Abstract
In November 2017, New Orleans elected the first woman, and first Black woman, mayor in the city’s history. Voters were unable to rely on gender, race, or partisanship to differentiate between the candidates in the race. How, then, do voters make decisions absent traditional heuristics? Using an analysis of campaign materials and two-wave panel survey, we show that the candidates sent ideological signals with endorsements and issue foci and that voters responded by placing the candidates ideologically. Those voters who could not differentiate between the candidates’ ideologies were less likely to turn out to vote and took longer to decide in the elections. Using a new measure of relative ideological distance adopted for multicandidate races, we show that the distance between each voter and the nearest candidates correlated with vote choice. Our results add to our knowledge of voting behavior and the use of ideology in local elections.
In 2018, New Orleans celebrated the tricentennial anniversary of the city’s founding. Presiding over this celebration was something new for the city: a woman mayor. Elected in what the local press called a “topical depression” of a mayoral race (Morris 2017), LaToya Cantrell’s selection as New Orleans’ mayor represents an opportunity to examine how candidate behavior and voter ideology interact to shape vote choice. Although political science often focuses on national elections, more than 95% of elections in the United States are local. These elections are also typically low-information affairs (Crowder-Meyer, Gadarian, and Trounstine 2015; de Benedictis-Kessner 2018b; Trounstine 2013) where voters search for easily identifiable cues that help them make a nonrandom choice (Fulton and Ondercin 2013; Ondercin and Fulton 2019).
Although voters rely on partisanship to help them make decisions in national elections, this cue is often weak, useless, or missing entirely in local elections (Bernhard and Freeder 2018; Schaffner, Streb, and Wright 2001). As a result, voters rely on other heuristics like incumbency, experience, and stereotypes based on the candidate’s identifiable demographic characteristics (Badas and Stauffer 2019; Bauer 2020; Bernhard and Anzia 2019; Crowder-Meyer, Gadarian, and Trounstine 2015; Trounstine 2011). What happens when voters must make decisions absent all traditional heuristics, that is, when electoral circumstances mean no variation in candidate party, gender, and race? Some opt out of voting when they are unable to make an easy decision (Barnes, Tchintian, and Alles 2017; Fulton and Ondercin 2013; Ondercin and Fulton 2019), but many voters routinely turn out to vote even in low-information elections. 1
We argue that candidates recognize this challenge and provide voters with a variety of ideological cues. Voters, then, assume the ideological placement of a candidate based on the candidates’ self-presentation and endorsements. To make sense of a narrow ideological space, voters will project more extreme ideological positions onto the candidates in the race and then compare the ideological distance between themselves and the candidates to select a representative.
To test these propositions, we examine candidate and voting behavior in the 2017 New Orleans mayoral election. LaToya Cantrell was elected as the city’s first woman and first Black woman mayor. The city selected Cantrell in a decidedly low-information environment, absent party, gender, and race variation across the candidates. As expected, we find that candidates sent strategic ideological cues. In turn, voters responded by projecting a more extreme ideological position onto the candidates than their positions warranted. Those voters who are unable to distinguish an ideological distance between the candidates are less likely to turn out to vote and take longer to choose a candidate in the race.
Using a new measure of relative ideological distance (Adams et al. 2017; Tausanovitch and Warshaw 2018) adapted for a multiperson race, we show that ideological distance shapes vote choice above demographic and socioeconomic factors. Our results suggest that when typical voting heuristics are missing, most voters receive candidates’ ideological signals and vote accordingly, but that this ideological placement is relative to the other candidates in the race. As the number of candidates shrinks, the relative distance becomes easier for the voters to assess and becomes a more powerful factor in shaping vote choice.
Low-Information Elections
The most common voting experience in the United States is making a choice in a low-information local contest (where “every day is election day somewhere”; Sumner, Farris, and Holman 2019). Partisanship and ideology play key roles in voting behavior elections (Boudreau, Elmendorf, and MacKenzie 2015; Hajnal and Trounstine 2014; Oliver and Ha 2007; Warshaw 2019), as voters seek to simplify their political choices using shortcuts. Party identification is often absent in local politics because so many local races are officially nonpartisan or feature candidates of the same party. Even so, Kirkland and Coppock (2018, p. 573) remind us that “elections that are nonpartisan in theory are not necessarily nonpartisan in practice” and voters use other pieces of information from candidates as if they had partisanship to guide their vote choices (Boudreau, Elmendorf, and MacKenzie 2015; Sances 2018), perhaps especially in low-information and nonpartisan contests. In these races, voters infer partisan and ideological placement of candidates, even if they often are inaccurate in these assessments (“Former Judge Bagneris Announces He’s Running For Mayor Again” 2017; Kalmoe 2020), and mayoral race offers an opportunity to understand how ideology operates in an environment where partisanship is held constant (de Benedictis-Kessner and Warshaw 2016; Kirkland and Coppock 2018).
