The Application of Big Data in Surveys to the Study of Elections, Public Opinion, and Representation

Abstract

INTRODUCTION Until recently, political scientists relied primarily on a small number of academic surveys that were extremely limited in size and scope. Their small sample sizes made them unsuitable for examining variation in public opinion across subpopulations, such as between Hispanics and Asian Americans. These surveys were also ill suited for examining public opinion at the subnational level. Recently, however, there has been a revolution in scholars’ ability to access large data sets of public opinion data and in their ability to leverage the full array of survey data at their disposal. Over the past decade, the amount of available public opinion has exploded due to the increasing availability of large archives of commercial polls that were collected over the past 75 years. In addition, advances in cooperative survey research have dramatically expanded the sample sizes in individual academic surveys (Ansolabehere and Rivers, 2013). At the same time, a broad array of statistical advances have improved scholars’ ability to make substantive inferences from the full array of available survey data. Researchers have developed new model-based weights to compensate for the unrepresentativeness of survey data from earlier time periods (e.g., Berinsky, 2006; Berinsky et al., 2011). Moreover, they have developed new techniques to estimate the attitudes of small geographic areas and demographic groups (Gelman and Little, 1997; Park, Gelman, and Bafumi, 2004). These methodological advances have enabled scholars to measure the mass public's ideology and policy views at a variety of geographic scales, such as states, congressional districts, state legislative districts, and even large cities and towns. Moreover, scholars have developed new methods to leverage historical survey data to examine changes in public opinion over time (Stimson, 1991; Caughey and Warshaw, 2015; Enns and Koch, 2013). These advances in the measurement of public opinion have enabled scholars to perform nuanced studies on political representation and accountability at the local (Tausanovitch and Warshaw, 2014), state (Lax and Phillips, 2009 a ; Pacheco, 2013), and federal levels (Ansolabehere and Jones, 2010; Bafumi and Herron, 2010; Clinton, 2006; Hill, 2014). Scholars have found that policy outcomes at each geographic level are responsive to public opinion. They have also examined whether particular political institutions, such as direct democracy or nonpartisan elections, enhance the relationship between public opinion and salient political outcomes (Canes-Wrone, Clark, and Kelly, 2014; Lax and Phillips, 2011). This work could eventually provide new insights into how institutional reforms might improve democratic governance.