Identifying Political Habits Beyond Voting

Abstract

Voting is repetitive and persistent, but these ideas have received scarce attention when it comes to other political behaviors. Do repetition and persistence characterize political acts beyond voting? Drawing together existing theories of participation with the idea of habit, we present an integrated model of repeated behavior. We then analyze the characteristics of 17 different political acts using two methodological approaches including a “conditions on observables” approach using the 2010-2014 Cooperative Election Study Panel and a “direct measurement” approach using a new survey instrument called the Self-Report Habit Index on two Qualtrics surveys and the 2020 Cooperative Election Study Panel. From these results, we identify political acts along a spectrum of habit, discuss the implications, and provide suggestions for future work on political habits.

Keywords

political participation voting habit

Voting is repetitive: many people who vote once vote again. Voting is also persistent: the act of voting increases the probability of voting in the future (Green & Ron, 2000; Holbien et al., 2023). As such, repetitive voting is partly due to its persistence. While scholars have come to refer to this stylized fact using an assortment of terms, such as consuetude (Green & Ron, 2000), habit (Dinas, 2012), habit-formation (Górecki, 2013), inertia (Plutzer, 2002), and persistence (Denny & Doyle, 2008), the prevailing wisdom is the same: “lure someone to the voting booth” and “increase the propensity for him or her to vote in a future election” (Green & Ron, 2000, p. 562). Lure enough citizens to the voting booth early in the life course—perhaps by lowering the voting age (Franklin, 2004)—and decrease political inequalities potentially for generations to come.

That said, voting is infrequent, scheduled, structured, and not especially influential. It is also only one way to engage in the political process. Other electoral behaviors, such as donating money, volunteering for a campaign, contacting an official, or attending a rally can be done with greater frequency and in less structured ways. Less formal ways to engage with politics—such as posting to social media, discussing politics with family and friends, or following the news—can occur multiple times a day as part of a person’s routine. These political acts are meaningful, yet we know much less about the extent to which they are repetitive and persistent.

Are most political acts repetitive or independent decisions? Which political acts exhibit persistence? And to what extent are decisions to participate in politics driven by habits? Answering these questions has both practical and theoretical importance. If most political behaviors are repetitive and persistent, the implications are far-reaching for policymakers and civics advocates. Interventions can be used to increase participation by targeting citizens who are typically left out of the political process and by focusing on political acts that are most likely to persist. Increasing political engagement via repetition across a broad range of acts contributes to a more inclusive society beyond the voting booth. Moreover, where political acts follow established behavioral patterns, strategies could be developed to change existing behaviors that undermine democratic norms. If certain acts—such as sharing misinformation on social media—are part of a daily practice, it may only take a small change in that routine to develop better, healthier, and more democratic political habits.

Understanding the characteristics of political acts is also theoretically valuable. Repeated behaviors are characterized by at least two stages: initiation and persistence (e.g., Robinson, Shaver, & Lawrence). What determines the initiation of an act may be different from what determines its persistence. In fact, research suggests this is exactly the case with voting (Plutzer, 2002). This raises the possibility that the persistence of political behaviors is largely determined by automatic processes—what psychologists think of as habits—rather than resources or motivation, which may be more influential at the initiation stage. Scholars must have an understanding about the behavioral patterns of political acts to test this thesis. If empirical analyses find that most political behaviors are repetitive and persistent, then our explanations must be more complex than what is currently offered.

We start by introducing an integrated model of repeated political behaviors and discuss how it fits with current literature. We then explore the repetition and persistence of political behavior using two non-experimental approaches. Using the 2010-2014 Cooperative Election Study Panel (CES), we take a “condition-on-observables” approach to present descriptive analyses about 9 acts. We supplement these analyses with a “direct measurement” approach by introducing and applying the Self-Report Habit Index (SRHI) to 11 acts across three surveys. Despite its widespread use in the field of psychology, the SRHI has only once been applied to voting (Cravens, 2020) and never to other political acts. The combination of the longitudinal analyses with the SRHI constitutes a novel approach to studying the characteristics of political behavior.

Both analyses suggest that voting is consistently the most repetitive and persistent political act, but that news consumption, including following the news, reading the newspaper and watching tv, also exhibit habitual tendencies. Participating in these acts once or twice may make future participation likely. On the other hand, our work suggests that working for a campaign, attending a meeting, and displaying yard signs are less persistent, although performed by repeat participants. We suspect that these acts are likely independent decisions—driven less so by habit and more by resources, motivation, or mobilization. In the middle of these two extremes are acts such as donating money or reading a blog. Our analyses suggest that these acts can in time be driven by habits, but it would require a long history of repeated behavior.

We conclude by noting somewhat of a paradox when describing the characteristics of political behavior: some acts are persistent with little repetition, while others are repetitive with little persistence. We discuss implications for the effectiveness of policy interventions in the conclusion as well as how future research can build on our exploratory study.

The Persistence Model of Political Behavior

The traditional model of political behavior focuses on resources, motivations, and recruitment (Schlozman et al., 2018). In recent years, however, scholars have turned their attention to a related question: why do some people repeatedly participate in politics while others only occasionally participate or repeatedly abstain? The logic of the traditional model holds that regular (or irregular) participation is a function of the relative constancy (or fluctuations) in the resources, motivations, and recruitment required for participation. For instance, according to the traditional model, the regularity with which people donate money is due to the continuity of income, political interest, and targeted fundraising efforts from one election cycle to the next.

