Introduction
There are approximately 90,000 local governments in the United States, for which citizens elect hundreds of thousands of public officials each year (Warshaw Reference Warshaw2019). Yet most studies of candidate preferences focus on the national level.Footnote 1 This is problematic for our understanding of elections and candidate choice because local electoral contests differ from national elections in important ways (Oliver and Ha Reference Oliver and Ha2007). Local elections are characterized by nonpartisan races (for example, 77% of city council races use non-partisan ballots (MacManus and Bullock III Reference MacManus and Bullock2003)), as well as lower information and lower attention on the part of both voters and the media. Our understanding of candidate choice and the tradeoffs that voters make from national contexts may therefore provide limited insight into local electoral contests. This study, therefore, addresses the following question: Which candidate attributes do voters value most in local political contests?
To explore how individuals evaluate different candidate characteristics in local elections, we use a conjoint experiment embedded in the 2022 Cooperative Election Study, which asks 1,308 respondents to choose between two candidates in five hypothetical nonpartisan local elections. We include candidate attributes previously explored in the literature – name, age, career, prior political experience, family, and endorsements – as well as two additional attributes that may be especially important in local elections: homeownership and the length of time the candidate has lived in the community. This allows us not only to compare our results on local elections to prior findings about key attributes in state and national contests but also to contribute new knowledge on the relative importance of “community embeddedness,” or a candidate’s roots and ties to a community.
We find that several candidate attributes are as important in local elections as they are in national and state elections. Voters are more likely to prefer younger candidates, those with previous political experience, business owners, and candidates with families. In addition, we find that attributes signaling “community embeddedness” are particularly attractive to voters. Our respondents are 4.0 percentage points more likely to vote for a homeowner than a renter, and 7.6 percentage points more likely to vote for someone who has lived in the community for a decade (compared to the base category of 2 years). While Democrats and Republicans are split on the importance of homeownership (greater Republican preference for homeownership), there is bipartisan consensus on the value of being embedded in a community for a longer period of time. Moreover, the magnitude of the residency duration effect is on par with the prior political experience effect, an attribute shown to be considerably important in prior work. Overall, the results imply that voters prefer their local elected officials to have strong local roots, and future work should explore the mechanisms through which community embeddedness matters to voters.
Candidate attributes and voter preferences
In an effort to combat the corruption and inefficiencies produced by party patronage that characterized many cities during the early part of the 20th century, progressive reformers championed reforms that included the secret ballot, direct primaries, and the nonpartisan ballot. While the former – secret ballots and direct primaries – have been almost universally adopted by cities, the nonpartisan ballot characterizes only about 75 percent of municipal elections and roughly one-half of all elections in the United States (see discussion in Wright (Reference Wright2008)). Without party affiliation as a low-cost information cue, voters must turn to whatever information they have or can infer from the ballot.
Heuristics in non-partisan elections
One such source of information is incumbency status or prior political experience. Those who do show up to vote in local contests often rely on incumbency (Schaffner, Streb and Wright Reference Schaffner, Streb and Wright2001; Squire and Smith Reference Squire and Smith1988), especially as this information is frequently indicated directly on the ballot. Both incumbency and a candidate’s political history can signal job experience or that a candidate is higher quality after successfully defeating challengers in a prior election. Prior work has documented incumbency advantages for mayors and city council members (Trounstine Reference Trounstine2011; Ferreira and Gyourko Reference Ferreira and Gyourko2014) and has shown that individuals use information about candidates’ political experience when party labels are absent (Kirkland and Coppock Reference Kirkland and Coppock2018).
Voters also draw on candidate demographics, including race and ethnicity (Pomper Reference Pomper1966), gender (Matson and Fine Reference Matson and Fine2006), and age (Eshima and Smith Reference Eshima and Smith2022). While these characteristics are not indicated directly on the ballot, voters may be able to infer some of these characteristics from candidates’ names. Individuals may even interpret race and gender as party cues in non-partisan elections (Huddy and Terkildsen Reference Huddy and Terkildsen1993; McDermott Reference McDermott1998), and there is evidence that voters prefer to elect women to “stereotype-congruent” positions like school boards (Anzia and Bernhard Reference Anzia and Bernhard2022).
