Andrés F.
Rodríguez

We present initial results from a randomized evaluation that varies the degree of decentralization within Zambia's Constituency Development Fund, which has expanded twenty-fold since 2021. In partnership with the government, we randomly assigned districts to a reform with two components: reducing central government influence over local community project selections, and expanding districts' discretion to use light-touch procurement processes. District officials internalized both reforms, obtaining project approvals substantially faster and shifting hundreds of tenders into simpler procurement processes. Increased discretion modestly improves the sectoral alignment of selected projects with local demand while, at a minimum, not exacerbating existing biases in the spatial allocation of resources. Reduced bidding on tenders appears to reflect a reduction in non-competitive bids rather than a consequential reduction in competition. Ongoing and future data collection seeks to evaluate downstream outcomes on project construction, utilization, financial management, and electoral accountability.

This paper investigates the relative merit of traditional leaders in promoting cooperative behavior when informally taxing citizens' labor in low-income states and whether this comes at the expense of relatively poor households. I designed a field experiment to measure whether citizens distort their behavior when facing informal taxation schemes by engaging in costly actions to avoid contributing their labor to a public good. I randomize communities into different methods to select contributors and compare the status quo of selection by chiefs versus two alternatives: random lotteries and progressive selection based on household surveys. Results show that chiefs' informal taxes distort individual behavior slightly less than the alternative mechanisms, revealing that chiefs can indeed promote pro-social behavior but not in a magnitude suggesting large efficiency gains. However, chief selection leads to less overall participant effort during the experiment, suggesting that chiefs' involvement does alter people's incentives during income-generating tasks. When focusing on differences by wealth, I also find chiefs do generate distortions among this group, which are more likely to accept lower wages to avoid being informally taxed by chiefs. On the other hand, progressive selection is slightly more distortionary than chief selection in terms of people avoiding informal taxes through costly actions, shifting costs toward people with intermediate levels of wealth.

Can structured cross socioeconomic status (SES) collaboration in elite universities build economic connectedness? We study this question in a field experiment embedded in a required undergraduate course at Universidad de los Andes during a period of rapidly rising low-SES enrollment. We map within-section friendship networks and elicit students' willingness to trade off working with a friend for working with a higher-performing non-friend, using randomized monetary incentives and randomized SES information about the potential teammate. We then randomly assign students to essay teams that vary in SES composition and instrumentally pair them with a random friend to work with. Baseline networks display strong SES homophily. Making SES salient increases high-SES students' preference to work only with friends, while low-SES students' choices become more homophilous at the top of the performance distribution. In realized teams, SES diversity reduces meeting and satisfaction and lowers willingness to work together again, with little effect on essay grades. These frictions persist in a follow-up survey, while friend-pairing mainly benefits low-SES students.

The high failure rate of new businesses in Colombia suggests a high entry rate among entrepreneurs with only low success potential. We test the hypothesis that this may be driven by the tendency to incorrectly estimate their success likelihoods. Specifically, we test the relative roles of ambiguity preferences, confidence, and optimism and whether there are heterogeneities in these biases with regard to the true underlying potential for success as an entrepreneur. We recruit current, past, and aspiring Colombian entrepreneurs to participate in an online experiment. As part of the study, we simulate the decision to become an entrepreneur by having participants make an "entry decision" about playing an entrepreneurship-related game for stakes or taking a certain payment. We vary the available information to disentangle the potential behavioral biases leading to over-entry. We find a high degree of entry at the baseline, with over 80% of participants choosing to play the game for stakes across the board that is statistically significantly higher than what risk preferences can account for combined with the expectation of 20% succeeding. Providing information about the average likelihood of success encourages the lower potential participants to enter more, closing the slight gap between them and their higher potential counterparts at baseline. Information about the individual-specific skill-based success likelihood discourages some lower potential individuals, in some circumstances, suggesting over-confidence. Sharing information about the role of luck may discourage the lower potential participants, providing evidence against high levels of optimism or risk-seeking in this group. Taken together, the findings show that overconfidence among the lower potential individuals generates over-entry in this group, but the resulting high failure rate does not deter high-potential individuals' entry.