Black Voters, Safe Districts, and Primary Elections

How do Black voters make choices in majority-Black congressional districts? One of the most well-known factors driving Black vote choice is partisanship. Black voters support the Democratic party at very high rates and have done so for decades (Carmines and Stimson Reference Carmines and Stimson1989). Evidence also shows that this party attachment is stronger than that of other racial groups (Dun and Jessee Reference Dun and Jessee2020). There is even a large body of literature investigating why Black conservatives support the Democratic party, despite their economic interests (Dawson Reference Dawson1994) and ideological preferences (Philpot Reference Philpot2017) being out of line with the party. Strong partisanship also allows strategic voting based on candidate electability (Simas Reference Simas2017). Since Black voters are such strong partisans, electing a Democratic candidate becomes the most important factor driving their vote choice in the primary election. Rather than voting for their sincere preference,Footnote ¹ many primary voters support the most “electable” candidate or the candidate with the best shot of beating the Republican candidate in the general election (Abramowitz Reference Abramowitz1989). Recent research shows that Black voters rely heavily on a candidate’s electability to choose candidates (e.g., Smith Reference Smith2022).

Partisanship explains Black vote choice in many electoral settings, but the growing number of safe congressional districts reduces its electoral significance. Safe congressional districts are electorally secure for one of the two major parties (Abramowitz, Alexander, and Gunning Reference Abramowitz, Alexander and Gunning2006). There is no competition in the general election because one party has far more political power than another. Scholars investigating this trend point to factors like gerrymandering (McDonald Reference McDonald2007) or the incumbency advantage (Abramowitz, Alexander, and Gunning Reference Abramowitz, Alexander and Gunning2006). Regardless of the exact cause, the lack of congressional competition is troubling because it can influence government responsiveness (Griffin Reference Griffin2006) or political participation (Stauffer and Fraga Reference Stauffer and Fraga2022), among other factors. This trend is most relevant for voters in majority-Black congressional districts because all are safe for the Democratic party.

To illustrate this, I calculated each district’s level of partisanship using the normalized presidential vote (e.g. Abramowitz, Alexander, and Gunning Reference Abramowitz, Alexander and Gunning2006; Stauffer and Fraga Reference Stauffer and Fraga2022). This measure is calculated by subtracting the Democratic party’s national presidential vote share from the Democratic candidate’s vote share in the district. Districts are considered safe if the district-level presidential vote share is ten points more than the national vote. This approach is ideal because it provides a measure of district-level partisanship that can be compared across districts and is independent of individual congressional election results (Abramowitz, Alexander, and Gunning Reference Abramowitz, Alexander and Gunning2006). I collected partisan and racial demographic data for every congressional district between 1992 and 2020. Appendix A describes the results in detail.Footnote ² The main findings show that districts with large Black populations are almost always safe. Between 2012 and 2020, one hundred percent of majority-Black districts were safe for the Democratic party, suggesting that Black voters in majority-Black districts never experience competitive general elections.

Since these districts are safe, the Democratic party has electoral control, and no Republican could win the general election. This means whoever wins the Democratic primary election will serve in Congress. Since all the candidates running in the Democratic primary are Democratic candidates, their partisanship cannot be used to differentiate between them. In addition, candidate electability, which applies specifically to vote choice in primary elections, falls short because there is no meaningful two-stage electoral process. Since the winner of the general election is effectively predetermined to be the Democratic candidate, Black voters do not need to consider who is the most electable because any Democratic candidate would win. This fact casts doubt on the role of partisanship and places great importance on the winner of the primary election.

A related body of literature uses group-based explanations to explain Black vote choice (Verba and Nie Reference Verba and Nie1972; Dawson Reference Dawson1994; White and Laird Reference White and Laird2020; Wamble Reference Wamble2025). Early work on Black political participation points to racial group attachments to explain high levels of voter turnout (Verba and Nie Reference Verba and Nie1972, 157). The literature is full of examples highlighting the role of group connectedness, like the role of descriptive representation in vote choice (e.g., Dovi Reference Dovi2002; Mansbridge Reference Mansbridge1999). Dawson (Reference Dawson1994) states that “as long as African Americans’ life chances are powerfully shaped by race, it is efficient for individual African Americans to use their perceptions of the interests of African Americans as a group as a proxy for their own interests” (61).