Furthermore, voters use information that they gather about candidates from campaigns, endorsements, and media coverage prior to heading to the ballot box. Profession, career history, and private sector experience provide valuable cues (Kirkland and Coppock Reference Kirkland and Coppock2018; Schaffner, Streb and Wright Reference Schaffner, Streb and Wright2001; Lim and Snyder Jr Reference Lim and Snyder2015).Footnote 2 Voters value candidate qualifications, relevant training, and functional competence for office and use cues in the form of candidate occupation to assess who is or is not fit for the job. For example, Atkeson and Hamel (Reference Atkeson and Hamel2020) find that voters prefer candidates with careers in education for positions on local school boards. In general, voters also tend to favor business owners and executives for the position of mayor. Kirkland (Reference Kirkland2021) finds that business owners and executives “make up the largest occupational category among US mayors – both over time and across regions of the country.” Republican voters especially prefer candidates with job experience (Kirkland and Coppock Reference Kirkland and Coppock2018), business experience in particular (Adams, Lascher Jr and Martin Reference Adams, Lascher and Martin2021).
Endorsements are another effective way for voters to overcome informational deficits. Endorsements from interest groups (Lupia Reference Lupia1994; Gerber and Phillips Reference Gerber and Phillips2003), co-ethnics (Benjamin Reference Benjamin2017), and newspapers (Ansolabehere, Lessem and Snyder Jr Reference Ansolabehere, Lessem and Snyder2006; Lieske Reference Lieske1989) all seem to influence voter preferences. For example, McDermott (Reference McDermott2006) finds that endorsements from groups with a shared common interest – such as unions and union members – effectively improve ideologically and policy-aligned voting. Similarly, Arceneaux and Kolodny (Reference Arceneaux and Kolodny2009) find that endorsements can help the least informed make decisions in a relatively low information real-world setting.
Cultural stereotypes surrounding marriage and children also play an important role in shaping perceptions of candidates. Teele, Kalla and Rosenbluth (Reference Teele, Kalla and Rosenbluth2018) find that voters and elites prefer candidates who are both married and have children. Moreover, candidates who are perceived as going against these traditional stereotypes are penalized. A large literature finds that these penalties are concentrated particularly among women as motherhood becomes more politicized (Deason, Greenlee and Langner Reference Deason, Greenlee and Langner2015), uneven child-rearing responsibilities persist (Iversen and Rosenbluth Reference Iversen and Rosenbluth2006), and women pursuing leadership roles are seen as too ambitious (Dittmar Reference Dittmar2015; Jamieson Reference Jamieson1995).
Community embeddedness: Homeownership and residency duration
Despite featuring prominently in campaign ads, less is known about how voters use information about candidates’ community embeddedness – their roots and ties to the community expressed through attributes such as homeownership and residency duration – to make vote choices. We know that homeowners are significantly overrepresented among public officeholders at all levels of government (Einstein, Ornstein and Palmer Reference Einstein, Ornstein and Palmer2022), but despite an extensive literature on how homeownership affects turnout and vote choice (Fischel Reference Fischel2002; Hall and Yoder Reference Hall and Yoder2022; Einstein, Glick and Palmer Reference Einstein, Glick and Palmer2020; Einstein, Palmer and Glick Reference Einstein, Palmer and Glick2019; Oliver and Ha Reference Oliver and Ha2007; Hankinson Reference Hankinson2018), we know little about whether this overrepresentation of homeowners is driven by candidate self-selection or voter preferences.
Homeownership and residency duration may impact voter preferences in local elections through three distinct mechanisms: as a signal of the candidate’s investment in the community, as a marker of a shared place-based identity, and through the web of personal relationships that candidates cultivate with their friends and neighbors.
First, these attributes may signal candidates’ investment of time and money into the community. Voters may prefer candidates with a stake in the long-term success of the community, and investments provide personal incentives to produce quality policies for the benefit of the community. Voters with large financial investments in a community tend to be more highly engaged in local politics (Fischel Reference Fischel2002); the same may be true for public officials as well.
Second, homeownership and residency duration may signal that a candidate shares the voters’ identity and policy preferences. Since voters themselves are more likely to be homeowners with strong ties to the community, they may prefer candidates with similar experiences and in-depth knowledge of concerns in the community (Mansbridge Reference Mansbridge1999). Community embeddedness involves symbolic or place-based identity (Munis Reference Munis2021), where voters reward candidates for living near them and identifying as part of the group. Work by Schulte-Cloos and Bauer (Reference Schulte-Cloos and Bauer2023) argue that “the local roots of political candidates act as social identity cues to voters” (pg. 695). Place-based social identity instills in voters confidence that their neighbors will represent them effectively, whether through substantive or descriptive representation (Campbell et al. Reference Campbell, Cowley, Vivyan and Wagner2019; Meredith Reference Meredith2013).
