villages in Kandal Province, near Phnom Penh, Cambodia. Households in these communities accessed a variety of water sources – rainwater harvesting, piped water, wells, vendors, and surface waters – that vary by season. We found that perceptions of water quality were significantly correlated with a variety of water storage and handling behaviors, and that perceptions of the safety of drinking water were negatively related to WTP for safe water, which was about $3 per month on average. We also observed that water quality perceptions were uncorrelated with measures of actual water quality. These results suggest that the mechanisms driving household behavior and demand for enhanced water quality may be related to erroneous perceptions of the safety of existing sources of drinking water.