Another explanation for repetition in political behavior is persistence. A behavior is persistent if doing it once makes it more likely to occur in the future, given the opportunity. As John Holbein and colleagues put it, “…studies have argued that voting treatments in one election do have transformative, durable effects in terms of the number of times individuals vote in the future” (2023, p. 1021). Persistence can take hold in two ways. The first is when undertaking a political behavior increases the resources, motivations, and recruitment available for future action. For instance, scholars find that participating once increases political efficacy (Leighley, 1991), political interest (Prior, 2018), and norms of citizenship (Quintelier & Jan, 2014), although recent research casts doubt on whether voting does in fact have such “transformative” effects (Holbein et al., 2023). Nevertheless, in theory, participation can persist if it reinforces the substance of the traditional model—the resources, motivation, and recruitment that underpin political action. We refer to this type of persistence as a participation feedback loop.

The second way behavior is persistent is through habit. In psychology, habits are defined as “actions that are triggered automatically in response to contextual cues that have been associated with their performance” (Gardner, 2012, p. 664). Putting on a seatbelt (action) after getting into a car (contextual cue) or saying amen (action) at the end of a group prayer in church (contextual cue) are two examples. Behaviors might begin because of motivations, rewards, or end goals, but habits are formed when initiation of an action is transferred to external cues (Wood & Rünger, 2016). Once transference occurs, the dependence on conscious attention or motivational processes is dramatically reduced and replaced by automaticity (Gardner, 2012).

Figure 1 presents what we call the persistence model of political behavior. It integrates the traditional model—indicated by the solid lines—with the concept of persistence—indicated by the dashed lines. A behavior is repetitive if it is undertaken with some regularity and persistence if doing it makes it correlated with future action. Persistence comes about through feedback loops—which increase the resources, motivation, and recruitment for future action—as well as habits—which make it more likely that future action will be automatically activated by a context cue.¹

It is important to note that the models presented in Figure 1 are not mutually exclusive. Related, our goal is not to distinguish one model as “better” than the other at explaining why individuals engage in repetitive political acts. Instead, we are interested in exploring the variations of repetition in political acts beyond voting as well as presenting innovative ways to theoretically and methodologically study political habits. To that end, it is useful to keep in mind a few theoretical implications of the persistence model:

1. Not all repetitive behavior is persistent. A person could regularly act because they make an independent decision to do so each time the opportunity arises or because the motivation, resources, and recruitment required for action are more-or-less consistent across opportunities.

2. Not all persistent behavior is repetitive. Acting once may make future action more likely, but that doesn’t mean future action will happen; persistence only leads to repetition when opportunities for action arise.

3. Not all persistent behavior is habitual. Persistence can arise because of participation feedback loops or because of habit-formation. Participation feedback loops make future action more likely by increasing the motivation, resources, and/or recruitment required for it; however, these changes do not necessarily mean action is automatically activated when presented with a context cue.

4. Habits and feedback loops necessarily lead to persistence but not repetition. Habits and feedback loops make future action more likely given the opportunity, which is to say it makes it more persistent. However, as we noted earlier, not all persistent behavior is repetitive, so by the same standard not all habits and feedback loops necessarily lead to repetition either.

Figure 1.

A persistence model of political behavior

Notably, if the context regularly supplies opportunities and cues for participation then all habitual behavior will be persistent and all persistent behavior will be repetitive. In this paper, we assume that the supply of opportunities and cues for participation are more-or-less regular across contexts (e.g., geography, year) and persons. This means that observing habits or feedback loops implies persistence and observing persistence implies repetitiveness.

Moving Beyond Voter Turnout

An abundance of research has shed light on the persistence of voting. The variety of statistical tests (e.g., latent growth curves) and research designs (e.g., field experiments) employed over the past two decades have made evidence of the persistence phenomenon unassailable (Denny & Doyle, 2008; Dinas, 2012; Gerber, Green, and Shachar, 2003; Górecki, 2013; Plutzer, 2002). Some studies suggest this persistence is driven in part by habit (Aldrich et al., 2011; Cravens, 2020). Other research indicates that participation feedback loops also play a role (see, for instance, Leighley, 1991; Quintelier & Jan, 2014; Prior, 2018).

What about behaviors beyond voting? We draw on the Cooperative Election Study 2010- 2014 Panel Study to highlight patterns of repetition across a variety of political behaviors. Figure 2 shows the percentage of respondents who never report each behavior across surveys in 2010, 2012, and 2014, who report each behavior in only some years of the study, and who report each behavior in all years. These patterns highlight variations across political acts. Voting is one of the most undertaken behaviors and most respondents who report voting at least once also report voting in the other years of the study. Working for a campaign is rarely undertaken. Among respondents who report working for a campaign at least once, only a fraction report working for a campaign in all years of the study. Overall, these patterns suggest that political behaviors exhibit some degree of regularity, but how this regularity varies across behaviors.

Figure 2.

The Regularity of Political Behavior.

To what extent is the repetition in political behaviors driven by persistence? We could generalize from extant research if all political behaviors shared the same characteristics as voting; however, political behaviors differ from one another in many ways and often quite dramatically. Liking the tweet of a preferred political group shares little in common with running for office, except that both occur in the political domain. Attempts at classifying political behaviors showcases the myriad ways they are different and similar. Henry Brady focuses his definition of political participation on whether a behavior is an action vs non-action, undertaken by ordinary citizens vs elites, intended to influence vs not intended to influence, and directed towards politics vs directed elsewhere (1993). Jan van Deth’s “conceptual map” highlights other characteristics: actions vs non-action, voluntary vs obligatory, amateurish vs professional, as well as the locus, target, and motivation (2014). Still other scholars focus on the medium of the behavior (Oser et al., 2013). For instance, voting can be done by mail or in-person. Consuming the news can be done by listening, watching, or reading; the latter of which can be digital or print. It thus remains an open question as to which and to what extent political behaviors beyond voting are characterized by regularity and persistence.

