Civic Virtues in Conflict?
Identifying “bridge populations” for reconciling tensions between American ideals of civic life
Advocates for a renewal of civic life in the United States have been working to reverse a decades-long decay of participation in voluntary organizations, political associations, and sustained community service. The curious phenomenon of “bowling alone” has turned into a full-fledged epidemic of loneliness, and the proliferation of cable news and social media has festered a generation of “political hobbyists” disengaged from the real grassroots work of democracy. The rot of civic life is not only an individualized spiritual crisis, but a republic-wide emergency of authoritarianism, polarization, and disinformation.
At interpersonal and municipal levels, projects such as interfaith roundtables, participatory budgets, and youth service corps have been moderately successful at bucking this trend. However, efforts to scale these projects have struggled to take root, evidenced by the present lack of a nationwide awakening or revolution of civic engagement. For social initiatives of all types, commonly known impediments to scale include financial constraints, institutional weaknesses, hostile politico-legal environments, and cultural incompatibility. We can assume that each of these impediments are at play in the ongoing suffocation of American civic life.
The question that guides this analysis focuses on the cultural impediment: as citizens prioritize certain civic ideals, do they simultaneously diminish other civic ideals for the sake of maintaining a coherent worldview? In other words, there may exist a myriad of inner logics governing how Americans delineate between civic virtue and civic vice that can be rooted in any mixture of their political commitments, philosophical reasoning, and/or cultural milieu. Interventions to promote civic life that do not account for these tensions risk contradicting their own message, siloing their audience, or disparaging their stakeholders.
Utilizing data on the perceptions American voters hold about civic language, this analysis identifies 10 complementary pairs and 10 antagonistic pairs of civic ideals. Together, these pairs give us a panoramic view of how Americans broadly delineate between civic virtue and civic vice. The antagonistic pairs merit special attention: advocacy for one ideal may unwittingly subvert (or be subverted by) a pre-existing high regard for an antagonistic ideal. To encourage projects that overcome these antagonistic pairs, this analysis continues to identify “bridge populations” that are a demographic profile of the top ≈10% of Americans who hold positive associations for both ideals of an antagonistic pair. These bridge populations arguably are an untapped, dynamic source of creativity for overcoming the tensions constraining civic life in the United States.
Data & Theory
During November 2021, Philanthropy for Active Civic Engagement (PACE) and their polling partner Citizen Data launched the “Civic Language Perceptions Project” (CLPP) with a nationally representative survey of 5,000 American voters. Motivating this endeavor was a desire to measure the gap between the professional usage and public perception of 21 specific terms used in the field of promoting democracy and civic life. The responses provide an unprecedented glimpse into how people in the United States view terminology related to civic participation.
Prior analysis of CLPP data has focused on cross-tabulations of the 21 terms against demographic variables such as age, gender, religion and race. This analysis takes a different approach, instead focusing on how perceptions of the 21 terms are interrelated with each other. That is: which pairings of terms are positively correlated, and which are negatively correlated? The most negatively correlated pairings, or “antagonistic pairs”, represent tensions in the culture of American civic life. Additional analysis reveals “bridge populations” using demographic cross-tabulations of respondents reporting relatively high positive associations of both terms in the antagonistic pair.
This line of questioning is informed by Yanni Loukissas’ approach to interpreting data, where data is “seen as a means with which to connect to and learn from those with actual local knowledge” (Loukissas, All Data is Local, 183). The correlation mapping used here is an example of what Loukissas calls “panoramic” views of data that “reveal perspectives at previously incomprehensible scales” (40). As a dataset co-created by surveyors and surveyed alike, this approach directs our attention towards outlier and liminal subpopulations. If we were to meet with people belonging to these subpopulations, we would encounter a “local knowledge” of how they have reconciled the broader tensions that frustrate American civic life.
