It’s Not All About Resources

Abstract

The dominant political science explanations of the causes of individual-level political participation converge on three sets of antecedents—resources/skills, recruitment, and political engagement. However, the overwhelming majority of the empirical tests of these antecedents rely on cross-sectional data, obscuring the fact that micro-level participation in the United States is more accurately characterized by instability rather than by stability. Using the American National Election Study and Jennings time-series data, we for the first time demonstrate the inability of traditionally examined antecedents to explain individual-level variation in political behavior over time. Finding extant theory inadequate in this regard, we propose a modification of participation theories that puts the concept of motivation in the foreground. We argue that a model that includes motivation may both pave the way for a better understanding of the variation in participation over time and suggest possible prescriptions to help alleviate representational biases at the individual level.

Keywords

political participation resource model motivation Civic Voluntarism Model behavioral instability panel data

Over a century ago, psychologist William James (1890) speculated about the number of things that could be attended to at one time and described the world as “one great blooming, buzzing, confusion.” People react to their complicated, ever-changing environment by deciding to focus their cognitive and behavioral energies on stimuli that they value, at the expense of less important ones. Within the realm of politics, a domain about which many people care very little, the question has been asked: Why do people participate at all? In better words,

It would clear the air of a good deal of cant if instead of assuming that politics is a normal and natural concern of human beings, one were to make the contrary assumption that whatever lip service citizens may pay to conventional attitudes, politics is a remote, alien, and unrewarding activity. Instead of seeking to explain why citizens are not interested, concerned and active, the task is to explain why a few citizens are. (Dahl, 1961, p. 279)

The question of why people participate in politics is of fundamental importance in a democracy, and as such there is a long tradition of trying to answer it in political science. Theory and research to date converges on three sets of predictors of participation—resources/skills, recruitment, and political engagement. However, the overwhelming majority of the empirical studies on this topic rely on cross-sectional (i.e., correlational) data. Providing strong support for the notion that those high in socioeconomic status (SES) are more likely to participate in political activities than their low SES counterparts, these cross-sectional studies illuminate the significant and troubling representational biases in our political system. This knowledge has indisputable value, but studying the antecedents solely at the cross-sectional level obscures the empirical fact that participation in the United States is more accurately characterized by instability, rather than stability, at the micro-level. Cross-sectional studies are also limited in their ability to provide prescriptions to alleviate these representational biases at the individual level. If the sole diagnosis is differences in SES, a cure is elusive; SES cannot be easily increased to amplify the participation levels of traditionally underrepresented groups.

In light of the fact that participation is characterized by instability more so than stability, we humbly submit, and with all due respect to Dahl, that the task is to explain why some citizens in the United States participate routinely, why others come and go, and why others never participate. In other words, we argue that scholars should be interested in identifying, and democratic systems should be interested in producing, the institutional, cultural, and educational conditions that facilitate not only a more diverse active citizenry, but also a more consistently active one.

We therefore focus on change in political participation over time, and are the first to empirically demonstrate that the variables that have been found to be highly correlated with participation at the cross-sectional level do not adequately explain individual-level variation. We show this to be the case because resources, mobilization, and political engagement vary too slowly over time, if at all, to be able to explain the variability in participation that exists. This calls into question the theoretical underpinnings of extant participation research.

We begin by briefly reviewing the findings of past research on the causes of individual-level political participation. Next, we lay out our argument as to why the variables most commonly examined as antecedents of participation at the cross-sectional level may not be adequate for explaining individual-level changes in participation over time. We then use the 2000-2002-2004 American National Election Study (ANES) panel and the Jennings Youth–Parent Socialization panel to test whether variables associated with resource-based theories of political behavior predict changes in participation over time. Finding resources lacking in this regard, we propose an alternative model that takes into account the role of motivation that we suspect will be better able to explain micro-level changes in participation.

