This paper examines inference on social interactions models in the presence of missing data on outcomes. In these models, missing data on outcomes imply an incomplete data problem on both the endogenous variable and the regressors. However, getting a sharp estimate of the partially identified...