Job Market Paper
- “Data Preferences in Firm Learning: Evidence from an Online Auction Platform”
Abstract
I examine whether firms prefer their own data over others' data when learning to enhance decision-making in a new environment. Recent government initiatives and technological advancements have enabled data sharing among firms, potentially expediting learning through broader insights from shared data. However, the benefits of data sharing become less clear if firms prefer their own data in their learning process. This paper investigates how auction houses (i.e., sellers) optimize their auction timing choices on China's largest online auction platform. Payoffs vary by hour, driven by differences in bidders' valuations and the number of residual bidders at different times. I first leverage state-of-the-art empirical auction literature to estimate how bidders' valuations change throughout the day. Then, I build and estimate a structural model of sellers' ending-hour choices that incorporates adaptive learning about the number of bidders and competition, based on both their own data and others' data. The model estimates suggest that sellers' preferences for their own data and others' data change as they gain more experience. Initially, inexperienced sellers rely heavily on others' data. This preference shifts as they accumulate experience on the platform. Specifically, at the average experience level, sellers weigh their own data at 90% compared to 10% for others' data. These findings have two implications for platforms. First, data sharing may not effectively guide sellers in choosing optimal auction ending hours, as sellers place an overall smaller weight on others' data. Second, platforms could guide new sellers to optimal ending hours early in their tenure, and this guidance can have a lasting impact. Overall, these implications suggest that platforms should play a coordinating role in helping sellers identify the best ending hours for their auctions.
Working Paper
- “Shaping the Influencers: The Role of Multi-Channel Networks” [SSRN]
- with Yulin Hao
- Revise & Resubmit at Journal of Marketing Research
- Presentation:
Abstract
Social media influencers are increasingly affiliating with multi-channel networks (MCNs), also known as influencer agencies. These MCNs recruit influencers and help them monetize their content. More importantly, MCNs are rumored to be directly involved in content creation. This paper provides the first empirical examination of the effects of MCN affiliation on influencer content. To this end, we construct a unique dataset tracking influencers' changes in their MCN affiliation on TikTok in China. Using a difference-in-differences strategy, we compare influencers who switched their affiliation status with observably similar non-switchers. The findings reveal that MCN affiliation enhances content engagement and leads to more homogeneous and focused content, steering influencers towards topics with higher advertising prices. However, the content quantity does not change. When influencers affiliate with MCNs, these influencers also have more sponsorships and charge a higher advertising price, which is predominantly driven by changes in content resulting from the affiliation. These results suggest that platforms and influencers can benefit from improved engagement and sponsorships resulting from MCN affiliation. Although advertisers face higher advertising prices charged by MCN-affiliated influencers, these prices are justified by more engaging and focused content, which may also be beneficial to the advertisers.
Work in Progress
- “Consumer as Helping Hands: The Value of Incentives in Mitigating Demand Fluctuations under Capacity Constraints”
- “Optimal Design of Credit Card Discounts: Monetizing Discount Synergy Across Products”
- with Bowen Luo and Ruiqi Wu
- Data access through collaborating with Wharton AI & Analytics for Business (AIAB)
- “Bargaining in New Product Launch (tentative title)”