Essays in Development Economics

This dissertation contains three chapters, which cover different topics that are of interest to development economists, and economists more generally.
The first two chapters make empirical contributions to policy relevant and highly studied topics in development economics: discrimination and cash transfers. Chapter 1 introduces a new source of discrimination, called retaliatory discrimination, whereby individuals are more likely to discriminate against members of a group after perceiving that they were previously discriminated against by other members of that group. Experiments in Uganda and the United States offer empirical support for retaliatory discrimination, while ruling out alternative explanations.
Chapter 2 conducts a Randomized Controlled Trial in Uganda among 861 refugee households that are eligible for a 7-month-long unconditional cash transfer. We introduce and randomize access to a mental accounting intervention, in the form of four labelled envelopes (see the front cover illustration). The labelled envelopes, which cost $1.78 and hence 0.46% of the total cash transfer’s value, result in 18% larger monthly earnings, and 22% greater savings one year after the end of the cash transfer program. This is driven by treated households spending 26% more on productive investments, particularly in “lumpy” investments that require large upfront investments (31% increase).
The third chapter makes a methodological contribution. Researchers are often concerned about the role of outliers in a dataset and hence use techniques to minimize their influence. Two of the most common techniques are to winsorize and trim outliers. Chapter 3 discusses two winsorizing/trimming techniques in the presence of sub-groups in the data. Monte Carlo simulations illustrate that the chosen winsorizing/trimming technique can affect treatment effect estimates, as well as Type I and II errors. Application to two published papers highlight that the chosen technique can affect the conclusions drawn from the data.