White and Laird (Reference White and Laird2020) argue that the Black community enforces support for the Democratic party, and other important norms, by using social sanctioning and positive and negative incentives. Similarly, Anoll (Reference Anoll2022) also highlights the role of group membership when discussing the “Racialized Norm Model (RNM)” (43). Adherence to two norms, “honoring ancestors” and “helping hands,” explains Black political behavior. Her interview data signifiy that Black respondents are most likely to link these norms to politics. This effect is most significant when the individual is embedded in Black spaces. Most recently, Wamble (Reference Wamble2025) points to the role of community commitment and finds that candidates who can signal that they sacrificed for the Black community will likely get large support from said community. Group-based explanations garner much attention in the literature and provide important insight into Black vote choice.

Despite the rich information provided by previous group-based explanations, two factors complicate their application to majority-Black congressional districts. First, Black voters cannot use race to distinguish between candidates. As the racial makeup of a district increases, the number of candidates from that racial group will also increase (Branton Reference Branton2009; Fraga Reference Fraga2018; Fraga et al., Reference Fraga, Shah and Gonzalez Juenke2020). Therefore, when almost all the candidates who emerge are of the same race, voters cannot use racial cues to distinguish between candidates. Second, voters in primary elections have limited knowledge about the candidates running. Congressional primary elections are considered low-information because voters have little information about the candidates running (Ahler, Citrin, and Lenz Reference Ahler, Citrin and Lenz2016; Bawn et al. Reference Bawn, DeMora, Dowdle, Hall, Myers, Patterson and Zaller2019; Hirano et al. Reference Hirano, Lenz, Pinkovskiy and Snyder2015). Research shows voters learn about candidates during gubernatorial and U.S. Senate primary elections but not in U.S. House primary elections (Hirano et al. Reference Hirano, Lenz, Pinkovskiy and Snyder2015).

Other scholars find that voters cannot recall important information about primary candidates. One study found that “47% of exit poll respondents could not recall the name of a single House primary candidate, despite having case ballots only minutes before. Only 16% were able to list all the candidates” (Bawn et al. Reference Bawn, DeMora, Dowdle, Hall, Myers, Patterson and Zaller2019). This lack of knowledge complicates many of the major theoretical explanations. When explaining how Black voters utilize heuristics, Dawson (Reference Dawson1994) claims that “determining what political and social policies would provide the most utility for each individual African American … was more cost-effective than the calculation of individual utility” (57). This calculation is straightforward when a Black voter chooses between Republican or Democratic candidates or between Black or White candidates. However, the lack of information in primary elections makes engaging in deeper cognitive reasoning difficult.

This point applies more to White and Laird’s (2020) theory of racialized social constraint, which only works in highly salient settings. When discussing the assumptions of their theory, they point to four necessary conditions: (1) a well-defined norm of Black political behavior, (2) the behavior is public, (3) racially homogeneous social networks, and (4) potential for sanctioning or rewards (White and Laird Reference White and Laird2020, 47). Majority-Black congressional districts create an opportunity for Black people to have racially homogeneous social networks and create a potential for sanctioning, but they do not meet the other conditions. Support for the Democratic party is the norm they reference, but partisan considerations are not relevant in primaries because all candidates are in the same party. There are also no well-defined norms for choosing between Black candidates. Wamble’s (Reference Wamble2025) theory suffers from a similar issue when applied to majority-Black congressional districts. It is difficult for voters to learn much about primary candidates, so I question if voters have the knowledge to determine which candidate sacrificed the most for the Black community.

A body of literature investigates Black vote choice in low-information environments, and scholars point to the importance of electoral cues. Evidence shows that primary elections have significantly lower turnout than general elections (Boatright Reference Boatright2014), less money is contributed and spent (Boatright Reference Boatright2014, 224), and campaigns have smaller ground games (Hughes et al. Reference Hughes, Levitt, Hill and Kousser2017). This means campaigns are less equipped to contact voters and spread their message. Since voters have little factual information about the candidates who emerge, voters look for signals in the environment to help them make choices. Black voters cannot use racial or partisan cues to choose candidates, but there are other cues that they can utilize. For example, Black voters could turn to incumbency (Benjamin and Carr Reference Benjamin and Carr2022), ballot cues (Matson and Fine Reference Matson and Susan Fine2006), endorsements (Benjamin Reference Benjamin2017; Benjamin and Miller Reference Benjamin and Miller2019), polling (Boudreau and McCubbins Reference Boudreau and McCubbins2010), yard signs (Green et al. Reference Green, Krasno, Coppock, Farrer, Lenoir and Zingher2016), and many others. This paper contributes to the literature on Black vote choice by highlighting the role of group consensus.

