Who are Persuadable: Assessing External Validity of Local Persuasion Rate

• In Event: Political Methodology Virtual Poster Session

Abstract

This project advances the empirical analysis of political persuasion when the consumption of political information is itself a political decision. In the quantitative study of political persuasion, researchers are typically interested in identifying a quantity known as the persuasion rate. The persuasion rate measures the percentage of individuals who do respond to political persuasive information and otherwise would not (Jun and Lee, 2019). When there is non-compliance, analysis is restricted to a local persuasion rate, which is limited to measuring the persuasion rate for compliers (Jun and Lee, 2019). Under the monotone treatment effect and standard instrumental variable assumptions, local persuasion rate is point identified from the data. However, local persuasion rate has limited external validity because it is a persuasion rate for an unidentified subpopulation: a subpopulation who will not take the action of interest without consuming persuasive information and are compliers. In this paper, we provide identification results for identifying the pre-treatment characteristics of this unidentified subpopulation. Our identification results share similar flavor with Abadie’s kappa: we use a specific weight to ‘’find’’ this unidentified subpopulation. Our identification results are especially useful when the pre-treatment variables have discrete support. When the pre-treatment variables are continuous, we can nonetheless use our identification results to estimate their densities. Moreover, we provide the asymptotically valid confidence intervals that account for multiple hypothesis testing of the pre-treatment characteristics. We apply our methods to assess the external validity of the local persuasion rate in GOTV campaign (Green et al., 2003).

Author

• Arthur Zeyang Yu, University of Chicago