Absent the partisan cue, factors, such as name recognition (Kam and Zechmeister 2013) and demographic characteristics of candidates, provide informational shortcuts for voters (Alvarez, Hall, and Levin 2018; Badas and Stauffer 2019; Bauer 2020). Race and gender stereotypes provide cues for voters to assess a candidate’s issue competencies, personal traits, and, ultimately, candidates’ ideology (Bauer 2015a; Cassese and Holman 2018; Holman 2015; McDermott 1998; White, Laird, and Allen 2014). The role of race is especially salient in cities (Kaufmann 2004), where the “typical urban election pits two racial groups against each other” (Hajnal and Trounstine 2014, p. 72) and gender and race can combine to shape assessments of women of color in local politics (Cargile and Pringle 2019).
Hutchings and Valentino (2004, p. 4) remind us that “racial concerns” are “often tied up” with political ideology in America. A candidate’s race or gender sends signals about ideology, especially in low-information contexts. Voters often believe women and African-American candidates to be more liberal than male and White candidates of the same party (Bauer 2015b; McDermott 1997; Schneider and Bos 2011b). To the extent that Black candidates and women are actually more liberal, these stereotypes provide a useful cue (Bauer 2018; Laird 2019). In elections featuring two members of minority groups, racial divides among voters are smaller than in biracial contests, but still dwarf all other demographic divisions (Hajnal and Trounstine 2014).
Studies of voting behavior in local elections often concentrate on how candidate race and voters’ race interact (Burns 2006; Hajnal and Trounstine 2014; Kaufmann 2004). Past elections in New Orleans demonstrate the importance of considering both the race of the voter and of candidates: In the 1986 and 2002 mayor’s races, when two Black male candidates faced off, the winning candidates in each election were the ones who carried White voters (Burns and Thomas 2015; Lay 2009; McBride and Parker 2008).
Elections that feature two African-American women candidates, as in the 2017 New Orleans mayoral race, offer an opportunity to study how candidates signal their ideological differences in an environment that is deprived of traditional information shortcuts and whether voters receive these signals and use them to select a candidate. We should note that most U.S. elections feature two candidates of the same gender and race, but they are usually White men. In these contests, cues about racial issues are used strategically by candidates and received by voters in certain circumstances (Lemi 2017; Mendelberg 2001). Two Black women frontrunners is a relatively rare phenomenon in American politics, and although these candidates share these characteristics, their mere presence sends racial and gender cues in ways that two White men do not. We do not argue that race and gender “do not matter” when two Black women face off, but rather that voters have a more difficult time using traditional racial and gender stereotypes as information shortcuts when the candidates share these characteristics.
Ideological Signaling in the 2017 New Orleans Mayoral Race
When filing for the 2017 mayoral election closed, New Orleans faced a slate of 17 candidates, including three clear frontrunners: LaToya Cantrell, a nonnative city council member who entered politics via post-Katrina community organizing; Desiree Charbonnet, who came from a prominent and well-connected family and had served in an elected judicial position; and Michael Bagneris, a former judge and a member of a well-connected family who had previous run for mayor. Other candidates were a mix of politicians, newcomers hoping to improve name recognition for future races, and the run-of-the-mill New Orleans weirdos. 2
The Progressive
Cantrell emerged from a grassroots background, having gotten her start in politics as a neighborhood organizer in the aftermath of Hurricane Katrina in 2005. Raising less than Charbonnet ($741,650 over the course of the election), Cantrell primarily relied on a community-engaged campaign, a savvy social media campaign, and support from some local political groups with clout (like BOLD: Black Leadership Organizational Development [Burns and Thomas 2015]) and national progressive groups, such as the Working Families Party, Indivisible, Our Revolution, and Step Up Louisiana. Local newspapers frequently referenced her liberal positions and “populist appeal” (Tidmore 2017).