The Community as Information Sources

Existing studies of Black vote choice already have group consensus components. For example, studies suggest that Black voters use positive and negative incentives to ensure that members of the Black community follow in-group norms (White and Laird Reference White and Laird2020; Wamble et al. Reference Wamble, Laird, McConnaughy and White2022). This shows that the Black community values group consensus and is even willing to take steps to enforce consensus. I agree that group-based explanations can provide important insights. Rather than having other Black people enforce social norms (e.g., White and Laird Reference White and Laird2020), or having individual Black people make rational choices (Dawson Reference Dawson1994), I suggest that Black voters use other community members and Black institutions to determine group consensus. Since congressional primary elections are low-information, Black voters have limited information about the candidates running. In these environments, all voters, including Black voters, will rely heavily on cues (McDermott Reference McDermott1998). However, the most important cues they traditionally use (e.g., partisan, electability, or racial cues) cannot be used in these settings.

Because of their strong group connection, Black voters will rely on in-group consensus cues to choose candidates. Consensus cues are simply signals that provide respondents with information about consensus on a topic, and they can effectively induce a change of opinion (Kruglanski and Webster Reference Kruglanski and Webster1996; Burnstein, Vinokur, and Trope Reference Burnstein, Vinokur and Trope1973). Since members of the Black community exhibit a strong connection to their racial group (Verba and Nie Reference Verba and Nie1972; Dawson Reference Dawson1994; White and Laird Reference White and Laird2020; Wamble Reference Wamble2025), it is reasonable to theorize that they will turn to group consensus cues in the right setting. Consensus cues can come in many forms. Repetition can work as a consensus cue. The more people hear about a particular position, the more likely they are to adjust their attitudes accordingly (Tesser Reference Tesser1978). Therefore, repetitively seeing a candidate’s yard sign or campaign flyer could influence vote choice.

Incumbency is one of the most well-studied examples of a consensus cue used in political science. Existing literature highlights that incumbents have built-in legitimacy, an existing fundraising network, and can garner endorsements from prominent party leaders (Banks and Kiewiet Reference Banks and Roderick Kiewiet1989; Boatright Reference Boatright2013; Reference Boatright2014). Data shows that incumbents virtually never lose, and fears of “getting primaried” by an ideologically extreme challenger are unfounded (Boatright Reference Boatright2014).Footnote ³ Once voters in the district see that most voters support a particular candidate, other voters will fall in line and support the incumbent. What is less clear is how Black voters utilize consensus cues if the incumbent is electorally weak (e.g., political scandal), or no incumbent is present (e.g., open seat elections). Which consensus cues are Black voters using to choose candidates in majority-Black congressional primary elections?

Two other important consensus cues that can significantly affect Black vote choice are elite endorsements and public opinion polling results. Both factors provide information about broad segments of the voting population. Endorsements are a way “that elites—celebrities, community leaders, interest groups—can vouch for the quality of a candidate” (Dominquez Reference Dominguez2011, 535). The individual or institution is putting its reputation on the line to support this candidate publicly. Public opinion data provide the clearest message of who other voters support and significantly improves the perception of the leading candidate (Ansolabehere and Iyengar Reference Ansolabehere and Iyengar1994). We also know that Black voters rely on these cues in low-information environments (Benjamin Reference Benjamin2017; Benjamin and Miller Reference Benjamin and Miller2019; Dominguez Reference Dominguez2011). For example, if a non-Black minority candidate receives an endorsement from a Black institution, Black voters are significantly more likely to vote for the candidate (Benjamin Reference Benjamin2017).

However, endorsements and public opinion polling data are not equally accessible to voters. Endorsements are far more readily available than public opinion data. In most elections, most candidates receive some form of endorsement. For example, Dominguez (Reference Dominguez2011) collected data on candidates running in the 2002 primary elections in forty-three districts and found that each candidate received 14.3 endorsements on average (538). Polling provides the best snapshot of consensus but is rare in congressional primary elections. Conducting a public opinion poll is a costly endeavor that many campaigns do not and cannot engage in during the primary election (Boatright Reference Boatright2014). Third-party firms also only get involved when the elections have national significance.Footnote ⁴ Polling results may not be as accessible to voters as endorsements, but they can have a meaningful effect on vote choice when combined with endorsements.