Extant Research on the Antecedents of Political Participation

Resources and Skills

Participating in politics can require significant financial resources and/or free time (Rosenstone & Hansen, 1993; Schlozman, Verba, & Brady, 2012; Verba, Schlozman, & Brady, 1995). In addition, civic skills are an important determinant of participation, as is education (W. E. Miller, 1980; Rosenstone & Hansen, 1993; Verba & Nie, 1972; Verba et al., 1995).¹

Recruitment

Rosenstone and Hansen (1993) argue that any theory of political participation is incomplete without a consideration of strategic mobilization—how, when, and why political elites choose to mobilize American citizens into political life. Consistent with their argument, people who are contacted by a political group or candidate are more likely to participate in politics than those who have not been contacted (Berelson, Lazarsfeld, & McPhee, 1954; Chong, 1991; Johnson, 1998; Lee, 2002; Schlozman & Tierney, 1986). More recently, Green and Gerber (2008) have shown that face-to-face get-out-the-vote (GOTV) efforts are more effective at encouraging people to turn out to vote than phone or mail communications. An important, sizable body of work further confirms that face-to-face messages that involve social pressure have a positive, statistically significant effect on turnout compared with messages that do not involve social pressure (Davenport et al., 2010; Gerber, Green, & Larimer, 2008; Gerber & Rogers, 2009).

Political Engagement

One of the most consistent predictors of who is likely to become politically active is political engagement. Political interest, knowledge, and efficacy, as well as concern about the election outcome, civic duty, and strength of partisan identification, are all positively correlated with a variety of political acts (see Campbell, Converse, Miller, & Stokes, 1960; Finkel & Opp, 1991; Milbrath & Goel, 1977; Verba et al., 1995).

Additional Predictors of Participation

Recent research has examined the role of the Big Five personality traits in predicting political behaviors (e.g., Gerber et al., 2011; Mondak, 2010). The emphasis of this research is on explaining the myriad ways personality affects political behavior and not on incorporating personality into a comprehensive theory of participation. Moreover, given the fact that personality traits are stable over time (Gerber, Huber, Doherty, & Dowling, 2012), personality, ipso facto, cannot explain participation instability.²

Putting It All Together

In an empirical test of their Civic Voluntarism Model (CVM), Verba et al. (1995) use a cross-sectional survey to show that resources/skills, recruitment, and political engagement are all positively correlated with a variety of political activities. In summarizing their model, Verba et al. (1995) argue that

all three components of the model are important. However, we place greater emphasis on the resources that facilitate participation and on the variety of psychological predispositions toward politics that we label “political engagement” . . . with respect to resources and engagement, for several reasons we place greater stress on the former . . . (p. 270)

The authors go on to argue that resources can be measured more reliably and validly, that the causal priority of resources can be more readily established, and that a resource theory has significant substantive relevance (see also Schlozman et al., 2012).³

The Quandaries Presented by Extant Research and the Assumption of Stability

Verba et al.’s (1995) full model, which regresses participation on resources, skills, political engagement, and recruitment, does an adequate job of explaining participation (the R² in the comprehensive test of the model is .45; p. 389). With that said, it is important to emphasize that the authors limit their theoretical emphasis to resources and skills at the expense of the other commonly studied antecedents included in their full model.

Moreover, Verba et al.’s data, as well as those of most past studies of the antecedents of political participation, are cross-sectional. Whether our theories of participation are constrained by available data, or whether our data collection efforts are constrained by our theories, we argue that dominant theories in the discipline conceptualize political behavior as if it were a one-time decision made anew every time. This would be acceptable if Time 2 (T₂) participation is uncorrelated with Time 1 (T₁) participation, and if participation is relatively stable at the individual level.

Although there seems to be an infinite number of studies of political behavior at the cross-sectional level, there are far fewer studies that allow for examinations of the micro-level stability (or instability) of participation in the United States. Rosenstone and Hansen (1993) make use of the 1956-1958-1960 and 1972-1974-1976 ANES panels to examine the stability of political participation (voting, trying to influence how others vote, contributing money to a party/candidate, attending a meeting/rally, and working for a party/candidate) at the individual level. They find that, outside of voting, very few people participate consistently across the 3 election years (and even across 2 presidential years), concluding that “put differently, only one participant in three participates regularly; over two-thirds of the participants take part sporadically. The core group of people that participates election after election . . . is remarkably small” (p. 55).

Investigating the Explanations of Instability

Despite the significant amount of instability in participation at the individual level, participation theories have only been tested using cross-sectional data. As such, we are left to assume that the predictors that affect behavior at T₁ will predict in the same way at T₂. There has been little discussion about the possibility that the behaviors themselves, or the criteria that go into the decision to participate (or not), are linked to one another across time (one exception is work that suggests that voting is habitual; Aldrich, Montgomery, & Wood, 2011; Cravens, 2013; Fowler, 2006; Gerber, Green, & Shachar, 2003). In the studies below, we test the theory that resources/skills, recruitment, and engagement (particularly resources) are less able to explain individual-level ebbs and flows in participation over time than they are at explaining cross-sectional participation.

