Voting Can Be Hard, Information Helps
Melody Crowder-Meyer, Shana Kushner Gadarian, and Jessica Trounstine
How do voters make decisions about which candidates to support? This isn’t just a question we study as political scientists – it’s a question we confront as voters as well. In 2016, one of us had to vote for a presidential nominee by picking convention delegates on a ballot that did not clearly indicate which presidential candidate each delegate supported. In 2017, another one of us was asked to select 5 names from a list of 13 candidates for town board on a ballot with no additional information about the candidates – not even their party affiliations. In contrast, in both of those years, one of us living in California chose among candidates – in both partisan and non-partisan races – on ballots including not just names but also “ballot designations” indicating candidate occupations and past experience in the office.
Our research asks whether voters make different choices depending on the kind of information they have available. Ideally, in a democracy, voting decisions are based on information about candidates’ political positions and experience. Yet, voters know little about candidates for many offices and the decline of localmedia makes it even harder for voters to be informed. Ballot design may help voters, but ballots don’t always allow voters to live up to this ideal. Instead voters glean what they can from the names they see, using traits like race or gender and their associated stereotypes to make their choices. We use a set of experiments to test how ballot designs that provide different levels of information to voters affect the choices they made. This research helps us better understand what to expect from elections as it becomes increasingly common for voters to have little information about the candidates.
Figure 1
Note: Each dot represents a regression coefficient. We ran separate regressions for each information condition including both race and gender of all candidates. Horizontal lines represent 95% confidence intervals around each coefficient. The coefficients can be interpreted as the probability of selecting a candidate with the given characteristic relative to the baseline characteristic (White/male).
When nothing but candidate names are available on ballots, voters can and often do make assumptions about the race, ethnicity, and gender of the candidates among whom they are selecting – and the skills, experiences, and ideologies of the candidates based on those traits. A wide body of research (and our own investigation of how Americans rated candidates from different groups) demonstrates that people hold a variety of stereotypes about men and women and candidates from different racial and ethnic backgrounds. Americans believe that men and White candidates have more political experience than female and minority candidates. People also rank Whites higher on qualities like being hardworking and intelligent, and believe that women and people of color are more ideologically liberal than White and male candidates. Consequently, when a sample of Americans in our experiments was asked to vote for candidates on ballots mimicking those common in local and primary elections across the country – with no party identification nor any details about the candidate other than their name provided – candidates of color lost support relative to White candidates. On the other hand, women candidates fared better than men, especially among liberal voters.
In other words, when voters are asked to make decisions in the kinds of low-information conditions very common in American elections, they use their assumptions about candidates’ race, ethnicity, and gender to help guide their choices. These assumptions result in large penalties for Black candidates, moderate penalties for Asian and Latino candidates, and slight advantages for female candidates, relative to White and male candidates. Among other things, voters appear to be making assumptions about candidates’ political ideologies due to their race, ethnicity, and gender, expecting that White and male candidates are more conservative than other candidates. In low-information elections, when voters are provided only names on ballots, liberal and Democratic respondents penalize candidates of color relative to White candidates less than conservative and Republican respondents. Additionally, in low-information elections, liberal and Democratic respondents vote more often for women than men candidates, while conservative and Republican respondents vote more often for men than women. Of course, ampleexamplesexist of politicians who do not fit within race- and gender-based ideological stereotypes – so voters making decisions based on such assumptions may end up supporting candidates with an ideology very different from what they intended.
Setting up elections in ways that reproduce prejudices and stereotypes seems problematic. However, there is a solution! When voters are offered easily available alternative information about candidates – such as their occupation and incumbency status – reliance on candidate race and gender decreases substantially. This effect can be seen in the graph below by comparing how respondents in our experimental elections voted in an election where they had only candidate names on the ballot (the blue triangles) versus an election where they were offered ballots like those in California, noting both candidate names and their occupations (the red circles). Points to the left of the 0 line indicate a candidate with that race receives less votes than a White candidate, while points to the right of the 0 line indicate a candidate with that gender receives more votes than a male candidate.
By adding just one little piece of information to voters’ ballots – as some states and localities already do – the penalties faced by Asian and Latino candidates relative to White candidates completely disappeared, and the penalties faced by Black candidates decreased substantially. This movement occurred even among those individuals in our study with high levels of racial resentment. While the effects of candidate race and ethnicity for racially resentful voters did not go away completely, adding information made the penalties they applied to candidates of color much smaller. Adding even more information (like incumbency status) to ballots further ameliorated race and gender-based voting, such that neither characteristic influenced voter’s choices in our “high information” condition. Instead, voters used characteristics more under candidates’ control and likely more relevant to their policymaking – candidate occupations and political experience – to make determinations about candidate quality and likely behavior in office. This is a rare case where a policy solution to help address inequality – simply adding a few words next to candidate names on the ballot – is both straightforward and inexpensive.
Melody Crowder-Meyer is Assistant Professor of Political Science at Davidson University. Her research investigates inequality in American politics and government, with a particular focus on women in politics, candidate emergence, the behavior of political elites, and local politics. Her research has been published in outlets such as Journal of Politics, Political Behavior, Politics & Gender, and Research & Politics. Website: http://www.melodycrowdermeyer.org/
Shana Kushner Gadarian is Associate Professor of Political Science at Syracuse University. Her research focuses on political behavior, political psychology, and political communication in the United States. She is the co-author of Anxious Politics: Democratic Citizenship in a Threatening World (Cambridge University Press) and her work has been published in the American Journal of Political Science, Journal of Politics, Political Behavior and others. Website: https://sgadaria.expressions.syr.edu/J
essica Trounstine is Associate Professor of Political Science at University of California, Merced. She is the author of Segregation by Design: Local Politics and Inequality in American Cities (Cambridge University Press) and Political Monopolies in American Cities: The Rise and Fall of Bosses and Reformers (Chicago University Press) as well as numerous articles. Her work is focused on representation, inequality, elections, and public services in American local governments. Website: http://faculty.ucmerced.edu/jtrounstine/