Introduction
The 2020 presidential election year provided unusual circumstances for political participation and, hence, for voter mobilization. The pandemic shifted the voting process, campaign strategies, and social context. The hotly contested presidential election was fought over high-stakes, highly polarized issues that resonated differently depending on one’s identity. Social justice unrest, racial reckoning and protests, uncertainty over reproductive rights, a candidate overtly appealing to white male voters, reactions to the pandemic and government responses, and institutional changes to how ballots could be cast created a charged political context, but one that varied widely across the social divisions of the electorate. In this environment, social pressures to vote (or to abstain) would be expected to carry at least as much weight as they did in past elections. But in 2020, we suspect that the nature and effectiveness of social appeals to turnout varied across the intersections of race, gender, and age.
The conditions of the pandemic forced more individuals into close quarters with their communities, heightening the possibilities for social pressure effects on turnout. As society shut down and people were forced to stay home, many took the opportunity to get to know their neighbors, even holding socially distant gatherings. Communities came together to provide mutual aid support, such as mask distributions and prescription and grocery pick-ups and drop-offs. At the same time, campaigns shifted to holding virtual rallies, increasing social media outreach, and emphasizing digital fundraising. In short, COVID-19 changed access to the ballot box and to elected officials while also bringing some communities together as people sought ways to maintain social relationships. The environment gives opportunities for people to be attuned to whether those they care about think that voting is important, and this perception may influence whether or not they themselves vote. Given this context, our paper asks: who is more likely to vote when they feel social pressure? Does social pressure vary by the intersections of gender, race or ethnicity, and age? Additionally, we expect that young people (those under 30 years old) will not be as susceptible to messages of the importance of voting, given their usual pattern of low voter turnout coupled with the protest environment of 2020. We begin with a brief discussion of intersectionality before moving on to summarizing what is known about who votes and the role of social pressure in participation. The paper then moves to the empirical test, first describing the data set, the 2020 Collaborative Multiracial Post-Election Survey (CMPS) (Frasure et al. Reference Frasure, Wong, Barreto and Vargas2021). After presenting some descriptive information regarding feelings of social pressure to vote, the paper focuses on assessing how the contribution of social pressure to explaining voting behavior in 2020 varied across intersectional identities of race, gender, and age cohort.
Intersectionality
The intersectionality literature has made clear the importance of considering the cross-cutting effects of gender with race or ethnicity. Women of color are not just “of color” or “women”: they are both at the same time. Espiritu (Reference Espiritu1997) states that “race, gender, and class, as categories of difference, do not parallel but instead intersect and confirm each other” (116). Simply adding identities together hides the interlocking contributions to identity formation for women of color. Intersectionality theory offers a conceptually powerful idea that women of color’s life experiences as women and as racial subjects affect their worldview, including their perception of political opportunity and appropriate actions. Crenshaw (Reference Crenshaw1991) notes, “The concept of political intersectionality highlights the fact that women of color are situated within at least two subordinated groups that frequently pursue conflicting political agendas” (1251–52). Women of color have the “double” experience of racism and sexism (Cassese, Barnes, and Branton Reference Cassese, Barnes and Branton2015; Gay and Tate Reference Gay and Tate1998). Arguably, as a result, the partisan gender gap is quite different among white women and women of color (Frasure-Yokley Reference Frasure-Yokley2018; Junn Reference Junn2017). Focusing on only one dimension at a time can obscure substantive differences by incorrectly averaging across an entire gender or across both genders in a racial/ethnic group. Single-dimensional approaches also promote asking whether gender or racial/ethnic identity takes priority. The intersectional approach suggests that separating dimensions is misguided and that instead one needs to separately consider the experiences of women and men of different races and ethnicities, especially for women of color.
In addition to this literature’s usual focus on race or ethnicity and gender, additional cross-cutting identities add further complexity. The experiences of the young differ markedly from those of older cohorts; younger people tend to be more dubious about voting (Beadle et al. Reference Beadle, Booth, Cohen, de Guzman, Wood, Hayat, Kei Kawashima-Ginsberg, Kiesa, Junco, Lanzilla, Lundberg, Medina and Soherr2020; Dalton Reference Dalton2021; Loader, Vromen, and Xenos Reference Loader, Vromen and Xenos2014). Generational experiences also differ by race, ethnicity, and gender. For example, until recently, young women took for granted reproductive rights that older women remember striving to obtain. For older Black Americans, the right to vote reflects a hard-fought battle, one in the past for the young.
Who Participates and Why
Arguably, the best overall account of the individual-level determinants of political participation is the Civic Voluntarism Model (CVM), developed by Verba, Schlozman, and Brady (Reference Verba, Schlozman and Brady1995). Individuals are more likely to participate to the extent they have more resources, are more psychologically engaged with politics, and are recruited into the activity. Resources include both those typically obtained individually, such as knowledge and money, and civic skills obtained in non-political adult life, notably through workplaces, religious affiliations, and organizational activity. Psychological engagement includes such well-established variables as efficacy, civic duty, and general political interest, as well as concern for particular political issues and partisanship. Recruitment, also referred to in the literature as mobilization (Rosenstone and Hansen Reference Rosenstone and Mark Hansen1993), refers to being asked to participate. Income and education, positively related to all three components of the CVM, generally correlate with higher activity (Conway Reference Conway2000; Milbrath and Goel Reference Milbrath and Goel1977; Verba, Nie, and Kim Reference Verba, Nie and Kim1978). Women in the United States vote at higher rates than men, especially women of color (Bejarano Reference Bejarano2014; Conway Reference Conway and Whitaker2008). In the United States, women generally participate less than men in activities other than voting, albeit only by small amounts recently (Schlozman, Brady, and Verba Reference Schlozman, Brady and Verba2018). Some variables affecting participation are linked with ethnicity or race, including both linked fate and factors that stem from immigration, such as citizenship status, English language facility, and time lived in the United States (DeSipio Reference DeSipio1996; Ramakrishnan Reference Ramakrishnan2005; Uhlaner, Cain, and Kiewiet Reference Uhlaner, Cain and Roderick Kiewiet1989; Wong et al. Reference Wong, Karthick Ramakrishnan, Lee and Junn2011). While generally citizens are more participatory, and citizenship is necessary to vote, it matters less for activities unrelated to elections.
