Working Papers
Making hidden fees visible informs sellers as well as buyers: after Airbnb displayed fee-inclusive prices to EU users, hosts without cleaning fees found that competing listings were costlier than they appeared and raised nightly rates by about 17%, while fee-charging hosts reduced their fees only modestly.
When YouTube tightened monetization after the 2017 “Adpocalypse”, exposed creators added member-only content on Patreon and accumulated recurring revenue at a rate about 9% higher, shifting value capture to an auxiliary platform (Patreon) while continuing to rely on YouTube for audience discovery.
Reputation disciplines sellers less effectively when market exit approaches: after a Los Angeles regulation forced some Airbnb hosts to anticipate leaving, effort-related ratings declined in their final transactions, especially among hosts with long, strong review histories, and less so where highly rated neighbors maintained competitive pressure.
Work in Progress
As generative AI is folded into general search engines, answers increasingly arrive without a click. We measure how this reshapes what users search for, how traffic is allocated across the web, and what it implies for the content providers whose material feeds those answers.
Recommender systems must explore to learn, and exploration slots give platforms a discretionary channel for favoring their own content. We compare steering through exploration with conventional ranking bias, and ask which does more damage to consumers per unit of platform profit and which is harder to detect from recommendation data.
Minimum pay rules set a floor on driver earnings, but the platform retains control of the prices and matching that determine how much work there is. We ask whether Uber's pricing algorithm adjusts in ways that blunt the protection these rules are meant to deliver.
A driver who declines a passenger may be acting on his own preferences or anticipating those of the passengers already in the car. Exploiting the sequential formation of shared rides, this project separates taste-based discrimination from beliefs about how passengers will react to one another.
Scientific Articles
Oscar nominations can raise expectations enough to reduce satisfaction: among viewers with similar pre-nomination tastes, the same film receives lower ratings after being nominated, especially from less experienced viewers, a pattern driven by disappointment rather than a changing audience.
When the DMA led Google to remove map links from EU search results, searches for "Google Maps" rose by more than 21%, but total visits to Google Maps were unchanged. With no traffic gains for Bing Maps or other rivals, the reform changed access paths rather than competition.
A unified framework decomposing online ratings into the stages that generate them, from experienced quality and prior expectations through strategic distortion and selection into reviewing, organizing findings from fake reviews to disappointment effects and clarifying which platform interventions can target which distortions.
During the COVID-19 pandemic, Airbnb hosts with distinctively Asian names lost about 20% of their guests and US$180–$330 in monthly revenue relative to hosts with distinctively White names, with no comparable decline for Black or Hispanic hosts, showing how rising societal bias can translate into marketplace discrimination.
Experienced users select higher-quality movies but rate more harshly than novices, compressing average ratings and penalizing high-quality titles. A simple debiasing algorithm reverses more than 8% of pairwise rankings and aligns ratings more closely with external measures of quality.
Competition disciplines sellers but also erodes the premium generated by a strong reputation. Using a San Francisco regulation that halved the number of Airbnb listings, the paper shows that the second force dominates: hosts facing more competitors receive lower effort-related ratings and respond to guests less promptly.
Chapters and Media
- “Research: How Top Reviewers Skew Online Ratings” with Tommaso Bondi and Ryan Louis Stevens. Harvard Business Review Digital Articles (2025)
- “Ensuring Your Products Aren’t Used for Discrimination” with Michael Luca and Elizaveta Pronkina. Harvard Business Review Digital Articles (2022)
- “Asymmetric Information and Review Systems: The Challenge of Digital Platforms” In “Economic Analysis of the Digital Revolution” edited by J.J. Ganuza and G. Llobet. Chapter 2. 37-74 (2018). Madrid: Funcas