Research

3. Saharsh Agarwal, Uttara Ananthakrishnan. Social Media and Well-being: Impact of TikTok on Screentime and Sleep Patterns

Major revision, Management Science

In this paper, we investigate the behavioral changes associated with TikTok adoption, focusing on its impact on screen time, nighttime phone usage, sleep patterns, and substitution across other apps. We employ a Difference-in-Differences (DiD) framework leveraging a high-frequency app usage dataset from nearly 7,000 TikTok adopters. Our findings reveal a significant increase in overall screen time—averaging 221 minutes per week for heavy TikTok users and 43 minutes for medium users—following TikTok adoption, alongside notable declines in usage of other social media and gaming apps, suggesting a strong substitution effect. The screen time effects are highest during late night hours. We also observe shifts in sleep patterns, including delayed bedtimes and a moderate reduction in sleep duration, among high-intensity users. We observe no effects for low-intensity users, who engage minimally with TikTok post-adoption. These results underscore the potential impacts of heavy usage of short-form video platforms on well- being. As a side result, our analysis provides clear evidence of substitutability between TikTok and traditional social media platforms, which has immediate implications for ongoing antitrust proceedings.

2. Saharsh Agarwal, Vivek Choudhary, Li Ding. Ka-Ching: The Impact of Cash-on-delivery on Gig Drivers' Behavior

The rise of food delivery platforms has transformed the gig economy, with Cash on Delivery (COD) remaining a dominant payment method. With COD, drivers temporarily retain the entire order payment on delivery, effectively receiving a temporary interest-free loan. While prior gig economy studies have focused on explicit incentives such as bonuses and wages, our study investigates the implicit liquidity benefit that COD provides, an important aspect that has received little attention in previous literature. COD provides drivers with immediate liquidity to cover expenses such as fuel and vehicle maintenance but also introduces risks of theft and cash-handling complexities. We examine how COD influences drivers’ decisions, particularly whether receiving cash impacts their intra-day shift ending decisions, the likelihood of returning to work the next day, and long-term attrition. Using data from a major Asian food delivery platform covering 8,371 drivers and about 4.1 million orders over six months, we exploit the exogenous lottery-like assignment of COD orders to analyze its effects. Our findings reveal that COD significantly shapes workforce dynamics. An additional US$10 in COD makes drivers 15.3% more likely to stop working in the next hour compared to non-COD payments, reflecting lower cash-based income targets. COD also increases drivers’ probability of returning to work the next day by 1.2% points, stabilizing weekly labor supply while reducing attrition and quitting rates, resulting in significant economic savings for the firm. These insights challenge platforms’ aversion to COD, showing it can serve as a non-monetary tool for workforce management by enhancing driver retention. 

1. Saharsh Agarwal, Uttara Ananthakrishnan, Catherine Tucker. Content Moderation at the Infrastructure Layer

Major revision, Information Systems Research

We examine the effects of infrastructural deplatforming through an analysis of Parler's shutdown following the January 2021 Capitol riot. Leveraging mobile phone touchstream data, we track user migration patterns and content consumption following this intervention. Our findings reveal that Parler's shutdown significantly amplified privacy-seeking behavior among its users, leading to a sustained migration to privacy-centric platforms and a notable increase in privacy-preserving browsers. Ideologically-aligned alternatives attracted initial interest but could not retain users over the long term. These results suggest that infrastructural deplatforming due to content violations may  inadvertently drive users toward encrypted communication channels that are harder to monitor. Users who migrated to these alternative platforms showed increased misinformation consumption post-shutdown, suggesting that deplatforming may intensify echo chambers by concentrating ideologically-aligned users within more insulated environments. These results provide critical insights for policymakers, technology providers, and regulators navigating content governance in an increasingly interconnected digital landscape. Our results also serve as the first long-run evidence of cross-category platform substitution, which can guide competition policy for social media applications.