We begin by introducing the generalized method of moments, and overidentification tests. We then discuss instrumental variables, a quasi-experimental methodology to estimate treatment effects. Under the assumption that the treatment effect is constant, the generalized method of moments can be used for estimation with instrumental variables, and overidentification tests can be used to test instruments? validity. In a model with heterogeneous effects, overidentification tests break down, but two-stage least squares regressions can be used to identify the treatment effect among units that comply with the instrument. Finally, we discuss recent developments in the instrumental variable literature.