Research

Much of Andrew's research is in the field of econometrics, and concerns instrumental variables. Instrumental variables are variables that affect one part of a co-determined system without affecting another part of the same system. For this type of estimation to be effective, an instrumental variable must satisfy the relevance condition, that is, it must affect one part of the system, and it must satisfy the exclusion restriction, that is, it must not affect the other part of the system. Andrews' research concerns situations in which either the relevance condition or the exclusion restriction hold only weakly. Together with Mikusheva, he has studied the properties of weak instruments in estimating dynamic stochastic general equilibrium models and other economic models that employ nonlinear Generalized Method of Moments statistics models. They have proposed methods to more accurately test hypotheses and construct confidence intervals under these statistical conditions. Together with Gentzkow and Shapiro, he has developed a measure of the potential bias in estimators due to violations of the exclusion restriction (sensitivity). Together with Kasy, he has developed a method to correct for publication bias in replication studies and meta-analyses.^[1]