Analytic Strategy

How can we observe persistence? The gold standard in voting research is to use an experimental or regression discontinuity design to isolate the causal effect of voting at one point in time on voting in the future. For example, Coppock and Green reanalyzed several get-out-the-vote treatments from field experiments conducted in the 2000s to show they had long-term impacts on turnout (2016). Although these studies often talk in terms of habit and habit-formation, the research designs leave open the possibility that the causal estimate is partly or wholly a function of feedback loops. These designs are ideal for studying persistence of voting but unfortunately are less feasible when it comes to non-voting behaviors (but see Holbein et al., 2023 for examples of these designs as applied to voting behavior).

We use two alternative approaches to explore the persistence of various political. We want to be clear that neither of our approaches are causal. Nonetheless, we believe that they are justified first steps in moving towards studying political habits beyond voting. The first is a “condition-on-observables” approach in which we use regression models to estimate the association of past behavior on future behavior while controlling for other variables that measure resources, motivation, and recruitment. The coefficient for past behavior is what we refer to as the “habit” estimate. The goal of conditioning on observables is to reduce the possibility that the correlation between behaviors over time (i.e., the repetition of behavior) is due to traditional explanations (i.e., non-persistent) or feedback loops (i.e., persistent but non-habitual). This approach was used effectively in early studies of voting habits (e.g., Green & Ron, 2000; Plutzer, 2002), which serve as useful benchmarks for the analyses we conduct below.

Importantly, this approach suffers from the possibility of omitted variables that could account for the correlation of behavior over time. Nonetheless, we leverage one important empirical implication of the persistence model to more rigorously deal with the threat of omitted variable bias. Specifically, the habit-formation process is behavior-specific, so doing one behavior should not affect other political behaviors in the future. When a current behavior affects another future behavior, we refer to this as “cross-over” effect. Traditional explanations and participation feedback loops should create such cross-over effects. Any behavior that increases resources, motivation, and recruitment will increase the probability of other behaviors in the future. We therefore re-estimate our models using other past political behaviors to generate “cross-over” estimates. We then compare these cross-over estimates to our habit estimates—as the magnitude of the cross-over estimates approach the magnitude of the habit estimates, we lose confidence that the estimates reflect habit rather than omitted variables reflecting traditional explanations or feedback loops. Even still, the results of this analysis can only provide evidence that is suggestive of habit but not demonstrative of it.

The second approach we use is to directly measure automaticity, the hallmark of a habitual behavior. We adapt a 12-item survey battery, known as the Self-Report Habit Index (SRHI), developed by psychologists to study the habitualness of political behaviors (Verplanken & Orbell, 2003). Scores on the SRHI—especially on a validated “automaticity” subscale consisting of the four starred items below (Gardner, 2012)—constitute our “habit” estimates. Although no studies to date have applied the SRHI to non-voting political behaviors, one study has used it to study voting habits (Cravens, 2020). We draw on this study as a benchmark for our analyses.

Many of the standard measurement problems that arise in survey research may also affect the SRHI: respondents may have different standards by which they evaluate these statements, be influenced by social desirability bias and overreport habits, or not attentively engage in the survey. Importantly, the SRHI has been validated in the field of psychology and used to demonstrate habits across a multitude of behaviors, ranging from diets and physical activity (Gardner et al., 2012) to listening to music and watching television (Verplanken & Orbell, 2003). This body of research somewhat allays our concerns, but we nevertheless explore the extent to which these problems may compromise our findings.

Data and Results

We start with the condition-on-observables approach before turning to the direct measurement approach. Recall that the goal of our analyses is two-fold: (1) to estimate the habitualness of a variety of political behaviors, and (2) to understand how the habitualness of political behaviors compares to one another. Table 1 lists the political acts we analyze and breaks them down into eight traditional forms of political participation, five forms of news consumption, and four non-traditional acts. Five of these behaviors—voting, attending a meeting, displaying a sign, working for a campaign, and donating money—are analyzed using both approaches, which allows us to highlight their consistency and complementariness on the one hand and where they produce divergent findings on the other.

Table 1.

Political Behaviors Analyzed in Studies

	CES panel2010-2014	Self-reportHabit index
Political participation
Vote	✓	✓
Attend meeting	✓	✓
Display sign	✓	✓
Work for campaign	✓	✓
Donate money	✓	✓
Contact an official		✓
Attend a protest		✓
Sign a petition		✓
News consumption
Read blog	✓
Watch TV	✓
Read newspaper	✓
Listen to radio	✓
Follow the news		✓
Other behaviors
Discuss politics		✓
Post on social media		✓
Boycott		✓

Note. Appendix A elaborates on the definition of political behavior and political participation as well as the selection of behaviors used in these analyses.

The “Condition-On-Observables” Approach

We first analyze the Cooperative Election Panel Survey from 2010-2014 (CES). The 2010 CES interviewed over 55,400 adults during the 2010 election. A subset of 9,500 respondents were re-interviewed during the 2012 election and again during the 2014 election. Respondents were asked about participation in 9 political acts over the past year. The CES also includes a measure of validated voter turnout in addition to the self-reported measure. Analyzing voting is thus useful because it not only allows us to establish a point of comparison for the characteristics of other political acts but also allows us to conduct a novel replication of existing research using new data and a validated measure. The Guide to the 2010-2014 CES Panel Study has more information about response rates, attrition, panel conditioning, and voter validation.

Table 2 reports the weighted percentage of respondents participating in each act across years. Voting is among the most common and repetitive political act with the highest turnout in the presidential election.

Table 2.

