Assessing the ideological extremism of American party activists

Introduction

A cottage industry of research finds that the American Congress has become substantially more polarized by party and ideology during the past three decades (see, e.g., Brewer and Stonecash, 2009: 28–32; Lee, 2015; Poole, 2015). This increase in polarization has had profound effects, ranging from more combative election campaigns and frequent congressional paralysis (Abramowitz, 2010) to increased party leadership power in Congress (Aldrich and Rohde, 2009).

To many analysts of American politics, party activists—those who participate in a variety of the standard party and campaign activities long examined by researchers—are the proximate cause of this polarization (e.g., Aldrich, 2011, Chapter 6; Layman et al., 2010; Miller and Schofield, 2008). If Democratic activists have moved toward the extreme left of their party’s distribution and Republican activists toward the extreme right, then they may pressure candidates and elected officials to move toward these poles and take more extreme stands on issues. Yet with one notable exception (Layman et al., 2010), there is little evidence for this argument beyond findings that, empirically, activists are usually more extreme in their policy views and ideological self-placement than candidates are (Enos and Hersh, 2015; Layman, 2014; Hershey et al., 2012; Francia et al., 2003, but see La Raja and Wiltse, 2012). Layman et al. (2010: 332–335) find that with respect to abortion attitudes in particular, party polarization among party activists occurred earlier in time than it did in either the parties in the electorate or the parties in Congress. They conclude, “It appears that the growth of religious and cultural differences among party activists helped to spur increases in religious and cultural polarization among the parties’ elected officials as well as the parties’ coalitions in the mass electorate” (Layman, 2014: 216).

Hershey et al. (2012) suggest that activists might cause elected officials to become more polarized for three reasons: resources, contact, and participation in nominations. In providing resources or information, activists may be incentivizing ideological extremity among politicians by subsidizing otherwise costly political activity. That is, by offering scarce resources to political candidates on the condition that the candidate supports a particular issue, ideologically extreme activists may be able to induce the candidates to adopt more extreme positions. Second, there may be an interactive effect: the act of donating money may allow an activist greater access to contact, and thus influence, their elected official (Kalla and Broockman, 2016).

Third, because party and policy activists are a large proportion of the voters in US primaries, and especially of caucus participants, they have a bigger say than other voters do in the selection of their party’s candidates. To win nomination, then, candidates might take non-centrist stands in order to attract activists’ support, particularly in low-turnout elections (Boatright, 2013; Cohen et al., 2008).

The problem with this plausible and widely repeated assumption—that extremist activists cause elected officials to polarize—is that very little empirical evidence has been presented to support it. First, there are reasons to question the conclusion that activists are in fact more ideologically extreme in the United States than are other party identifiers. Although many studies report this finding (see Abramowitz and Saunders, 2005: 6–8; Stone and Pietryka, 2010: 339–341), Broockman (2016: 181–183), for instance, shows that measuring individuals’ attitudes with indices encompassing several policy areas is likely to classify as “extreme” those respondents who have highly consistent attitudes, regardless of their degree of extremism, while misclassifying others as moderates. He argues that activists have often been characterized as more extremist when in reality their attitudes are simply more internally consistent. Carmines et al. (2013) show that although those with consistent liberal or consistent conservative views are more likely than other voters to participate in campaign-related activities, those with atypical combinations of views (e.g., libertarians, who are left-wing on social issues and right-wing on economics) are often mischaracterized as moderates in studies that assess extremism using only the traditional left-right ideological divide.

Others question this finding for different reasons. Bartels (2016: 155–159), for instance, finds that at times, American presidential candidates take more extreme stands than their core supporters do, though he defines “core supporters” more broadly than activists alone. Moreover, the size of the presumed “extremism gap” between activists and other partisans has varied across time (Miller and Jennings, 1986; Saunders and Abramowitz, 2004). To complicate matters further, much of the existing research uses conflicting definitions of party activism. As Bartels (2016: 148) points out, “we have no real scholarly consensus regarding the political views of party activists in contemporary America, much less a persuasive analysis of how they matter.” The polarization research often refers to activists as though they were an undifferentiated group. Yet, researchers have defined “party activists” in many different ways, and there is no single, widely accepted definition of the term. It could include many different groups of people: those who give money to candidates or political groups, attend political conventions or events, go door-to-door for political causes, display political signs, or initiate political discussions with others.

