科学研究

We propose an adjusted 2SLS estimator for social network models when some existing network links are missing from the sample. We show missing links add additional sources of regressor endogeneity, thus invalidating conventional estimators used in the literature. We resolve these issues by rescaling peer outcomes with estimates of missing rates and constructing instruments that exploit properties of noisy network measures. We apply our method to study peer effects in household decisions to participate in a microfinance program in India. We find that conventional instruments that ignore our missing link correction yield sizeable upward biases in peer effect estimates.