Essays on Applied Political Methodology for Political Campaigns

Abstract

Quantitative social science techniques are increasingly used and misused by American political campaigns. In their efforts to influence a nationalized and polarized electorate, practitioners collect ever-larger amounts of data and conduct experiments to understand the position and malleability of the electorate. This rise in quantitative techniques has led to efficiency---increasing the concentration of spending and attention on the most persuadable voters in the most pivotal states---as well as stagnation. We now confront a political landscape crowded with analysis that can be critically informative but is also often incomplete, biased, and reliant on largely private data. A critical issue is the frequent use of quantitative social science and machine learning tools for inference without adequate attention paid to uncertainty and generalizability. More transparency and emphasis on the fundamentals of inference would allow practitioners to better understand the electorate, how and when to take action, and to appropriately consider uncertainty and the range of possible outcomes in political scenarios. To do this, both applied researchers and practitioners need to move toward an iterative use of research to build knowledge over time, rather than relying on single studies to answer big questions across broad contexts. In this work, I contribute to this agenda with new methods campaigns can use to understand the electorate and offer an overall approach to iterative research that helps resource-constrained actors optimize their influence.