Weighted Percentages of Respondents Participating in 9 Political Acts

All Respondents				Only Respondents Who Misreport Voting
Political Act	2010	2012	2014	2010	2012	2014
Vote (Validated)	58.5%	61.4%	58.3%
Vote (Self-reported)	85.9%	94.3%	87.7%
Attend meeting	14.3%	13.3%	11.0%	16.3%	13.1%	10.8%
Display Sign	19.2%	22.0%	13.6%	20.5%	19.6%	14.9%
Work for a campaign	6.7%	8.2%	4.5%	6.3%	7.5%	4.9%
Donate money	22.0%	29.2%	19.7%	21.8%	26.6%	20.3%
Read blog	25.8%	24.4%	21.2%	28.0%	21.2%	24.1%
Watch TV	74.2%	73.7%	66.7%	78.4%	73.2%	69.7%
Read newspaper	61.9%	54.6%	50.5%	65.9%	52.5%	56.6%
Listen to radio	42.2%	40.5%	34.8%	46.2%	34.1%	40.9%

The other patterns of participation are largely as expected. Roughly 10-30% of the public report attending a meeting, displaying a sign, and donating money, generally in any given year. We describe these behaviors as occasional political acts—neither common nor rare. Working for a campaign occurs less frequently, with only about 4-8% of the public reporting this behavior. Notably, political participation is generally higher in 2010 than 2014, which might reflect greater political activism on the right due to the rise of the Tea Party following the election of Obama in 2008. There is also substantial variation in how respondents consume news. Most respondents report TV and newspaper usage while only a minority report blog and radio usage. All forms of news consumption decline over the years of the study.²

Descriptive patterns of participation over time offer initial insights into repetition and persistence. Figure 3 reports the weighted percentage of respondents undertaking each act in 2014 based on whether respondents undertook that same act in 2010, 2012, both years, or not at all. If participation in each year is random than it shouldn’t vary across groups. In fact, we see greater activity among respondents who participated in a prior year and the most activity for respondents who participated in both prior years. These patterns indicate repetition but are not by themselves evidence of persistence. There could be any number of explanations for why the same people participate or abstain year-to-year. Notably, voting displays the largest gap between those who never participated and those who participated at least once and the smallest between those who participated at least once and those who participated in both prior years. This pattern contrasts to other political acts, which show a more “linear” trend. This suggests that voting may be more intensely shaped by persistence compared to other acts.

Figure 3.

The activity gap in 2014 based on participation in 2010 and 2012

To analyze the correlation between past and current behavior (e.g., persistence) more rigorously, we estimate recursive and non-recursive regression models. The recursive models are logistic regressions in which past political action is the key predictor. They do not include any statistical correction for potential spuriousness between the lagged and current political acts beyond the inclusion of an extensive set of control variables. The non-recursive models estimate a two-step probit regression. The first step uses the lagged control variables to predict the lagged political act for each year separately. The second step estimates current political action using the predicted political action from the first step for each lagged year as well as the current control variables. In short, the non-recursive models add a statistical correction that helps account for possible sources of spuriousness.

We include an extensive set of control variables in our models: gender, age, race, education, employment status, marital status, church attendance, parenthood, home ownership, union membership, partisan strength, interest in the news, party contact, and presidential approval. These variables attempt to capture one or more of the alternative explanations—resources, motivation, and mobilization—laid out in Figure 1.³ For instance, union membership taps into both resources (e.g., the development of civic skills), motivation (e.g., the pressure from other union members to participate), and mobilization (e.g., formal efforts by the union to initiate participation among members). There is some trickiness around the timing of control variables in the recursive models. Including controls from 2010, 2012, and 2014 provides the most robust check on potential sources of spuriousness but may also artificially deflate the key quantity of interest if control variables from latter years mediate the relationship between lagged and current action. We therefore estimate three recursive models for each political act: a “proximal” model with controls for 2010, 2012, 2014, an “intermediate” model with controls for 2012 and 2014, and a “distal” model with controls for 2010. The time-invariant control variables (gender, race) are only included once in the proximal and intermediate recursive models. Probability weights are included in the recursive models. Details on the measurement and descriptive statistics of these variables are reported in the Appendix.

Table 3 reports our estimates from the models of all 9 political acts, broken into three blocks: voting, non-voting political acts, and news consumption. The results of the recursive models are similar regardless of whether we include control variables that are distal (Model 1), intermediate (Model 2), or proximal (Model 3). In all cases, the association between lagged and current participation is statistically significant at traditional levels. The non-recursive model with full control variables (Model 4) suggest that some political acts may not be persistent. The key coefficients for lagged participation are not statistically significant in Model 4 for attending a meeting or working for a campaign. All other behaviors exhibit some persistence, although these results show variation in whether earlier participation (2010) or later participation (2012) matter. For voting, donating money, reading a blog, and reading a newspaper, only the coefficients for participation in 2010 are statistically significant, whereas the opposite is true for displaying a sign. For watching television and listening to the radio, the coefficients for both lagged participation (2010 & 2012) variables are significant. The Appendix reports the results of the recursive and non-recursive models when control variables are excluded.

Table 3.