Electoral rules can also affect participation. Some of these are subtle, but in recent years in the United States, there has been more focus on obvious impediments, such as voter identification requirements (Darrah-Okike, Rita, and Logan Reference Darrah-Okike, Rita and Logan2021; Hajnal, Lajevardi, and Nielson Reference Hajnal, Lajevardi and Nielson2017) and closure of polling places (Abramson and Thrower Reference Abramson and Thrower2025), or obvious facilitators, such as easing registration and use of absentee ballots (Absher and Kavanagh Reference Absher and Kavanagh2023; Garnett Reference Garnett2022; Yu Reference Yu2019). Citizens in different states and even in different counties face different combinations of these conditions and thus varied ease of voting (Blais, Massicotte, and Yoshinaka Reference Blais, Massicotte and Yoshinaka2001; Li, Pomante, and Schraufnagel Reference Li, Pomante and Schraufnagel2018).
While these measures offer knowledge of characteristics of the politically active, they do not speak to the motivations for voting, nor do they address either the social context of participation decisions or the social mechanisms leaders can use for recruitment. Past authors have noted that sociability and similar desires arising from social embeddedness can motivate participation.Footnote 1 These include, for example, solidary gratifications (Wilson Reference Wilson1973) and selective social gratification (Verba, Schlozman, and Brady Reference Verba, Schlozman and Brady1995). Olson (Reference Olson1965) includes interpersonal interaction as a mechanism to overcome the free rider problem.
Our study contributes to the sociability literature by examining social pressure as a mechanism for political action. In situations involving intimate contacts, such as friends and family, people not only observe what behavior these contacts expect but also often explicitly discuss opinions about desirable and undesirable actions and thus can expect positive outcomes if their actions conform with these opinions (Bhatti and Hansen Reference Bhatti and Hansen2012; Blais, Galais, and Coulombe Reference Blais, Galais and Coulombe2019). Social pressure can arise from being in a relationship with valued others and perceiving that they expect certain behavior. Social pressure may arise from observing the norms and expectations of valued others. These observations and discussions can create a feeling of the need to conform (i.e., social pressure) to perceived expectations, shaping behavior to align with what they believe others consider important (Le Reference Le2013; Le Reference Le2014; Uhlaner Reference Uhlaner1989; Uhlaner Reference Uhlaner2017; Uhlaner Reference Uhlaner, Congleton, Grofman and Voight2019). Individuals also experience the absence of opinions about some types of activity, from which they can infer that neither action nor inaction will have much effect. If their significant others do not care about the action, they may not experience pressure to act in a certain way.Footnote 2 One might ask why people would vote instead of just claiming that they did, that is, why not just lie? Note that the closer someone is to others, the harder it is to successfully sustain a lie about an observable behavior. We point out that even persons who feel compelled to lie about their activity demonstrate the importance of these expectations, as they evidently care about the opinions of others.
The Census Bureau (2021) and Fabina (Reference Fabina2021) reported turnout in 2020 was the highest in the 21st century. We believe that this was partly due to circumstances during this election year that created opportunities to transmit messages to vote more easily, heighten identity politics, and activate community belongingness—all potentially amplifying the role of social pressure. We investigate whether social pressure was a potent source of political action given the peculiar circumstances of 2020. However, seeking social approval, and thus the effects of social pressure, is not uniform: only individuals who feel it is important to be a part of and identify as a member of a certain group will seek approval from that group and thus feel social pressure to behave accordingly.
Leaders can use mechanisms of social pressure to mobilize action by influencing whether people think voting is important to their chosen community (i.e., fellow worshippers, feminists, veterans, etc.) and whether voting (or abstaining) is expected in order to be perceived as a member in good standing of the group (Uhlaner Reference Uhlaner1989; Uhlaner Reference Uhlaner2014). Some people may not care, but we believe many do feel pressure to stay aligned with their community. Both survey evidence and field experiments have shown that recruitment increases participation (Garcia Bedolla and Michelson Reference Bedolla and Michelson2012; Gerber and Green Reference Gerber and Green2000; Green, McGrath, and Aronow Reference Green, McGrath and Aronow2013; Green and Gerber Reference Green and Gerber2015; Rosenstone and Hansen Reference Rosenstone and Mark Hansen1993; Verba, Schlozman, and Brady Reference Verba, Schlozman and Brady1995). Some of the results, both from surveys and from field experiments (Gerber, Green, and Larimer Reference Gerber, Green and Larimer2008; Gerber, Green, and Larimer Reference Gerber, Green and Larimer2010; Le Reference Le2013; Le Reference Le2014; Uhlaner Reference Uhlaner2017), suggest that social pressure plays a significant role.