Our reasoning is straightforward—these antecedents are unlikely to vary enough at the individual level to be able to explain the instability in participation that exists. Nowhere is this more evident than with regard to resources and civic skills, the very set of variables that Verba et al. (1995) argue should be front and center of any comprehensive model of political participation. Logic alone tells us that adult education levels are incredibly stable. To the extent that education levels change, they only do so in a positive direction. A similar logic applies to civic skills.⁴ We also suspect that income, recruitment, and engagement are too stable at the individual level to explain instability in participation that we see, even over short time periods.⁵

Rosenstone and Hansen (1993) make a similar argument, theorizing instead that participation instability is due to changes in mobilization efforts by elites. They show that political participation is characterized more by instability than stability at the individual level. However, they do not put their theory that instability is due to changes in elite mobilization to an empirical test at the individual level.⁶

We depart from Rosenstone and Hansen in that we theorize that, at the individual level, mobilization efforts are also unlikely to explain participation instability, as we suspect that candidates and parties will continue to mobilize the people they have been successful at mobilizing in the past. Our contribution, therefore, is to provide the first empirical test of whether the variables that have been the focus of much of the cross-sectional work on the antecedents of political behavior have the same power to explain individual instability in participation over time.

In the sections that follow, we first replicate Rosenstone and Hansen’s findings regarding participation instability using the 2000-2002-2004 ANES panel and the second, third, and fourth waves of the “youth” component of the Jennings Youth–Parent Socialization panel. We then, for the first time, put resources/skills, recruitment, and political engagement to the test in models aimed at explaining participation instability. Finding them lacking in explanatory power, we argue for a revision of resource-based theories that incorporates motivation, citing recent research that has begun to address the motivational underpinnings of political participation.

Study 1: ANES Panel

Method

Data

We begin with the ANES 2000, 2002, and 2004 Panel Study (Interuniversity Consortium for Political and Social Research [ICPSR], 21500). These data are advantageous because they include measures of nine political behaviors and many of the proposed covariates, including resources, civic skills, recruitment, engagement, and demographics, at three time points. Our main interest is in testing whether change in traditional antecedents of participation explains change in political behavior at the individual level. We do these analyses by looking at change over three wave pairs: 2000-2002, 2000-2004, and 2002-2004.⁷

Dependent variables

Our dependent variables consist of nine political behaviors that were available over all three strobes of the 2000-2004 panel: voting, trying to influence others, displaying a yard sign/button, going to meetings/rallies, doing other campaign work, contributing money to candidate, contributing money to party, contacting a public official, and attending a protest or march.⁸ Respondents were asked at each time point whether they had or had not engaged in each behavior during the current election season (coded such that 0 = no; 1 = yes).

We are mindful of two types of measurement error that could arise when respondents are asked to report whether they engaged in particular political behaviors over the course of an election season. The first is random measurement error, which could occur because respondents’ memories are imperfect. As such, they may mistakenly report that they had (or had not) engaged in a particular behavior. Some of the variability in behaviors over time may therefore be an artifact of random error and could interfere with our ability to detect significant effects.

The second type of measurement error that could arise is systematic; respondents may consistently misreport that they had engaged in a political behavior to portray themselves as good, civically engaged citizens (i.e., socially desirable responding). Systematic overreporting (or underreporting) of behaviors would create the appearance of less individual-level variability and could also interfere with our ability to detect significant effects. There is strong evidence that people do, in fact, systematically overreport whether they have voted (e.g., Ansolabehere & Hersh, 2011, 2012; Bernstein, Chadha, & Montjoy, 2001; Silver, Anderson, & Abramson, 1986; Vavreck, 2007). One of the reasons people may say they voted when they did not is because voting is a socially desirable action in the United States. As such, individuals may overreport as a way to “save face” with an interviewer or possibly themselves (as in the case of an Internet or mail survey).⁹

Evidence of systematic self-report bias of other political behaviors is more sparse. Verba et al. (1995) argue that overreporting is less of a problem for other political activities besides voting, but do not provide direct evidence to support their claim. In contrast, Enos and Hersh (2015) speculate that there is overreporting in self-reports of working for a party or campaign, but also do not provide direct evidence in this regard. Volgy and Schwarz’s (1984) analysis of engagement in local activities (petition signing, participating in public hearings, and contacting officials) found that 57 of 204 people surveyed (28%) misreported that they had participated when they had not. Interestingly, Pierce and Lovrich (1982) found that people systematically underreported a seemingly socially acceptable behavior of signing an initiative petition.

