Research

Working Papers:

  1.  Network of Loyalty Programs: A Sequential Formation [under review]

    Abstract: By forming partnerships with hotel chains and airlines, major credit card issuers in the U.S. allow customers to transfer points from their loyalty programs to the partners'. Hotel chains form similar partnerships with airlines, and airlines with other airlines. Viewing their partnerships as a network and using a sequential network formation model, this article studies how credit card issuers and hotel chains strategically choose to add, delete, or maintain partnerships with airlines. Using a novel dataset that involves 3 credit card issuers, 7 hotel chains, and 43 airlines from 2014 to 2018, the estimation result suggests that a credit card issuer is more likely to form a partnership with an airline that (1) is a partner of another credit card issuer (2) is a partner of its hotel chain partner (3) better complements its existing portfolio of airline partners, after accounting for key characteristics of firms. The first two results are similar when hotel chains choose airline partners.

  2.  Dyadic Regression with Block-Specific Fixed Effects and Application to Input-Output Matrix

    Abstract: Viewing the Input-Output matrix as a network of industries, this article studies how exogenous shocks, such as changes in tax rates and import duty rates, can affect relationships between industries within the U.S. economy. It also proposes a method for making counterfactual predictions on the Input-Output Matrix. A key innovation is utilizing a stochastic blockmodel to group industries into blocks, so that pairs of industries that belong to the same pair of blocks share common pairwise block-specific fixed effects. It uses dyadic regression model to estimate model parameters, together with pairwise block-specific fixed effects. An empirical analysis using the 2017 U.S. Input-Output Matrix shows that the supplier-buyer relationship between commodities (or industries) are sensitive to changes in net tax rates and import duty rates.

  3.  Communities and Centralities of Supplier-Buyer Network of the U.S. Tech Sector

    Abstract: Analysis of the U.S. tech sector's supplier-buyer network shows a strengthening of ecosystems around major firms from 2003 to 2014. Community detection methods reveal four distinct ecosystems, each led by AAPL, DELL, IBM, and MSFT. Supplier-buyer interactions within each ecosystem is far greater than across ecosystems, indicating exclusivity in suppliers and buyers around each of the four firms. Moreover, the substantial growth of AAPL and MSFT's significance, or centrality, in the network and their ecosystems underscore their rising dominance within the U.S. tech sector.


Work in Progress:

  1.  Quantification of Consumer Responses to Product and Non-Product Characteristics

    Abstract: Online social shopping platforms empower users to share “deals”. Users share information on price discount, product quality, and their opinion on overall value. It is evident that overall value of a “deal” depends on product characteristics such as product type (e.g., durable v.s. nondurable) and non-product characteristics such as price history (e.g., frequently on sale v.s. rarely on sale). This paper models consumer’s response to “deals” as a function of product and non-product characteristics and empirically quantifies it.

  2.  Linear-in-Means Peer Effects with Mismeasured Linkages (with Suyong Song)

    Abstract: We consider linear-in-means peer effects model with a social network in which individuals may systematically misreport friendships (“measurement error”) due to unobserved heterogeneities. Both the outcome variable of the linear-in-means model and the measurement error could covary with unobserved heterogeneities, so that peer effects are not identified. Moreover, instruments that depend on the social network may be invalid in the presence of measurement error. We propose conditions for identifying peer effects in the presence of measurement error and a consistent estimator.