We seek to answer Bartels’ call and address this broad, diverse literature by explicitly examining the relationship between ideological characteristics and political activism in American politics. We wish to accomplish two goals. First, we intend to determine if it is reasonable to treat those who donate money, display yard signs, try to influence others’ vote choice, as one ideologically homogeneous group. Second, we hope to push the literature toward a more theoretically motivated, empirically grounded, manner of characterizing activists by the content of their action.

Why we should distinguish among types of activism

Perhaps reflecting the lack of scholarly consensus on how to treat activists, there is no consistent approach on how to categorize them. ¹ Some (e.g., Jackson and Green, 2010; Layman et al., 2010; McClosky et al., 1960; Miller and Jennings, 1986) focus on presidential convention attendees, while others (e.g., Bruzios, 1990; Carmines and Woods, 2002) look at survey respondents who report participating in various political activities. Still others focus only on particular types of participation, including people who report trying to influence others’ votes (Baumer and Gold, 2010), canvass for a specific candidate (Enos and Hersh, 2015), or give money to campaigns (Aldrich, 2011: 194; Francia et al., 2003: 60–63; Koger et al., 2009; see also Barber, 2016).

Each approach has its own merits, but they each have important limitations as well. Modern American presidential convention delegates often have only a symbolic role in the nominations, and include members of Congress and other elected officials, and may not represent the whole population of party activists. ² The approach of developing a dichotomous indicator of “campaign activists” from survey responses may conflate a host of different factors into one class. Lastly, the piecemeal approach can make it difficult to put the findings about campaign donors or canvassers into a broader context of political activism.

Much of the research on American political activists, then, treats the various individual forms of political participation as essentially interchangeable. Yet, many of the various aspects of activism—from donating money to canvassing—require different levels of engagement, different tolerance levels for conflict, and different resources. Verba et al. show that while voting is driven by political interest, financially supporting a campaign is primarily driven by income, and that canvassing is driven by church involvement and free time. They conclude (1995: 368) that different modes of political activism are not interchangeable, either in their origins or their consequences. ³

If different types of political activities attract different types of participants and convey different messages to American public officials, then they may differ in their levels of polarization. Those who take part in some types of activities may hold more extreme views than others. Polarization among, say, campaign activists could affect candidates differently from polarization among financial contributors. ⁴

Research questions/hypotheses

We start by asking whether the people who engage in these different forms of participation are similar to one another in their socioeconomic characteristics and their political views. In short, is it justifiable to generalize about “political activists”? To evaluate the extent to which activists are similar to one another, we predict forms of activism ⁵ as a function of demographic and ideological positions. If it is appropriate to generalize about the ideological positions of political activists, we should see commonalities in how ideological extremism predicts activity.

On the other hand, if different types of activism engage different types of people, we similarly expect that extremism will also vary by type of activity.

Second, we test whether the three possible pathways of activist influence on elected officials—direct contact, resources, and nominations—are more likely to be traveled by extremists. Here we operationalize extremism as does much other research on polarization in American politics (e.g., Abramowitz, 2010; Layman et al., 2010; Thomsen, 2014): taking non-centrist issue or ideological positions on survey questions—positions at the poles of the options available —rather than positions close to the center of those options. Are the types of activities that require a lot of personal contact (e.g., trying to influence others’ votes vs. clicking on PayPal) more likely to be associated with extremism in political attitudes than are the types of activities that require the use of resources such as time or money, or activities involving nominations? We expect that some forms of activism will be more appealing to extremists than others. For example, extremists may be particularly motivated by the purposive benefits of working on a campaign, while they may be less apt to participate in private forms of activism such as voting or donating money.

