Essays on behavioral economics: applications to agricultural surveys and decision making

Abstract

Chapter 1 I estimate the prevalence of social desirability bias in childhood feeding reports in a UNICEF nutrition cash-plus program in Sri Lanka. Social desirability bias occurs when respondents give the socially “correct” answer, rather than the true answer. While cash benefits were not explicitly conditioned on meeting childhood feeding targets, the training, or “plus” component, made the ideal dietary outcome explicit. We test whether participants misreport the consumption of vitamin A rich foods among young children in this context using list experiments. We find households overstate adherence to program advice by 23 percentage points. The mismeasurement of one feeding component passes through and affects aggregate measures of dietary diversity. The magnitude of the findings suggests that social desirability bias could serve as a potential explanation for the persistent gap between recalled dietary intake and anthropometric outcomes in cash-plus program evaluations. The findings of this study bring together the broader measurement error and program evaluation areas of literature. Chapter 2 List experiments utilize indirect survey questions to reduce social desirability bias in measures of sensitive behaviors and sentiments. While often used to assess retrospective behavior or opinions of respondents, list experiments have not been widely applied to assessing “deep” parameters of economic models, such as willingness to pay. Common stated preference methods of estimating willingness to pay may be impacted by social desirability bias, particularly when a product has been provided to survey recipients for free. List experiments can uncover the share of respondents willing to pay a given price while reducing social desirability bias. Repeating the method at a variety of prices recovers a partial demand curve. This study discusses the conditions required to satisfy the list experiment validity assumptions and demonstrates the method in an e-extension platform randomized control trial in Sri Lanka. We show that the “no design effect” assumption for list experiments requires that the budget constraint for a household be nonbinding. Under conditions where that assumption is likely to hold, we find direct estimates overstate willingness to pay at low prices. Our findings suggest list experiments may provide a cheap method of more accurately assessing the typically large share of respondents unwilling to pay any non-zero-sum (extensive margin), but are less effective at reducing bias from exaggerated demand (intensive margin). Chapter 3 Recent events influence decision-making in many domains, including insurance. Pasture, Rangeland, and Forage (PRF) insurance is a novel product that is relatively underused compared to other federal crop insurance programs. As such, we may expect producers to use recent events as a heuristic in decision making. We estimate the impact of recent events on PRF insurance decision making on enrollment decisions and interval selection. We use a county-level two-way fixed effects model to identify the impact of a recent PRF experience, a Livestock Forage Disaster Program (LFP) payment, and weather events on PRF enrollment and the share of acres enrolled in the growing season. A recent positive experience with PRF (defined as an increase in indemnity payments relative to premiums) is associated with increased participation in PRF and but has no relationship with interval selection. LFP payments on the other hand, are consistently positively related to PRF enrollment and selection of growing season intervals. These findings suggest that producers may view PRF and LFP as complements rather than substitutes.