Cantrell’s campaign highlighted traditional progressive and women’s issues to differentiate herself from the other frontrunners. Cantrell was the only candidate that focused on issues like LGBTQ (lesbian, gay, bisexual, transgender, and questioning) rights, equal pay, and racial equality on her website, along with discussing crime, the economy, housing, and flooding (see Table 1). Endorsements on her website came from mainstream White newspapers, local leaders, and progressive national organizations like Democracy for America. This issue portfolio was less evident in her direct mail advertisements, where ads were most commonly focused on crime and the economy. However, direct mail from third-party organizations and endorsements focused on progressive issues, including race relations, women’s issues, and affordable housing.
Issue Foci and Endorsements by Candidates.
Note. Websites: a checkmark means that the candidate specifically mentioned the issue on their website or included an endorsement from an individual with a background in the area or from a group that focused on this issue. Websites coded by the authors one week before the general election and one week prior to the runoff election. Direct mail: A checkmark indicates at least one direct mail piece that mentioned the issue or included an endorsement from an individual with a background in the area or from a group that focused on this issue. Data were from candidate flyer archive developed by the authors; only ads that mention a mayoral candidate are included. Each ad could mention multiple issues. LGBTQ = lesbian, gay, bisexual, transgender, and questioning.
The Moderate Frontrunner
Charbonnet’s frontrunner status was cemented early in the race by nearly a million dollars in her campaign fund and deep connections to political elites in the city. Her prominent family was strongly connected to labor unions and features several political leaders. She had ties to U.S. House Representative Cedric Richmond, whose “powerful political organization . . . has all but supplanted the traditional political groups in the city” (Williams and Adelson 2017) and to other political leaders, including former Mayor Sidney Barthelemy, in whose administration her father had served. She also had the backing of major labor groups, the Orleans Parish Democratic Executive Committee, several local Black newspapers, and a “horde of city contractors” (Vanacore 2017). Her campaign was polished and professional, including a formal announcement party, glossy policy proposals, practiced debate answers, and a chauffeur to drive her around. She raised $1.4 million for the general election.
Her background as a lawyer and a judge made a focus on criminal justice a natural fit for her campaign. Table 1 shows the issues that appeared on the websites and in direct mail pieces: Charbonnet’s website primarily focused on crime, the economy, and housing. She trumpeted her judicial experience, while still tapping into liberal (affordable housing) and conservative (economic development) issues of importance.
The Conservative(ish) Democrat
Michael Bagneris focused his narrative on his American Dream–like story, from a humble upbringing, the son of a cleaning lady and a janitor, to attending Yale and Tulane Law (“Former Judge Bagneris Announces He’s Running For Mayor Again” 2017). A former Civil District Court Judge, Bagneris raised $705,000 from his base, which included the business community and Frank Stewart, a local businessman who opposed the city’s controversial removal of Confederate monuments (Vanacore 2017). Focusing on crime and safety, and with the backing of pro-monument White voters, Bagneris represented the most conservative positions early in the race. 3 Bagneris saw an opportunity to position himself to the right of Charbonnet, focusing on economic development and infrastructure.
Although being far from a conservative by most measures of the term in American politics, his positions were the most conservative of the top-three candidates and fairly conservative for liberal New Orleans. He actively courted an endorsement from the Orleans Parish Republican Executive Committee, whose support he had received in previous elections (Tidmore 2017). These issues and his moderate approach appeared on his website and in direct mail pieces, which focused on economic development, crime, flooding, and infrastructure. His endorsements from organizations like the Central City Democratic Association and trade unions also followed this vein. He also focused extensively on corruption, attempting to remind voters of the differences between him and Charbonnet and Cantrell. Ultimately, Bagneris came in third in the general election, with 19% of the vote. He endorsed Cantrell in the runoff held a month later.
We use this three- and, then, two-person race that contained very few information shortcuts to examine the role that ideological signaling might play in low-information elections. To do so, we rely on the literature on how voters use ideology to evaluate candidate positions, particularly in the context of a two-candidate race with partisan cues. We, then, adapt these models to a multicandidate race where partisan identification is not a useful shortcut.