People will vary in the degree to which they feel social pressure to vote, partly depending upon their identity or intersectionalities. The amplification of identity politics in 2020 meant that people were hearing different messages, which may have activated different identities. With intersecting identities, multiple cross-pressures came into play, further diversifying who felt social pressure and which identities were linked to voting. People who felt threatened, such as women by the erosion of reproductive rights and minority communities by threats to civil rights, may have seen the election as more consequential and experienced a heightened sense among their primary contacts that voting matters. The resurgence of the Black Lives Matter movement, increasing concerns among Asian Americans surrounding reports of anti-Asian sentiment, and threats of deportation of immigrants may have served to amplify the activation of racial identity, whereas other aspects of the identity politics of 2020 focused on the grievances of men and of white men in particular. In light of this sociopolitical context, we propose the following hypotheses:
H1a: Feeling social pressure to vote in 2020 will have been felt most acutely by persons under threat or in a group targeted by the campaigns. The political environment likely made women of color feel doubly threatened and may have heightened their feelings of social pressure to vote.
H1b: The effect of social pressure on turnout will vary in strength depending upon intersectional identities.
Even though young voter enthusiasm increased in 2020 (“Harvard Youth Poll: Election 2020”), younger voters have been more skeptical than their elders about the usefulness of electoral activity in recent years (Beadle et al. Reference Beadle, Booth, Cohen, de Guzman, Wood, Hayat, Kei Kawashima-Ginsberg, Kiesa, Junco, Lanzilla, Lundberg, Medina and Soherr2020; Dalton Reference Dalton2021; Loader, Vromen, and Xenos Reference Loader, Vromen and Xenos2014). Perceived importance of voting among family and friends has been shown to influence whether college students (mean age = 20.73) expressed their intention to vote (Glynn, Huge, and Lunney Reference Glynn, Huge and Lunney2009). However, work by Panagopoulos and Abrajano (Reference Bejarano2014) found that older cohorts tend to be more responsive to social pressure to vote than younger ones. The protests of 2020 may well have accentuated this age divide in views of voting and, given the content, may have had an uneven impact across racial and ethnic groups. The Summer of Unrest was especially meaningful for young Black Americans when people took to the streets demanding justice for Breonna Taylor and George Floyd; thus, young Black Americans may have felt the futility of voting more so than most. Given the protest environment coupled with COVID-19 restricting youth interaction, particularly on college campuses, we expect that individuals under 30 years old were less likely to receive messages from their friends and family about the importance of voting, less likely to perceive social pressure to turnout, and less likely than older people to translate social pressure into voting. Taken together, these considerations lead to our next set of hypotheses:
H2a: Individuals under 30 years old were less likely to feel social pressure to vote.
H2b: Individuals under 30 years old are less likely to translate social pressure to vote into turnout.
To address our hypotheses, we consider ethnicity/race, gender, and age in examining the effect on turnout when people perceive that their network thinks that voting is important. Before we examine our primary question, namely how social pressure affects turnout, we must first establish whether the prevalence of social pressure differs across groups. The next section describes the data we use to test these hypotheses.
Data and Methods
This study uses the primary samples from the 2020 Collaborative Multiracial Post-Election Survey (CMPS) (Frasure et al. Reference Frasure, Wong, Barreto and Vargas2021)Footnote 3 . The survey was structured as separate samples of Latino/Latina, Black American, Asian AmericanFootnote 4 , and white, non-Hispanic U.S. residents, age 18 and above, both citizens and noncitizens. There are just under 15,000 completed interviews, approximately 4,000 from each of the first three groups and 3,000 from the last. Data are weighted within each racial/ethnic group to correspond to the 2019 Census American Community Survey. Each ethnic/racial population made up an independent sampling frame, so the respondents are representative of each group but not, collectively, of the U.S. population. Respondents completed the survey online between April 2, 2021, and August 25, 2021. Appendix C contains further technical details, including a more detailed description of the sampling methodology. Note that an augmented data file, containing respondents from oversamples, and a file from a sample of youth aged 16 or 17 were released subsequently, but these data are not analyzed in the present research. Appendix A contains detailed information on the items used to construct the variables used in this study, including some question wording and variable numbers in the data file.
Operationalizing Social Pressure
We suggest people would feel social pressure to take an action if two conditions hold: 1) there are persons or a group whose good opinion they value, and 2) they think that these others expect the action. In other words, if a person believes that the people they care about think it is important to participate in politics, they should theoretically anticipate receiving some form of positive feedback by participating and thus feel social pressure to do so.
The instrument contains questions modified from Le’s (Reference Le2013, Reference Le2014) study of social pressure and political participation in a particular ethnic enclave, Vietnamese Americans in Orange County, California (Little Saigon). She asked people about their sense of belonging in the group along with two additional questions important for assessing the role of social pressure. These questions asked whether the person cared about being considered a member of the group and whether the person believed that the group had a norm of political participation specific to several different types of activities. Since Le’s research was restricted to one ethnically homogeneous enclave, she could include detailed group membership questions. Because the CMPS includes multiple ethnic groups across the United States, a more general prompt was required.