This paper contributes to this literature by focusing on how Black voters balance partisan and racial consensus cues. Which cue will significantly influence vote choice in majority-Black congressional districts? Will an endorsement from a prominent Black Institution or a national Democratic Institution have a more significant effect on vote choice? In addition, what will the combination of consensus cues play in a candidate’s vote share? Will endorsements and polling together have a significant effect on vote choice?

As mentioned in previous sections, partisanship plays a dominant role in Black vote choice, but because majority-Black districts are safe, many partisan signals are not helpful to voters. However, partisanship can still play a significant role through endorsements. A Democratic institution or individual can put their full support behind a candidate, and partisan endorsements can significantly affect vote choice (Dominquez Reference Dominguez2011). However, when two endorsers are both members of the Democratic party, I expect racial signals will have a significant effect on vote choice. Since Blacks were excluded from the larger political and social world in the United States, they developed parallel institutions to advance Black interests. This includes major activist organizations, like the NAACP, and Black-focused media outlets or newspapers (Jet, etc.), but it also includes common spaces, like churches, barber shops, or hair salons, where people regularly meet. These institutions allow Black people to engage with other Black people about important topics. Harris-Lacewell (Reference Harris-Lacewell2004) refers to these encounters as “everyday talk.” These institutions help solidify a strong group connection and work to spread information relevant to the Black community (Harris-Lacewell Reference Harris-Lacewell2004).

Currently, many of these Black institutions align with the Democratic party. Much of this can be traced to the Civil Rights Movement in the middle of the 20^th century. The Democratic party passed critical legislation and solidified support from the Black community. However, Black institutions take the fight for equality further than the Democratic party. The Democratic party has a much broader issue base, and Black institutions focus on issues relevant to the Black community. This means that Black voters can be confident that an endorsement from a Black institution will signal that the candidate will advocate for the Black community. Therefore, I argue that endorsements from prominent Black institutions will positively affect a candidate’s vote share among Black respondents voting in a majority-Black congressional primary compared to a national Democratic institution.

While polling is less readily available, it should have a significant positive effect, regardless of the racial or partisan endorsement. Reliable polling in majority-Black congressional districts specifically signals Black voters’ preferences. In other words, polling data directly signal consensus in the Black community. When polling numbers are high, Black voters can be confident that large portions of fellow Black voters support a particular candidate. Therefore, I argue that high polling numbers will positively affect a candidate’s vote share among Black respondents voting in a majority-Black congressional primary compared to a candidate with no polling information.

A natural corollary is that bad polling numbers will hurt candidates’ support. While high polling numbers signal mass support, poor polling numbers raise doubts about the candidate’s potential success. This is especially true in a majority-Black congressional district. If a candidate signals that a small number of Black voters are supporting them, then it will signal to other Black voters that they may not be a viable option. Therefore, I argue that negative polling numbers will have a negative effect on a candidate’s support among Black respondents voting in a majority-Black congressional primary when compared to a candidate with no polling information.

While I expect polling and endorsements to have an independent effect, I contend that candidates who can signal multiple consensus cues will have the most support. In particular, the candidate who can signal a racial endorsement and high polling numbers in a majority-Black district should get the highest support. Therefore, I argue that the combination of high polling numbers and a racial endorsement should have the largest effect on a candidate’s vote total.

Data and Methodology

I tested my theory using a mixed methodological approach with quantitative and qualitative methods. Combining both empirical approaches will provide a comprehensive understanding of Black vote choice and allow me to explore the topic more deeply. I use an Explanatory Sequential Design with two phases (Arezina Reference Arezina2023; Driscol et al. Reference Driscoll, Appiah-Yeboah, Salib and Rupert2007). In the first phase, the researcher conducts quantitative analysis. I conducted a randomized online survey experiment in September 2022, a few weeks before the 2022 midterm election. Experiments are excellent because they allow researchers to isolate specific factors and measure their effects. Since respondents are randomly assigned to treatment groups, and each respondent has an equal chance of being in any treatment, we can make causal claims about how viability affects likely support (Gerber and Green Reference Gerber and Green2012). However, one limitation is that we cannot gather in-depth information about why voters decide.

