A Causal Framework for Intervention Evaluation with Survey Data

• Activity Number: 141
• Type: Contributed
• Date/Time: Monday, August 5, 2013 : 8:30 AM to 10:20 AM
• Sponsor: Social Statistics Section
• Author(s): Robert Ashmead*+
• Keywords: Causal Inference ; Propensity Scores ; Complex Survey Design

Abstract

In recent years propensity score methods such as stratification, inverse-weighting, and matching have become common practice in estimating treatment effects from observational studies with a causal interpretation. The methodology of these estimators is based on the assumption of a simple random sample from the population. However, probability samples based on survey designs that are not simple random samples represent a sizable collection of observational data to which researchers are often interested in applying causal inference methods. Through the literature we found that researchers either use mostly ad-hoc procedures to account for the survey design or ignore it all together. We develop a framework from which to apply propensity score methods to survey data. We also discuss the practical implications of the proposed framework with a large health survey.