Estimating Spillovers from Sampled Connections

TL;DR

This paper demonstrates that common sampling methods bias spillover effect estimates in network data and provides biased-corrected estimators, enabling more accurate analysis of network effects in empirical research.

Contribution

It introduces bias correction techniques for spillover effect estimation under various sampling schemes and link measurement assumptions.

Findings

Sampling schemes bias estimates upwards

Derived estimators correct for sampling bias

Applied methods to climate shock propagation among firms

Abstract

Empirical researchers often estimate spillover effects by fitting linear or non-linear regression models to sampled network data. We show that common sampling schemes bias these estimates, potentially upwards, and derive biased-corrected estimators that researchers can construct from aggregate network statistics. Our results apply under different assumptions on the relationship between observed and unobserved links, allow researchers to bound true effect sizes, and to determine robustness to mismeasured links. As an application, we estimate the propagation of climate shocks between U.S. public firms from self-reported supply links, building a new dataset of county-level incidence of large climate shocks.