Through this panoramic exploration of correlations and deviations, we catch glimpses of the worldviews that underpin American civic life. Each respondent’s pattern of responses is neither random nor purely a function of demographics, but rather the respondent’s own worldview where certain ideals of civic life are perceived as either complementary or contradictory to each other. As Mona Kanwal Sheikh articulates, a worldview consists of both “its social, ethical, political, and spiritual aspects and how they come together in a coherent whole” (Sheikh, “Worldviews and Conflict Analysis”, 2022). Although most worldview analysis is done ethnographically (from the bottom-up with qualitative data), worldviews can also be investigated statistically (from the top-down with quantitative data). While cross-tabulation of survey data reveals the “social, ethical, political, and spiritual aspects” of worldviews by comparing attitudes with demographic variables, a correlation mapping allows for exploration of how these aspects cohere by showing which attitudes trend together and which work against each other. A statistical approach alone cannot tell us why these correlations happen, but it can show how prevalent these correlations are. In particular, this analysis reveals how Americans broadly demarcate civic virtue and civic vice. For advocates of a reinvigorated civic life in the United States, such awareness might mean the difference between pouring effort into designing an idiosyncratic project that only makes an impact in a particular community, and a robust scalable initiative capable of easing a major tension in American civic life.
Aspects of the CLPP data that provide additional context for this analysis can be found in the appendix.
CLPP respondents were asked to indicate whether they have a positive or negative impression of seven of the twenty-one civic terms surveyed. In this analysis, “Negative” responses are scored as −1, “Positive” responses are scored as +1, and neutral responses such as “Not familiar with this word” or “Neither positive nor negative” are scored as 0.
Many respondents ranked words similarly and/or positively. This created an interpretive challenge: for each pairing, did the comparative score accurately reflect the respondent’s relative view for each word? Was the comparative score skewed by acquiescence bias? What about respondents who systematically gave a positive association for each word because they generally have a positive outlook on life, rather than a specific sentiment about that particular word?
This challenge was addressed through the standardization of data for each respondent. To illustrate, consider respondent #4142’s answers, as seen in the below table:
Notice that each of the 5 positive impressions are given absolute scores of +1, but standardized scores of only +0.784. “Patriotism” only has an absolute score of −1, but because it is uniquely negative for this respondent, the standardized score is a whopping −1.961. Notice that even the "neutral" impressions have negative standardized scores of −0.589: this is so that the standardized scores sum to zero.
The absolute score for each word is independent of how a respondent scored the other six words, while the standardized score is overly influenced by those other scores. To achieve a measure with balanced levels of influence, a principal component was calculated using both the correlations of the absolute and standardized scores. This principal component was used to generate the following correlation matrix.
The correlation matrix of course omits the perfect correlations between the same term, resulting in the diagonal line of white squares. The other missing correlations point towards the second interpretive challenge, which arises from the structure of the survey. Respondents were randomly assigned one of three non-overlapping sets of terms from the full list. Instead of having 210 possible pairings, the non-overlapping structure of the random assignment means that there were only 63 possible pairings to investigate.
Using the principal component as a least biased measure of ranked correlation, the antagonistic pairs were identified as the 10 most negatively correlated pairings distributed among the term sets. Similarly, the 10 complementary pairs were identified as the most positively correlated pairings. Because Term Set A had the most variance, an additional complementary pair and antagonistic pair were included from Term Set A. A table showing these pairings is available to view in the Results section below.
After identifying the 10 antagonistic pairs, the next step was to identify “bridge populations”. This was done by aggregating and applying weight vectors for respondents whose sum of standardized scores for each word in the pair is greater than average. The threshold was set at 1.33 standard deviations greater than the mean for all standardized sum pairings, which resulted in a subset of the sample representing those who held the top 10% of positive associations for a pairing seen as antagonistic by the overall sample.
To make meaning out of these “bridge populations”, the demographic profile of each bridge population was compared to the overall sample. Generally, gaps between the two demographic proportions that came in at least ±1.96 standard deviations (i.e. no more than a 5% chance of occurring randomly) were seen as noteworthy. Numerical variables (such as income and age) where gaps in adjacent values trended similarly were given additional attention.
Antagonistic and Complementary Pairs
Out of 63 observable civic word pairings in the CLPP data, the table below shows the 10 most representative examples of “antagonistic pairs” (i.e. negatively correlated) and the 10 most representative “complementary pairs” (i.e. positively correlated). An antagonistic pair represents a pair of words that, if one word is rated positively (at best, or neutral at worst) by a survey respondent, then that respondent likely rates the other word negatively (at worst, or neutral at best), and vice versa.
