Job Market Paper

Abstract: Bans on digital products from foreign tech giants have become a policy staple. Yet the welfare implications of these bans are unknown. Policy-targeted products are typically large and highly differentiated, leading to complex but essential substitution patterns for welfare analysis. I empirically reveal these substitution patterns through an event study of India's ban on Chinese apps: pairwise elasticities of substitution are correlated with pairwise similarities between product descriptions.  

Motivated by this finding, I develop a general equilibrium model in which granular products with heterogeneous productivities and hedonic attributes engage in Bertrand oligopolistic competition. This model is purpose-built to feature the pairwise elasticities of substitution, which map to the cosine similarity constructed using the large language model (LLM). I use the estimated model to evaluate the impact of China's Great Firewall (GFW) policy. I find that while the GFW increased Chinese real incomes, the loss in leisure utility from using inferior apps overwhelms the benefits, resulting in a net welfare loss of 7.5%. This framework is easily implemented and widely applicable to research questions where pairwise substitutions are essential and rich text is available.

Working Papers

(Latest WP | NBER WP link | BibTex)

Publication

(Paper link | Online appendix | Replication files | BibTex)


(Book Chapter link | NBER WP link | BibTex)

Work in Progress