Data and measures

To answer this first set of questions—whether it is reasonable to generalize about American party activists—we use the ANES cumulative file for the years 2000–2012, in order to combine the need for large numbers of respondents with a desire to tap political attitudes during the current polarized period. ⁶ While the demographic variable construction is straightforward, the construction of an ideological extremism variable is less so. Here, we opt for an issue-based measure of ideology rather than the respondent’s self-reported (liberal/conservative) ideology for several reasons. First, the greater American cultural approval of the terms “conservative” (Iyengar et al., 2012: 414) and “moderate” than of the term “liberal” may skew responses to the self-reported measure. Respondents’ understanding of the meaning of the terms “liberal” and “conservative” may not always coincide with researchers’ assumptions. Second, the general measure of ideology is less able to predict an individual’s responses to a particular issue than is a measure referring specifically to that issue (Zaller, 1992: 27). ⁷ In fact, some argue (Mason, 2015) that an individual’s ideological self-description is a social identity that should be seen as separate from his or her views on issues. Third, extremist attitudes on one issue could affect behavior differently from extremist attitudes on a different issue. For example, activists’ level of extremism on abortion may be more closely associated with their level and type of participation than their level of extremism on social welfare. In turn, this may affect candidates’ behavior differently, and would be wholly missed with the use of a single, general self-reported measure of a respondent’s ideology.

To construct our issue-based measures of ideology, we use confirmatory factor analyses ⁸ of the ANES data to estimate measures of ideology on a single ideological dimension. ⁹ The measure is constructed using issues ranging from government’s provision of social services and health insurance to abortion to affirmative action or government aid (Specifics on the construction of this measure are included in the Online Appendix.).

The values of these measures range from −1 (liberal) to +1 (conservative), with a mean of 0. To construct the measure of ideological extremism, we fold the scores at the median value. Accordingly, lower values of the ideological extremism variable are more moderate, while higher values are more extreme (either liberal or conservative).

Findings: Are different types of activities interchangeable?

In the following analysis, we analyze the various forms of political activity—voting (“vote”), attempting to influence others (“influence”), attending a meeting or rally (“rally”), working for candidate or party during campaign (“work”), displaying a button or sticker (“button”), and donating money to a party or candidate (“money”)—that are consistently asked in the ANES over time, and are most frequently utilized by the literature on American party activists. ¹⁰ For each dependent variable, we estimate a series of logistic regressions as a function of ideological extremism, income, education, gender, partisan affiliation, race as well as year controls (not reported). ¹¹

The results of the logistic regressions are presented above in Figure 1. Each of the panels reports the results from a distinct regression, so that the first panel shows results from predicting the likelihood of voting, the second panel shows results from predicting the likelihood of attempting to influence someone's vote decision, and so on. ¹² Within each panel, the coefficient estimate and 95% confidence intervals are presented by the black circles and black lines, respectively. The vertical red line in each panel allows us to visualize whether the coefficient is statistically significant or not. The further the point estimate is from the red line, the stronger the relationship between the form of activity and the variable of interest, controlling for all other variables in the regression.

OPEN IN VIEWER

Figure 1.

Ideological and demographic models of political activism.

As expected, we find that party identifiers and those with more education are more likely to participate—in all forms—than “pure” independents or the less educated. On the other hand, income, gender, and race each have mixed effects on participation: in some forms of participation, race, gender, or income differentiate between activists and non-activists, but in other forms, they do not. ¹³

With respect to extremist attitudes, we find that ideological extremism is associated with a greater likelihood of participation in each form of political activism (see Figure 2). ¹⁴ Although the size of the magnitude differs, extremism is consistently one of the most important factors in predicting when an individual will engage in politics. While the logit results presented above show that ideological extremism has a significant relationship with each activity, we are unable to tell just how much extremism matters (Long and Freese, 2015). Accordingly, we present predicted probabilities from each model, showing the effects of ideological extremism and education on likelihood of political activity. We focus on ideological extremism and education here because they tend to be the most influential variables in explaining each form of activity.

OPEN IN VIEWER

Figure 2.

Predicted probabilities of political activity by ideological extremism.

The predicted probability of each form of activity is shown on the y-axis, across the range of ideological extremism on the x-axis from 0 (Moderate) to 1 (Extremist).