If one reason people overreport that they engaged in a political activity is because they want to look and/or feel like good democratic citizens, then we might expect overreporting of other socially desirable political actions (such as, possibly, contacting an elected official) but less overreporting of political behaviors that are not as strongly linked to civic duty (such as, possibly, contributing money to or volunteering for a candidate). Socially desirable responding could even work in reverse; people may systematically underreport their engagement in political activities that are looked down upon in the United States (e.g., protesting may be one such activity).

Unfortunately, we do not have variables in either of our datasets that would allow us to model the tendency to engage in socially desirable responding independent of the variables of theoretical interest in our analyses (and neither panel included any validated behavior measures).¹⁰ However, we attempted to mitigate concerns about random and systematic measurement error in a few ways (recognizing, of course, that we cannot completely rule out measurement error as a possible explanation for the results reported below).

First, given that indices of like measures have higher measurement reliability than single items (e.g., Spector, 1992), we constructed three behavior indices by averaging the total number of behaviors for each respondent at each wave (the resulting indices range from 0 to 1; the alphas for the behavior indices for the 2000, 2002, and 2004 waves are .60, .62, and .65, respectively). If our results are consistent across the nine individual behaviors (some of which we presume would be easier to recall or more subject to socially desirable responding than others) and the behavior index, we will be more confident that the findings are not primarily the result of random or systematic measurement error. Second, we also relaxed the threshold for rejecting the null hypothesis (by using 90% confidence intervals) to increase the chances of detecting significant effects. Finally, as we discuss below, our results replicated when we re-ran the models after removing respondents whom we suspected were the most likely to have been “overreporters.”

Independent variables

With regard to resources, we use a measure of income and a dichotomous measure of employment (coded 1 if the respondent is currently employed and 0 if the respondent is not employed). Education was only assessed in 2000 (but, as we noted above, we expect the least amount of change in education levels, compared with the other predictors, in the short 4-year time span).

With regard to civic/institutional skills, we have measures of whether respondents had (in the previous 6 months) (a) given a presentation or speech at their place of worship, (b) planned a meeting for their place of worship, (c) given a presentation or speech for their job, and (d) planned a meeting for their job. We averaged these four yes/no measures to construct an index of institutional skills (ranging from 0-1). The skills measures are only available for 2002 and 2004.

We assessed recruitment with measures of whether respondents had been contacted by a political party and whether anyone had contacted them about registering or turning out to vote (coded such that 0 = no; 1 = yes). Finally, with regard to engagement, we use measures of how often respondents follow politics and public affairs, political efficacy, and strength of partisan identification.¹¹ All of the independent variables were coded to range from 0 to 1.

Constructing the change measures

For both the dependent and independent variables, we construct change variables by subtracting the T₁ score from the T₂ score (for each of the three wave pairs in the panel). For the dichotomous dependent variables, −1 indicates that the person engaged in the behavior at T₁ but not at T₂, 0 indicates that the person either engaged in the behavior in both years or in neither year, and + 1 indicates that the respondent engaged in the behavior at T₂ but not at T₁. For the behavior index and independent variables, the resulting change score ranges from −1 to +1, with negative numbers indicating “less” of the construct at T₂ than T₁, 0 indicating the same amount of the construct at both time points and positive numbers indicating “more” of the construct at T₁ than at T₂.

Control variables

We also included the following control variables: education, age, retirement, Catholic religious affiliation, gender, race, ethnicity, frequency of religious attendance, and non-political organizational affiliation. We are treating these variables as “static,” either because they are, in fact, static (e.g., gender), or because they were assessed only once during the three-wave panel (e.g., education), and/or because they are best viewed as antecedents of resources/skills or political engagement, our main change variables of interest (e.g., religious attendance, non-political affiliations; see, for example, Verba et al., 1995).¹²

Results

Variability in participation

Table 1 displays the distributions of each of the behavior-change variables (for the index and each of the individual behaviors) for each of the three wave pairs to examine whether the instabilities Rosenstone and Hansen (1993) uncovered still hold.¹³ As Table 1 shows, the change variables for the behavior index are approximately normally distributed and show considerable variability across the 3 year pairs. Between 63% and 69% of respondents evidenced change on the nine-behavior index, with between 21% and 28% of them engaging in less than two or more, or greater than two or more, behaviors between the year pairs.