The 2020 CMPS operationalized social pressure through a battery of questions assessing the respondent’s perception of the importance of different types of participation for the people they care about.Footnote 5 The items tap the respondents’ views on what political behaviors the people close to them value and likely reward. For this paper, we focus on voting behavior. Respondents were asked, “For each of the following activities, indicate if the people you care about, like your friends and family, think doing each activity is important or unimportant, or do they disapprove?” followed by a list of activities including “voting.” The cue of “people you care about” addresses the first condition, that these are persons whom the respondent values, while “friends and family” suggests direct benefits. The respondents were offered the following response options: very important, somewhat important, not that important, not at all important, disapprove, strongly disapprove, and I don’t know what they think about this. See Appendix A for these and subsequently discussed variables for the variable numbers in the CMPS data file, coding, and text of the instrument questions. The social pressure variables were each coded as follows: a “strongly disapprove” response was coded as 1, “disapprove” as 2, “don’t know” as 3, “not important at all” as 4, “not that important” as 5, “somewhat important” as 6, and “very important” as 7.
It is worth further clarifying the link between this measure and the theory. We assume that when a respondent believes that other people think that voting is important, the respondent will conclude that these others look positively upon other people—including the respondent—turning out. More generally, in conversation, when we tell someone it is important to do x (brush your teeth, signal when turning, vote), that usually carries an implicit injunction for x to be done along with implicit disapproval of failure to do x. Certainly, endogeneity may be a contributing factor: respondents may be projecting their own preferences onto their friends and family. That is not a problem for our hypotheses; what matters for influencing behavior is the respondent’s perception, what the respondent believes about the opinions of others, even if that perception is factually inaccurate. The “don’t know” responses are also meaningful, as people cannot be influenced by what they do not perceive.
The social pressure variables just described will help us identify any main effects of perceived social pressure. We are particularly interested in the distribution and effect of social pressure for intersectional identities, given the focus on identity politics in this election. While candidates may be appealing to specific races or genders, those with intersectional identities may experience conflicting messaging, thus feeling cross-pressures that may be sending a message of needing to “choose” one identity (e.g., gender) over another (e.g., race or ethnicity). The political climate may also contribute to compounding messaging for those with intersectional identities, where campaign messaging may activate one identity (e.g., gender) and the protest environment can activate another (e.g., race or ethnicity). To explore this, we estimate separate models for men and women, as well as for the young and the older, thereby allowing the entire model to vary.
Operationalizing Other Variables
The dependent variable is drawn from the CMPS questions on voting in the 2020 election. Appendix A provides detailed wording and coding information; a summary is presented below.
The vote in 2020 question was asked early in the interview to respondents who indicated that they were registered to vote: “This year a lot of people said they did NOT vote in the election, because they were just too busy, not that interested in politics, or frankly don’t like their choices. How about you? Would the official vote records for {INSERT STATE} indicate that you voted in the November 2020 election, or like some people, did you skip this one?” People who were not registered to vote were coded as non-voters.
The CMPS also includes items tapping many of the generally recognized demographic and attitudinal predictors of participation. Most independent variables follow standard coding (see Appendix A for details). Binary variables—age 65+, homeownerFootnote 6 , femaleFootnote 7 , unemployed, married/partnered, and citizen—are coded as 1 if the respondent fits the category. For some analyses, respondents under 30 are separated out. Family income ranges from 20 to 200 (in $10,000 increments), and education is measured in years (7 to 19). Immigration experience is measured by the percentage of life spent outside the United States—zero for U.S.-born individuals, and for others, calculated as years abroad divided by age (e.g., 0.5 for someone who immigrated at 25 and is now 50).Footnote 8
The attitudinal independent variables include standard measurements of partisan strength and political interest. Partisan strength is coded from independent to leaner through weak to strong partisan. Political interest goes from not at all interested in politics through two intermediate levels to very interested. A few other variables are less standard. Linked fate uses a somewhat nonstandard question; all participants were asked, “How much do you think what happens to the following groups here in the U.S. will have something to do with what happens in YOUR life?” Respondents answered this question for each group separately. We use the answer to the person’s own racial/ethnic group to construct the linked fate variable, with five levels from “nothing” to “a huge amount to do with what happens in my life.”
Two independent variables were taken from the battery of items tapping political participation: organizational membershipFootnote 9 and political discussion with friends and family. Typically, organizational activity is correlated with more activity, arguably because it provides an indirect indicator for civic skills (Verba, Schlozman, and Brady Reference Verba, Schlozman and Brady1995), not otherwise measured in the CMPS. The item was introduced with “Since January 2020, have you?” followed by “participated in one or more social, cultural, civic, political groups or unions?” If respondents responded affirmatively, they were then asked how many groups they regularly participate in. Responses were coded as 3 for three or more groups, 2 for two groups, 1 for one group or just a yes to the first question, and 0 otherwise. People who discuss politics with friends and family are more likely to have an opinion about their expectations for activity. Those who answered “Yes, I’m certain I did that last year” were coded as 1; all others were coded as 0. Although efficacy is a key factor in participation research, the external efficacy measure was only significant for non-Hispanic white women—and even then, it had a negative effect, possibly due to confusion caused by a double negative in the question wording. Including the measure did not affect other results. Therefore, we excluded it from our final models. The CMPS did not include a measure of internal efficacy.
