Research

Publications:

•  Wages, Hires, and Labor Market Concentration, with Ioana Marinescu and Ivan Ouss  (2021, Journal of Economic Behavior and Organization)

Media :  Libération , Concurrences , Resolution Foundation,  Groupe d'Experts sur le SMIC

• Personalized content, engagement, and monetization in a mobile puzzle game,  with Christian Helmers, Alessandro Iaria, Julian Runge, and Stefan Wagner  (Volume 98, January 2025, International Journal of Industrial Organization)

• Is Competition Only One Click Away? The Digital Markets Act Impact on Google Maps, with Michelangelo Rossi - Accepted at Marketing Science (2025)

Media :  Forbes ; The Conversation ; StartMag ;   Vogon ; Department of Product ; Snowball ; Google's Response to the Australian Competition and Consumer Commission's Digital Platform Services Inquiry (2024).   

Working Papers :

• Platform Information Provision : Evidence from an Online Auction Platform, with Pierre-Francois Darlas

Draft available on SSRN

Abtract: Digital platforms have reduced search costs, fostering niche product markets. However, these markets often suffer from limited and asymmetric information between buyers and sellers due to a lack of consumer feedback. This paper examines Catawiki's solution, where over 240 experts provide value estimates for rare collectibles being auctioned on a digital platform. Using data from 57,000 listings, we analyze the impact of these estimates on final prices and seller behavior. By leveraging both minimum and maximum expert estimates, we isolate the effect of increasing the maximum estimate while holding the minimum estimate fixed. Our findings indicate that higher expert estimates increase final bidden prices, suggesting buyer trust. Sellers also adjust their behavior by setting fewer reserve prices for items with high estimates, leading to more bids. Despite potential conflicts of interest stemming from the platform's dual role as matchmaker and advisor, our results show that expert estimates are influential even when potentially overinflated. This study underscores the critical role of platform-provided information in enhancing market efficiency.

• How to improve both the pay and quality of microtasking platforms ? (Working Paper Arriving Soon with Chiara Belletti)

Abstract : Micro-tasking platforms enable the collection of data used to train machine learning algorithms and artificial intelligence. However, a classical Principal-Agent problem may limit the quality of the data produced. As firms do not always monitor the quality of the work done with sufficient frequency, a moral hazard problem may arise. We develop a structural model of equilibrium demand and supply of effort to measure quality and monitoring behavior. We estimate the parameters of this model using proprietary data from a leading micro-tasking platform. We find that metrics relying on observed task rejection underestimate quality. We discuss several mitigation strategies. We suggest a more accurate back-of-the-envelope correction based on a firm’s own monitoring rate to increase the employers awareness about the potential data quality. Finally, we discuss incentive schemes to induce higher quality work. Using counter-factual simulations, we show that charging penalties for workers with a rejected task could induce higher effort and require less monitoring from the firms.

• Study of Deemed Suppliers on VAT Tax Fraud on French Digital Platforms (Paquet TVA, DAC 7)  (joint work with French Ministry of Public Finances (DGFIP, Matthieu Chtioui).  

• No Poaching Agreements and Salary Arbitration : Quasi-experimental Evidence from Major League Baseball (Resubmitted to Review of Industrial Organization )

Abstract: Workers subject to a non-compete clause can suffer wage loss, raising the question of the adequate policy response. This article shows that eligibility to salary arbitration, by which a third party can be called upon to resolve a wage dispute, can be an effective way to raise incomes. Exploiting a quasi-random discontinuity in the rule determining eligibility for players in Major League Baseball subject to non-compete clauses, this article leverages a Regression Discontinuity Design to identify the effects of salary arbitration eligibility. Workers right above the eligibility threshold have wages over 50% higher than those right below. No severe side effects were detected, including compression of the wage distribution, lower productivity, or higher unemployment.

•       How to Detect and Measure Labor Market Collusion? (Job Market Paper)  
  Abstract: With the aim of expanding the set of tools available to antitrust practitioners, this paper develops two new econometric methods to detect and measure the effects of labor market cartels. The first method is reduced form and aims to estimate wage loss. It exploits the inter-percentile difference between high earners and low earners within a difference-in-differences framework. This approach is simple to implement,  can easily be explained to non-economists, measures heterogeneous effects, and requires no additional data compared to that necessary for a before-after analysis. The method is illustrated by revisiting the 1986-8 case of collusion in Major League Baseball, measuring an average yearly income loss of 26%. Second, this paper  develops a structural model of labor market competition for the purpose of detecting collusive behavior.  Applied to the data, it reveals that at the beginning of the cartel, there were heightened barriers to mobility across firms, rising profits, and a decreasing labor share of income. Surprisingly, these patterns sustain past the end date of the cartel, suggesting important and underestimated long-run effects. Finally, the structural model is used to simulate counter-factual wages, revealing that the yearly average wage should have been at least 30% higher.  

  • • • • slides 

  • Dealing with Logs and Zeros in Regression Models, with Christophe Bellego and David Benatia (submitted to the Journal of Econometrics) 

Abstract: Log-linear models are prevalent in empirical research. Yet, how to handle zeros in the dependent variable remains an unsettled issue. This article clarifies it and addresses the "log of zero'' by developing a new family of estimators called iterated Ordinary Least Squares (iOLS). This family nests standard approaches such as log-linear and Poisson regressions, offers several computational advantages, and corresponds to the correct way to perform the popular log(Y+1) transformation. We extend it to the endogenous regressor setting (i2SLS) and overcome other common issues with Poisson models, such as controlling for many fixed-effects. We also develop specification tests to help researchers select between alternative estimators. Finally, our methods are illustrated through numerical simulations and replications of landmark publications. 

Among the Top 525 most downloaded papers of SSRN.

Stata Software available from : https://github.com/ldpape/iOLS-i2SLS

ASSA 2022 Poster

Media : Empirical Legal Studies, Eviews, David Giles' Blog.

• Judge (Ideology) Shopping, with Christian Helmers and Brian Love (submitted) 

Abstract: In his first term in office, President Donald Trump appointed 174 judges to U.S. District Courts. Weanalyze the impact of these appointments on the ideology of the judges that helm the trial courts ofthe U.S. federal judiciary, as well as the ability of ideologically-motivated litigants to engage in “judgeshopping.” Using judge-level information on case assignment rates and ideology in an instrumenteddifference-in-differences approach, we show that Trump’s appointments both shifted the ideology ofthe U.S. district court bench significantly to the “right” and increased litigants’ ability to select a veryconservative judge with high probability. The latter effect is driven by Trump’s appointment of veryconservative judges to seats in rural court divisions where all or substantially all cases have traditionallybeen assigned to a single judge. We present evidence that greater ability to engage in ideologically-motivated judge shopping led to a large increase in civil rights case filings in impacted divisions, particularlycases concerning subjects with a recognized ideological valence, such as reproductive rights, religiousliberty, and immigration. A simple counterfactual exercise reveals that randomizing case assignment atthe judicial district level would reduce the number of politically motivated case filings by approximately 62% over a four-year period. 

Stata Software for Instrumented Differences-in-Differences (IVDID) based on Miyaji (2025) available at : https://github.com/ldpape/ivdid 

Policy Reports :

DGFIP Analyse n° 9 - "Près de 40 Md€ de ventes sur les plateformes en 2022 : une diversité de profils de vendeurs" available here : https://www.impots.gouv.fr/dgfip-analyses