Working papers

Abstract: Observed performance in knowledge-intensive tasks reflects both genuine effort and substitution of effort to artificial intelligence (AI), neither of which is directly observed. This paper shows that relative substitution of effort to AI is identifiable from observed performance measures. Identification relies on matching tasks where the use of AI is allowed to comparable tasks performed in controlled environments where the use of AI is reliably restricted. As a proof-of-concept, the framework is implemented in a small higher education setting but can be extended to any knowledge-intensive workplace. The results suggest that the use of AI in knowledge-intensive tasks creates a mismatch between observed performance and underlying human capital. This introduces a measurement error in performance, compressing its observed distribution relative to the underlying distribution of human capital, and carrying implications for the design of tasks in environments where effort is unobservable, but performance outcomes depend on the use of AI. 

Download the replication package

Under review at the Journal of Human Capital

Abstract: The use of Generative Artificial Intelligence (Gen AI) in economics classrooms has the potential to both improve learning-democratizing education, personalizing tutoring, and summarizing content-as well as harm it-reducing student effort, offloading critical thinking, and encouraging cheating. It thus behooves instructors to understand how Gen AI is used in their classes. To that end, we design comprehensive surveys on the use of Gen AI by both economics instructors and their students. Surveys are completed at the start and end of the 2025 Spring semester and provide a snapshot of perceptions, policies, and usage of Gen AI tools. Based on our results, we find a disconnect between instructor Gen AI policies and student perceptions of those policies. This implies the need for better communication. In addition, survey results show how students use Gen AI to help with both studying and completing assignments. While sentiment about usage of Gen AI for studying aligns between instructors and students, more instructor guidance in how best to use Gen AI for learning is recommended. Unsurprisingly, we find divergence between instructor and student sentiment regarding the use of Gen AI for completing assignment.

Available here

R&R at the Journal of Economic Education

Abstract: We study how populist governance has affected sovereign yields in Europe since 1960. Using fixed effects and synthetic control methods (SCMs), we find comprehensive evidence that bond markets impose a populist risk premium, which we define as an increase in sovereign yields in periods of populist governance. We also show that the populist risk premium (PRP) varies across the term structure of bond yields and the ideological position of the incumbent populist. While short-term yields seem unaffected, the longer end of the yield curve sees a positive PRP, particularly for left-wing populists. SCMs confirm the substantial heterogeneity in the PRP across the term structure, as do several robustness checks. This is consistent with recent evidence of a lasting impact of populist governance on income levels, triggering substantial long-term fiscal risks. There are two primary mechanisms underlying the observed heterogeneity in the PRP: fiscal and monetary policy convergence in Europe, and stronger internal constraints on executive power relative to those in South and North America. 

R&R at the Journal of Comparative Economics

Abstract: We build a theory of politically optimal environmental regulations and test its predictions using a new individual-level data set of climate reform votes in the European Parliament (EP). This informs the most comprehensive analysis to date of how Members of the European Parliament (MEPs) differ in their legislative support for climate reforms. In line with the theory, we find that ideological polarisation among MEPs does not follow a traditional left-right divide. Instead, MEPs at the ideological extremes appear similarly averse to pro-climate measures, while the political centre emerges as the strongest advocate for climate reforms. The apparent extreme convergence against centrist compromises emerges naturally from the theory. However, it does not imply a shared aversion to climate reforms by the ideological extremes. We discuss several mechanisms that may generate these results. Our results identify legislator profiles likely to advance climate reforms in the EP, and they are mostly found in the ideological centre. 

Supplementary data and code here

Abstract: We estimate the effects of California’s cap-and-trade program (CCTP) on the greenhouse gas (GHG) emissions of California’s manufacturing, waste, and electricity-generation (EG) sectors 8 years after the programme has taken effect. We also estimate its effects on California’s relative change in carbon intensity of its thermal EG plants and renewable-energy’s share of total electricity generated. Using a difference-in-differences approach, we are unable to identify a reduction of GHG emissions of California’s affected manufacturing plants relative to the closely matched control group of facilities in other states, although we do find a 12.5% abatement effect on California’s waste sector. In addition, we estimate higher CO₂ emissions and carbon-intensity of its EG sector relative to that of the control group of states. Contrary to theory predictions, our evidence suggests that the first 8 years of the CCTP were largely ineffective in delivering a simultaneously low-cost and accelerating path to industrial net zero. Therefore, while alternatives exist and changes to the scheme might yield results in the future, the CCTP does not offer a convincing blueprint for industrial net zero policies thus far.

Supplementary material (data and code) here