Ideological Signaling from Campaigns and Voter Reactions
We contend that candidates signaled their ideology through their campaign messages, the issues on which they focused, and their endorsements. Local media tend to cover campaigns only sparingly (McDermott 1997; Napoli et al. 2018), giving candidates and interest groups a great deal more independent influence in shaping voters’ perceptions. Candidates behave strategically in their campaigns to attract voters, signal competence and expertise, and mobilize supporters to turn out to vote (Dittmar 2015; Holman, Schneider, and Pondel 2015; Vavreck 2001). Research on candidate behavior suggests that, in presidential and congressional elections, candidates engage in both issue- and identity-based targeting to prime particular issues or persuade particular voters of the supremacy of one candidate over another (Herrnson, Lay, and Stokes 2003; Hillygus and Shields 2008; Holman, Schneider, and Pondel 2015).
Community, issue-based, labor, and local party organizations all also engage in strategic behavior to assist their preferred candidate’s campaign or damage a less preferred candidate (Benjamin 2017; Benjamin and Miller 2019). Group endorsements operate much like candidate demographic characteristics—giving voters a cue about whether a candidate’s interests are in line with theirs. In nonpartisan contests or races in which candidates of the same party are facing one another, endorsements can give voters information about ideological leanings (Benjamin and Miller 2019), providing cues to voters about whom to vote for (McDermott 2006) or against (Arceneaux and Kolodny 2009).
Group endorsements are particularly powerful in off-cycle elections, where voters have fewer partisan cues to rely on in voting (Anzia 2013). Race, ethnicity, and gender all interact with both endorsing organizations and candidate characteristics to shape the efficacy of the information signal. For example, in contests where race or ethnicity becomes salient and a coethnic is not running, endorsements from Black organizations serve as an important cue to Black voters (Benjamin 2017). Research also suggests an increased “nationalization” of local elections (Hopkins 2018), where voters apply their knowledge of national political affairs to make local decisions. Given what we know about strategic behavior by candidates and groups, we expect that the candidates in the New Orleans mayor’s race will jockey for a recognizable ideological position on a relatively narrow ideological space. To do so, candidates will strategically feature issues and endorsements associated with a specific ideological position they have chosen (Hypothesis 1).
Voters are motivated reasoners that seek out consistency in their evaluations of candidates (Ditonto, Hamilton, and Redlawsk 2013; Taber and Lodge 2006). Ideology and partisanship stereotypes help individuals process and organize additional information and fill in gaps in their knowledge (Adams et al. 2017; Rahn 1993). And, voters behave “as if” they have complete information about the candidates when they only have partisanship (Tausanovitch and Warshaw 2018).
In the context of local elections, voters may use information shortcuts (including gender and race) about candidates strategically to make decisions about which candidate to select (Bauer 2020; Crowder-Meyer, Gadarian, and Trounstine 2020; Keele et al. 2017). We argue that, if a voter is able to place a candidate ideologically, they should project that ideology to fill in gaps in their knowledge about the candidate, creating ideological space between the candidates. Furthermore, we contend that this is a relative projection, such that voters assess candidates’ ideologies relative to one another. In a context with several candidates, we expect that once the candidates engage in ideological signaling, voters respond by separating the candidates further ideologically than the candidates actually have positioned themselves. Specifically, we expect that voters will see the slightly more liberal candidate as much more liberal, the moderate candidate as solidly moderate, and the slightly more conservative candidate as much more conservative (Hypothesis 2). In the two-stage process in Louisiana, voters will carry these views over into the runoff election, creating even more ideological space between the winning candidates (Hypothesis 3). These pieces of ideological information assist voters in making decisions. As a result, those voters who cannot discern between the candidates’ ideologies will be more likely to be undecided and less likely to turn out to vote (Hypothesis 4).
Finally, we argue that these pieces of ideological information about the candidates will be useful for voters in making a decision in the election. To assess this, we evaluate relative ideological distance between the candidates. In doing so, we assume a linear space and examine each respondent’s relative ideological proximity to the candidates (Adams et al. 2017; Tausanovitch and Warshaw 2018). Although previous research on ideological positioning has concentrated on selection between a Democratic and a Republican candidate, local elections often feature more than two potential candidates in the general election. As such, we define the relative distance in three ways: between the liberal and moderate candidates, between the moderate and conservative candidates, and between the liberal and conservative candidates. For each, relative proximity is defined as the distance between the respondent’s distance from candidate 1 and candidate 2. For example, the liberal–moderate (L–M) distance is as follows:
where vie represents the self-placement of respondent i voting in election e, and Le and Me represent the voter’s placement of the most liberal and moderate candidates, respectively. When |vie – Le| < |vie – Me|, the most liberal candidate is closer to the voter and the value will be negative. We repeat this relative measure for Relative Proximity M–C, or the relative distance between the voter and the moderate and conservative candidate, and Relative Proximity L–C, or the relative distance between the voter and the liberal and conservative candidate.