A cursory glance at the antagonistic pairs shows that some reflect political cleavages. One example is the “Patriotism & Racial Equity” antagonistic pair, where Republicans are nearly twice (1.8×) as likely to have a positive association with the word patriotism relative to Democrats, whereas Democrats are more likely (1.6×) to have a positive association with the term racial equity. Other pairings seem to reflect philosophical paradoxes. For example, “Privilege & Common Good” seemingly contradict in that a special advantage available only to a particular person or group (privilege) is opposed to the benefit or interests of all people and groups (common good). Arguably, it is possible to resolve this paradox: the strategic granting of privileges, such as legal exemptions for elected officials to carry out their duties in the public interest, or tax credits for clergy to sustain their ministry that weaves together the social fabric, are meaningful contributions to the common good. But it is understandable that some Americans might still emphasize the importance of the common good over the necessity of granting privileges (and vice versa).
Complementary pairs are those where a positively rated word increases the likelihood that its paired word is likely to be rated relatively positively. Some complementary pairs represent opposite ends of the political spectrum. Take for example the “Activism & Diversity” pairing, where Democrats are (relative to Republicans) 2.1× more likely to view activism positively and 1.6× more likely to view diversity positively. Other complementary pairings, such as “Civic Health & Civil Society” or “Common Good & Common Ground” are logically consistent insofar as a robust civil society is a sign of civic health, and striving towards the common good requires at least some common ground — if only to agree on what that common good consists of.
The complementary pair “Liberty & Justice” is particularly interesting in that neither represents a political extreme, nor are they necessarily logically consistent. One inconsistency arises if someone understands liberty as “freedom” and justice as “equality,” be it equality under the law or equal economic opportunity. A society that enforces strict equality will have to compromise on some freedoms, and a society that promotes absolute freedom will have some inequality. However, Americans have a deep-rooted sense that both liberty and justice have more expansive definitions that are mutually positive and interdependent; a pairing that is reinforced with every recitation of the Pledge of Allegiance: “…with liberty and justice for all.” The positive correlation between liberty and justice suggests that, in addition to political pairs and logical pairs, there exists a third pairing type: the socially-constructed dyad that entangles two independent and opposed concepts so closely together that it becomes increasingly impossible to imagine one flourishing without the other.
The existence of these socially-constructed dyads motivate the next part of the analysis, exploring populations that can serve as bridges for the antagonistic pairs. These bridge populations represent centers of creativity where socially-constructed dyads are already in the process of being formed. To cultivate a civic engagement culture where each of the 21 CLPP terms are viewed positively, engagement with these bridge populations can help de-antagonize pairings by promoting their unique perspective among the wider population.
Each “bridge population” below represents roughly the top 10% of survey respondents who, contrary to the general population, view both words in an “antagonistic pair” positively. Using demographic information in the CLPP data, it is possible to create statistical portraits of these bridge populations. Some bridge populations have demographic profiles that are similar to the overall population, while others are wildly divergent. An interpretation of the bridge population is provided for each antagonistic pair.
Some helpful tips as you review these bridge populations:
- There can be multiple subgroups in a bridge population independent of each other. For example, if people in a bridge population are both more likely to be “Muslim” and more likely to be “Rural”, that does not necessarily mean people in the group are more likely to be “Rural Muslim.”
- A demographic might characterize a bridge population, but that does not mean the average person belonging to that demographic is part of the bridge population. For example, it might be more likely a person who has relatively high perceptions of the words “Civic Health” and “Belonging” will identify as a “Republican”, but that does not mean the average “Republican” has relatively high perceptions of the words “Civic Health” and “Belonging.”
1Patriotism & Racial Equity
The pairing of patriotism and racial equity was the most negatively correlated in the entire CLPP dataset, yet, race is not driving the wedge between patriotism and racial equity; the distribution of racial identifications are nearly indistinguishable between the bridge and overall population. As discussed earlier, partisanship contributes to the antagonism, with Republicans favoring patriotism and Democrats favoring racial equity. Therefore, it is to be expected that the bridge population for these terms was significantly more likely to be “not sure” about their party identification (19% of the bridge population, compared to 8% overall).
