Telescope Matching: A Flexible Approach to Estimating Direct Effects

(2018)

(with Anton Strezhnev)

Estimating the direct effect of a treatment fixing the value of a consequence of that treatment is becoming a common part of social science research. In many cases, however, these effects are difficult to estimate standard methods since they can induce post-treatment bias. More complicated methods like marginal structural models or structural nested mean models can recover direct effects in these situations but require parametric models for the outcome or the post-treatment covariates. In this paper, we propose an alternative approach, which we call \emph{telescope matching}, to estimating direct effects. The method combines matching and regression to impute missing counterfactual outcomes in a flexible manner. Using simulation and empirical studies, we show how this approach weakens model dependence for researchers estimating direct treatment effects.