The survey addresses representation and recruitment by first assessing whether respondents feel represented by someone. Since leaders can increase feelings of social pressure as part of mobilization, we include this in our analyses. Approximately one-third of respondents indicated that they thought some leader or organization looked out for his or her concerns; this is the leader variable. These respondents were then asked whether they had been asked to vote, asked to protest, or asked to take some other political action in the last two years by the leader or group. The two-thirds of respondents who answered “no,” there was not such a leader or group, were asked whether anyone in the last two years had asked them to vote, to protest, or to take some other political action. Another recruitment question elsewhere in the interview asked respondents whether a candidate, party, or community organization had asked them to register or to vote. All three of these questions are coded 1 for those who said yes and 0 otherwise. These recruitment variables create a rigorous test of the influence of social pressure in our analyses.
Finally, we include the Cost of Voting Index (COVI) (Li, Pomante, and Schraufnagel Reference Li, Pomante and Schraufnagel2018; Schraufnagel, Pomante, and Li Reference Schraufnagel, Pomante and Li2020; Schraufnagel, Pomante, and Li Reference Schraufnagel, Pomante and Li2022; costofvotingindex.com/data) to capture differences across states in the difficulty of voting. COVI is positive for states where voting is costly and more negative for states where voting is easier. The pandemic led to major changes in the rules for casting a ballot and thus the conditions for turnout (Morris and Miller Reference Morris and Miller2022); thus, Schraufnagel et al. (Reference Schraufnagel, Pomante and Li2022) provide a cost of voting measure for 2020, which captures the possible temporary conditions of the 2020 election. We use this measure in the estimations, labeling it “voting costs.” It was constructed from multiple measures, including ease of voter registration (deadlines, restrictions, youth pre-registration), voter convenience (early, absentee, and mail voting, polling stations, drop boxes, poll hours), and voter identification laws. The measure does present one disadvantage: the authors did not construct it for the District of Columbia, so those 64 (unweighted) respondentsFootnote 10 are missing from models that include voting costs. This omission most seriously impacts the Black American sample, with 42 (unweighted) residents of D.C. But as noted below, estimations that drop the voting costs variable and include these cases produce essentially the same results.
Results and Discussion
Does perceiving social pressure translate to increased turnout, and if so, for whom does it translate? Do intersectional identities raise the susceptibility of feeling social pressure, and if so, in what way? We motivate the examination of social pressure by first noting that it does appear to be related to voter turnout. In all four racial/ethnic groups, the correlation coefficient between vote turnout and social pressure is positive and significantly different from zero, with p-values of less than .001. The correlation coefficients are also substantial: 0.29 for Asian Americans, 0.34 for Latinos/Latinas, 0.38 for Black Americans, and 0.39 for non-Hispanic whites. The distribution of responses shows that voters are more likely than abstainers to sense that valued others think voting is important. Indeed, in each racial/ethnic group, voters were more than twice as likely as abstainers to say the people they cared about think it is “very important” to vote. Table B.1 in Appendix B reports the full distribution of responses for voters and abstainers.
To answer our primary questions on the effect of social pressure on turnout, we tabulated the percentage of citizen respondents who gave each answer to the social pressure question in each racial/ethnic group, split by gender. Interesting patterns emerge once we take account of gender and age cohort, presented in Figure 1 and in more detail in Table 1. The “young” are defined as persons 18 through 29 years old, while the “old” are people age 30 and above.
Percent of citizens responding that people they care about think it is “very important” to vote by race/ethnicity and age cohort.
Percent giving each response to the social pressure to vote question, by gender, age, and racial/ethnic group (citizens only)
Figure 1 splits the sample by age and gender in addition to race or ethnicity and shows the percentage of people in each group who give the response “very important” when asked what the people they care about think about voting, that is, the people who are likely to feel social pressure to vote. Figure 1 highlights that feeling social pressure to vote is not uniform, and it merits further investigation to examine who may be more likely to feel it. First, Figure 1 illustrates that gender clearly matters. For both young and old and for all four groups, more women than men perceive this social pressure; however, the magnitude of the gender gap depends on both age and ethnicity. It is especially wide for youth of color and also for older Black Americans. The gap is still there, but too narrow to be significant for non-Hispanic white men and women, regardless of age cohort. Young women of color are less likely than young non-Hispanic white women to feel social pressure to vote, while older Latinas and Asian American women are about equally as likely as non-Hispanic white women to report this. Black American women, however, do stand out, with seventy percent reporting voting is “very important” for their significant others, consistent with H1a’s expectation that social pressure would be strongest among those who felt most targeted. The 2020 political environment likely made women of color feel doubly threatened, amplifying their sense of pressure to vote. Next, Figure 1 shows a stark contrast by age. Across all ethnic/racial groups, the older cohort is more likely than the youth to say that the people they care about think voting is very important, consistent with H2a. The difference between youth and their elders is far larger for Latine, Black American, and Asian American respondents than for non-Hispanic whites, reinforcing literature that intersectional identity matters.
Table 1 reports the full distribution of responses to the social pressure question by gender, racial/ethnic group, and age for citizens. The other responses of most interest are the reports that significant others disapprove of voting or strongly disapprove. Few people report perceiving disapproval, but there is variation. Those under 30 are more likely than older people to say that their significant others disapprove of voting. Young Latines differ little, but in the other groups, young men are more likely to see disapproval than young women, markedly so for young Black men. Nonetheless, these percentages are small; few citizens thought the people they cared about disapproved of voting. The results in some investigations not reported here suggest disapproval tapped a different dimension than importance. Given that issue and in light of the very few disapproval responses, we handled them in several ways in the analyses that follow, as discussed later. The different approaches produced very similar results, as would be expected from the few people affected.Footnote 11 The results for the main independent variables of interest were very similar, with few differences in other variables.