When there are three candidates in the race, we argue that votes for the most liberal candidate will relate to the relative distance between the liberal and moderate candidates and less related to the relative distance between the moderate and conservative candidates or the liberal and conservative candidates (Hypothesis 5a). Similarly, votes for the conservative candidate will relate to the relative position between the conservative and moderate candidates and less so to the relative position between the moderate and liberal candidates or the distance to the most liberal candidate (Hypothesis 5b). Finally, support for the moderate candidate should relate to both distance to the liberal candidate (L–M) and distance to the conservative candidate (M–C) (Hypothesis 5c). When the pool of candidates shrinks in the runoff election, the relative distance between the candidates should be easier for voters to evaluate and exert a stronger effect on vote choice (Hypothesis 6). To test our hypotheses, we turn to two sets of data: candidate behavior and voter reactions.
Data and Methods
Assessing voting behavior in local elections provides a specific set of methodological challenges (Sumner, Farris, and Holman 2019). To understand how candidates and groups behaved in the election, we use data on candidate behavior, endorsements, local media coverage, and voter attitudes via focus groups and surveys.
Candidate Behavior
We use website, endorsement, and direct mail data to evaluate how campaigns signal ideological positions. 4 We coded the content of the direct mail pieces, including information on the featured candidate(s), sponsors, the issue focus, and their timing (Spenkuch and Toniatti 2018). We also collected versions of the candidate’s websites from before the general and runoff elections and coded information about their issue positions (Schneider 2013), endorsements, and information about their appeals targeted at specific groups (Holman, Schneider, and Pondel 2015).
We look at three pieces of evidence: Which conservative, moderate, and liberal issues did each candidate focus on in their websites and direct mail; which conservative, moderate, and liberal organizations endorsed each candidate; and did third-party endorsements and mailers focus on conservative, moderate, and liberal issues?
Voter Preferences
We evaluate voter decision making through two waves of survey data. Participants in the survey data were recruited via a proprietary list generated by a New Orleans public opinion firm. Survey participants were recruited via a series of short emails asking them for their opinion, with the incentive of winning one of 10 $100 Amazon gift cards. We began recruitment for the first wave 18 days before the general election, right after early voting had begun. In this wave, 525 respondents completed the entirety of the survey. We recruited for the second wave of the survey starting 14 days before the runoff election, with 483 participants in this wave of the survey. The majority of the respondents (383) filled out both waves of the survey. We only use data from the 383 participants who took both waves of our survey. Benchmarking of the racial and gender breakdown of survey participants against New Orleans voters (see Supplemental Appendix) shows a close match of our sample to voters who participate in local elections. In addition, the overall vote choices in our sample closely resembled actual voter outcomes in both the general and runoff elections.
Ideological Positioning
To evaluate how ideology shapes vote choice in the election, we asked voters to place themselves and each of the top candidates onto an ideological scale from extremely liberal to extremely conservative.
Ideological Signals from the Candidates
Not all issues in New Orleans local politics are easy to place on an ideological spectrum. For example, corruption in New Orleans is always a major issue in local elections and does not track easily with an ideological position (Crawford 2019). We rely on scholarship on New Orleans politics to classify liberal, moderate, and conservative issues in local politics (Burns and Thomas 2015; Chambers and Nelson 2014; Lay 2016). Liberal issues are those that involve clear examples of redistributional efforts (such as affordable housing and raising the minimum wage) and the rights of marginalized groups (like women’s issues or LGBTQ advocacy), or are connected with national liberal efforts (like an endorsement from an organization like Indivisible). 5 Moderate issues are those with broad appeal and no specific ideological tilt: education, infrastructure, and, in New Orleans, organized labor. Conservative issues (again, relatively conservative, given the context of a liberal New Orleans) include crime, corruption, and economic development. In Table 1, we present the issue foci and endorsements of the candidates’ websites and direct mail.
