Parametric and Semiparametric IV Estimation of Network Models with Selectivity

We propose parametric and semiparametric IV estimators for spatial autoregressive models with network \ndata where the network structure is endogenous. We embed a dyadic network formation process in the \ncontrol function approach as in Heckman and Robb (1985). In the semiparametric case, we use power \nseries to approximate the correction terms. We establish the consistency and asymptotic normality for \nboth parametric and semiparametric cases. We also investigate their finite sample properties via Monte \nCarlo simulation