Finally, voters may prefer candidates with whom they have a personal relationship (Sinclair Reference Sinclair2012; Panagopoulos, Leighley and Hamel Reference Panagopoulos, Leighley and Hamel2017), and candidates embedded in their communities are able to form deep personal networks over many years. Long-term residents have greater name recognition and a deeper understanding of the community and its voters’ priorities (Hunt Reference Hunt2022). Furthermore, personal relationships play a significant role in politics, perhaps even more so in local contests than in statewide or congressional races due to the general lack of salient policy information. V. O. Key’s concept of friends and neighbor voting is based on candidates winning support not due to their political positions but because they reside near voters (Key Reference Key1949). Voters tend to “back the home-town boy” (Key Reference Key1949, p. 41).
For these reasons, we expect that voters will be more likely to select candidates who are homeowners, and those who have lived in the community for a longer period of time, over renters and newer residents. A key feature of the conjoint design used in this study is that we are able to estimate the average marginal effect of these attributes on vote share and compare the magnitude of these estimates with those of other candidate attributes with documented importance for voters’ preferences. Although this design cannot definitively identify why voters prefer candidates with these attributes, it can provide some suggestive evidence distinguishing the three mechanisms outlined above. Because our respondents are asked to evaluate fictional candidates, we can isolate the effect of community embeddedness from the effect of personal relationships. And by comparing preferences across different groups of respondents, we can assess whether our respondents are more likely to prefer candidates who share their attributes. For example, if homeowners and renters diverge on whether they prefer homeowner candidates, this would provide evidence in favor of the shared identity mechanism. But if all voters prefer homeowner candidates, it would provide suggestive evidence in favor of the investment mechanism.
Experimental design
We surveyed 1,308 respondents from the post-election survey module of the 2022 Cooperative Election Study (CES).Footnote 3 Of these respondents, 39% were men, 71% White, 12% Black, and 8% Hispanic. 27% live in urban areas, 40% live in suburbs, and 60% reported owning their own home. Our survey instrument is provided in Appendix Section A.1.
The survey included a conjoint choice task to assess the impact of candidate attributes on respondents’ preferences (Hainmueller, Hopkins and Yamamoto Reference Hainmueller, Hopkins and Yamamoto2014). Conjoint designs have been used to study candidate preferences in a variety of contexts (Carlson Reference Carlson2015; Franchino and Zucchini Reference Franchino and Zucchini2015; Horiuchi, Smith and Yamamoto Reference Horiuchi, Smith and Yamamoto2020; Carnes and Lupu Reference Carnes and Lupu2016; Kirkland and Coppock Reference Kirkland and Coppock2018; Sung Reference Sung2022). In our survey, each respondent completed five pairwise comparisons between hypothetical candidates, like the example in Figure 1. Seven attributes were provided for each candidate, drawn uniformly from the distributions in Table 1 with no restrictions on combinations. Footnote 4,Footnote 5 The bolded attribute – Community Ties – represents our “community embeddedness” variables of homeownership and residency duration, randomized independently. Note that homeownership is not a perfect measure of community embeddedness and may suffer from an information equivalence problem (Dafoe, Zhang and Caughey Reference Dafoe, Zhang and Caughey2018) in that it can signal other information like socioeconomic status, race, or that a candidate “has their act together,” especially when combined with other attributes in the conjoint experiment. We expect that this is less of a concern for residency duration.
Figure 1. Example conjoint choice.
Table 1. Conjoint task attributes and levels
Our analyses follow our pre-registration.Footnote 6 For each attribute, we estimate the average marginal component effect (AMCE), which in a forced-choice conjoint experiment has a straightforward, politically meaningful interpretation: it estimates the marginal effect of an attribute on a candidate’s vote share (Bansak et al. Reference Bansak, Hainmueller, Hopkins and Yamamoto2022). We also estimate conditional AMCEs by respondent homeownership and political party identification, discussed below, and by other respondent demographics (gender, race, and urban/suburban/rural place of residence), reported in Appendix Section A.3.
Results
Standard candidate attributes
Figure 2 displays the estimated AMCEs and 95% confidence intervals for each level of the candidate attributes. Many of the results are consistent with our expectations and prior research, with a few notable exceptions. Consistent with Kirkland and Coppock (Reference Kirkland and Coppock2018), we find that voters are more likely to prefer candidates with previous political experience: respondents were 7.0 percentage points more likely to choose candidates that had previously been elected to political office than those with no prior political experience, all else equal. Respondents were also somewhat more likely to prefer younger candidates (4.3 percentage points less likely to choose a 60-year-old candidate compared to the base category of 30, holding all else constant). Moreover, respondents preferred candidates who are married with children (+6.3 percentage points compared to single candidates), consistent with other candidate preference studies.