Table 1.

Distribution of Respondents (Percentages) at Each Level of the Behavior-Change Dependent Variables, ANES 2000-2004.

Political behavior index (−9 to +9)	Index-change score
	Neg. numbers = less behaviors at T₂			0 = No change	Pos. numbers = more behaviors at T₂			Total change	Valid N
	≤−3.0	−2.0	−1.0	0.0	+1.0	+2.0	≥+3.0	Total change	Valid N
2000-2002	3.8	7.8	23.3	37.1	18.2	6.5	3.1	62.9	807
2000-2004	2.2	3.7	14.0	34.9	26.2	10.2	9.2	65.1	632
2002-2004	1.4	3.4	15.3	30.8	25.6	14.2	8.5	69.2	692
Individual behaviors			Behavior-change score
			−1	0	1
			(Did behavior at T₁, did not at T₂)	(No change)	(Did not do behavior at T₁, did at T₂)			Total change	Valid N
Voting
2000-2002			11.5	86.9	1.6			13.1	792
2000-2004			2.9	92.2	5.0			7.8	625
2002-2004			2.0	84.2	13.7			15.8	692
Influence others
2000-2002			19.5	65.4	15.1			34.6	807
2000-2004			12.0	62.3	25.6			37.7	632
2002-2004			9.4	63.3	27.3			36.7	689
Sign or button
2000-2002			7.6	84.9	7.6			15.1	806
2000-2004			5.9	78.6	15.5			21.4	632
2002-2004			4.8	79.2	16.0			20.8	692
Meeting or rally
2000-2002			4.8	91.6	3.6			8.4	807
2000-2004			3.5	87.6	8.9			12.4	631
2002-2004			2.5	88.7	8.8			11.3	690
Campaign work
2000-2002			2.9	93.3	3.8			6.7	806
2000-2004			3.2	92.6	4.3			7.4	632
2002-2004			4.1	92.0	3.9			8.0	691
Contributing US$ to candidate
2000-2002			5.2	90.4	4.3			9.6	805
2000-2004			4.9	85.2	9.8			14.8	630
2002-2004			3.9	86.2	9.9			13.8	690
Contributing US$ to Party
2000-2002			4.2	89.2	6.6			10.8	806
2000-2004			3.5	86.4	10.1			13.6	632
2002-2004			4.3	88.0	7.7			12.0	690
Contact official
2000-2002			9.7	77.8	12.5			22.2	806
2000-2004			11.1	73.2	15.7			26.8	631
2002-2004			11.6	76.1	12.3			23.9	690
Protest
2000-2002			2.5	95.2	2.4			4.8	807
2000-2004			2.5	93.5	4.0			6.5	631
2002-2004			2.2	95.1	2.7			4.9	691

Source. ANES 2000, 2002, and 2004 Full Panel Study, Study Number 21500, ICPSR.

Note. The sum of changes may not add up to 100% due to rounding error. Each cell entry represents the percentage change over each year pair for each dependent variable, moving from T₁ to T₂. The “0” column is equal to the percentage of respondents demonstrating no change over the year pair for each behavior and the index. For the individual behaviors, +1 indicates the percentage of respondents moving from not doing the behavior in the first year of the wave pair to doing so in the second, while −1 is the converse. The behavior index, then, is an aggregation of these nine behavior-change scores within the wave pair, so the positive and negative scores range from −9 (doing every behavior at T₁ and not doing any at T₂) to +9 (doing every behavior at T₂ while not having done any at T₁). ANES = American National Election Study.

Scanning the total change column for the individual behaviors (see Table 1), we see a good bit of stability (although the bulk of this stability comes from people who never engage in the behavior). But there is also a non-trivial amount of instability, which varies from behavior to behavior. Attempting to influence how someone else is going to vote, displaying a yard sign or wearing a political button, and contacting a public official demonstrate the greatest instability, and working for a political candidate and participating in a protest show not only the least variability but also the smallest percentage of participants.¹⁴

Explaining the variability in participation

Our interest is in testing whether the most commonly studied predictors of participation at the cross-sectional level can explain variations in behavior over time. To do this, we estimate change models in ordinary least squares (OLS). Here, the dependent variables are change in participation from T₁ to T₂ across the three 2-year wave pairs (for the index and each of the separate behaviors), as described above. As independent variables we include the resources, skills, recruitment, and engagement change measures, as well as the aforementioned controls. For ease of interpretation, we report only the direction and significance level of the statistically significant coefficients for the variables of interest (see Table 2; the coefficients table appears in Online Appendix Table 1).

