UC Berkeley

This dissertation examines the decision-making of poor rural Kenyan households with respect to the adoption of point-of-use (POU) safe water technologies designed to expand access to safe drinking water in the developing world. Low-cost POU products such as chlorine and filters substantially reduce diarrhea, which kills two million children in poor countries each year. Nevertheless, POU products remain little used in many parts of the developing world, even when they are widely available at subsidized prices. This dissertation presents results from a six-month field experiment conducted in rural western Kenya that provided all participating households exposure to a variety of free POU products. The design of this study allows me to compare competing safe water products as well as to explore the primary factors that determine consumer preferences for water treatment.

In chapter 1 I consider relative consumer preferences for, and the use of, three competing POU products to understand the role of product design in adoption. My study cycled 400 households through three successive, randomly ordered two-month trials of three competing POU products. I find that households' stated preferences for products often deviate from their revealed usage behaviors. I find suggestive evidence that a product's market value plays a role in stated, but not revealed, preference. In particular, the cheapest of the three products, a liquid chlorine product branded as WaterGuard, was consistently used at the highest rates by households. Nonetheless, when households were asked to choose a six-month supply of one of the three products as a parting gift at the end of our study when all households had experienced all products, WaterGuard was chosen at the lowest rates. This divergence of stated and revealed preferences could have important implications for the scalability of all three POU products: If households will use what they won't choose (in a market setting) and vice versa, reducing disease and achieving market scalability may be two distinct problems to solve. Of course, these findings ignore the role of price, an issue we consider in chapter 3.

In chapter 2, I consider the common decision-making barriers to the adoption of any safe water product or behavior. I hypothesize that incomplete information and behavioral biases may constrain a household's decision to use a POU technology. To test these hypotheses, households were randomly assigned to receive the results of water quality tests, as well as marketing messages designed to appeal to well-known decision-making heuristics. Sharing water quality information increased water treatment by 8-13 percentage points, representing a 12-23% increase over base values. Social marketing messages that harnessed findings from behavioral economics increased water treatment by an additional 9-11 percentage points in total. In particular, framing safe water products as both increasing health and avoiding disease (not just increasing health) increased usage on the order of 4-6 percentage points. This finding is consistent with a story that loss aversion will spur greater action. A public commitment to treat water regularly had similar results, and even larger effects at households that showed "present-biased" responses to hypothetical questions about future payoffs.

Chapter 3 is split into two main parts. In the first part, I explicitly model and quantify the role of experience in changing households' valuations for POU technologies. I find that experience with a safe water product can increase households' stated willingness to pay for a safe water product, suggesting that these private health products are experience goods and offering potential insight into ways to increase demand for these products. The second part of chapter 3 considers various anomalies from the collected Kenyan data that were unexpected at the outset. I find that households are more likely to adopt a safe water technology when they receive an unannounced visit from a survey enumerator during a product trial, suggesting strong Hawthorne effects as well as a potential channel for social pressure to increase adoption. I also find that relative product preferences are strongly influenced by the order in which a household experienced a product, and these ordering effects do not fully dissipate with market experience. This finding does not support the neoclassical assumption of stable and innate preferences behind the rational model.

In total, this dissertation aims to better understand how households form preferences over safe water products as well as choose to adopt safe water behaviors. Its findings can offer promising avenues for incremental improvements in the market for safe water (and other) technologies. More broadly, it suggests that economic models of decision-making that fail to account for the accompanying psychology behind a decision can often fall short.