Title:Inference in Linear Dyadic Data Models with Network Spillovers

Abstract:When using dyadic data (i.e., data indexed by pairs of units, such as trade flow data between two countries), researchers typically assume a linear model, estimate it using Ordinary Least Squares and conduct inference using "dyadic-robust" variance estimators. The latter assumes that dyads are uncorrelated if they do not share a common unit (e.g., if one country does not appear in both pairs of trade flow data). We show that this assumption does not hold in many empirical applications because indirect links may exist due to network connections, e.g., different country-pairs may have correlated trade outcomes due to sharing common trading partner links. Hence, as we show in Monte Carlo simulations, "dyadic-robust" estimators can be severely biased. We develop a consistent variance estimator appropriate for such contexts by leveraging results in network econometrics. Our estimator has good finite sample properties in numerical simulations. We then illustrate our message with an application to voting behavior by seating neighbors in the European Parliament.