Table 2.

Covariates of Political Activities, American National Elections Full Panel Study, 2000, 2002, 2004.

Time span	Behavior index			Vote			Influence others			Sign or button			Meeting or rally			Campaign work			Contributing US$ to candidate			Contributing US$ to party			Contact official			Protest
Time span	1	2	3	1	2	3	1	2	3	1	2	3	1	2	3	1	2	3	1	2	3	1	2	3	1	2	3	1	2	3
Resources
Income																							p
Employment			n	N					N						n						n				P
Institutional skills			P												P			P
Recruitment
Contacted by party	P			P	P		P						P			P			P			p		p				P	p
Contacted to vote/register								N																	P					p
Engagement
How often follow politics	P		P				P	P	P	p									P
PID strength			n								N													N
Efficacy	P						P		P							P													P

Source. ANES 2000, 2002, and 2004 Full Panel Study, Study Number 21500, ICPSR.

Note. Table entries represent signs (P = positive, N = negative) on coefficients from OLS multiple regression analyses. Empty cells correspond to p values > .10. Capital letters (P or N) indicate p ≤ .05. Italicized lowercase letters (p or n) indicate .05 < p ≤ .10. Time spans are indicated below activity type, and are indexed as follows: 1, 2000-2002; 2, 2000-2004; 3, 2002-2004. Coefficient estimates with standard errors and other model statistics are reported in Online Appendix Table 1. Institutional skills measures available only in 2002 and 2004, and are therefore only included in Time Span 3. ANES = American National Election Study; OLS = ordinary least squares; PID = Party Identification

The simplest way to sum up these results is that neither change in resources, nor skills (for the one wave pair in which we have measures of skills), nor recruitment, nor political engagement is a consistent predictor of micro-level changes in participation. Most notably, the resource variables (income, employment, and institutional skills) are underperformers. Income is uniformly non-significant (income change does not even predict change in financial contributions to a candidate or political party, save for a marginally significant coefficient for change in contributions to a party in 2000-2004). Another resource variable, employment, is also a poor predictor, and is negatively associated with change in participation in 5 of the 10 models in 2002-2004. Finally, in 2002-2004, the one year pair in which we have a measure of change in institutional skills, the variable is positively associated with change in only 3 of the 10 models.

There are also a few patterns in the results to note. First, change in being contacted by a political party predicts 8 of the 10 behavior-change variables in the 2000-2002 wave pair; we hesitate to make too much of this pattern because it does not carry over to the 2000-2004 or 2002-2004 analyses. Second, and perhaps more interestingly, change in the engagement variables (following politics and efficacy, but not strength of partisan identity) is consistently and positively related to behavior change for the index and for influencing others. One way to think about the engagement variables is that they capture citizens’ general propensity toward following, caring about, and being efficacious with regard to politics. Viewed in this way, it makes sense that change in engagement would be consistently (albeit not always) and positively related to change in the behavior index—a general measure of citizens’ propensity to participate in politics. Moreover, changes in following politics and in efficacy are also consistently and positively related to the social and expressive act of attempting to influence others to vote a certain way. Again, we hesitate to make too much of the pattern of association between the engagement variables and influencing others, because the pattern does not replicate with other social and/or expressive behaviors such as displaying a political sign/button and attending a meeting/rally.

The lede, however, is the remarkable reliability of the non-findings. We find essentially no evidence that the most oft-theorized about and empirically confirmed antecedents of participation at the cross-sectional level (especially resources and skills) have any sizable and/or reliable explanatory power with regard to micro-level variability in political participation between election years. This is the case across wave pairs that both include and do not include a presidential election, across the nine separate behaviors (some of which are more prone to overreporting or recall errors than others) and the index (which has higher measurement reliability), and across multiple measures of the same underlying construct.¹⁵

Addressing systematic measurement error

From our theoretical standpoint, these null findings make sense (and were predicted)—resources, skills, and engagement are unlikely to vary enough at the individual level to be able to explain instability in participation over time. However, an alternative explanation for the null findings has to do with measurement error. In particular, self-reports of political behaviors may be subject to social desirability bias, such that respondents will overreport their activity levels to look like good democratic citizens to themselves or an interviewer. Such overreporting would give the appearance of greater participation stability and artificially decrease the amount of variability in participation that could potentially be explained by variation in the independent variables.