In the second phase, the researcher conducts qualitative research to supplement the initial results. I conducted two online focus groups with Black primary voters in January 2023 to get in-depth accounts directly from voters. Focus groups reveal insight and nuances that other approaches cannot, and they are the ideal way to discover a person’s (or voter’s) hidden feelings and motives. The questions used in this approach are open-ended so the researcher can capture a wide range of responses. Researchers can hear people’s word choices when discussing these topics and uncover common terms that voters use when discussing this topic. This setting allows me to get details about why voters choose the candidate they do. Once I determined important factors, I returned to my experimental analysis to investigate subgroup effects. The sections below detail my quantitative and qualitative approaches in chronological order.

Randomized Experimental Design

To quantitatively test my hypotheses, I conducted a randomized experiment embedded in an online survey. I used the Qualtrics online panel to collect a sample of African American primary voters. While this is a convenience sample, Qualtrics produces more demographically representative samples than other popular online recruitment companies (Boas, Christenson & Glick Reference Boas, Christenson and Glick2020). This fact makes it the best choice to obtain a sample of African American respondents. 1,548 respondents were initially recruited for my survey experiment. Following the informed consent, respondents were screened pretreatment to verify that they were African American, they identified as Democrats, and they were primary voters.Footnote ⁵ Following these screener questions, I was left with 1,046 respondents. I then randomly assigned the remaining respondents to their treatment groups.

I take steps to compare the demographics of my sample with the demographics of national African American primary voters. To accomplish this, I utilize data from AP Vote Cast, who conducted over 130,000 interviews with registered primary voters before the 2020 general election. Their sample covered 17 states, and the interviews were conducted via phone or the web. I looked at three critical demographics: gender, age, and income. Table 1 reports those results. Overall, my sample is similar to the national sample in terms of age and income. The most significant difference can be found with gender. My sample has 14% more men than the national sample. However, results for age and income are very similar. The most significant age variable is among those above fifty years old. My sample has 6% fewer respondents over fifty than the national sample.Footnote ⁶ The results for income are almost identical. This suggests that my sample has some level of generalizability to actual Black primary voters.

Table 1.

Comparing Survey Sample to National Sample

^a The measure for the Age variable is different for my survey and AP Vote Cast. My survey measures Age from 18–34, 30–55, and 55 and above. As Table 1 indicates, AP Vote Cast measures Age from 18–29, 30–49, and 50 and above.

Experimental Results

This section reports the results of my experiment. My first hypothesis suggests that an endorsement from a predominantly Black institution will positively influence vote choice compared to an endorsement from a Democratic institution. I compare the vote choice for the candidate with a DLC endorsement and no polling information, 0.68, to a candidate with an endorsement from the CBC and no polling information, 0.69. This is only a one-point difference, and it is not statistically significant.

Next, I investigate the effect of racial endorsements on the entire sample. In other words, I compare the vote share of the candidates with a DLC endorsement, regardless of polling numbers, to the vote share of the candidates with a CBC endorsement, regardless of polling numbers. Figure 1 reports the results. The candidate with an endorsement from the CBC has a mean level of support of 0.73, and the candidate with an endorsement from the DLC has a mean level of support of 0.72. There is no significant difference between the groups. This finding suggests that an endorsement from a Black institution does not affect a respondent’s level of support when compared to a generic Democratic institution.

Figure 1.

Respondent’s Support by Racialized Endorsement.

The next factor I investigate is the role that polling can have on Black vote choice. Figure 2 below shows a candidate’s level of support with varying polling results, regardless of endorsement type. The results suggest that high polling numbers significantly affect the candidate’s support. The results indicate that informing voters about high polling numbers is associated with almost a ten-point increase in support, and this difference is statistically significant. Respondents in the high-polling treatment group have a mean level of support of 0.78, and respondents in the no polling group have a mean level of support of 0.69. On the other hand, there is no significant difference for those with low polling numbers. The level of support for those in the low-polling group is 0.72, only two points higher than the no-polling group, and the result is not statistically significant. This result suggests that there is no difference between bad polling information and no polling information.

Figure 2.

Respondent’s Support by Level of Polling.

