Research
Research
Works in Progress
Do Non-Monetary Incentives Increase Output? Evidence from the Online Creator Economy [draft]
(Job Market Paper)
Ongoing
Abstract: Non-monetary incentives are understood to shape worker effort and output, but isolated, large-scale variation in such incentives is rare. This paper exploits YouTube's March 31, 2025 redefinition of a ``view'' for Shorts, which inflated a salient, publicly displayed measure of attention while leaving platform-based monetization unchanged. Using channel-level data on roughly 36 million videos from about 182,000 U.S. channels (October 2024–December 2025), this analysis compares Shorts-specializing and long-form-specializing channels in a difference-in-differences design, accounting for brand-sponsorship status -- through which the change plausibly shifted monetary incentives -- in a triple-difference design. The change in view accounting immediately raised non-sponsored Shorts output by about 9% before declining gradually to baseline in the following six months. The response was larger among sponsored channels, and additional analysis of a likes-per-view engagement measure suggests that additional output may have come at the cost of quality. The findings offer novel evidence that nominal, non-pecuniary platform features can meaningfully, if temporarily, move real creator effort in a monetized environment.
Willingness to Play: Online Platform Participation After Exposure to Abusive Peers [draft]
Ongoing
Abstract: Online social platforms constitute a key public forum and generate significant economic value through advertising and peer-to-peer commerce. Unfortunately, such platforms are frequently plagued by antisocial behavior such as bullying and harassment. This paper leverages random matchmaking on Chess.com, a large online chess community and gaming site, to examine how exposure to toxic peers affects user engagement. Results suggest that playing a game against a randomly assigned opponent prone to abusive behavior increases short-run engagement by 1-2%. This effect does not translate to the long-run, where cumulative exposure to abusive peers is negatively correlated with platform participation.
Cable Improvements and Fiber Deployment (with Kangoh Lee) [draft]
Submitted
Abstract: Fiber technology is considered the future of the internet, as it can handle the growing demand for internet services in the future, offers faster and symmetric speeds between download and upload, and improves reliability. However, fiber is much less available than legacy technologies, namely cable and digital subscriber line. This paper studies whether increases in cable speeds discourage fiber deployment by exploiting the advent of DOCSIS (Data Over Cable Service Interface Specification) 3.1 as an identification strategy. Using the Federal Communications Commission’s fixed broadband deployment data during 2014 - 2021, the analysis shows that increases in cable speed decrease the availability of fiber internet technology. This result may be interpreted as the undesirable effect of directed technological change toward the cable sector on fiber deployment, and the effect is stronger for local internet service providers and in nonmetro areas.
Publications
Abstract: Undocumented immigrants are more than five times as likely as US citizens to be uninsured. Before 2020, undocumented 19-25-year-olds in California were eligible for restricted-scope Medi-Cal, which only covers emergency services. This cross-sectional study used a triple difference analysis to compare coverage changes among noncitizens ages 19-25 and 26-32 in California to this same difference in control states before and after full-scope Medi-Cal was expanded to 19-25-year-old noncitizens in 2020. Medi-Cal expansion was associated with a 4.2 percentage-point increase in Medicaid coverage and a 3.5 percentage point increase in any coverage among 19-25-year-olds relative to 26-32-year-olds in California compared to the control states. Estimates for private insurance were small in magnitude and not statistically significant at conventional levels. We estimated larger impacts for Hispanic young adults compared with other race and ethnicity groups. Estimates by age and sex were similar. In post-hoc analyses, our estimates translated to increases of 24 and 20 percentage points in Medi-Cal and any coverage, respectively, among newly eligible undocumented young adults.
Abstract: Vertical identification laws (VILs)—which require state drivers' licenses issued to individuals under age 21 to be vertically oriented—have been adopted widely across the United States. We find no evidence that VIL adoption has a statistically significant or economically important effect on teenage drinking and smoking. This result persists using Callaway and Sant'Anna estimates and extends to binge drinking, marijuana use, and drinking and driving. Finally, we show that prior estimates reported in the literature are sensitive to sample period studied, choice of Youth Risk Behavior Survey dataset, and accounting for heterogeneous and dynamic treatment effects.
Abstract: Using Federal Communications Commission internet data during the period 2014–2021, this paper studies high-speed internet access across metro and nonmetro areas by technology types. The paper develops the measures of the tract-level availability of a technology, based on the number of internet service providers of the technology in each census block. The analysis shows that all technologies had been less available and slower in nonmetro areas than metro areas even after controlling for relevant factors such as household incomes and labor market conditions. In addition, a new technology, namely Cable Modem-DOCSIS (Data Over Cable Service Interface Specifications) 3.1, had replaced an old one, namely Cable Modem-DOCSIS 3.0, in nonmetro areas at much slower rates than in metro areas. The paper then discusses policy implications of the high-speed internet gap.