Having established who is more likely to feel social pressure to vote, we now examine our central question: to what extent—and for whom—is this pressure associated with actual voter turnout? Recall H1b states the effect of social pressure on turnout will vary in strength depending upon intersectional identities, whereas H2b states individuals under 30 years old are less likely to translate social pressure to vote into turnout. As discussed above, many other factors also affect participation and must be considered before concluding whether social pressure contributes to the decision to vote. We use ordinary least squares regression with robust standard errors for these estimations.Footnote 12 Separate regressions are run for each racial/ethnic group both because of the sampling design and because the models can then differ by group.
Are There Gender Differences?
Gender issues were front and center in 2020, and the intersection of gender with race and ethnicity is clearly important. Access to reproductive care was further threatened by a conservative shift in the Supreme Court at the same time as women opposed to abortion were energized (Frey Reference Frey2022). Furthermore, George Floyd’s final words, calling out for his mother, may have resonated deeply with many women, perhaps more so with Black women, who may have had an instinctual, empathetic reaction (Miller Reference Miller2021). Burdens of the pandemic, from childcare to healthcare, fell hard upon women (Ayittey et al. Reference Ayittey, Dhar, Anani and Bennita Chiwero2020; Flor et al. Reference Flor, Friedman, Spencer, Cagney, Arrieta, Herbert and Stein2022; Power Reference Power2020). Additionally, women may respond to social expectations differently than men (Ridgeway Reference Ridgeway and England1993). To explore this, Table 2 shows the results of the estimations separately for men and women. The table entries are the unstandardized coefficientsFootnote 13 and p-values. Both men and women in each racial/ethnic group are more likely to vote if they perceive that the people they care about think that voting is important, captured by the “social pressure” variable; these coefficients are highly significant, with p values of .000.
Vote in 2020 election estimations (OLS) by ethnic/racial group and gender, citizens only
^ = sig at 0.10;*= sig at 0.05; **= sig at 0.01; ***= sig at 0.001
When we compare across groups, Table 2 indicates that feeling social pressure to vote influences turnout more for non-Hispanic white men (B = 0.08, SE = 0.011) than for men of color, including Latino men (B = 0.048, SE = 0.010), Black men (B = 0.038, SE = 0.008), and Asian American men (B = 0.049, SE = 0.011). This is contrasted with the little variation in the effect of social pressure among women across all four racial/ethnic groups. When we consider gender differences within each group, Black women and Latinas are more likely than their male counterparts to turnout if they feel social pressure to do so, whereas non-Hispanic white men are more likely to turnout if they feel social pressure than non-Hispanic white females. These results support H1b; social pressure to vote has varying effects on turnout depending on intersectional identities. We argue that in 2020, the Summer of Unrest, threats of deportation, and the rise of anti-Asian hate activated intersectional identities and placed a larger burden on Black women. Additionally, campaign rhetoric frequently emphasized perceived grievances of non-Hispanic white males.
We also note an intriguing gender difference in the effects of voting cost, which—as previously mentioned—reflect the ease or difficulty of voting during the pandemic. The coefficient for voting costs is significant at .05 and negative for non-Hispanic white women and Latinas and significant at .10 and negative for Black women and Asian American women. In contrast, voting costs make no significant difference for men regardless of ethnicity/race. These findings suggest that reductions in voting barriers during the pandemic may have had a disproportionately positive impact on women, highlighting the gendered and racialized nature of how structural constraints shape electoral participation.
Does the Effect of Social Pressure Vary by Age Cohort?
As outlined earlier, older cohorts are more likely than younger cohorts to perceive that voting is valued by the people they care about. We now turn to whether these age cohorts differ in their response to this perception. Our next model incorporates age cohort into the estimation to assess whether social pressure functions similarly for the young and their elders. Results are presented in Table 3. The left pair of columns under each race/ethnicity heading presents unstandardized coefficients and p-values for men; the right pair presents results for women. Within each pair, the left column presents results for people under 30 years old, and the right column presents results for those 30 and older. As expected, the numbers of cases drop, especially for the young cohort. Nonetheless, some interesting differences emerge in the effectiveness of social pressure.
Vote in 2020 election estimations (OLS) by ethnic/racial group, gender, and age, citizens only
^ = sig at 0.10; *= sig at 0.05; **= sig at 0.01; ***= sig at 0.00
Perhaps the most striking result is the overall greater significance of social pressure for the turnout of older people than for the young, consistent with H2b. For ease of viewing, Figure 2 provides a visual depiction of the significant effects of social pressure reported in Table 3. In contrast to the older cohort, fewer younger people report feeling social pressure, but when they do, the effects on turnout are generally weaker than for their elders in all groups, with the exception of young Asian American men. The difference is slight for non-Hispanic whites but substantial for Latinos and Latinas, for Black men and Black women, and for Asian American women. This is consistent with scholarship finding that youth are disillusioned about the efficacy of electoral activities. Since our measure of social pressure is based on a question referencing friends and family, it may not fully capture the sources of social pressure that young adults are more sensitive to, such as a broader peer group or messages gleaned from social media. In other words, social pressure may play a more important role in shaping youth turnout than we find: the measurement used in this study may not reflect the channels through which young adults experience it. Nonetheless, the analysis shows that social pressure to vote is significant for youth of color, except among young Asian American women, even if the magnitude is lower than that of the older age cohort. These generational patterns offer an important context as we turn to examining voting costs among age cohorts and race and gender.