Lastly, I test whether high polling numbers and a racialized endorsement boost support for a candidate. Figure 3 reports the results. The graph on the left reports results for an endorsement from the DLC, and the graph on the right reports results for the CBC. When I compare the vote share for those with a DLC endorsement and no polling information, 0.68, to those with a CBC endorsement and high polling numbers, 0.79, there is an eleven-point difference. However, the results in Figure 3 suggest no added benefit for racialized endorsements. We can see that across both graphs, the point estimates are virtually identical. For example, the level of support for respondents in the high-polling treatment group with an endorsement from the DLC has a level of support of 0.78, and respondents in the high-polling treatment group with an endorsement from the CBC also have a level of support of 0.79. This suggests no added benefit exists for a racialized endorsement, even for candidates with high polling numbers.

Figure 3.

Respondent’s Support by Polling and Endorsement.

The results from the experimental design point to numerous take-aways. Evidence suggests that Black voters rely on consensus cues. On average, all candidates received over seventy percent vote support, and there was a significant and positive effect on high polling. There was no effect for low polling numbers. These findings suggest that high polling numbers improve a candidate’s vote percentage. I also find that Black voters support a candidate with low polling information at the same rate as a candidate with no polling information. The biggest surprise was that racial endorsement had no effect. The evidence shows that the vote share was the same for candidates endorsed by the CBC and the DLC. While these preliminary results provide important insights, it is critical to provide first-hand accounts describing why Black primary voters utilize consensus cues. I conducted two focus groups to get open-ended responses to questions about vote choice from Black primary voters.

Focus Group in Georgia’s 5^thCongressional District

The second stage of my methodological approach takes advantage of qualitative methods. I use the focus group setting to understand Black political decision-making in primary elections held in majority-Black congressional districts. I conducted two virtual focus groups with Black voters in Georgia’s 5th congressional district. Georgia’s 5th congressional district is an excellent setting to host this discussion for multiple reasons. First, it has a long history of being safely Democratic. Figure 4 shows the Democratic vote share and the Black population for Georgia’s 5th congressional district from 1980 to 2020.Footnote ⁸ We can see that the Democratic candidate has gotten no less than 70 percent of the vote for this period. Vote share for Democrats only dropped below 70 percent once, when it was 69.1 percent. This fact illustrates complete Democratic dominance in the district.

Figure 4.

Democratic Vote Percentage and Black Percentage in Georgia’s 5^th (1980 – 2020).

Second, Georgia’s 5th has been a majority Black district for decades. Figure 4 indicates that the percentage of Black voters in Georgia’s 5th district has dropped below 50 percent only once, down to 49.2 percent in 2008. The only exception to this was the 1980 election, which occurred before it changed to majority-Black following the 1982 redistricting cycle. Finally, this district has been represented by a Black representative for a long time. John Lewis, the lauded civil rights activist, served in this district for over 20 years. He won the election in 1986, assuming the seat held by his white predecessor, Wyche Fowler. Before Fowler, Andrew Young, another prominent Black leader, served in this district. This point shows that Black leadership has served in this district for a long time. Therefore, Georgia’s 5th congressional district has critical long-term trends that make it prime for my focus groups.

Finally, Georgia’s 5th congressional district is a good option because of election-specific factors. The primary took place on May 24, 2022, consisting of three Black women. The incumbent, Nikema Williams, won 86.6 percent of the vote, defeating Charlotte Macbagito (3.9) and Valencia Stovall (9.3). However, Williams was not a traditional incumbent. John Lewis actually won the 2020 Democratic primary election but passed away before the November election. Williams, who had political experience at the state level, was appointed to fill his seat following the death of John Lewis. This made Williams the incumbent, but 2022 was the first time she actually ran in a Democratic primary election. This fact suggests that she still had an incumbency bump but did not have the same name recognition as her predecessor. These three important factors make Georgia’s 5th congressional district a good option for my focus groups.

Focus Group Questions and Responses

Responses from this focus group come from a larger project, and the responses provide important insight into the importance of consensus cues. I am interested in responses on four subjects, and the exact wording of the questions is in Appendix C.2. First, I ask respondents about their sincere preferences for the optimal candidate. In other words, I ask them what are the most important characteristics of a candidate. This question is crucial because it sheds light on the candidate a voter would choose if they used their sincere preferences. Second, I ask respondents how important the opinions of other voters in your district are. This question is crucial because it will show voters how relevant polling cues will be. Third, I asked respondents how vital prominent Black institutions’ opinions are. This question is crucial because it will provide important feedback on the relevant political endorsements from prominent institutions. Finally, I ask respondents about the importance of the media and attempt to provide insight into how they utilize this source to make decisions.