The issue foci and endorsements of the candidates show that each candidate wanted to engage in as broad appeal as possible while still communicating a specific ideological position. Cantrell’s public presentation focused on a broad set of issues, with a particular focus on liberal issues like affordable housing, race relations, and women’s issues. Her mailers reflect this attempt to bridge liberal and mainstream issues: For example, in one direct mail flyer, Cantrell focused on her successes on infrastructure improvements, affordable housing, health, safety, and economic development. There was very little change over the entire course of Cantrell’s campaign: She focused on issues like affordable housing early in the race and continued to discuss the issue throughout the entire course of the election. She also focused on endorsements from liberal organizations throughout the course of the election.
Looking next at the moderate candidate, we find that Charbonnet’s issues and endorsements span the ideological spectrum, with a clustering on moderate issues. Part of this emerged organically for Charbonnet—as a former judge, her focus on crime allowed her to highlight both her experience and her policy proposals. In one of her central pieces of direct mail, the candidate focused on safety, noting the need for “more police officers per shift,” and streets, pledging to engage in “an aggressive paving program from Day One.” Charbonnet’s issue and endorsement foci changed considerably over the course of the election. Although Charbonnet focused on crime throughout the election, she shifted to more conservative policy stances during the runoff, focusing on economic development and shifting crime discussions toward policies like more policing and crime cameras. In the last days of the election, Charbonnet’s mailers turned decidedly conservative, including one featuring an endorsement from Steve Scalise, then-Majority Whip in the U.S. House, and a very conservative Republican representative from Louisiana. In this way, Charbonnet’s campaign reflects an awareness that Cantrell had successfully staked out the liberal ideological position and an attempt to woo more conservative voters.
As the more conservative candidate, Bagneris focused almost entirely on conservative and moderate issues, with a particular emphasis on crime and corruption. In one of his flyers, he notes, “The city is in crisis. The pumps don’t work, the police are understaffed, and the roads are worse than those found in a third world country.” Bagneris’s conservative position was also amplified by an endorsement by Frank Stewart, a local businessman who opposed the city’s controversial removal of Confederate monuments (Vanacore 2017).
Taken together, the narratives of the campaigns and their behavior in self-presentation and endorsements all suggest that the candidates strategically positioned themselves in the race in the position where they thought that they would receive the most votes, but do so in the context of the other candidates in the race. In evaluating the candidates’ behavior, we find support for Hypothesis 1. It is also worth noting that we found very little evidence of microtargeting (Holman, Schneider, and Pondel 2015) based on income, race, or gender of the voter. The only circumstance where we see any clear form of attempted microtargeting is that Charbonnet tried to activate racism by distributing a mailer in Lakeview, a largely White, middle-class area of the city, that noted that Cantrell was “straight out of Compton.” 6
Voters’ Ideological Projections
Did voters respond to these ideological signals? Recall that we anticipated that voters would ideologically distinguish the candidates by placing room between the candidates, such that Cantrell would be seen as very liberal, Charbonnet as a moderate, and Bagneris as conservative, above and beyond the candidates’ positions on issues.
In Figure 1, the left panel presents voters’ ideological placement of Cantrell (blue dash-dot line), Charbonnet (green dot), and Bagneris (red solid) and self-placement (black dash) of ideology in the first wave of the survey. As shown in the figure, voters distinguish between them, seeing Cantrell as the most liberal of the candidates, Charbonnet as the moderate, and Bagneris as the most conservative. 7 The ideological placement of Cantrell as firmly liberal is counter to many of her own mainstream policies; as one local newspaper noted, “The vast majority of LaToya Cantrell’s economic policy proposals fall into the moderate mainstream, hardly constituting the Bernie Sanders Socialism” (Tidmore 2017). Similarly, although Bagneris was more conservative than Cantrell (and maybe more so Charbonnet, although some locals and pundits would disagree), he still focused on mainstream issues like infrastructure and crime. By the runoff, presented in the right panel of Figure 1, voters see Cantrell as slightly more liberal (p < .10) and Charbonnet as slightly more conservative (p < .10), suggesting that initial ideological placements are sticky and hard for candidates throughout the course of the election. Thus, we find evidence in support of Hypothesis 2, or that voters will respond to candidate cues and place the candidates ideologically, creating space between them on the ideological spectrum. At the same time, we find less support for Hypothesis 3, or that voters create more space between the candidates between the general and runoff elections.

Ideological placement of candidates and respondent.
What are the consequences for voters who are unable to either place the candidates or differentiate between the candidates ideologically? If ideological placement facilitates voter choices (Ondercin and Fulton 2019), then we would expect that political decisions will be more costly for those voters who cannot differentiate between the candidates ideologically. We look at those voters who see no ideological distance between any pair of candidates in the race. As dependent variables, we examine the voters who say that they will not turn out to vote and those who had not yet decided two weeks prior to the elections. Simplified results of these models are presented in Table 2.