Estimates of Persistence Across 9 Political Acts

Act	Year	Recursive Models			Non-Recursive Model
Act	Year	1	2	3	4
Vote	2010	1.395* (.122)	1.378* (.120)	1.311* (.120)	1.657* (.335)
Vote	2012	1.504* (.120)	1.499* (.119)	1.449* (.117)	0.398 (.470)
Attend meeting	2010	1.365* (.157)	1.373* (.156)	1.295* (.152)	1.117 (.816)
Attend meeting	2012	2.015* (.157)	1.946* (.157)	1.956* (.151)	1.234 (1.048)
Display sign	2010	1.392* (.136)	1.400* (.135)	1.368* (.130)	0.367 (.553)
Display sign	2012	1.737* (.141)	1.696* (.140)	1.676* (.139)	2.030* (.501)
Work for campaign	2010	1.406* (.308)	1.439* (.300)	1.520* (.282)	0.672 (1.201)
Work for campaign	2012	2.490* (.329)	2.389* (.298)	2.473* (.258)	0.782 (1.444)
Donate money	2010	1.238* (.129)	1.208* (.128)	1.255* (.119)	2.395* (.494)
Donate money	2012	2.118* (.138)	2.107* (.130)	2.101* (.126)	0.389 (.472)
Read blog	2010	1.161* (.133)	1.204* (.132)	1.082* (.131)	2.069* (.657)
Read blog	2012	2.128* (.131)	2.103* (.128)	2.235* (.132)	0.918 (.620)
Watch TV	2010	1.201* (.153)	1.183* (.151)	1.166* (.154)	2.520* (.599)
Watch TV	2012	1.700* (.155)	1.713* (.159)	1.694* (.151)	1.077* (.533)
Read newspaper	2010	1.174* (.132)	1.220* (.131)	1.235* (.132)	2.167* (.761)
Read newspaper	2012	1.685* (.121)	1.678* (.121)	1.684* (.120)	0.674 (.726)
Listen to radio	2010	1.180* (.123)	1.217* (.122)	1.251* (.122)	1.244* (.412)
Listen to radio	2012	1.843* (.125)	1.797* (.131)	1.796* (.123)	1.105* (.439)

Note. Entries are logit coefficients for the lagged dependent variable models and probit coefficients for the instrumental variable models. Standard errors are reported in parentheses. Model 1 is the distal model, Model 2 is the intermediate model, Model 3 is the proximal model, Model 4 includes control variables for 2014. Estimated coefficients and standard errors for the control variables are not reported here. Full results are available upon request.

Our results for voting largely mirror what Green & Ron (2000) find in their analysis of the American National Election Studies from 1972-1976 and 1992-1996. Their non-recursive models, like ours, show that voting in the more recent election (1974/1994) is less predictive of current turnout than is voting in the more distant election (1972/1992). In fact, their estimated coefficient for voting in 1994 (the more recent election) falls just shy of a conventional level of statistical significance (p < 0.05). They very reasonably point out that earlier elections in their data are presidential elections, which likely exert a stronger effect on subsequent participation.

We also find that voting in the earlier election is more influential on subsequent turnout, but in our data this election is a midterm. An alternative explanation that accounts for both these findings is that voting early in life is more important to persistence than voting recently. This explanation makes sense considering how voting habits develop over the life course (Plutzer, 2002).

To illuminate the magnitude of the associations we see in Table 3, we generate predicted probabilities of participation in 2014 using the estimates from Model 4 (recursive with full controls). Figure 4 plots these values based on prior activity—whether a respondent didn’t participate in 2010 or 2012, participated in only one of these years, or participated in both years—while all other variables are set at their mean value. The results are organized from the lowest to the highest predicted probability for those who participated in both prior years. The pattern looks remarkably like the descriptive statistics we reported in Figure 3, albeit somewhat attenuated.

Figure 4.

Predicted probability of participation in 2014 based on prior activity

The results so far—which we summarize in Table 4—suggest that many political acts exhibit some persistence. To further scrutinize this finding, we conduct placebo tests that swap out the lagged political activity variables from Model 4 with alternative lagged behaviors (e.g., predicting voting in 2014 using attending a meeting in 2010 and 2012). We divide our tests into two groups—traditional participation and news consumption—and iterate through all combinations within each group. The idea is that if a political act is persistent—doing it once leads you to do it again—then participation in these other activities should not have as large an effect because they do not induce future action. Of course, we don’t expect that political activities are entirely unrelated. A person displays a sign because they are passionate about a candidate and plan to vote for them. A person who plans to vote may also donate money in order to amplify their voice. Nevertheless, our assumption is that persistence should be stronger “within acts” (i.e., lagged A -- > current A) than “between acts” (i.e., lagged B -- > current A), so null or weaker effects in these placebo tests will increase our confidence in the results.

Table 4.

Summary of Statistical Significance in Models With Full Controls

	Recursive Models		Non-Recursive Models
	2010	2012	2010	2012
Vote	✓	✓	✓
Attend meeting	✓	✓
Display sign	✓	✓		✓
Work for campaign	✓	✓
Donate money	✓	✓	✓
Read blog	✓	✓	✓
Watch TV	✓	✓	✓	✓
Read newspaper	✓	✓	✓
Listen to radio	✓	✓	✓	✓

Note. Checkmark indicates coefficient is statistically significant at p < 0.05 in Model 4 (recursive model with proximate controls) and Model 6 (non-recursive model with full controls) from Table 3.

Figure 5 reports the estimated coefficients for the placebo tests of political participation, with the top panel displaying the coefficients from 2010 and the bottom panel displaying the coefficients from 2012. The black markers represent the coefficients from the original model. What becomes immediately clear is that the original coefficients are consistently larger than the placebo coefficients. While several “cross-behavior” associations are statistically significant, the magnitude of the coefficient is always substantially attenuated in comparison to the original estimate.

Figure 5.