They also were more likely to be younger (42% were 18-34 years old, compared to 29%), make less money (59% made less than $50,000, compared to 48%), and be less educated (46% had a high school education or less and 11% had only a Bachelors, compared to 33% and 20% overall respectively). They were less likely to identify as living in the suburbs (30%, compared to 44%); and more likely to respond they were "not sure" if they lived in rural areas or urban centers (13%, compared to 3%).
The CLPP data suggests that a vision of civic life that simultaneously celebrates patriotism and racial equity is highly elusive. The bridge population respondents who ranked both words highly were significantly more likely to respond “not sure” for six of the seven categories. This high degree of self-uncertainty and negated identity makes it difficult to positively describe this bridge population.
A hypothesis worth further exploration is that this bridge population overlaps with the millennials who have been moving to the "exurbs" — communities that are too remote to be called suburban, but too dense to be called rural. This would explain the high amount of uncertainty regarding their geographic location within the rural-suburban-urban trichotomy, and is consistent with the findings that this bridge population is younger, poorer, and less educated, yet representative of the racial makeup of the United States.
2 Patriotism & Civic Engagement
The bridge population for patriotism and civic engagement is less likely to self-identify as rural (13%, compared to 22%), Democrat (26%, compared to 37%) or with any party affiliation (17% “Not sure”, compared to 7%). They are more likely to self-identify as Buddhist (4%, compared to 1%) and as Asian-American or Pacific Islander (9%, compared to 5%).
3 Racial Equity & Liberty
Similar to the “Patriotism & Racial Equity” bridge population, the bridge population for racial equity and liberty is less suburban (31%, compared to 44%), more likely to be unsure about their urbanicity (13%, compare to 3%), and younger, with 44% aged 18–34 (compared to 29% overall). However, the similarities end there: the racial equity and liberty bridge population is more likely to be Hispanic (30%, compared to 20%) and less likely to be White (29%, compared to 38%). Ideologically, they are significantly more likely to identify as “very liberal” (25%, compared to 11%) and have a slight partisan skew towards the Democratic Party.
4 Civic Engagement & Liberty
The bridge population for civic engagement and liberty is full of paradoxes. These respondents are significantly more likely to be younger (45% are 18–34 years old, compared to 29% overall; only 13% are 65+, compared to 22%), but they also skew towards higher levels of income (17% earn $100,000-$149,999 annually, compared to 11%, and 6% earn $250,000+ annually, compared to 2%) and are significantly more likely to have a postgraduate degree (20%, compare to 13%). They are more likely to self-identify as “very conservative” (22%, compared to 12.5%) but also as a Democrat (46%, compared to 37%). They are also twice as likely to identify as a god-denying Atheist (8%, compared to 4%) yet the Protestants were significantly more likely to identify as a God-proclaiming “evangelical or born-again Christian” (70%, compared to 54%). The only statistical pattern that makes intuitive sense is that they are more likely to identify as Black (43%, compared to 32%) and also less likely to identify as White (22%, compared to 38%).
How do we explain these inconsistencies? One plausible explanation is that this bridge population tends towards strong political opinions. Notably, 30% of this group define civic engagement strictly as “influencing government functions” (compared to 17% overall) while only 12% define civic engagement strictly as “making communities better” (compared to 24%). What may underlie this bridge population is a fierce debate about the meaning of liberty: what it is and who it is for. A potential way of engaging this bridge population is through asking them to elucidate their various conceptions of liberty, perhaps in a panel, podcast, or documentary. The final product would illustrate how many of our fiercest political debates have at their root different understandings — and occasional misunderstandings — over what constitutes our constitutional rights and freedoms.
5 Privilege & Common Ground
When considering “Privilege & Common Ground” and the next antagonistic pair, it is worth highlighting that privilege is the only term in the CLPP dataset that respondents gave more negative associations (37%) than positive (21%). The only pairing for privilege that was more positively correlated within its term set was “Privilege & Civic Health”, yet it was only slightly above average and too small a margin to qualify as a complementary pair.
The bridge population for privilege and common ground is eclectic. They are more likely be urban (41%, compared to 31%), earn less than $50,000 annually (60%, compared to 48%), be “somewhat conservative” (23%, compared to 16%), belong to the Latter Day Saint Church (5%, compared to 1% overall), and identify as Black (40%, compared to 32%) or Middle Eastern (2%, compared to 0.7%). They are less likely to be “very liberal” (5%, compared to 11%) or Catholic (17%, compared to 23%).
