Research 

Discrimination remains pervasive, yet little is known about how past experiences of discrimination shape future discriminatory behavior, and hence the dynamics of discrimination. This paper introduces and empirically documents retaliatory discrimination, whereby individuals are more likely to discriminate against a group after perceiving past discrimination from that group. Guided by a theoretical framework that models retaliatory discrimination alongside taste-based and statistical discrimination, I conduct experiments with refugees in Uganda and men in the United States. These provide empirical support for retaliatory discrimination, driven by motivated beliefs, while ruling out other mechanisms. The findings reveal that retaliatory discrimination increases the number of discriminators, rather than the intensity of discrimination. I experimentally show that retaliatory discrimination can generate unintended consequences when affirmative action policies are removed, as well as providing a micro-foundation for anticipated discrimination and the formation of discriminatory tastes. Finally, I discuss and experimentally evaluate a novel mitigation strategy designed to reduce retaliatory discrimination by highlighting the salience of future rounds. 

Academic Presentations: Graduate Student Society Tilburg University Seminar, Tilburg University Internal Seminar (scheduled), UCL Enter Seminar (scheduled)

AEARCTR-0015358 and AEARCTR-0016047

Can mental accounting help economically disadvantaged people accumulate capital and grow their income? We conducted a field experiment with 861 refugee households in Uganda, who received unconditional cash transfers over seven months. Treatment households could divide their monthly transfers among four labeled envelopes (Education, Health, Investments,  Other), while control households received cash in a single, unlabeled envelope. Demand for the labeled envelopes was high: 93\% of treatment households opted in, and 37\% were still using them a year after the program ended. Compared to the control group, treatment households significantly increased investments, particularly in lumpy assets, leading to higher income and savings after one year. Effects were larger among households that kept using the envelopes, who also reported improved budgeting, planning, and spending discipline.

Academic Presentations: KVS New Paper Series (2023); Summer School in Development Economics (2023); Maastricht University Interdisciplinary Brownbag Seminar of Behavioral Sciences (2024); Development PhD Workshop in Lund (2024); Tilburg University Economics Department Internal Seminar (2024); Development Economics & Economic History Research Group (2024); Tilburg University Zero Hunger Lab (2025); UC Berkeley Development Lunch (2025), PacDev Conference (2025); MWIEDC Conference (2025); LIC/path2dev/BREAD Conference on Development Economics (2025).

Policy Presentations: Danish Refugee Council; IMPACT Initiatives; UNHCR; WFP; 100Weeks; ZOA; GiveDirectly.

AEARCTR-0010472 

Winsorizing and trimming are used to minimize the effects of outliers and measurement errors on estimated treatment effects. The most common approach winsorizes/trims the tails of the whole sample, even if there are heterogeneous subgroups within the sample – for example a treatment and control group in a Randomized Controlled Trial. An alternative approach – called Stratified Winsorizing/Trimming – winsorizes subgroups separately, ensuring that an equal proportion of observations are winsorized/trimmed from each subgroup. Monte Carlo simulations of a Randomized Controlled Trial illustrate that Stratified Winsorizing/Trimming reduces the bias on the estimated treatment effect and the risk of Type II errors compared to the traditional approach of winsorizing/trimming, albeit at the cost of a greater likelihood of Type I errors. Applications to Angelucci et al. (2023) and Jack et al. (2023) illustrate that the choice of winsorizing/trimming technique can affect both the magnitude and statistical significance of treatment effects. The paper concludes by discussing practical implications for different empirical strategies and Pre-Analysis Plans when researchers want to winsorize/trim a sample that consists of subgroups.

Large Language Models (LLMs) have the potential to transform the labor market, including hiring. This paper assesses their impact on signals that job-seekers send to potential employers and how this affects labor market matching. Through two field experiments, focusing on cover letters and involving job-seekers and recruiters, we document that LLMs enhance the quality of signals, particularly benefiting lower-quality applicants. However, these improvements do not translate into increased interview invitations because the improvements are concentrated in standardized, less influential sections of the cover letters. When recruiters are explicitly informed of candidates’ use of LLMs, they place greater value on high-quality cover letters crafted without AI assistance. Our findings indicate that LLMs reduce the informativeness of signals, potentially leading to increased inefficiencies in labor market matching.

AEARCTR-0013355 and AEARCTR-0014784

Climate scientists are in agreement that the timing of greenhouse gas emissions plays a crucial role in climate change mitigation, with early reductions being more effective than equally sized emissions reductions in the future. This paper incorporates this scientific fact within Climate Protection Games, a multi-round collective risk threshold public goods game that simulates countries’ greenhouse gas emissions reductions over time, by introducing temporal dynamics that emphasize the importance of early action at mitigating climate change. A laboratory experiment with 300 participants compares two treatments: the status quo where emissions reductions are equally weighted across time (Linear), and another where earlier emissions reductions have a greater impact than equally sized later emissions reductions, reflecting the greater environmental benefits of early action (Step). Results indicate that while the likelihood of reaching the necessary threshold to avert disastrous climate change does not differ significantly between treatments, the Step treatment leads to higher early emissions reductions and increased individual payoffs, suggesting welfare improvements. Evolutionary Game Theory simulations corroborate these findings, showing that early emissions reductions are welfare-enhancing. By highlighting the significance of early emissions reductions, this paper can contribute to the discourse on climate policy, suggesting that strategies incentivizing early emissions reductions can result in better environmental and welfare outcomes.

Academic Presentations: GSS Seminar Tilburg (2022); RGS Doctoral Conference (2023); TIBER Symposium (2023)

AsPredicted #99114