Estimated coefficients from placebo tests for political engagement

The “Direct Measurement” Approach

The analysis of panel data revealed that many political acts demonstrate persistence. We now turn to a second test of this possibility—an analysis of the Self-Report Habit Index (SRHI) of 12 political acts across three surveys. The SRHI is a reliable and valid 12-item survey battery developed by psychologists (Verplanken & Orbell, 2003). The SRHI is a survey instrument that measures habits by having respondents evaluate twelve statements on a standard Likert scale:

“BEHAVIOR X [e.g., voting] is something…

1. I do frequently.

2. I do automatically.

3. I do without having to consciously remember.

4. That makes me feel weird if I do not do it.

5. I do without thinking.

6. That would require effort not to do it.

7. That belongs to my (daily, weekly, monthly) routine.

8. I start doing before I even realize I’m doing it.

9. I would find hard not to do.

10. I have no need to think about doing.

11. That’s typically “me.”

12. I have been doing for a long time.”

We adapt the SRHI to political participation on two Qualtrics surveys and a team module of the 2020 Cooperative Congressional Election Study (CCES). The first Qualtrics survey (Qualtrics #1) was administered to 1,112 respondents in March 2018. The SRHI instrument was adapted to five political acts: voting, donating money to a campaign or candidate, contacting an elected official, discussing politics with friends or family, and following the news. All respondents were asked about voting and then randomly assigned two additional acts. The second Qualtrics survey (Qualtrics #2) was administered to 1,014 respondents in August 2019. Here, the SRHI instrument was adapted to eight political acts: voting, posting political content on social media, attending a protest, attending a political meeting (e.g., a city council meeting), displaying a political yard sign, boycotting a company for political purposes, signing a petition, and working for a political campaign.⁴ The third study is a team module of 1,000 respondents from the 2020 Congressional Election Study (CES), which includes a pre-election survey administered in September 2020 and a post-election survey administered in November 2020. An abridged version of the SRHI instrument was adapted to two political acts on the pre-election survey: voting and following the news. Table 5 summarizes the key features of these three studies.

Table 5.

Key Features of the SRHI Studies

	Qualtrics #1	Qualtrics #2	CCES
Time	March 2018	August 2019	September 2020
Sample size	1,112	1,014	1,000
Political acts	Voting Donating money Contacting an official Discussing politics Following the news	Voting Posting on social media Attending a protest Attending a meeting Displaying a yard sign Boycotting Signing a petition Working for a campaign	Voting Following the news
Response options	7-Point likert scale including middle position	5-Point likert scale, including middle position but excluding “agree” and “disagree”	6-Point likert scale excluding middle position
Notes	All respondents asked voting and then randomly assigned two other political acts	All respondents randomly assigned four political acts	Respondents randomly assigned one political act; only asked the automaticity subscale

Our analysis of the SRHI is made up of two parts. We first examine descriptive statistics for the SRHI across different political acts; we compare these patterns to the results of the panel data analysis. Next, we conduct a series of tests that provide evidence for the convergent and nomological validity of the SRHI. These analyses bolster our confidence that the SRHI provides valid and reliable estimates of persistence when applied to political behavior.

We calculate the persistence of each political act in two ways: first as the mean response across all items and second as the mean response to items 2, 3, 5, and 8 which constituent an “automaticity subscale” (Gardner, 2012). Notably, the CES only includes these items in the subscale. The measures are then transformed to range from 0 to 1 to ease comparisons across studies. The Cronbach alpha score for the SRHI ranges from 0.85 (for following the news in the CES) to 0.98 (for several different political acts in Qualtrics #1 and Qualtrics #2). These scores indicate a high reliability of the SRHI when studying political behavior. Appendix C (“Statistics and Distributions of Habit Strength Scores”) reports the mean, standard deviations, and alpha scores for the SRHI of each political act across studies.

The average persistence (and 95% confidence interval) for each political act is reported in Figure 6. Following the news and voting are the only items where the average score is higher than simply saying “neither agree nor disagree” to each item (i.e., a value of 0.5). These political acts are therefore more persistent than not on average. The average persistence for all other political acts falls below the threshold of responding neither agree nor disagree to each item on average. Specifically, the SRHI scores are all between 0.20 and 0.30, which is roughly equivalent to responding “somewhat disagree” to each item. Discussing politics stands apart from this group. While it falls below the threshold of 0.5, its score of 0.45 means it is characterized by a comparatively stronger level of persistence. These results largely comport with our analysis of the panel data. Voting and news consumption are consistently the most persistent and working/volunteering for a political campaign is consistently the least persistent. Displaying a sign, attending a meeting, and donating money also demonstrate low levels of persistence—relative to other political acts—with both methodological approaches.

Figure 6.

Persistence across political acts with 95% confidence interval

We now turn to the connection between persistence and participation. To assess this connection, we asked respondents in Qualtrics #2 whether they completed each political act in the past year. Figure 7 plots the probability of each act across values of the persistence subscale using a fractional polynomial model with a 95% confidence interval. The results indicate a strong correlation. The probability of voting is nearly 1.0 for those with a strong voting persistence. For many political acts, the probability of action falls short of 1.0 for even those with the highest level of persistence, which indicate that even the highest level of persistence is not a guarantee of participation for some political acts. Of course, the causal relationship between persistence and participation will likely be bidirectional—as we note in our integrated model of political participation—so analyses of cross-sectional data cannot establish the magnitude of the effect of persistence on participation. Nevertheless, this correlation helps establish the (convergent) validity of the SRHI in the study of political behavior.

Figure 7.

Habit strength and political participation (Qualtrics #2)

One benefit of the SRHI is that it directly measures persistence, which allows us to assess variations more easily. Turning first to voting, research finds that age and education correlate with voting (Leighley & Nagler, 2014) and the formation of a voting habit (Plutzer, 2002), but that gender does not have an association with voting (Smets & van Ham, 2013). Table 6 reports the overall and subscale scores for voting persistence across values of gender, age, and education. We use a one-way analysis of variance test to assess the statistical independence between each demographic factor and each measure of persistence. The results follow our expectations: we see persistence consistently associated with age and education, but we do not see consistent differences in persistence between men and women. Age is significantly correlated with persistence in all studies, while education is correlated in all cases except for the subscale in Qualtrics #2. In contrast, gender is only statistically significantly correlated with persistence in the Qualtrics #1 study and exhibits no association in either the Qualtrics #2 or CCES studies. That we see large, consistent, and statistically significant differences in persistence across values of age, education but not gender adds to the (nomological) validity of the SRHI.