Social pressure coefficients by intersections of race/ethnicity, gender, and age.
* = sig at 0.05; ** = sig at 0.01; *** = sig at 0.001
Once we take account of age, the results for the voting costs become more mixed. Table 3 shows that the negative impact of the costs falls mainly on the older cohort. Non-Hispanic white men do not seem to be affected by voting costs. Voting costs have a highly significant (at p < .01) negative effect on the voting rates of older white women, young Latinas, older Black Americans, and older Asian Americans. It also negatively affects the turnout of older Latinas, but with a p-value of 0.10. And it has a positive effect (but again at a p-value of 0.10) on turnout of young Black males. The positive effect for young Black males combined with the negative effect for older Black males to produce the zero effect overall for Black men when we did not take age into account.Footnote 14
This analysis underscores the importance of the intersectional perspective in understanding political participation. Examining age as it intersects with race and gender, we observe meaningful generational differences. Younger citizens experience social pressure differently than their elders, and the effect on turnout varies when considering race and gender. These findings challenge the notion that youth are uniformly disengaged from electoral participation and highlight how intersecting identities shape patterns of political behavior.
Conclusion
We have provided evidence that social pressure plays an important role in moving people to the polls. Social pressure to engage in an activity often powerfully influences many types of behavior, notably including a decision to vote in an election or abstain. Many factors contributed to the high turnout in the 2020 election, and many circumstances created fertile ground for social pressure to operate. The response to the pandemic upended normal campaigning and reduced barriers to voting. The intense partisan polarization coupled with identity politics enhanced opportunities for mobilization on identity at the same time as the pandemic intensified close interpersonal connections. Different people experience varying levels of social pressure and, moreover, differ in their response to that pressure. We focused in this study on social pressure stemming from individuals’ close contacts: their friends and family. We find strong supporting evidence for our key hypotheses: People are more likely to vote the more they think that the people they care about consider voting is important, and this will vary depending on intersectional identities. In sum, individuals observe and engage with people they care about and can infer desirable or undesirable behavior to remain in good standing with valued others.
Identity politics played a pivotal role in the 2020 election, with race/ethnicity and gender being prominent in the campaign. This may have served to activate multiple identities at once. Those with marginalized intersectional identities may have been more attuned to how the election outcome may affect them both as a woman and a person of color. Leaders can craft messages that increase social pressure to vote by tapping into people’s desires to be accepted by others and linking turnout to acceptable behavior.
The protest environment fomented by police brutality and frustration with the government’s response to the pandemic may have served to further distance youth from electoral politics. Our findings suggest that most young people did not feel social pressure to vote. Moreover, among the majority of those who did, social pressure did not translate to turning out. These observations about youth may have been more pronounced in 2020 than in previous election years. During the pandemic, youth may have found themselves on social media more often and had constant exposure to protest and activism, which are more emotionally engaging than traditional GOTV campaigns. Rather than seeing voting as a primary avenue for change, many may have viewed direct action as more influential. In addition, distance learning also made it so that young people were less engaged with their campus peers and were less available for campus-based mobilization efforts, thus decreasing opportunities for political discussion and exposure to norms.
Every election will present unique contexts with varying issues and events. Perceptions of the importance of voting may vary depending on how competitive the election is, the visibility of the election, and overall media attention to the election. For example, midterm elections tend to have lower turnout; thus, they may not be discussed as much among friends and neighbors, and they have low feelings of social pressure to vote. However, a hotly contested local election may garner intense discussions and visibility of lawn signs, thus raising the perceptions of social pressure to vote. Each election also presents its own mix of salient issues and of candidates looking for support. That mix can magnify the effects of social pressure for targeted citizens, which would lead to corresponding variations in the perception of social pressure to vote and in its effects on turnout. The specific differences we found in social pressure and its effects across gender, race, or ethnicity, and even age are reflections of the political moment. These moments, with help from leaders, shape people’s expectations and thus who perceives that their community thinks voting is important and the degree to which that shapes their behavior. Our study supports previous scholarship that suggests social pressure is influential in voting regardless of election contexts; we further expand the existing literature by showing that feelings of social pressure vary by identities, particularly intersectional identities.
The circumstances of 2024 were unique in the incumbent president stepping down just fifteen weeks before the election to make way for a Black female candidate. Moreover, 2024 was the first election post-Dobbs. These circumstances may have created cross-pressures for Black women and increased social pressure for them. While the specifics of elections and campaigns are dynamic, what we think remains constant is the responsiveness of individuals to these perceptions of the opinions of valued others. Future elections will continue to be shaped by evolving issues, and fostering a thriving democracy requires strategic efforts to ensure that all eligible voters feel both empowered and compelled to participate. By recognizing how social pressure operates across different identities and contexts, leaders, campaigns, and advocacy groups can craft messages that highlight the importance of voting and why this action is important for the welfare of the community.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/rep.2025.10037.
Funding statement
The authors gratefully acknowledge funding support for this project from the Russell Sage Foundation, grants 2008-27809 and 2008-27923 and from the Center for the Study of Democracy, University of California—Irvine. We thank the conference discussants, especially Tabitha Bonilla, and the anonymous reviewers for their helpful suggestions.
Competing interests
Authors declare none.