The first relevant finding is that Black voters have opinions about the qualities a candidate should have. The most common response pointed to either a personality characteristic or a specific policy position. For example, the most common personality characteristic was trustworthiness or authenticity. Respondent 1 suggests that his most important trait is having “the heart for people” or having a community-centered mindset. He claims that “leadership requires a commitment to serve, it requires a commitment to others, and then also having balance. You must be able to fight for the district’s people but also be diplomatic. Respondent 12 made a similar point by highlighting those who have sacrificed for the Black community. She claims that “many Black leaders have been arrested for supporting Black interests. Most leaders will not sacrifice for their community, so those that do should garner additional support.” Therefore, we see that trustworthiness and authenticity are important traits of a candidate.

Many respondents also point to the importance of social welfare policies. For example, Respondent 7 argues that her deal breaker is whether they show up for poor people. She says, “Do they speak up for poor people? That is my litmus test right there. You know, for the people who do not necessarily have a voice? Are they ever a voice for those types of individuals … and when I see that, that speaks volumes to get my support.” Also, Respondent 3 highlights the importance of social issues in her vote choice. She argues that “social issues do weigh in on some of this. Like women’s rights.”

However, the most important factor is a candidate’s level of electability. Respondent 5 claims, “I may like this candidate over there, but knowing people as I do, does this person have the best chance of winning? If the answer is no, then I will vote for someone who has a better chance.” This sentiment was repeated by almost all respondents in the focus group. For example, Respondent 6 claims that she ultimately votes for the person who “will do the least amount of damage” and Respondent 9 claims “sometimes I will support a candidate early in an electoral cycle, but if they do not perform well, I will look for other options.” Therefore, respondents all point to the importance of sincere preferences, but these preferences come second to a candidate’s ability to win an election.

The second important finding is that respondents discuss using consensus cues to influence their vote choice, and some directly point to the importance of polling. For example, Respondent 11 discusses the importance of the opinions of others. He says, “If it is a candidate that I was not thinking about, but it seems like an overwhelming majority of voters are leaning that way, it would not make me vote that way, but it would make me question what is going on. What are they about? Did I miss something?” Respondent 3 claims, “In real life, in this real election, this real vote is too valuable for me to throw it away on somebody who cannot win.”

Other respondents also directly pointed to the importance of polling, and Respondent 5 provides the clearest example of how she uses consensus cues strategically. She claims, “If you are of the mindset that I do not want to waste my vote, right…I wonder if there is a psychological trigger that says if it is close, I may need to pay closer attention… if this person has got 80 percent of the vote and has 20 percent, why would I need to go out and do much research? We are pack animals, so if all these folks think she is cool, I will vote for her. Then I can go on with the other things in my life.” After others agree, she continues, “Is there a psychological trigger when the gap is big enough, you say, I am going to vote because that is what we have to do. After all, this person will probably win, so I will not do much research and just support that candidate.” Therefore, respondents value group consensus and suggest polling results are enough to sway their vote toward a candidate.

Third, respondents argue that Black sources are more reliable than Democratic sources. Most respondents point to the importance of Black institutions. However, they claim that information from a Black source is not enough. Respondent 12 claims she utilizes information from “the Congressional Black Caucus, the local Democratic party, and the state Democratic party… Although I will probably read both, I pay more attention to the Congressional Black Caucus.” She highlights the importance of information coming from Democratic and Black Democratic sources, but she places more importance on information coming from Black Democratic sources. Respondent 10 has a similar sentiment, claiming when comparing these two institutions, “I would need to do more research if the information came from a traditional Democratic organization. She claims, “Black institutions, specifically the Congressional Black Caucus, are more trustworthy.”

However, respondents are also skeptical of all information coming from political institutions. For example, Respondent 11 “says he goes to Black sources first but needs to fact-check everything. Just because a Black source says, it does not mean the information is accurate.” In addition, Respondent 7 concurs that she is far more trusting of Black institutions and claims that mainstream sources are far too conservative. Finally, Respondent 3 shares skepticism of celebrity endorsements. She argues that “Killer Mike was introduced to progressive politics by Bernie Sanders, but [Killer Mike] did not do Stacy Abrams any factors. So we should avoid listening to or being motivated by celebrities.” Therefore, the group sentiment is that information from prominent Black institutions weighs more than information coming from generic Democratic institutions. However, respondents were skeptical of signals coming from both institutions.