Table 6.

Demographic Patterns of Voting Habit Strength

		Qualtrics #1		Qualtrics #2		CCES
		Overall Scale	Sub-scale	Overall Scale	Sub-scale	Sub-scale
Gender	Women	0.52	0.50	0.60	0.53	0.50
	Men	0.61	0.57	0.65	0.54	0.51
	F-statistic	26.53**	16.86**	1.02	0.07	0.36
Age	18 to 24	0.45	0.43	0.47	0.48	0.35
	25 to 39	0.53	0.50	0.59	0.53	0.46
	40 to 49	0.54	0.51	0.55	0.45	0.49
	50 to 59	0.60	0.57	0.71	0.58	0.56
	60 to 69	0.70	0.65	0.70	0.57	0.55
	70+	0.65	0.56	0.82	0.67	0.59
	F-statistic	20.41**	10.94**	10.57**	3.58**	4.99**
Education	High school or less	0.50	0.48	0.57	0.50	0.46
	Some college	0.58	0.54	0.67	0.56	0.51
	Bachelor’s degree	0.65	0.60	0.65	0.54	0.51
	Graduate degree	0.63	0.59	0.69	0.57	0.60
	F-statistic	16.24**	9.24**	3.52**	1.04	2.11*

Note. **p < 0.05; *p < 0.10. Cell entries are the mean habit strength and automaticity score for each group. F-statistics are generated from a one-way analysis of variance test that assesses the statistical independence of the groups (e.g., men/women) within category (e.g., gender).

We report the demographic patterns of persistence for non-voting acts in Appendix D. These results offer auxiliary evidence of nomological validity as well as novel insights. Young adults, compared to older citizens, have the strongest persistence for posting online, protesting, boycotting, and signing petitions, but the weakest persistence when it comes to following the news, which is consistent with generational analyses (Zukin et al., 2006). The resource model of participation leads us to expect educational differences in persistence only for activities that are cognitively demanding (Verba et al., 1995). In fact, we see that education is associated with stronger persistence when it comes to following the news, discussing politics, boycotting, and signing a petition, but not for posting online, attending a protest or meeting, or volunteering for a campaign. Displaying a yard sign reveals a pattern in which the less educated report greater persistence. These patterns are broadly consistent with the resource model of participation. In terms of gender, we find that men report greater persistence when it comes to following the news and discussing politics, which is consistent with research that finds a gender gap in political interest (Fraile & Raul, 2017). Other research suggests that women engage more in private political acts while men engage more in public political acts (Coffé & Bolzendahl, 2010), which is also generally consistent with our findings: men report greater persistence for protesting, volunteering for a campaign, and contacting an official.

Discussion

The novelty that voting could be habitual—an idea pioneered by scholars in the 1990s and early 2000s—is now twenty years later a mainstay of academics, journalists, democratic activists, and anyone with an educated interest in electoral politics. We built on this work by applying the ideas of repetition and persistence to other political behaviors. We used two methods—a conditions on observables approach and direct measurement approach—to describe 16 political acts. We found that all political behaviors exhibit repetition and persistence to some degree—but some more than others. We also introduced the Self-Report Habit Index (SRHI)—a tool adapted from the field of psychology to directly measure behavioral automaticity. Here the SRHI can play a role: abridged versions of the SRHI could be added to surveys for a low cost and thus supplement existing experimental and longitudinal methods in the study of political behavior. At the same time, our results make clear that automaticity accounts for only some of the voting habit phenomenon and so other mechanisms likely matter too.

These results are a first cut at identifying the repetition and persistence of political acts beyond voting. More research is needed before we can be fully confident in the findings, and so future studies should seek to overcome some of the limitations of our methods. The condition on observables approach can never fully overcome omitted variable bias, and so one possibility is that we overestimate the persistence of political acts using this method. Indeed, we cannot be sure that the control variables we included in our model fully account for the effect of motivation, resources, and mobilization. On the other hand, error in the measurement of self-reported political participation could lead us to underestimate their persistence over time. Leveraging field experiments and validated measures of participation, such as Coppock & Green (2016) did with voting, will help overcome these problems. The SRHI approach likewise has limitations—specifically, the items were not designed with political acts in mind, and so they may be especially prone to measurement error. However, our findings serve as proof of concept for the direct measurement approach, and so future research should seek to redesign this index to better capture the automaticity of political behavior.

To the extent that the psychological principles of habit-formation apply to politics, political advocates who want to build an active electorate might draw lessons from behavioral interventions used by clinicians to encourage healthy habits, potentially including boosting voter turnout. Indeed, the most recent presidential elections showcase the need for innovation. About 1 out of 3 eligible Americans did not vote in the 2020 presidential election. In some ways, this number is unsurprising and confirms what we already know about turnout in the United States— many people don’t participate. In other ways, this number is shocking given that over $14 billion dollars were spent on the election, including hundreds of millions of dollars on political advertising. It seems unthinkable that in the polarized context of the 2020 election—and during a pandemic that forced states to make voting easier than ever before—so sit it out. Clearly there are limits on the degree to which money, advertising, get-out-the-vote efforts, and low barriers to voting can induce turnout and other forms of political activity.

Political activists often advocate for electoral reform and civics education as a remedy to low civic engagement, but they might also consider pairing the principals of habit with a “small changes” approach (Lally et al., 2010). For example, citizen-participants could select a self-determined behavioral goal (e.g., read a fact-checking website), initiate a plan of when and where to do the goal (e.g., Tuesday mornings at home), and stick to the plan for a required length of time (e.g., receive regular reminders that taper off over the course of a year). Citizens who rarely participate might focus on small changes at first since failure can be discouraging and because simpler actions become habitual more quickly (Gardner, 2012). Forming one small political habit (e.g., consuming news) may increase self-confidence for working towards other healthy political habits (e.g., voting). These novel interventions will only be successful if political acts follow the psychological principles of habitual behavior—thus, we encourage others to continue understanding the behavioral characteristics of political behaviors, especially beyond voting.

