Thinking about instrumental variables (in Russian)
We take a decision-theoretic view on the question of how to use instrumental variables and method of moments. Since prior beliefs play an inevitably strong role when instruments are possibly "weak", or when the number of instruments is large relative to the number of observations, it is important in these cases to report characteristics of the likelihood beyond the usual IV or ML estimates and their asymptotic (i.e. second-order local) approximate standard errors. IV and GMM appeal because of their legitimate claim to be convenient to compute in many cases, and a (spurious) claim that they can be justified with few "assumptions". We discuss some approaches to making such a claim more legitimately.