We do not have a direct way to control for respondents’ tendency to engage in socially desirable responding in the ANES (or the Jennings) dataset. However, we conducted a robustness check by re-estimating our models after removing respondents who we speculated were most likely to be engaging in socially desirable responding. Specifically, we know from past research on self-monitoring (e.g., Snyder, 1979) that socially desirable responding is situationally specific, based in part on the respondent’s perceptions about the social norms governing each behavior. Based on this research, we can assume that respondents who overreport an activity at T₁ as a result of wanting to make a good impression will consistently engage in overreporting of that same behavior at subsequent time points. Therefore, ANES respondents who report that they engaged in a particular behavior in all three panel waves are the ones who are most likely to be engaging in systematic overreporting, which would artificially decrease the amount of variance in our dependent variables. We therefore re-estimated our models after removing all respondents who said that they engaged in a behavior in all three waves (we did not re-estimate the model for the behavior index, as the “consistent-yes” measure is not comparable for the sum of all the behaviors).

The analyses reported above assume that there is no systematic measurement error (due to socially desirable responding) in our data—a strong assumption to be sure. In re-estimating our models, we are making the opposite strong assumption that all respondents who reported that they engaged in a behavior in all three waves are providing inaccurate data (this assumption is especially dicey in the case of voting, which has an habitual component; for example, Aldrich et al., 2011; Gerber et al., 2003). The truth is undoubtedly in between these two extremes; in the absence of measures that would enable us to model socially desirable responding directly, we see dropping the “three-yes” respondents as the next-best alternative. To the extent that the results of both extremes converge, we can be more confident (with appropriate caveats) that our null findings are not an artifact of systematic measurement error.

It is telling to note the percentage of “three-yes” respondents for each behavior. At 69%, voting is the behavior with the highest percentage of consistent yes-responders across the three waves. This is to be expected, given both the habitual nature of voting (e.g., a good number of consistent yes-responses are, in fact, truthful) and the evidence that voting is subject to a considerable amount of overreporting). Interestingly, the drop-off beyond voting is dramatic. The behavior with the second highest percentage of consistent yes-responders is influencing others (17%), followed by contacting an official (10%), wearing a button and contributing money to a party (each 3%), attending a meeting/rally and contributing money to a candidate (each 1.5%), protesting (1%), and working for a candidate/party (0.5%).¹⁶ The paucity of respondents who report that they participated in an activity in all three waves is consistent with Rosenstone and Hansen’s (1993) analyses of the 1956-1958-1960 and 1972-1974-1976 ANES panels.

Given the small percentage of respondents dropped from each analysis (besides the ones for voting), it is not surprising that the results of the re-estimated models are remarkably similar to the results reported above. In sum, across the analyses for all behaviors and all three wave pairs, the re-estimated models yield two statistically significant coefficients that are not present in the original analyses and six coefficients that drop from statistically significant to not significant (complete results available upon request). Although these new analyses are not definitive proof that systematic measurement error is not at least partly responsible for our null findings, they do give us more confidence that our findings are not an artifact of social desirability.

Variability in antecedents

One could argue that another reason for our null findings is that the 4-year time period of the ANES panel is not long enough for variables such as income, institutional skills, and recruitment to have a chance to vary. The distributions of the change variables across the three wave pairs can be found in Online Appendix Table 2. There is a good amount of variation in the antecedents; the level of variability is quite consistent across the three wave pairs and tends to be skewed in the positive direction.¹⁷ To check whether our null findings are an artifact of the shorter time periods of the ANES panel, we replicate our analyses using another panel study that was conducted over a longer time period and with longer time spans between waves.

Study 2: Jennings Panel

Method

Data

We turn now to the Jennings Youth–Parent Socialization Panel Study, 1965-1997: Four Waves Combined (ICPSR 4037), which is a panel survey of high school seniors and their parents that began in 1965; for this study, we are only using the youth component, and only the 1973, 1982, and 1997 waves—the only ones that contain the participation variables of interest.

