We hope you can join us this Wednesday, September 14, 2011 for the Applied Statistics Workshop. Matt Blackwell, a Ph.D. candidate from the Department of Government at Harvard University, will give a practice job talk entitled "A Dynamic Causal Inference Approach for Estimating the Effectiveness of Negative Campaigning". A light lunch will be served at 12 pm and the talk will begin at 12.15.

"A Dynamic Causal Inference Approach for Estimating the Effectiveness of Negative Campaigning"
Matt Blackwell
Government Department, Harvard University
CGIS K354 (1737 Cambridge St.)
Wednesday, September 14th, 2011 12.00 pm

Abstract:

Traditional single-shot causal inference models investigate the effect of a single action at a single point in time and are an invaluable tool for political scientists. Often, however, actions unfold over time, with political entities reacting to a shifting environment. Accordingly, single-shot methods leave researchers unable to extract meaningful causal inferences about these dynamic processes. This stems from a fundamental tension: in dynamic settings, regression and matching force a choice between omitted variable bias on the one hand, and post-treatment bias on the other and are unable to simultaneously correct for both. To avoid these problems, I introduce a framework for dynamic causal inference and utilize marginal structural models to estimate dynamic causal effects. The effectiveness of "going negative" serves as a motivating example---an apt illustration since candidates change their strategy as the campaign unfolds. Furthermore, I introduce novel diagnostics and a sensitivity analysis for the model.

Posted by Konstantin Kashin at September 12, 2011 2:42 AM