Research

Research interests

Digital Marketing, Quantitative Marketing, Applied Economics, Digital Economics, Economics of Information Systems, Industrial Organization, Platform Strategy.

Working Papers

• “Rating Systems and the End-Game Effect: When Reputation Works and When it Doesn’t”, with Elizaveta Pronkina (Amazon) and Michelangelo Rossi (HEC Paris), 2025.
  • Draft: Latest Version
  • Abstract: Do rating systems provide incentives when sellers are about to exit a market? Using data from Airbnb, this paper examines how end-game considerations influence hosts’ effort decisions and how these effects depend on accumulated reputation capital. We exploit the implementation of the Home-Sharing Ordinance in Los Angeles, which forced ineligible hosts to leave the platform, to identify sellers who could anticipate their imminent exit. Effort is measured through guest ratings in categories directly tied to host behavior (communication, check-in, and cleanliness) and compared to ratings on location, which are unaffected by effort. In a Difference-in-Differences framework, we find that effort-related ratings decline significantly in the final transactions of exiting hosts, with larger reductions among listings holding stronger reputational capital or longer review histories. At the same time, competitive pressure from neighboring listings with strong reputation mitigates these declines: when nearby competitors have high ratings, hosts face higher reputational costs from underperforming even as they approach exit. These findings reveal that the effectiveness of reputation systems erodes near market exit, yet the competitive pressure from highly rated nearby listings can partially sustain service quality and trust on digital platforms.

Selected Work in Progress

• “How to Assess and Improve the Quality of Crowd-Sourced Data Work”, with Louis Daniel Pape (Télécom Paris).
  • Draft: available upon request.
  • 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 by micro-taskers as firms do not always monitor the quality of the work done with sufficient frequency. 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 severely underestimate the quality/effort with which data annotation tasks are performed. This suggest AI is being built with mis-annotated data. We discuss several mitigation strategies. We find that increasing the pay of micro-taskers along with more frequent monitoring could help improve the quality of the data. Finally, we discuss incentive schemes to induce higher quality work by relying on 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.
• “Visual Clutter and Marketing Cues: Experimental Evidence from the OTA industry”, with Damien Mayaux (Université Paris Dauphine - PSL), Bruno Chaves Ferreira (CNRS, Université Paris Dauphine - PSL) and Calvin Pan (University of Virginia).
  • Experiment design stage.
  • Abstract: E-commerce platforms rely on visual cues—such as popularity labels or recommendations—to influence consumer choices, while simultaneously shaping the level of visual clutter in which these cues appear. Visual clutter has an ambitious effect on the cue: it can increase reliance on simple heuristics while reducing the salience of individual cues. These mechanisms may have diverging implications on business metrics and consumer welfare. We study how visual clutter moderates the impact of cues on consumer choice in a laboratory experiment with real incentives. Using a web browser extension to manipulate the interface of a hotel reservation platform, we experimentally vary the presence of a cue within subjects and the level of visual clutter between subjects. Across levels of clutter, we measure the causal effect of cues on choice behavior as well as subjective perceptions of cue salience, cognitive load, and visual complexity. In addition, we analyze search behavior to characterize decision-making strategies across conditions. Our design allows us to disentangle the mechanisms through which visual clutter changes the effect of the cue in realistic online environment, providing insights for platform design and regulatory action regarding e-commerce interfaces.
• “Picky Drivers: How Reputation Capital Induces Selectiveness on a Matching Market”, with Dianzhuo Zhu (University of Lille).
  • Draft available soon.
  • Abstract: Reputation systems are designed to build trust and reward quality, but do they also make high-reputation providers more selective about whom they serve? We study this question using proprietary booking-request-level data from BlaBlaCar, the leading European intercity carpooling platform, covering over 100,000 requests across three years. Exploiting a rating-display rounding rule that creates sharp discontinuities in the reputation signal visible to drivers, we implement a regression discontinuity design to identify the causal effect of reputational standing on acceptance behavior and other selectivity dimensions. We find that drivers who cross the rounding threshold that makes their displayed rating exceed the platform average become significantly less likely to give up screening control and more likely to reject or ignore incoming requests. This selectiveness is concentrated at the bottom of the passenger rating distribution: higher-rated drivers’ selectiveness is mitigated by the rating of incoming passengers. Drivers are significantly more likely to screen out low-rated passengers — generating a reputation ladder effect in which access to high-quality supply becomes progressively stratified by passenger standing. These findings have direct implications for the design of reputation systems on matching platforms.
• “Replaceables? Profile Restarting on Digital Platforms. Evidence from the Short-Term Rental Market”, with Joerg Claussen (LMU Munich) and Michail Batikas (NOVA Lisbon).
  • Draft available soon.

Policy Reports and Other Publications

• “Crowdworking in France and Germany”, with Ulrich Laitenberger, Daniel Erdsiek, and Paola Tubaro, 2021.
  • ZEW expert brief Nr. 21-09.
• Country chapter Italy, with Riccardo Norbiato (PSE) in Social Protection of Non-Standard Workers and the Self-Employed During the Pandemic, Spasova S., Ghailani D., Sabato S. and Vanhercke B.
  • Available here.