In general, the figure demonstrates that while extremism is statistically significant in all cases, from a substantive standpoint, the forms of activism are not uniformly populated by ideological extremists. In the cases of voting, influencing others’ vote choices, displaying campaign paraphernalia, or donating money to a campaign, we see that extremists are generally more likely to participate than moderates are, and that these differences are substantively meaningful. An extremist is 21 percentage points more likely to report influencing someone else’s’ vote than a moderate is, and is 16 percentage points more likely to display a yard sign, all else equal.

By contrast, ideological extremism is less strongly related to the individual’s likelihood of attending a political meeting or rally, or likelihood of working on a campaign. Extremists are only about four percentage points more likely to participate in either activity than moderates are. Thus, we find that while ideological extremism is significantly and positively associated with activity, the effect is quite small, and it is misleading to suppose that rally attendees and campaign workers are likely to be extremists.

So, what distinguishes a campaign worker from a campaign donor? Quite a bit. Donors tend to be more ideologically extreme, and are well educated and have high incomes, relative to non-donors. By contrast, campaign workers are far less likely to hold extreme views and have lower levels of educational attainment and income, relative to non-campaign workers.

In short, not all forms of political activity are strongly associated with extremism. Rather, some methods of activity—such as influencing other’s vote choice—are strongly correlated with issue extremism, while others—such as working for a campaign—are only modestly related to extremism. Moreover, extremism is not associated with participation equally across demographic categories: individuals with a college education are generally likely to vote irrespective of their ideological position. Not so among those without a high school diploma: less educated extremists are considerably more likely to vote than less educated moderates are. Thus, we find that generalizing about “political activism” by grouping all of the commonly investigated types of political activities into one single dimension is not only theoretically inappropriate, due to the differences among types; it is also empirically misleading.

Here, we pause briefly to explore the relationship between extremism and activity when utilizing a multidimensional, rather than one-dimensional, measure of extremism. Given the multidimensionality of mass policy preferences (Carmines et al., 2013; Layman and Carsey, 2002), we explore whether extremism in general is linked with political activity, or if activism is related to only one subset of extreme issue positions. That is, we investigate whether American cultural extremists (those concerned about abortion access, gay marriage, and the like) are more likely to participate in certain activities than are social extremists (those concerned with the role of the government in creating a social safety net).

To do so, we use confirmatory factor analyses of the ANES to estimate measures of ideology on social, cultural, and racial dimensions. ¹⁵ Issues loading on the social extremism scale include government’s provision of social services, health insurance, social security, welfare, and child care. Issues loading on the cultural scale include abortion, school prayer, gay marriage, women’s roles in society, and similar questions. Those on the racial scale include affirmative action; government aid to Blacks; and whether the pace of civil rights has been too fast.

We execute a similar analysis as in the unidimensional measure of ideology approach outlined above, though substituting these multidimensional measures of ideological extremism for the unidimensional measure used previously. All other aspects of the model and estimation procedure are the same. The results from these logit models are presented below in Figure 3.

OPEN IN VIEWER

Figure 3.

Ideological and demographic models of political activism: Multidimensional measures of extremism.

These findings offer modest support for the assertion that individuals with extreme views on different issues are drawn to different forms of political activity. For example, campaign donors appear less likely to hold extreme positions on racial issues (affirmative action, aid to Blacks, etc.) or cultural issues (same-sex marriage, school prayer, etc.) and more inclined to hold extreme positions on government spending programs and the social safety net. As such, we should expect to see candidates staking out more ideologically extreme positions on these issues, relative to those on school prayer or government aid to Blacks.

Table 1, below, shows the average marginal effect of shifts in ideological extremism for each form of political activism. These results are calculated from the same models presented in Figure 3, holding continuous variables at their sample means and categorical values at their sample medians. ¹⁶ Column 1 (“Vote”) first reports the baseline probability that a respondent will report voting. It then shows the change in probability of voting as we increase each type of ideology from a moderate value (0) to an extreme value (1), while holding all other factors, and the other forms of ideology, constant.

OPEN IN VIEWER

Table 1.

Marginal effects of shifts in ideological extremism on political activism.