Supplemental Material

Supplemental material - Identifying Political Habits Beyond Voting

Supplemental material for Identifying Political Habits Beyond Voting by Christopher Ojeda, Julianna Pacheco in American Politics Research.

Footnotes

ORCID iDs

Christopher Ojeda

Julianna Pacheco

Funding

The authors received no financial support for the research, authorship, and/or publication of this article.

Declaration of Conflicting Interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Supplemental Material

Supplemental material for this article is available online.

Notes

References

Aldrich

J. H.

Jacob

M. M.

Wood

(2011). Turnout as habit. Political Behavior, 33(4), 535–563.

Coffé

Bolzendahl

(2010). Same game, different rules? Gender differences in political participation. Sex Roles, 62(5-6), 318–333. https://doi.org/10.1007/s11199-009-9729-y

Coppock

Green

D. P.

(2016). Is voting habit forming? New evidence from experiments and regression discontinuities. American Journal of Political Science, 60(4), 1044–1062. https://doi.org/10.1111/ajps.12210

Cravens

M. D.

(2020). Measuring the strength of voter turnout habits. Electoral Studies, 64, 102117.

Denny

K. J.

Doyle

O. M.

(2008). Does voting history matter? Analysing persistence in turnout. American Journal of Political Science, 53(1), 17–35. https://doi.org/10.1111/j.1540-5907.2008.00355.x

Dinas

(2012). The formation of voting habits. Journal of Elections, Public Opinion, and Parties, 22(4), 431–456. https://doi.org/10.1080/17457289.2012.718280

Fraile

Raul

(2017). Why does Alejandro know more about politics than Catalina? Explaining the Latin American gender gap in political knowledge.” British Journal of Political Science, 47(1): 91-112.

Franklin

M. N.

(2004). Voter turnout and the dynamics of electoral competition in established democracies since 1945: Cambridge University Press.

Gardner

(2012). Habit as automaticity, not frequency. The European Health Psychologist, 14(2), 32–36.

10.

Gardner

Abraham

Lally

de Bruijn

G.-J.

(2012). Towards parsimony in habit measurement: Testing the convergent and predictive validity of an automaticity subscale of the self-report habit index. International Journal of Behavioral Nutrition and Physical Activity, 9(102), 102. https://doi.org/10.1186/1479-5868-9-102

11.

Górecki

M. A.

(2013). Election closeness, habit formation, and voter turnout: Evidencefrom sixteen Swedish elections. Political Studies, 61(1), 234–248. https://doi.org/10.1111/1467-9248.12017

12.

Green

D. P.

Ron

(2000). Habit formation and political behaviour: Evidence of consuetude in voter turnout. British Journal of Political Science, 30(4), 561–573. https://doi.org/10.1017/s0007123400000247

13.

Holbein

J. B.

Rangel

M. A.

Moore

Croft

(2023). Is voting transformative? Expanding and meta-analyzing the evidence. Political Behavior, 45(3), 1015–1044. https://doi.org/10.1007/s11109-021-09746-2

14.

Lally

Van Jaarsveld

C. H. M.

Potts

H. W. W.

Wardle

(2010). How are habits formed: Modelling habit formation in the real world. European Journal of Social Psychology, 40(6), 998–1009. https://doi.org/10.1002/ejsp.674

15.

Leighley

(1991). Participation as a stimulus of political conceptualization. The Journal of Politics, 53(1), 198–211. https://doi.org/10.2307/2131728

16.

Leighley

J. E.

Nagler

(2014). Who votes now? Demographics, issues, inequality, and turnout in the United States: Princeton University Press.

17.

Oser

Hooghe

Marien

(2013). Is online participation distinct from offline participation? A latent class analysis of participation types and their stratification. Political Research Quarterly, 66(1), 91–101. https://doi.org/10.1177/1065912912436695

18.

Plutzer

(2002). Becoming a habitual voter: Inertia, resources, and growth in young adulthood. American Political Science Review, 96(01), 41–56. https://doi.org/10.1017/s0003055402004227

19.

Prior

(2018). Hooked: How politics captures people’s interest. Cambridge University Press.

20.

Quintelier

Jan

W. V. D.

(2014). Supporting democracy: Political participation and attitudes: Exploring causality using panel data. Political Studies, 62(1), 153–171.

21.

Schlozman

Brady

Verba

(2018). Unequal and unrepresented: Political inequality and the people's voice in the new gilded age. Princeton University Press.

22.

Smets

van Ham

(2013). The embarrassment of riches? A meta-analysis of individual-level research on voter turnout. Electoral Studies, 32(2), 344–359. https://doi.org/10.1016/j.electstud.2012.12.006

23.

Verba

Schlozman

K. L.

Brady

H. E.

(1995). Voice and equality: Civic voluntarism in American politics. Harvard University Press.

24.

Verplanken

Orbell

(2003). Reflections on past behavior: A self-report index of habit strength. Journal of Applied Social Psychology, 33(6), 1313–1330. https://doi.org/10.1111/j.1559-1816.2003.tb01951.x

25.

Wood

Rünger

(2016). The psychology of habit. Annual Review of Psychology, 67(1), 289–314. https://doi.org/10.1146/annurev-psych-122414-033417

26.

Zukin

Keeter

Andolina

Jenkins

Delli Carpini

M. X.

(2006). A new engagement? Political participation, civic life, and the changing American citizen. Oxford University Press.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

1.49 MB