In line with previous research, we find that our respondents prefer candidates who are business owners more than any other career we included in the survey. Compared to the base category of business owner, respondents were less likely to choose unemployed candidates (−12.5 percentage points) and police officers (−5.1 percentage points), though we will see momentarily that there are large partisan differences in AMCE for these attributes. Notably, our respondents’ preference for business owners does not extend to real estate developers, (−12.3pp compared to a generic business owner). Antipathy towards real estate developers – particularly among liberals (Manville Reference Manville2021) – is an interesting recent development in US local politics (Monkkonen and Manville Reference Monkkonen and Manville2019), and we were surprised to find that it was one of the strongest estimated effects from our survey experiment, regardless of respondent’s political party, homeownership status, and other demographic characteristics. These negative perceptions of real estate developers persist when looking at endorsements as well. Compared to no endorsement, respondents reacted negatively to an endorsement by an association of real estate developers (−2.9pp), but positively to an endorsement by a teachers union (+3.7pp), chamber of commerce (+4.7pp), or local newspaper (+3.1pp). In the aggregate, we find no significant impact of a police union endorsement on hypothetical vote choice (although we will again see partisan differences below).
Finally, we find small and statistically insignificant differences in choices based on candidate race and gender. Though this could be the result of a weak signal (we did not explicitly list candidate race and gender, but included race and gender cues in the candidates’ names), subgroup differences suggest that respondents are correctly identifying candidate race. For example, Black respondents are 7.6pp more likely to select Black candidates than White respondents (Figure A15). These results are broadly consistent with those from other conjoint experiments on racial discrimination (Butler and Homola Reference Butler and Homola2017). In all, we find little evidence that the average respondent is discriminating on the basis of race or gender when selecting candidates.
Community embeddedness
For our novel “community embeddedness” attributes – homeownership and residency duration – we find large and statistically significant impacts. All else being equal, respondents were 4.0 percentage points more likely to choose a homeowner over a renter, and 7.7 percentage points more likely to choose a candidate who had lived in the community for a decade (compared to the base category of 2 years). Strikingly, the magnitude of the residency duration effect rivals that of prior political experience, and the effect of homeownership is similar in magnitude to an endorsement from a local newspaper. In fact, the magnitude of the residency duration effect was the third largest of all of the candidate attributes that we examined (exceeded only by our respondents’ distaste for real estate developers and the unemployed). This suggests that community embeddedness is quite important to voters in local elections, both in its own right and in comparison to other candidate features.
To examine the proposed investment and shared identity mechanisms, we estimate conditional AMCEs for homeowners and renters separately, as shown in Figure 3. These results provide suggestive evidence in favor of the shared identity mechanism – homeowner respondents prefer homeowner candidates, but respondents who are renters do not share this preference. It is notable, however, that renters do not appear to prefer renter candidates over homeowner candidates.Footnote 7
Figure 2. Average marginal component effects (AMCEs) and 95% confidence intervals by candidate attribute.
Notes: The figure displays estimated AMCEs with 95% confidence intervals. The reference categories for each attribute are endorsement – no endorsements, family – single, political experience – no previous experience, age – 30, residency duration – 2 years, homeownership – renter, career history – business owner, race – Black, gender – Female.
Figure 3. Conditional effects by homeownership.
Notes: The figure displays estimated average marginal component effects (AMCEs) with 95% confidence intervals by respondent homeownership status.
When considering the residency duration attribute, we instead see suggestive evidence in favor of the investment mechanism – respondents of every kind prefer candidates who have lived in their community for a longer period of time, one of the few estimated effects that holds regardless of the respondent’s political party, demographics, or homeownership status. The estimated effect is somewhat smaller among respondents who reported living in their community for less than one year (see Figures A4 and A17 in the Supplementary Materials), but these differences are not statistically distinguishable from zero. We acknowledge that these are imperfect tests of these mechanisms and encourage future work to explore more directly why voters consider candidates who have lived in their community longer to be of higher quality.