Prediction			Vote	Influence	Rally	Work	Button	Money
Baseline probability			77%	42%	6%	3%	16%	12%
Marginal effect of shift in
	Social extremism		12%	21%	3%	4%	17%	20%
	(95% CI)		(7%, 17%)	(13%, 28%)	(−2%, 7%)	(0.1%, 8%)	(10%, 24%)	(13%, 27%)
	Racial extremism		5%	10%	4%	4%	13%	5%
	(95% CI)		(−0.2%, 9%)	(4%, 16%)	(0.1%, 8%)	(1%, 7%)	(8%, 19%)	(1%, 9%)
	Cultural extremism		6%	13%	1%	3%	6%	8%
	(95% CI)		(1%, 11%)	(7%, 20%)	(−2%, 5%)	(0.1%, 6%)	(1%, 11%)	(3%, 13%)

Note: Table entries are average marginal effects calculations of moving from moderate to extreme positions on social, racial, and cultural issues with 95% confidence intervals in parentheses. These are calculated from the models presented in Figure 3. Statistically significant coefficients are depicted in bold. Marginal effects calculated with continuous variable at sample means, and categorical variables at sample medians.

Here, we see that social (cultural) extremists are 12% (6%) more likely to report voting than moderates are, all else equal. On the other hand, racial extremists are no more likely to report voting than those who hold moderate views on racial policy. ¹⁷ Put another way, those with extreme positions on same-sex marriage or abortion are more likely to vote than are those with extreme positions on affirmative action or aid to Blacks.

Moving to Column 2 (“Influence”), an extremist on social policies is 21 percentage points more likely to report trying to influence someone’s vote than a moderate is. To put this in context, the baseline probability for the average respondent is 42%, which means that a 21 point shift in social extremism increases the likelihood of attempting to influencing someone else’s vote by 50 percentage points. Racial and cultural extremists are somewhat less likely (10% and 13%, respectively) to report trying to influence another person’s vote than moderates are.

Column 3 shows that while extremism on cultural or social issues does not have a significant relationship with attending a political rally or meeting, racial extremism does. Moreover, extremism on racial issues makes a substantive difference in predicting one’s likelihood of attending a rally. Although the marginal effect of extremism is 4%, given the baseline probability is 6%, the effect of extremism is to increase the probability of participating by two-thirds.

Columns 4–6 show that extremists of all varieties are more likely to work for a campaign, wear campaign buttons, or donate money than moderates are. Column 4 shows that while the baseline probability of working on a campaign is low, ideological extremism doubles one’s propensity to do so. The feature that stands out in Columns 5 and 6 is the impact of social extremism. The marginal effect of social extremism is slightly larger than the baseline probability of displaying a campaign button or yard sign, and considerably larger than the probability of donating money. Moreover, marginal effects of social extremism are much larger than they are for other issue areas. Those with strong views on abortion or affirmative action are more likely to donate money, but not as much as those with extreme views on issues such as Social Security or government provision of health care.

Thus, we find that different forms of activism are not interchangeable: they are differently related to ideological extremism, and even the activities most closely associated with extremism are linked with different sets of issues. Voting is associated with social and cultural extremism more than it is with racial extremism, while rally attendance is linked only with racial extremism. Social extremism has a dramatic effect on one’s propensity to donate money or display campaign materials—a magnitude that is not matched by either cultural or racial extremism.

Assessing the effect of extremism on the mechanisms of activist influence

Next, we return to the set of arguments cited early in this article: that activists’ role in the current political polarization could result from the resources they have to offer prospective candidates (such as the measures of individuals’ donations of time, money, and campaign effort), from their contact with candidates (which gives them the chance to convey information, such as contacting a House member), and/or from their central role in nominating candidates (measured by voting in a primary or caucus). We can begin to distinguish among these alternatives by examining the level of extremism among the individuals who populate them. Because the ANES dataset did not have enough measures of two of these pathways in election surveys since the 1980s, we used the data from the 2008 Cooperative Congressional Election Study (CCES; see Ansolabehere, 2011) which was unique in that it offered questions on whether the respondents had contacted a member of Congress (rather than inquiring whether the respondent had been contacted by a member of Congress, which is more common).

