« Back to Results

Structural Estimation of Network Models: Recent Advances and Applications

Paper Session

Friday, Jan. 5, 2018 2:30 PM - 4:30 PM

Marriott Philadelphia Downtown, Meeting Room 414
Hosted By: Econometric Society
  • Chair: Seth Richards-Shubik, Lehigh University

Approximate Variational Estimation for a Model of Network Formation

Angelo Mele
Johns Hopkins University
Lingjiong Zhu
Florida State University


We study an equilibrium model of sequential network formation with heterogeneous
players. The payoffs depend on the number and composition of direct connections, but
also the number of indirect links. We show that the network formation process is a potential
game and in the long run the model converges to an exponential random graph (ERGM).
Since standard simulation-based inference methods for ERGMs could have exponentially
slow convergence, we propose an alternative deterministic method, based on a variational
approximation of the likelihood. We compute bounds for the approximation error for a given
network size and we prove that our variational method is asymptotically exact, extending
results from the large deviations and graph limits literature to allow for covariates in the
ERGM. A simple Monte Carlo shows that our deterministic method provides more robust
estimates than standard simulation based inference.

Social Networks and the Stepping Stone Effect: The Case of Tobacco and Marijuana

Anton Badev
Federal Reserve Board


I study the role of the social network in generating a stepping stone effects where the consumption of one drug is often associated with consumption of another drug. The equilibrium framework of Badev (2013) allows to isolate the influences of the (endogenous) social environment, e.g. a tobacco smoker social environment may be more tolerant to drug use and/or may provide an easier access of marijuana supplies, and to empirically evaluate the equilibrium response of the social network to policies narrowly targeting specific drug consumption, e.g. the effect of an increase in tobacco prices may, through altering the social fabric, penetrate to a wider set of risky activities such as marijuana consumption. The empirical analysis of the case of tobacco and marijuana smoking quantifies the importance of distinguishing the social environment channel from complementarities in consumption, where tobacco price may have unintended effects on marijuana smoking prevalence.

Dependence-Robust Inference Using Randomized Subsampling

Michael P. Leung
University of Southern California


We study hypothesis tests and confidence intervals constructed using a randomized subsampling procedure first proposed by Song (2016). We show that these procedures are robust to general forms of weak dependence in the sense that they are asymptotically valid under the weak requirement that the target parameter can be consistently estimated at the root-n rate by an asymptotically linear estimator. The implementation of randomized subsampling does not depend on the correlation structure of the data and is computationally simple. We consider applications to constructing clustered standard errors when the level of clustering is unknown and the number of clusters is potentially small, inference on network data when the network is partially observed or the network-formation model is unknown, and testing for power laws with dependent data. We also develop randomized subsampling tests of moment inequalities.

Collaborative Production in Science: An Empirical Analysis of Coauthorships in Economics

Katharine A. Anderson
Carnegie Mellon University
Seth Richards-Shubik
Lehigh University


This paper studies productivity and preferences in scientific research. We apply a recently developed method for estimating network formation models to the network of coauthorships in economics. Our model focuses on skill complementarities among researchers with different specialties (e.g., econometrics or macroeconomics) and on costs such as coordinating among multiple coauthors and communicating across multiple locations. Once estimated, the model can be used to simulate aggregate research productivity under policy counterfactuals such as changing the distribution of skills in the workforce (e.g., by subsidizing certain fields in PhD programs) or decreasing the cost of inter-university collaboration (e.g., by funding travel or telecommunication equipment).
Luis Candelaria Barrera
University of Warwick
Salvador Navarro
University of Western Ontario
Eric Auerbach
University of California-Berkeley
Michael D. D. König
University of Zurich
JEL Classifications
  • C5 - Econometric Modeling
  • L1 - Market Structure, Firm Strategy, and Market Performance