Appendix A: Variable Definitions
The 2020 Collaborative Multiracial Post-Election Survey (CMPS) (Frasure et al., 2021) provided all but one of the variables. (The exception is voting costs, from the COVI data (Schraufnagel, Pomante, and Li 2022).) The variable numbers below refer to the numbering in the CMPS data file (some versions of the codebook differ). Respondents’ racial/ethnic group (race) was defined by the primary sampling frame.
Dependent variable:
Vote in 2020 election: Q12: Asked if Registered (S6 and S6_non): This year a lot of people said they did NOT vote in the election, because they were just too busy, not that interested in politics, or frankly don’t like their choice. How about you? Would the official vote records for (INSERT STATE) indicate that you voted in the November 2020 election, or like some people, did you skip this one? Recoded as (1) “Yes, I am 100% certain I voted,” (0) “I think I did NOT vote” or “I am 100% certain I did NOT vote” or not registered so not asked the question.
Demographic independent variables:
Citizen: (1) if born in the mainland U.S. or Puerto Rico or if born elsewhere but answered naturalized (1) on Q807 OR if answered other (8) on Q807 but open-ended response indicated R is a citizen (e.g. born abroad but child of U.S. citizens); noncitizens coded (0).
Gender: Recoded from S3b. Non-binary or “something else” coded as missing (115 respondents, unweighted; less than 1% of each racial/ethnic group, weighted)
Age: 2021 – birthyear; birthyear = S5
Age > 65: recoded from age as (0) for those under 65 and (1) for those 65 or older
Family income: Q813 recoded to midpoint of range in tens of thousands, 20 to 200
Education: S13 recoded to approximate number of years of schooling, range from 7 to 19
Homeowner: Q639 recoded as (1) if “own my home” OR age <=21 and “live with people who own the home”, (0) otherwise (rent, live with people who rent, other housing arrangement, or over 21 and live with people who own)
Unemployed: Q814 recoded as (1) if R currently unemployed, otherwise (0) (employed full or part-time, student, retired, homemaker)
Percent of life outside [mainland] United States: Required fixing Q806, year came to US or mainland, by adding 1899 to the values (as the years were labels and the values were numbers from 1 to 122). (0) for native born. For others, 100 times (years not in US divided by age). More precisely, 100 times ((year came to mainland minus year born) divided by age). 71 cases who claimed to immigrate more than 5 years before they were born were set to missing.
Married: Q741 coded (1) if married or “in a relationship and living together” coded (0) otherwise (divorced, widowed, single not in a relationship, in a romantic relationship but not living together)
Other independent variables:
Social Pressure: “For each of the following activities, indicate if the people you care about, like your friends and family, think doing each activity is important or unimportant, or do they disapprove?” Q162r1: “Voting in elections?” Response options and values assigned:
very important (7), somewhat important (6), not that important (5), not important at all (4), I don’t know what they think about this (3), disapprove (2), strongly disapprove (1).
Strength of party identification: Q21, Q22, Q23: “Generally speaking, do you think of yourself as a Republican, a Democrat, an independent, or something else?” “Do you consider yourself to be a strong [Dem/Rep] or not?” “If you had to choose, do you consider yourself closer to the Republican party or to the Democratic party?” Recoded as (0) independent, (1) leaners, (2) weak identifiers, (3) strong identifiers.
Interest: Q29 “Some people are very interested in politics while other people can’t stand politics, how about you? Are you”… (0) Not at all interested in politics, (1) Not that interested in politics, (2) Somewhat interested, (3) Very interested in politics.
Perceives linked fate: Q551_Q559rx. “How much do you think what happens to the following groups here in the U.S. will have something to do with what happens in YOUR life? What happens to [respondent’s race/ethnicity] will have …” (1) Nothing to do with what happens in my life, (2) Only a little to do with what happens in my life, (3) Something to do with what happens in my life, (4) A lot to do with what happens in my life, (5) A huge amount to do with what happens in my life. NOTE: This operationalization differs from the standard questions for linked fate.
Two recruitment questions branch from Q170 [leader]:
Q170: “Is there any political leader or any group or organization that you think looks out for your concerns, even if you are not a member of the organization?” (1) yes (0) all else
Recruited to vote by leader: [If Q170 yes] “Did this leader or someone in authority in the group ever suggest, in the last two years, that you:” Q172r1 “Personally vote for or against certain candidates in an election for public office?” (1) yes (0) all else
Recruited to vote, no leader:[If Q170 no] “Did anyone ever suggest, in the last two years, that you:” Q175r1 “Personally vote for or against certain candidates in an election for public office?” (1) yes (0) all else
Recruited to vote by party or org.: Q38r1 to Q38r4 “During the 2020 election and campaign, did anyone from a campaign, political party, or a community organization contact you to ask you to vote, or register to vote?” (1) if answered “yes” to any, (0) if answered “no” to all
Organization activity and Discusses politics: From the participation grid. The grid began with “Since January 2020, have you…?” followed by an activity, with the response categories coded (1) “Yes, I am certain I did that last year,” (0) All other responses (“not sure” and variants of “no”).
Discusses politics: Q77r1 “Discussed politics with family and friends?”
Organization activity: Q84: “Participated in one or more social, cultural, civic, political groups or unions?” [If Q84 yes] “In how many groups do you regularly participate?” Responses coded (3) three or more groups, (2) two groups, (1) one group or no answer but “yes” to Q84, (0) all else, including any answer to Q84 other than “yes, I am certain …”
Appendix B: Percent Giving Each Response to Social Pressure to Vote Question, by Abstainers and Voters by Racial/ethnic Group, (citizens only)
Open access