We then estimated the relationship between each mechanism of activism and respondents’ levels of extremism using an issue-based scale of extremism. As in the main analysis above, we construct a unidimensional measure of ideology using confirmatory factor analysis. The specific questions are listed in Online Appendix Section 1. As before, we fold the issue scale at the median, with lower values as more moderate and larger values more extreme (either liberal or conservative) and control for partisanship, education, income, race, and gender.

For each model, we predict whether someone took part in offering resources (either by volunteering for campaign, giving money, or offering advertisement by putting up a yard sign), conveying information (attending a meeting or contacting a member of the House), or participating in the nomination (voting in the primary or caucus). The specific measures used are listed in Online Appendix Section 1. ¹⁸

Given our findings above, we expect that those offering resources (the resource pathway) or information (the contact pathway) to candidates will see themselves as more ideologically extreme. Following Norrander (2010), we do not expect to find primary voters (the nomination pathway) to be any more extreme than non-primary voters.

The results are presented below in Figure 4. As before, the circle indicates the point estimate and the horizontal line the 95% confidence interval. Here, we see clearly that individuals who are more ideologically extreme are more likely to engage in each mechanism of influence.

OPEN IN VIEWER

Figure 4.

Determinants of pathways for influence.

Indeed, not only is ideological extremism a significant predictor of each form of political participation, it is the most important predictor—even more so than education, income, or identifying as a partisan. For example, moving from the lowest income to the highest income category increases the probability of providing resources to a candidate by 19 percentage points. By contrast, moving from the most moderate to the most extreme ideological position increases the probability of providing resources by 39 percentage points.

The relationship with ideology is even more noticeable when considering the “contact” pathway. Education is, aside from ideological extremism, the most influential predictor of contact: people with advanced degrees are 9% more likely to contact their elected officials than are high school dropouts, all else equal. On the other hand, ideological extremists are 31 percentage points more likely to contact their elected officials than are moderates, all else equal.

We also find that ideological extremists are more likely to take part in the candidate selection process: moderates are 39 percentage points less likely to participate in the primary or caucus than are extremists. By contrast, those who have not completed high school are only 22 percentage points less likely to participate than are those with advanced graduate degrees. We note, further, that this similar relationship holds when predicting voting in the general election as well: extremists are 28 percentage points more likely than moderates to vote in the general election, while advanced graduate degree holders are 19 percentage points more likely than high school dropouts to report voting in the general election. Thus, while we find that primary voters are likely to be ideological extremists, so are general election voters.

Here, we see clear evidence that ideological extremists are more engaged than moderates are. Moreover, they are more engaged in ways that directly influence elected officials—taking part in their selection, directly contacting them, or offering resources to their campaign. In each pathway, not only is ideological extremism a strong predictor of activity, it is the strongest predictor of activity by a wide margin. Interestingly, the other forms of activism that are commonly used to identify political activists but do not offer as clear a clear pathway for influencing a particular candidate (voting in the general election, seeking to influence someone’s vote) are the ones least affected by respondent’s ideology. Thus, previous analyses identifying activists as ideological extremists may well have been underestimating the extent of their extremism.

Further analysis is needed to parse out the issue preferences of those who take part in these different types of political activities and the ways in which these preferences are articulated to candidates. Which of the alternative pathways has the greatest effects on candidates and elected officials? To what extent does the polarization of current American politics stem from activists influencing candidates, and to what extent does it arise because of the nature of the candidates themselves?

Conclusion

This study adds to the research on activists and polarization in several ways. First, while we find that ideological extremism, income, and education are most commonly associated with political activism, the relationships are not always straightforward. While differences in education matter a great deal in determining who votes and who does not, educational differences are substantively unimportant in determining who works for a campaign. Similarly, ideological extremism is strongly related to the likelihood of attempting to influence someone else’s vote choice, while extremists are no more likely than other people to attend a rally. Moreover, extremism works differently within one form of participation: across the ideological spectrum, the well educated are likely to report voting. This is not the case among the poorly educated, whose likelihood of voting increases dramatically as their extremism increases. These findings raise important questions about the dominant approach in the literature, which treats different forms of political activities as interchangeable and equivalent in terms of the types of participants they draw.