Fourth, respondents also highlight the importance of the media. For example, Respondent 9 suggests that she follows “different news and media outlets,” highlighting that she relies on “trusted news sources for information.” Respondent 1 agrees and says he relies on the media for information on candidates who are running for Congress and other higher offices. There is also evidence that respondents also rely on horse race coverage. For example, Respondent 5 suggests that her vote choices change depending on whether it is clear “who is going to win anyway, versus when the media has told you it’s a tight race.” However, it must be stated that many respondents raised concerns about the reliability of the media. For example, Respondent 1 suggests that “a lot of the information we get is certainly through the media, but it is not always the most concise, and for lack of a better word, honest or accurate.” The responses indicate that voters use the media to learn about elections, despite being skeptical of many sources. Therefore, how much attention someone pays to the media could affect how primary voters use consensus cues. To test these implications, I conduct an exploratory analysis of my experimental results to determine if varying levels of media attention coincide with a higher use of consensus cues.

The Intervening Effect of Media Attention

This section returns to the experimental design to investigate subgroup media effects. In particular, the role of the media is critical for learning about consensus cues. Cues are only helpful to voters if they are accessible and allow them to differentiate between options (Chaiken, Liberman, and Eagly Reference Chaiken, Liberman, Eagly, Uleman and Bargh1989), and access to the media allows voters to learn more about the electoral setting (Hayes and Lawless Reference Hayes and Lawless2015). This point is particularly valid for horse-race coverage that discusses the winners and losers in an election (Bartels Reference Bartels1988). This type of coverage allows voters to reassess their own opinions and move them toward the opinion of consensus. When voters know that the information comes from other members of a closely attached group, that information will move opinions significantly. Therefore, I argue that voters who consume large amounts of media will be more likely to rely on consensus cues.

To investigate the role of the media, I included a pretreatment measure of media attention. Respondents were asked: “How often do you consume political news?” Respondents who indicated “very often” were coded as high media attention, respondents who indicated “often” were coded as medium media attention, and respondents who indicated “not very often” or “not at all” were coded as low media attention. This section will focus on vote share differences for the three polling groups because I found no statistically significant difference between racial and partisan endorsements. I calculated the subgroup effects for the DLC and CBC endorsement by level of media attention, and the results can be found in Appendix D.2. Figure 5 reports the candidate’s vote share for each polling treatment group by the level of media attention. The chart on the left shows respondents with low levels of media attention, the chart in the middle represents respondents with medium media attention, and the chart on the right represents respondents with high media attention.

Figure 5.

Support for Candidate by Level of Media Attention.

The results indicate that respondents with higher levels of media attention utilized cues at higher rates than those with low media attention. Among low media consumers, candidates with high polling numbers have a slight two-point advantage compared to those without polling, and the difference is not statistically significant. The difference in support between the low polling treatment and the control group is slightly higher at four points, but this difference is not statistically significant. This finding shows that varying levels of polling data do not influence respondents with low media attention. When looking at respondents with medium levels of media attention, we see a significant positive effect for high polling numbers. The candidate with high polling numbers has an eleven-point higher level of support when compared to a candidate without polling. The candidate without polling has a 0.67 mean level of support, and the candidate with high polling numbers has a 0.78 mean level of support. This difference is statistically significant. The difference in support between the candidate with low polling numbers and the candidate without polling is seven points, but the difference is not statistically significant.

As expected, the high media attention respondents had the most significant change from no polling to high polling. Among these respondents, the candidate without polling received a 0.78 level of support, and the candidate with high polling numbers had a 0.91 level of support. This is a thirteen-point increase, and the difference is statistically significant. This point suggests that high media attention respondents are the most likely to respond to consensus cues. The difference in support between the low-polling treatment and the no-polling group was negative, following the hypothesized pattern. The candidate with low polling has a mean support of 0.76, and the candidate with no polling has a mean support of 0.78. However, the difference is not statistically significant. The results from this section suggest that Black voters who consume more media are more likely to utilize consensus cues to choose candidates.