Attitude Activation and the Study of Political Campaigns

• In Event: How Campaigns Shape Voters

Abstract

Researchers commonly use “activation” or “priming” analyses to understand the determinants of voters’ choices across elections. The most common research design to assess activation involves regressing vote choice on (potentially pre-election) attitudes. An attitude is said to be “activated” in an election if this regression coefficient is larger in magnitude than in other elections. We document substantial confusion over what activation means substantively and how to measure it statistically. We provide a rigorous foundation to study three potential types of activation: causal, prediction, and description. Causally, we clarify the assumptions necessary to assess whether campaigns “activate” attitudes and show they are considerably stronger than just assuming attitudes are (as-if) randomly assigned. For prediction, we demonstrate how to assess how much predictive power a variable provides. Finally, for description, we show how to calculate which voting blocs make the biggest changes in their vote contributions. We also show that use of panel studies potentially avoids one source of bias in studying activation, but introduces others. We apply these frameworks to data from recent presidential elections.

Authors

• William Marble, Princeton University

• Justin Grimmer, Stanford University

• Cole Tanigawa-Lau