Research

Working Papers: (all working papers are available upon request)

Can Gender-Blind Algorithmic Pricing Eliminate the Gender Gap?

Insurance companies frequently use consumer attributes, such as gender or age, when setting a price for their services. Young male drivers, for example, are often charged more than young females for car insurance, as they are expected to be riskier. In 2019, California banned auto insurance companies from using information on gender in their pricing algorithms. I study how this ban affects the gender gap in prices, using a difference-in-differences strategy with older individuals and other states as control groups. I find that the ban reduced the gender gap in the insurance premiums paid by young drivers by around 55 percent, but it failed to eliminate it completely. My analysis of the pricing algorithm of a large insurance company in California indicates that algorithms are adjusted in a way that gender proxies receive larger weights after the policy. For instance, drivers using specific car models associated with young males were charged up to 22 percent more after the ban. My findings illustrate the limitations of anti-discrimination policies that impose group-blind pricing, with implications for the design of fairer regulations for algorithms.

Who Is AI Replacing? The Impact of Generative AI on Online Freelancing Platforms   (with Jonas Hannane and Xinrong Zhu) 

This paper studies the impact of generative AI technology on the demand for online freelancers using a large dataset from a leading global freelancing platform. We focus on how the release of ChatGPT affects various freelance jobs that require different skills or software. Our findings indicate a 21 percent decrease following the ChatGPT introduction in the number of job posts for automation-prone jobs when compared to jobs requiring manual-intensive skills. We also find that the introduction of image-generating AI technologies led to a significant 17 percent decrease in the number of job posts related to image creation. Furthermore, we utilize prior evidence on AI exposure to different occupations and Google Trends to demonstrate that the more pronounced decline in freelancer demand within those specific occupations is related to their heightened exposure to AI technology, as well as higher general public awareness of ChatGPT's substitutability.

Reducing Discrimination with Information: Evidence from Online Freelancing Platforms

Freelancing platforms connecting millions of employers and freelancers worldwide have become enormously popular. These online platforms enable workers to supply their labor easily by reducing transaction and search costs. Although online platforms reduce certain costs by making global labor markets more accessible, some information asymmetries persist. Such informational frictions may disadvantage freelancers from developing countries, particularly when employers are from high-income countries. In this paper, I study the wage gap between freelancers from high-income and developing countries in an online freelancing platform and explain the mechanisms driving this gap. I find that freelancers from developing countries earn 22 percent less than freelancers from high-income countries after controlling for the job and country-specific characteristics. However, the penalty on wages decreases as contractors provide information about themselves over time. Experience, reputation scores, and more information on previous earnings or standardized test scores benefit freelancers from developing countries more. My findings have implications for the role of information in achieving fairer outcomes in digital platforms.

Hybrid vs. In-person: How Does Online Access to Lectures Affect Student Behavior? 

Online education technologies are becoming increasingly popular in higher education. One widely used method is recording lectures during face-to-face classes. Access to online lectures allows students to review the content, whereas knowing that the lecture recordings will be available online can decrease incentives to attend classes. Besides, lecture recordings can have heterogeneous impacts on specific groups of students. It can be helpful as a revision tool, particularly for non-native English speakers. In that sense, the overall effect of online education in the hybrid setting is not easy to predict. This paper studies the impact of online access to recorded lectures on student performance, attendance, and satisfaction. I use a difference-in-differences strategy by exploiting the staggered implementation of the recorded lecture system in different courses in a top UK university. I find that (i) online lectures decrease the performance gap between native and non-native speakers significantly, (ii) they increase the achievement of high-performer students and decrease it for low-performers and (iii) they reduce attendance to the main lectures and have no statistically significant impact on student satisfaction. My findings have significant implications for designing policies on access to distance learning technologies and overcoming learning disparities across different groups.

Work in Progress: 

Do Privacy Policies Increase Data Sharing?  (with Eva Ascarza and Ayelet Israeli)


This study investigates the impact of privacy policies and nudges on consumers' data-sharing behavior. We focus on a leading US-based reward app and leverage a comprehensive dataset with over 50 million transactions from a quarter million consumers over time. First, we analyze how privacy policies influence the propensity of consumers to share their data within the app ecosystem. Furthermore, we examine the effectiveness of nudges, specifically those sent after prolonged periods of inactivity, on promoting data-sharing behavior. We employ difference-in-difference and regression discontinuity designs using the variation of privacy policies across US states. We aim to provide insights into how regulatory policies and behavioral interventions can effectively shape data-sharing behavior and customer engagement in online platforms.

Do Online Food Delivery Platforms Help or Harm Restaurant Businesses?

Online food delivery platforms (e.g., DoorDash and Uber Eats) have become highly popular, fundamentally changing how restaurants operate. However, whether these platforms benefit or harm restaurants remains unclear. This study investigates the impact of online food delivery platforms on restaurants. I use an extensive geospatial dataset containing 4 million geographic features in the UK over time, including information about all restaurant businesses, their location, capacity, and neighborhood properties. By exploiting the staggered entry of online food platforms in each postcode, I examine how restaurants, their locations and store sizes change over time as online food delivery companies start serving in the same postcode as the restaurant. This research aims to shed light on the evolving relationship between restaurants and online food delivery platforms, providing valuable insights into the dynamics of the restaurant industry in the digital age.