Boston University
Essays on digital platforms and applications of partial identification
Abstract
dc:description.abstractThis dissertation explores topics in marketing problems on digital platforms and econometric methods for partial identification. The first two chapters investigate content creator behavior on digital platforms and the implications of policy interventions, while leveraging a partial identification framework to address empirical challenges. The third chapter introduces a robust framework for discrete choice modeling, relaxing traditional assumptions in nested logit analysis. Chapter 1 examines the effects and economic consequences of a platform policy banning gambling content, leveraging a panel dataset of 6,000 top streamers on a leading livestreaming platform. To identify banned content and streamers affected by the policy, I leverage video analysis on historical video clips, high-frequency stream titles, and in-stream chat analysis. To tackle key identification challenges, I use both a two-way fixed effects DiD and a Synthetic DiD estimator, and propose a network analysis to address potential interference issues. Additionally, I propose a sensitivity analysis following partial identification in DiD to show that the results are robust to potential misclassification. Results reveal that the policy had a spillover regulatory effect on unbanned gambling livestreams due to streamers’ reputation concerns, disproportionately affecting highly popular creators and non-loyal viewers. This chapter highlights how platforms and policymakers can design content moderation policies while cautioning unintended supply- and demand-side consequences. Chapter 2 shifts focus to collaboration strategies among content creators, a growing trend with understudied economic implications. Using a game-theoretic model, I analyze how creators with heterogeneous content preferences and bargaining power negotiate collaborations. The model demonstrates that joint content creation moderates competition while expanding audience reach, increasing content diversity and monetization. In general, collaboration among creators has an effect of increasing the profits of creators while reducing consumer surplus. However, when creators differ significantly in content quality, lower-quality creators free-ride on collaborators, inadvertently benefiting consumers by diverting them from low-value content. This finding underscores the nuanced welfare effects of collaborative practices. Chapter 3 develops a partial identification framework for incomplete discrete choice models with nest structures, whereas the incompleteness arises from two types of unobserved heterogeneity: first, the model may have unobserved heterogeneity that is shared by alternatives within the same nest. Second, individuals may have different perceptions about the underlying nest structure. Based on the random set theory and Artstein's Inequality, I show that the sharp identified sets of model parameters can be characterized by a finite number of conditional inequalities, given a set of ``consideration nests". I illustrate this method in the context of an incomplete discrete choice model with three alternatives, in which the utility function of each choice contains an unobserved nest-specific fixed effect that varies across different nest structures. Numerical Examples show that the partial identification approach generates sharp identified sets that correctly cover the true parameter values at a cost of lower identifying power.
Degree
thesis:*- Name thesis:degree_name
- Doctor of Philosophy
- Level thesis:degree_level
- doctoral
- Discipline thesis:degree_discipline
- Economics
- Grantor
- Boston University
- Year dc:date.issued
- 2025
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Han, Qifan
- Advisor dc:contributor.advisor
-
- Kaido, Hiroaki
Subjects
dc:subject × 6Rights
- Language dc:language.iso
- en_US
Identifiers
dc:identifier.*- Handle dc:identifier.uri
- https://hdl.handle.net/2144/53281