Yajie Xu

University of South Florida · Economics · 2026–27 job market

Fields

Econometrics, Theory, Innovation

Job market paper

Bayesian Estimation of Endogenous Network Effects in SAR Models: Application to R&D Collaboration

Selection biases pose a big challenge in estimating spillover effects in spatial autoregressive (SAR) models, primarily because network formation often depends on unobservable individual preferences affecting outcomes. This paper proposes a novel Bayesian framework to jointly model the dynamics of network effects and their formation and extends it to panel data. Specifically, we introduce an explicit latent structure, which preserves conjugacy, connects outcomes and network selection equations in the SAR model through the error terms and allows computationally more efficient estimation than previous studies. This novel structure simultaneously models endogenous formation and captures the associated spillover effects. Estimation is based on a Markov Chain Monte Carlo (MCMC) algorithm, all steps of which except one are the Gibbs sampler, a benefit of the modified model specification. We first validate the model through a simulation study and then apply it to data on R&D collaboration networks to estimate technological spillovers and assess the role of immigrant participation in network formation. The paper intends to contribute a novel general framework for estimating endogenous spillover effects and to promote broader, computationally simpler and more efficient application of Bayesian methods in network modeling.