Correct Voting in 2010-2016 U.S. House of Representatives Elections

• In Event: Voting with All Five Senses

Abstract

Representative democracy is based on the idea that citizens choose the parties or candidates who most closely share their own policy views and concerns. All citizens do not vote, of course, but most political scientists are perfectly willing to grant that assumption at least for the subset of the population who make the effort to show up at the polls and vote. The concept of correct voting is designed to put these notions to the test. It is defined as the extent to which voters, inevitably under conditions of imperfect information, nonetheless vote for the candidate they should have voted for if they had perfect information. The concept has been extensively tested and validated with both survey and experimental data, but almost always with national-level federal elections. In the 12 U.S. presidential elections since 1972, for example, correct voting has varied between 59% and 88%, with a mean of 76%.

Using data from the 2010, 2012, 2014, and 2016 CCES surveys, this paper will extend the analysis of correct voting to elections for the U.S. House of Representatives. In each election year, we targeted races from four different types of electoral climates: (1) states with House races but neither gubernatorial nor Senate elections; (2) states with House and gubernatorial elections but no Senate race; (3) states with House and Senate elections but no gubernatorial race; and (4) states with House, Senate, and gubernatorial elections. Two of our election years included a presidential election, while two of them were “off-year” elections with no presidential race. We tried to select House races with at least 100 voters where a Democrat was facing off against a Republican. Within each election year and each type of race, roughly half of the selected races were competitive and half noncompetitive elections. All told we can examine 126 House races across these four election years. To estimate correct vote choices, we then searched candidate web pages to locate (and subsequently code) policy positions on the same issues respondents to the CCES surveys were asked about.

In past research (at the national level) individual-level predictors of correct voting include political interest and motivation, political knowledge/sophistication, and the availability of important political heuristics (e.g., strength of partisan identity). We expect those same factors to predict correct voting in House elections as well. More interesting hypotheses involve institutional-level variation across these 126 elections. We will test two primary institutional-level hypotheses:

H1 Office Matters: Overall levels of correct voting will, on average, be lower in House elections than they typically have been in presidential elections, because most voters will think the presidential election is more important, they will care more about its outcome, and they will have more information about the two major-party candidates.

H2 Finite Political Attention: The more simultaneous vote decisions a citizen must make in any given election, the less attention that citizen will be able to devote to any of them, and thus the lower the probability of casting a correct vote. Thus, we predict that general levels of correct voting will be higher in House races in states without simultaneous Senate or gubernatorial races on the ballot, than in states that have such simultaneous elections.

Authors

• Richard R. Lau, Rutgers University, New Brunswick

• Kathleen M. Rogers, Rutgers University