Near/Far Matching: Building a Stronger Instrument

• Activity Number: 402
• Type: Topic Contributed
• Date/Time: Tuesday, August 3, 2010 : 2:00 PM to 3:50 PM
• Sponsor: Health Policy Statistics Section
• Author(s): Mike Baiocchi*+ and Dylan Small and Scott Lorch and Paul Rosenbaum
• Companies: University of Pennsylvania and University of Pennsylvania and Children's Hospital of Philadelphia and University of Pennsylvania
• Keywords: Causal inference ; instrumental variables ; strength of instrument ; matching ; binary outcomes

Abstract

Instrumental variables (IV) is a framework for making causal inferences about the effect of a treatment based on an observational study in which there are unmeasured confounding variables. In many policy settings the objective is to estimate the effect of a treatment on a binary outcome. We develop a nonparametric matching technique - "near/far matching" - which is capable of estimating population level treatment effects when the outcome is binary. We provide a test statistic, with standard errors, and a method for sensitivity analysis. Many researchers assume the strength of a particular instrument is fixed. It is not. A novel feature of our method is that it allows us to manipulate the strength of a particular instrument. We illustrate our method using a study of neonatal intensive care units (NICUs) treatment effect on premature babies (preemies) born in Pennsylvania.