RESEARCH

Abstract: We describe the design, implementation, and evaluation of a low-cost and scalable program that supports women in Poland in transitioning into jobs in the information technology sector. This program, called “Challenges,” helps participants develop portfolios that demonstrate capability for relevant jobs. We conduct two independent evaluations, one focusing on the Challenges program and another on a one-to-one mentoring program. We exploit the fact that both programs were oversubscribed to randomize access among applicants and measure the impact of the programs on the probability of finding a job in the technology sector within four months. We estimate that the mentoring program increases the probability of finding a job in technology by 13 percentage points and the Challenges program by 9 percentage points. The benefit of Challenges can be compared to the program cost of approximately $15 per person. Next, we show that treatment effects vary with individual characteristics, and we estimate gains from optimally assigning applicants across the two programs. We find that optimal assignment increases participants’ average probability of finding a job in technology by approximately 13% compared to random assignment. Finally, we analyze the counterfactual impact of expanding the available spots in Challenges from 15% to 50% of applicants, while assigning applicants to programs using the proposed targeting rule. Considering the entire applicant pool as the baseline, this generates a 30% increase in technology sector jobs.

Abstract: Pay-for-delay patent settlements, in which the incumbent patentee pays a potential entrant to withdraw a patent challenge and stay out of the market, cost patients and taxpayers billions of dollars in higher pharmaceutical prices. We show that in markets with one incumbent and several entrants, the possibility of conditioning such settlements on litigation outcomes against other entrants results in the exclusion of all entrants from the market. When conditional contracts are infeasible, the incumbent licenses the patent or fights entry in court: the resulting competition benefiting consumers. Prohibiting all pay-for-delay settlements increases litigation and may harm consumers by reducing licensing.

ADC Competition Policy Award 2018

Abstract: Using data collected from one of the most popular ridesharing platforms, we illustrate how mobility has changed after the exit from the Covid-19 induced confinement. We measure the impact of the Covid-19 outbreak on the level of mobility and the price of ridesharing. Finally, we show that the pandemic has exacerbated ethnic discrimination. Our results suggest that a decision-maker encouraging the use of ridesharing during the pandemic should account for the impact of the perceived health risks on ridesharing prices and should find ways to ensure fair access.

Abstract: Online platforms often face challenges being both fair (i.e., non-discriminatory) and efficient (i.e., maximizing revenue). Using computer vision algorithms and observational data from a micro-lending marketplace, we find that choices made by borrowers creating online profiles impact both of these objectives. We further support this conclusion with a web-based randomized survey experiment. In the experiment, we create profile images using Generative Adversarial Networks that differ in a specific feature and estimate it’s impact on lender demand. We then counterfactually evaluate alternative platform policies and identify particular approaches to influencing the changeable profile photo features that can ameliorate the fairness-efficiency tension.

Abstract: We study the impact of personalized content recommendations on the usage of an educational app for children. In a randomized controlled trial, we show that the introduction of personalized recommendations increases the consumption of content in the personalized section of the app by approximately 60% and that the overall app usage increases by 14%, compared to the baseline system of stories selected by content editors for all students. The magnitude of individual gains from personalized content increases with the amount of data available about a student and with preferences for niche content: heavy users with long histories of content interactions who prefer niche content benefit more than infrequent, newer users who like popular content. To facilitate the diffusion of personalized recommendation systems, we provide a framework for using offline data to develop such a system. 

Abstract: Labor economists regularly analyze employment data by fitting predictive models to small, carefully constructed longitudinal survey datasets. Although modern machine learning methods offer promise for such problems, these survey datasets are too small to take advantage of them. In recent years large datasets of online resumes have also become available, providing data about the career trajectories of millions of individuals. However, standard econometric models cannot take advantage of their scale or incorporate them into the analysis of survey data. To this end we develop CAREER, a transformer-based model that uses transfer learning to learn representations of job sequences. CAREER is first fit to large, passively-collected resume data and then fine-tuned to smaller, better-curated datasets for economic inferences. We fit CAREER to a dataset of 24 million job sequences from resumes, and fine-tune its representations on longitudinal survey datasets. We find that CAREER forms accurate predictions of job sequences on three widely-used economics datasets. We further find that CAREER can be used to form good predictions of other downstream variables; incorporating CAREER into a wage model provides better predictions than the econometric models currently in use.

Abstract: We study the impact of personalized content recommendations on the usage of an educational app for children. In a randomized controlled trial, we show that the introduction of personalized recommendations increases the consumption of content in the personalized section of the app by approximately 60% and that the overall app usage increases by 14%, compared to the baseline system of stories selected by content editors for all students. The magnitude of individual gains from personalized content increases with the amount of data available about a student and with preferences for niche content: heavy users with long histories of content interactions who prefer niche content benefit more than infrequent, newer users who like popular content. To facilitate the diffusion of personalized recommendation systems, we provide a framework for using offline data to develop such a system. 

  1. Media coverage: La Tribune, Boston University TPRI
  2. Best paper award, Workshop on Digitization, Telecom ParisTech, May 2018
  • Information and price dynamics in online marketplaces, with Rossi Abi-Rafeh, 2022

Abstract: We study the entry and pricing decisions of sellers in a market with a reputation system. First, we present a puzzling empirical observation: in rich observational data from two major online marketplaces, the number of reviews and prices are negatively correlated. This finding is robust to a rich set of controls. However, studying a within sellers price variation, we show that sellers gradually increase their prices. Second, we provide a model of pricing and entry with heterogeneity in marginal and opportunity costs and individual reputation as a state variable. We show that new sellers are generally less likely to reenter the platform than incumbents and sellers who have a lower chance of entering in subsequent periods set on average higher prices. Finally, we counter-factually decrease the reputation state of incumbent sellers and estimate how much they would need to be compensated for the loss of their reputation signals. 

Draft available upon request

  • Competition-Innovation Nexus: Product vs. Process, does it matter?, 2018

Draft available upon request

policy reports