6 Privilege & Common Good
The bridge population for privilege and common good is not particularly distinctive. They are more likely to be “not sure” about their religious identity (11%, compared to 4%) or Hindu (2%, compared to 0.6%), but are also less likely to be agnostic (0.7%, compared to 4.3%) which may be interchangeable with “not sure”. They skew towards being less suburban (34%, compared to 44%) and less educated (41% attained high school or less, compared with 33% overall; 7% have at least a master’s degree compared to 13% overall). When compared to the general population, they skew more Republican (26%, compared to 20%) and less Democrat (29%, compared to 37%).
7 Social Justice & Common Ground
There is an apparent political bias in the bridge population for social justice and common ground. 49% identify as Democrat and 24% identify as “somewhat liberal” (compared with 37% and 16% overall), while only 13% identify as Republican and 4% identify as “very conservative” (compared with 20% and 13% overall). Furthermore, there is a slight skew towards people of Middle Eastern descent (2.5%, compared with 0.7%) and residents of urban environments (41%, compared with 31%). There is a slight skew away from people of Hispanic origin (14%, compared with 20%).
8 Activism & Civility
There is a political divide undergirding this antagonistic pair, so it is unsurprising that the population with positive perceptions of activism and civility skews away from being Republican (13%, compared to 20%) or Democrat (29%, compared to 37%), but rather are more likely to be nonpartisan or independent with a partisan lean (51%, compared to 35%). This group is significantly younger, with 41% being 18–34 years old (compared to 29% overall), and are more likely to identify as multiracial (2.8%, compared to 0.9%). Furthermore, they appear less likely to be “upper middle class”, defined here as making between $100,000 and $149,999 (5%, compared to 11%).
9 Pluralism & Diversity
This bridge population has a unique gender skew, with 57% of the bridge population identifying as male and only 42% identifying as female. This bridge population also skews significantly younger (49% are aged 18–34 years old, compared to 29% overall; only 7% are aged 65+, compared to 22%). Furthermore, they are more likely to identify as Orthodox Christian (4%, compared to 1%), Hispanic (30%, compared to 20%), Middle Eastern (2.3%, compared with 0.7%), “somewhat conservative” (23%, compared to 16%), or have postgraduate education (19%, compared to 13%). They are less likely to be White (26%, compared to 38%).
This bridge population might represent a unique opportunity to engage younger, center-right men of various ethnic backgrounds who are sometimes seen as “missing” from a civil society that skews older, female, White, and politically left. Notably, this bridge population is twice as likely than average to say that civic engagement is strictly about “influencing government decisions” (34%, compared to 17%). What might it look like to engage these men around an expanded vision of civic engagement that includes making communities better and centers ideas of pluralism and diversity in the process?
10 Pluralism & Unity
The Founding Fathers would be appalled that pluralism and unity make an antagonistic pair. The traditional motto of the United States, “E pluribus unum” (or “out of many, one”), was a rallying cry in the early days of the republic and remains alive through our Great Seal and on our currency. Yet pluralism and unity are in the bottom-fifth of all pairs the data allows us to correlate.
This bridge population has some similarities to the previous “Pluralism & Diversity” bridge population. They also are more likely to be Hispanic (30%, compared to 20% overall), ideologically more likely to be “somewhat conservative” (26% versus 16%), lean Republican (18% compared to 9%) and less likely to be “very liberal” (3%, compared to 11%). They also tend to see civic engagement as strictly being about influencing how government functions (29%, compare with 17%).
There are plenty of unique features to this group as well. Although there is no age or education skew, their income is more likely to be in the $50,000 — $149,999 range (61%, compared to 43%). They also are more likely to be evangelical Christian (14.1%, compared to 6.7%; exclusive of other Protestants), multiracial (3.4%, compared to 0.9%) and gender non-binary (2.4%, compared to 0.5%). They are more likely to perceive themselves as suburban (54%, compared to 44%) and less likely to self-perceive as urban (17%, compared to 31%).
