Estimating the Impacts of Program Benefits: Using Instrumental Variables with Underreported and Imputed Data