I am an applied microeconomist specializing in human capital formation, the economics of artificial intelligence, and development. I am currently a Researcher at the Leuven Economics of Education Research (LEER) center at KU Leuven, and a Visiting Fellow at the London School of Economics and Political Science (LSE) within WISPPRH.

I obtained my Ph.D. in Economics (joint degree) from KU Leuven and Pontificia Universidad Javeriana (defense: March 2026), under the supervision of Professor Kristof De Witte and Professor Gloria Bernal.

Jaime Polanco-Jimenez

My research investigates the behavioral and institutional mechanisms that govern how individuals accumulate, deploy, and adapt their human capital in response to technological transitions and structural economic shocks. Methodologically, I combine large-scale Randomized Controlled Trials (RCTs) and quasi-experimental designs with modern computational methods to produce actionable, evidence-based policy insights.

A distinctive feature of my research agenda is that I do not simply study technology adoption from the outside: I engineer purpose-built experimental AI systems from the ground up. Building proprietary experimental software enables me to experimentally decouple technology access from deliberate algorithmic design choices, directly addressing empirical identification challenges that commercial platforms cannot solve. I am a certified Google Cloud Professional Machine Learning Engineer and Data Engineer.

Job Market Paper

Accepted at JEBO

The Tailoring Premium: AI Design, Behavioral Engagement, and Human Capital Formation

with Kristof De Witte. Accepted for publication at the Journal of Economic Behavior and Organization.

Standard human capital models assume educational technology automatically expands the cognitive production frontier by lowering marginal delivery costs, yet real-world adoptions frequently yield null or unequal returns. We provide the first large-scale causal evidence on why. In a randomized controlled trial with 2,440 students across secondary schools, we evaluate a purpose-built, curriculum-tailored AI chatbot against a generic AI tool and conventional instruction. We find that generic AI produces zero learning gains over standard instruction, whereas curriculum-tailored AI increases learning by 0.126 standard deviations (0.309 SD LATE). Crucially, the underlying economic mechanism is behavioral: tailored design resolves participation frictions, raising module completion by 13.2 percentage points, which explains the entire cognitive gain. Our findings demonstrate that the returns to AI in human capital formation are strictly bounded by behavioral engagement constraints, which only purposeful pedagogical and algorithmic design can resolve.

Work in Progress: AI & Cognitive Productivity

My research on technology extends from educational foundations to cognitively intensive professional work. Through registered randomized trials, I study how AI decision-support tools interact with human expertise and task allocation in knowledge-intensive sectors.

Artificial Intelligence and Productivity in Cognitively Intensive Work

with Almudena Sevilla (LSE) and Iván Vicente. Registered at AEA RCT Registry (AEARCTR-0017749).

We evaluate whether access to structured, multi-agent AI systems raises the quality and velocity of research output among early-career economists. Using a randomized design, treated researchers interact with a purpose-built diagnostic tool delivering deep evaluative feedback on research drafts. We test whether AI complements or substitutes specialist knowledge, whether effort reallocates toward judgment-intensive tasks, and whether productivity gains concentrate among researchers closer to the knowledge frontier or democratize access for those acquiring tacit expertise.

Interactive Demo: The Treatment Platform

Below is a live preview of the Referee AI platform built for this study, which simulates the diagnostic review process of top-tier economics journals.

Research Access: To test the full multi-agent environment or discuss the experimental protocols, please contact me at jaime.polancojimenez@kuleuven.be.

Human Capital, Shocks & Development in Colombia

My empirical agenda also investigates how external, regulatory, and infrastructural shocks redirect educational trajectories and labor market outcomes in Colombia:

  • Schooling Transitions & the STEM Gender Gap: Does Coeducation Close the STEM Gap? Evidence from a Single-Sex Schooling Transition in Colombia (with Gloria Bernal and Kristof De Witte). Under review at Journal of Development Economics.
  • Infrastructure & Human Capital: Road to the Future: Identifying Impacts of Roads on Education in Colombia. Published at International Journal of Educational Development, 114, 103173 (2025).
  • Energy Transitions & Youth Choices: Decisiones educativas de los jóvenes en tiempos de transición (with Brigitte Castañeda), in Retos económicos de la transición energética en Colombia, edited by Andrés Álvarez, Jorge Guerra-España, and Marc Hofstetter (Ediciones Uniandes, 2026).
  • European Policy Evidence: Factors Influencing STEM Participation and Effective Intervention Strategies (with Kristof De Witte). European Commission (DG EAC) / NESET Policy Report (2025), doi: 10.2766/4854833.
  • Forthcoming Book: The Algorithmic Ladder: RCTs, Artificial Intelligence, and the Future of Human Capital (with Kristof De Witte). Under review at Leuven University Press.

Research Fields & Methodological Focus

Primary Research Fields

  • Human Capital Economics & Education: Behavioral constraints to skill acquisition, curriculum design, and ed-tech evaluation.
  • Economics of Artificial Intelligence: Task complementarities, AI design architecture, and cognitive productivity.
  • Development Economics & Applied Micro: Natural resource windfalls, infrastructural investments, and gender equity.

Methodological Toolkit

  • Field Experiments (RCTs): Multi-country experimental design, field trials across Belgium, Estonia, and Colombia.
  • Quasi-Experimental Methods: Difference-in-differences (staggered DiD), Regression Discontinuity (Geographic RDD).
  • Proprietary Tool Engineering: Multi-agent LLM systems, full-stack cloud data pipelines (GCP), and machine learning.