No Ideological Distance.
Note. No distance between candidates means a value of zero when the ideology of the liberal equals the moderate candidate or the moderate candidate’s ideology equals conservative candidate or liberal ideology equals the conservative candidate’s ideology; the results are from logistic regression models. Standard errors are given in parentheses. Controls include race, gender, income, age, home ownership, and news consumption. Full results with controls are available in Supplemental Appendix.
p < .05. **p < .01.
We find evidence in support of our hypothesis in three of the four models: Voters who see no distance between Cantrell, Charbonnet, or Bagneris in the general election are more likely to report that they will not vote in the election or that they had not chosen a preferred candidate in the weeks leading up to the election. By the runoff, voters who could not distinguish between Cantrell and Charbonnet ideologically were not less likely to say that they will turn out, but the ideological indistinction is associated with a slower decision in the race, suggesting general support for Hypothesis 4. Although research suggests that those who hold racist or sexist views might be reluctant to participate in an election that features all African-American candidates and two women in the runoff (Cassese and Barnes 2019; Cassese and Holman 2019; Piston 2010), we see no racial differences in who will not vote in the election or has not chosen a candidate yet. We do find that women took a longer time to decide in the runoff election, providing more evidence for a very complex gender affinity effect in low-information elections (Badas and Stauffer 2019; Cargile and Pringle 2019).
Ideological Distance and Vote Choice
Next, we examine whether ideological positioning translates into vote choices. We constructed the ideological distance measures discussed in the “Data and Methods” section for Cantrell, Charbonnet, and Bagneris and, then, for Cantrell and Charbonnet in the runoff. We estimate vote choice for each candidate separately, 8 controlling for the relative distance and a variety of demographic, socioeconomic, and political factors. Table 3 presents the coefficients, with substantive effects presented in figures following the table. A negative value for the L–M relative distance means that the liberal candidate (Cantrell) is relatively closer to the voter’s ideological position than the moderate (Charbonnet) candidate, whereas a negative value for the M–C relative distance means that the moderate candidate is closer to the voter than the conservative candidate (Bagneris).
Relative Distance in the General Election.
Note. Rel. Prox. L–M = distance between (voter’s ideology–liberal candidate ideology) and (voter’s ideology–moderate candidate ideology); Rel. Prox. M–C = distance between (voter’s ideology–moderate candidate ideology) and (voter’s ideology–conservative candidate ideology); Rel. Prox. L–C = distance between (voter’s ideology–liberal candidate ideology) and (voter’s ideology–conservative candidate ideology); the results were from logistic regression models. Standard errors are given in parentheses. Controls include race, gender, income, age, home ownership, and news consumption.
p < .05. **p < .01.
As shown in Table 3, we find evidence for most of our expectations for the liberal, moderate, and conservative candidates. To refresh, we first expected that the relative distance between the liberal and moderate candidates would matter for votes for the liberal and moderate candidates, but not for the conservative candidate (Hypothesis 5a). Looking at the first line of results in Table 3, we see a significant (negative) coefficient for Cantrell and Charbonnet, but no effect for Bagneris. In short, liberal voters evaluate the distance between themselves and the most similar candidates and do not use the distance to the more conservative candidate in their vote calculus.
We expected that the relative distance rule would apply for conservative voters: that the distance between the moderate and conservative candidates would shape votes for the moderate and conservative candidates, but not the liberal candidate (Hypothesis 5b). The second line of results shows some support for this hypothesis: the relative distance for voters between the moderate and conservative candidates is associated with votes for Charbonnet but not Bagneris. And finally, as we expected in Hypothesis 5c, the distance between the most liberal and most conservative candidates would not shape the vote.
The evaluation during the runoff is easier to calculate and assess, as the candidate pool shrinks to only liberal and moderate candidates (see Table 4). Again, we find evidence that the relative distance shapes vote choice in both of the models as the relative distance between the candidates as assessed during the runoff election shapes votes for Cantrell and Charbonnet.
Relative Distance and Vote Choice, Runoff.
Note. Rel. Prox. L–C = distance between (voter’s ideology–liberal candidate ideology) and (voter’s ideology–conservative candidate ideology); the results were from logistic regression models. Standard errors are given in parentheses. Controls include race, gender, income, age, home ownership, and news consumption.
p < .05. ***p < .001.