Second, we find that even when controlling for extremism, different issues are associated with different types of political activists. Campaign donors are more likely to be concerned with the social safety net, while button wearers are more concerned with racial and cultural issues. Those motivated to political activity may participate in one way but not another, depending on the issues linked with their activity.

Third, when focusing on the pathways (resources, information, and nominations) where activists are most likely to influence elected officials, we find that overall, extremists are more politically active than moderates are. Accordingly, politicians are more likely to hear from, and depend upon for campaign resources, the more extreme segments of the population. While the CCES does not allow for the creation of a three-factor measure of ideology, it follows from the ANES analysis that candidates are particularly exposed to extremism on social, rather than racial and cultural issues.

Fourth, our analysis shows that we need to carefully specify the relationship between activism and issue/ideological extremism. Much of the literature in this area compares those who engage in several types of activities with those who engage in none. We compare types of activists with one another (including with those who only vote), and we find that there are complex patterns of difference that are not necessarily characterized by the statement that activists are more extreme. It is worth noting that the most striking findings of a strong relationship between activism and extremism have been reported in surveys of national convention delegates, a population that may not be highly representative of the larger population of political activists, and in studies that examine “core partisans” (Bartels, 2016) or the “engaged public” (Abramowitz, 2010, Chapter 2)—those who not only take part in political activities but who also report political knowledge and interest, concern about election outcomes, and a much higher level of enthusiasm for their own party than for the other party.

Finally, with the notable exception of Layman et al. (2010: 342–343), studies asserting this causal relationship do not demonstrate the direction of the influence. Do extremist activists recruit and support more extreme candidates and push elected officials to take more extreme issue stands, or is it that entrepreneurial candidates can cause polarization among activists (see Abramowitz, 2010: 7) or encourage citizens who share their beliefs to become active in politics? If the emergence of a compelling candidate in an election affects the types of people who become active in that election, then activists would best be viewed as endogenous to the polarization process. Similarly, a “movement” in an election, such as the Tea Party since 2009, could stimulate the mobilization of a new set of party activists. To test this, ideally researchers would identify a pool of potential activists and then follow which ones were mobilized to active roles in various years and by which stimuli. Most of the studies to this point have been cross-sectional, however, so the causal direction of the relationships they find remains unclear.

It may also be that, as Mason argues (2015), the current polarization is based more on social identity than on issue differences and issue extremity. It is possible, then, for activists and other citizens to have developed stronger emotional reactions against the opposition party, its candidates, and its proposals, without having become more extreme in their issue attitudes—or, at least, no more extreme on the left-right dimension of ideology that political scientists often test for. When social identity and issue attitudes align, however, the intensity of the individual’s social group identity should increase as well as his or her anger toward the out-group. It may be that this greater intensity of social polarization is more true of activists than it is of less politically engaged citizens.

Another fertile area for research stems from Miller and Saunders’s finding (2016) that participation is not necessarily stable over time; many people who are active during one election do not take part in the next election. Only a relatively small core of activists remains engaged over the long term. Examining panel data on activism could then tell us whether the in-and-out activists—those whose involvement was presumably stimulated by a particular candidate, issue, or event—are as likely to communicate with candidates and elected officials and to get their voices heard as are the long-term activists. It would be intriguing to demonstrate whether the former are more endogenous to the polarization process than the latter.

Our findings in this article, then, help to clarify both the theory and the data on the relationship between activists and political polarization. We offer empirical support for the contention that political activists are more likely to hold extreme attitudes than are other citizens. At least in American politics—and it would be instructive to test whether these findings apply as well in other large industrialized democracies—people with extreme attitudes toward political issues are more likely to take part (though to different degrees) in voting, trying to influence others’ views, attending a rally, working for candidate or party, displaying a button or sticker, and donating money to a party or candidate. This evidence is consistent with the argument that partisan activists may be vital culprits in the greater polarization of American politics in recent decades. But our findings also show that it is unwarranted to approach the study of political participation as though different types of political activities are interchangeable. We find clear and theoretically interesting differences among types of political involvement with respect to the characteristics of the individuals involved and the specific issues on which they hold extreme views. Thus, political activists are not as uniformly ideologically extreme as has often been assumed.

Supplemental material