Applied Econometric Studies on Network Effects
Abstract
This dissertation contains three essays which examine network effect study in econometrics and its applications. In the first essay, I study two research questions: how underwriter network affects firm network through IPO process and how firm network affects issuer post-IPO stock market performance. To address the two research questions, I construct firm and underwriter network based on sharing the same institutional shareholders and cooperating in the same syndicate, respectively. The networks are measured by network centrality measures from social network analysis (SNA): degree, eigenvector, closeness and betweenness. I then adopt regression models with these centrality measures. Empirical results indicate that an issuer is significantly more central in its public firm network if this issuer is led by a bookrunner through IPO process that is more central in its underwriter network. The effect of firm network on issuer post-IPO market performance is significant. An issuer with higher network centrality will achieve higher holding period return and higher monthly average trading volume in the first post-IPO year, while an issuer with higher eigenvector centrality has lower holding period return in the third post-IPO year. To analyze the source of underwriter network effect on firm performance, I run a specification by adding underwriter network centralities along with firm network centralities to analyze the firm post-IPO market performance. I also find that the effect of underwriter network on firm performance that is documented in literature can be mainly attributed to the effect of underwriter network on firm network.
In the second essay, we study identification and estimation of peer effects in observed networks where the problem of mismeasured links arises. In applied work, researchers generally construct networks that contain mismeasured links because of the data they use or the method of construction they adopt. Failing to deal with these mismeasured links, that using the observed networks as the true networks of interest, can result in biased estimates of peer effects parameters. Our paper provides sufficient conditions to identify peer effects in a generalized linear-in-mean model in which networks of interest contain mismeasured links. With the help of repeated observations of networks, we can identify peer effects in a generalized method-of-moments (GMM) model through an observed conditional moment model, instead of knowing the latent network structures. Based on the identification strategy, we propose a three-stage semiparametric estimator and apply our method to analyze peer effects among firms on their financial policies. Empirical results indicate that the peer effect is significantly positive and it becomes insignificant when mismeasured links are not addressed. Empirically studying firm networks with mismeasured links provides us an indepth look on the biased estimates of peer effects if mismeasured links are not properly taken care of.
In the third essay, we conduct a reduced-form analysis of first-price sealed-bid auction with affiliated private values under a network perspective. In recent literature, researchers generally adopt symmetric dependency among bidder’s private values through affiliation. Failing to take the embedded network structure of those bidders into consideration, that assuming the dependent structures of every pair of bidders be the same, can result in biased estimates in dependency parameters. Our paper provides some empirical evidence on the dependency of private values among bidders based on their linked status and proposes a structural model for estimation. The primary data we use are collected from the detailed bid summary files provided by the California Department of Transportation (Caltrans). We construct contractor networks based on if two contractors share a same subcontractor in one project. In the end, reduced-form results indicate that being linked has a significantly negative effect on the difference of bid amount between two contractors (bidders). In other words, linked contractors tend to submit close bid prices.
Citation
Li, Shixuan (2020). Applied Econometric Studies on Network Effects. Doctoral dissertation, Texas A&M University. Available electronically from https : / /hdl .handle .net /1969 .1 /191740.