We then evaluate the substantive effect of relative distance on votes for Cantrell and Charbonnet in the general and runoff elections to test Hypothesis 6; results are presented in Figure 2. The effect of the relative distance grows between the general (solid line) and runoff (dashed line) elections. We argue that this is because voters have a simpler task of placing candidates ideologically and orienting themselves in comparison in the runoff election (with two candidates) compared with the general election (with 17 candidates). The stronger effect (as evidenced by the steeper line of the effect) of relative distance in the runoff election (as compared with the general election) suggests support for our final hypothesis, that voters will have an easier time assessing ideological distance in an election with fewer candidates.

Liberal–moderate relative distance for Cantrell and Charbonnet votes.
Cueless in New Orleans: Candidate and Voter Ideology in Low-Information Elections
Information shortcuts, like the party, gender, and race of the candidate, allow “voters to emulate the behavior of relatively well-informed voters” (Lupia 1994, p. 63; Ondercin and Fulton 2019). These cues can allow voters to estimate that their decisions are in line with their interests, even absent more complete information (Bauer 2015b; Conover and Feldman 1982; Crowder-Meyer, Gadarian, and Trounstine 2020). In this article, we ask how do voters make decisions in an environment devoid of these information shortcuts.
We posit that candidates will act strategically to send ideological signals to voters. Voters will respond by placing the candidates on an ideological spectrum and by separating the candidates to create ideological space. Those voters who cannot differentiate between the candidates’ ideologies have a more difficult time deciding in the election (and may not vote at all). The relative distance between the voter’s ideology and the candidates’ ideologies should shape vote choice. We find support for these hypotheses, suggesting the importance of ideology in understanding local elections.
We combine a two-wave panel survey of New Orleans residents conducted during the electoral cycle with an analysis of the campaign content (including campaign mailers, online presence, and endorsements) to examine the degree to which campaigns seek to establish themselves on an ideological spectrum and whether voters respond to these cues. We take advantage of the two-stage nature of the political campaign in New Orleans, where an open primary (featuring 17 candidates) preceded a runoff election featuring the two top vote-getters, who were both African-American Democratic women with extensive political experience.
Our results contribute to scholarship on gender in local politics (Bauer 2020; Bernhard and Anzia 2019; Crowder-Meyer 2013; Holman 2017), the importance of understanding voting in low-information contexts (Crowder-Meyer, Gadarian, and Trounstine 2020; de Benedictis-Kessner 2018a; Kirkland and Coppock 2018), and the need to evaluate local political factors when evaluating election returns (Benjamin and Miller 2019; Burns and Thomas 2006, 2015). In doing so, we invoke the “uniqueness of cities” (Trounstine 2009, p. 614) to speak to understanding when campaigns matter, how gender and race continue to shape voter behavior even if they are not useful as voting shortcuts, and how voters behave in ordinary elections. Finally, we also add to the scholarship on how ideology works in low-information environments where partisanship is unavailable as an easy cue to determine ideological placement of candidates. In a context where “everyday is election day in America” (Sumner, Farris, and Holman 2019), understanding how voters behave in these elections helps illuminate how voters behave in the majority of elections.
Supplemental Material
Appendix – Supplemental material for Are You Picking Up What I Am Laying Down?: Ideology in Low-Information Elections
Supplemental material, Appendix for Are You Picking Up What I Am Laying Down?: Ideology in Low-Information Elections by Mirya R. Holman and J. Celeste Lay in Urban Affairs Review
Footnotes
Acknowledgements
Thanks to Jessica Trounstine, Melody Crowder-Meyer, and Tiffany Barnes for their comments on the paper, Janna Mangasep for her research assistance, and the students, alumni, and friends in New Orleans who collected and forwarded campaign flyers and materials to us.
Authors’ Note
This paper has previously been presented at the 2018 American Political Science Association meeting, Boston, MA and the 2019 Women in Subnational Politics workshop at Tulane University.
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: Funding for this study was provided by the Levin-Bernick Faculty Grants, the Tulane Faculty Council on Research, the Newcomb Institute at Tulane University, and the New Orleans Center for the Gulf South.
Supplemental Material
Supplemental material for this article is available online.
Notes
Author Biographies
References
Supplementary Material
Please find the following supplemental material available below.
For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.
For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.