The most distinctive feature of this bridge population is the likelihood that they define democracy as “a form of representation in government that ensures the voices of those least likely to have access to political power are prioritized in policies and decisions” (37%, compared to 20%). This population appears to desire a more systematic representation of minorities within the United States (beyond, of course, the accommodations given to low population states through the Senate and the Electoral College). Albeit not politically neutral, a meaningful way of engaging this population that bridges pluralism and unity could be through the electoral reform campaigns to replace our “winner take all” system with proportional representation that empowers minority political groups through official representation in the political process.
After discussing “Pluralism & Diversity” followed by “Pluralism & Unity”, it is worth noting that the “Diversity & Unity” pairing was just shy of making the cut of being a complementary pair. Given that diversity and unity could logically be construed as opposite objectives akin to “divergence” and “uniformity”, the fact that they are instead nearly complementary implies a socially-constructed dyad synthesizing the two. Advocates of pluralism should be aware that as they advance pluralistic visions of civic life, they need to explain why “pluralism” is needed in addition to “diversity” and “unity” without subverting the cultural balance that has been struck to situate “diversity” and “unity” as twin goals.
No terms were closely correlated, therefore there is no reason to recommend eliminating any term in future survey iterations on the basis of being highly collinear and redundant. That said, the structure of the initial CLPP survey only allows for 63 out of the 210 possible pairings to be investigated. A follow-up to this survey would benefit from being structured to get statistically significant samples for as many of the remaining 147 pairings as possible, or better yet all 210 pairings simultaneously. Furthermore, reaching a critical threshold of data on approximately two-thirds of all possible pairings would open up the possibility for more advanced statistical techniques, such as unsupervised cluster analysis.
This analysis was designed to practically inform the future work of civic engagement researchers and professionals alike. Below are some general suggestions to catalyze action.
For researchers, these “bridge populations” are relatively small sub-samples with a tendency to fall off the tail end of the bell curve. While larger samples would certainly be welcome, these liminal groups are also a sweet spot of ethnographers, who could conceivably design a project that seeks people who regard any of the antagonistic pairings relatively highly. While we can reasonably speculate why certain civic terms tend towards negative correlations more than average, face-to-face interviews where exemplars share the inner logic to their civic life worldview may offer more illuminating results.
The opportunities to leverage the insights shared here are numerous for practitioners in the field of civic engagement. For funders in particular, each bridge population represents an investment opportunity to build a more robust culture of civic life and an opportunity to combat the toxic polarization afflicting the United States. For organizers, each bridge population represents a group that otherwise would be overlooked during outreach and whose gathering might be an inherently newsworthy event. Artists and journalists alike could embed within these bridge populations and discover a counter-cultural narrative that would nonetheless be deeply resonant among different parts of the public.
The takeaway from this analysis is not simply an imperative to engage with the various bridge populations. Guiding this analysis was an assumption that there exist widely shared worldview commitments that govern how Americans delineate between civic virtue and civic vice. Promotion of one ideal may come at the cost of another ideal — at least if certain assumptions causing tension between the two ideals remain unchallenged. Advocates of any particular ideal of civic life should be sensitive to these inner logics, deconstructing false antagonisms when prudent and imagining new synergies whenever possible. Working together and with data-driven awareness, it is possible to achieve new visions of civic participation within the United States and reinvigorate civic life for the 21st century.
Thanks to PACE for making this data and funding available through the CLPP Creative Mini-Grant program. Additional thanks to Anita Joshi for lending her sharp proofreading and editorial guidance. The views expressed here are strictly my own, and any mistakes remain mine.
In addition to attitudes regarding the 21 terms, the CLPP data contains both demographic variables and other qualitative features that were used in this analysis to describe bridge populations for the antagonistic pairs. Income was asked on a per household basis. Race and ethnicity were consolidated under a single “Race” multiple choice question, which allowed for a single response of “Asian American & Pacific Islander”, “Black”, “Hispanic”, “Middle Eastern”, “Multiracial”, “Native American”, “White”, and “Not Sure”. For religion, both “Protestant” and “Evangelical Christian” were among the many options; respondents who answered “Protestant” were given a follow-up question asking if they considered themselves “an evangelical or born-again Christian”. Half of respondents were also asked to give a definition of “democracy” and another half of respondents were asked to give a definition of “civic engagement”. The possible answers and corresponding percentages are provided in the table below.