4.3 Conditional effects by respondent characteristics
In Figure 4, we estimate conditional AMCEs by the political party of the survey respondent. We find a few substantial differences when comparing Democratic and Republican respondents. For example, the relatively small average treatment effects for teachers’ union and police union endorsements in Figure A5 mask much larger conditional treatment effects by the party. All else equal, Democratic respondents are 16 percentage points more likely to choose a candidate that has been endorsed by the teachers union (compared to no endorsement), while Republican respondents are 11.7 percentage points less likely. The reverse is true for endorsements by police unions: Republicans are 12.8 percentage points more likely to choose a police union-endorsed candidate (compared to no endorsement), while Democrats are 8.5 percentage points less likely. Similarly, Democrats are 10.9 percentage points less likely to choose a candidate who is a police officer (compared to a business owner).
Figure 4. Conditional effects by political party.
Notes: The figure displays estimated average marginal component effects (AMCEs) with 95% confidence intervals by political party ID of the respondent.
Looking at the community embeddedness attributes, we find a strong difference between Democratic and Republican respondents in their attitudes toward homeowner candidates. All else equal, Republicans are 9.6 percentage points more likely to choose a homeowner, while Democrats are just as likely to choose a renter candidate. In contrast, there are no partisan difference in preferences to residency duration. Both Democrats and Republicans are more likely to choose a candidate who has lived longer in the community.
In Appendix Section A.3, we present conditional AMCEs and marginal means by respondent gender, race, place of residence, and residency duration. Though there is broad consensus among our respondents about most of these candidate attributes, there are a few notable subgroup differences. We find that women respondents have a stronger preference for both prior political experience and longer residency duration, and rural respondents are the most strongly opposed to candidates with real estate development backgrounds.
5 Discussion and conclusion
Millions of voters cast ballots for local government representatives every year. These elections occur in highly varied institutional, demographic, and political contexts and have important consequences for the day-to-day lives of residents. In this study, we explored how individuals evaluate different candidate characteristics in local elections. Beyond previously studied demographic characteristics, career history, and endorsements, we examined the extent to which community embeddedness leads to greater support for local candidates.
We find that voters hold strong preferences for attributes that signal community embeddedness: living in the community for an extended period of time, and to some extent, owning a home in the community. We argue that this in part reflects a desire for descriptive representation – homeowners prefer homeowner candidates while renters do not – but it also reflects a desire for candidates who have invested significant amounts of time into their community – a preference that holds across all respondent subgroups.
We emphasize that this study cannot definitively explain why voters prefer candidates with these attributes. Like other studies in this area (Manville and Monkkonen Reference Manville and Monkkonen2024), our survey uses proxy measures of community embeddedness – residency duration and homeownership – and such measures may convey information about a multitude of candidate attributes. Therefore, future work should develop more extensive, validated measures of community embeddedness, including through employing careful survey, observational, and mixed-methods research. Moreover, we welcome future research on how the treatment effects from signaling community embeddedness differ across jurisdictional types and geographic context. We also encourage consideration of other theoretical mechanisms beyond signals of investment in the community and shared identity. We ultimately need a more comprehensive theory of community embeddedness, social ties, and deep roots in local politics.
In addition to our contribution to understanding the role of community embeddedness at the local level, our results confirm and bring additional nuance to existing insights in the literature on candidate choice. In line with prior findings, our results suggest that voters in local elections prefer candidates with previous political experience and do not favor candidates associated with real estate development. We also observe significant differences between Republicans and Democrats regarding candidate profession and local endorsements, comporting with recent research that finds attitudes toward the police vary by political ideology (Navarro and Hansen Reference Navarro and Hansen2023). Our work is similar – in spirit – to that of Hunt (Reference Hunt2022) and Hunt and Rouse (Reference Hunt and Rouse2023) who find that candidates with deep local roots have electoral advantages in Congress and state legislatures. We see our work as extending this argument, finding electoral advantages in more local contests as well. Altogether, the findings suggest that voters evaluate candidates in local elections similarly to those in national and state elections, with additional emphasis also placed on local roots and community ties.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/XPS.2024.16.
Data availability
Support for the 2022 Cooperative Election Study (CES) was provided by the National Science Foundation (Award no. 2148907). The data, code, and any additional materials required to replicate all analyses in this article are available at the Journal of Experimental Political Science Dataverse within the Harvard Dataverse Network, at https://doi.org/10.7910/DVN/PPMGWL (Ornstein et al. Reference Ornstein, Heideman, Moy and Jackson Schiff2024).
Competing interests
The authors have no conflicts of interest to declare.
Ethics statement
This study was approved with exempt status by the IRB at the University of Georgia (#00006053). This study adheres to APSA’s Principles and Guidance for Human Subjects Research.
Open access