Dependent variables

Our dependent variables consist of 10 political behaviors that were available across all three waves of the Jennings Panel Study: voting, trying to influence others, displaying a sticker/button, going to meetings/rallies, doing any other campaign work, contributing money to a party or candidate, contacting a public official, attending a protest or march, writing a letter to the editor, and engaging in community action.¹⁸ Respondents were asked at each of the three time points whether they had or had not engaged in each behavior during the most recent election cycle (coded such that 0 = no; 1 = yes).

Random measurement error due to respondents’ inability to accurately recall whether or not they had engaged in a behavior may be more of a concern in the Jennings Panel, given that the waves were not timed to be concurrent with an election. Therefore, as with the ANES study, we constructed three behavior indices (with higher measurement reliability than the individual items) by averaging the total number of behaviors for each respondent at each wave; the resulting indices range from 0 to 1. The alphas for the behavior indices for the 1973, 1982, and 1997 waves are .74, .74, and .78, respectively.

Independent variables

With regard to resources, we measure employment (coded 1 if the respondent was employed and 0 if the respondent was not employed at the time of the survey), education (coded 1 if respondents obtained additional education between the waves and 0 if they had not), and income. No measures of recruitment/institutional skills are available for Study 2.

With regard to political engagement, we have two indicators of political interest: the standard measure of self-reported political interest, and how often respondents said they follow politics and public affairs, as well as measures of political efficacy and strength of partisan identification (all recoded to range from 0-1). The Jennings panel also included measures of political knowledge. To maintain measurement consistency, we created a summary index of the five knowledge questions that were worded the same way in all three waves (thus creating a measure of the percentage of correct answers, which ranges from 0-1).

Constructing the change measures

Similar to Study 1, we construct change variables by subtracting the T₁ score from the T₂ score for both the dependent and independent variables (for the wave pairs in the panel—1973-1982, 1973-1997, and 1982-1997). All range from −1 to +1 (with 0 indicating no change) except for the education change independent variables, which have only two values (0 for no change and 1 for positive change).

Control variables

We also included the following control variables: Catholic religious affiliation, gender, race, frequency of religious attendance, and non-political organizational affiliation, parents’ education, exposure to politics at home, and high school political activity. We treat these variables as “static” for the same reasons we do so in Study 1.¹⁹

Results

Variability in participation and its antecedents

Table 3 displays the distributions of the behavior index and individual behavior-change variables (see Online Appendix Table 3 for the distributions of the antecedent change variables). As would be expected, given the longer time span of the Jennings panel compared with the ANES panel, there is more instability in the Jennings behavior index (with between 79% and 83% of respondents engaging in more or less behaviors than the preceding wave for each of the three year pairs compared with 63%-69% in the ANES). Between 42% (in the 1973-1982 wave pair) and 55% (in the 1973-1997 wave pair) of the respondents engaged in two or more (or two or less) behaviors out of 10 (see Table 3). This is a sizable amount of change. As Table 3 also indicates, the individual behavior variables also show a good bit of instability. Writing a letter to an editor and participating in a protest showed the least variability (ranging from approximately 9% to 18% change in either direction), whereas influencing others, displaying a yard sign or wearing a campaign button, contacting a public official, and participating in a community action showed the greatest instability (ranging between 29% and 40% change in either direction). Similarly, there was much more instability in antecedents in Study 2 than in Study 1 (see Online Appendix Table 3), most notably with regard to income (but see Note 17).

Table 3.

Distribution of Respondents (Percentages) at Each Level of the Behavior-Change Dependent Variables, Jennings Panel, 1973-1997.

Political behavior index (−10 to +10)	Index-change score							Total change	Valid n
	Neg. numbers = less behaviors at T₂			0 = No change	Pos. numbers = more behaviors at T₂
	≤−3.0	−2.0	−1.0	0.0	+1.0	+2.0	+≥3.0
1973-1982	7.7	8.6	17.7	21.6	18.7	11.0	14.9	78.6	509
1973-1997	6.2	7.7	10.5	17.1	17.3	15.6	25.7	83.0	591
1982-1997	4.4	9.1	11.5	21.5	20.0	14.4	19.3	78.7	517
Individual behaviors	Behavior-change score							Total change	Valid n
			−1	0	1
			(Did behavior at T₁, did not at T₂)	(No change)	(Did not do behavior at T₁, did at T₂)
Voting
1973-1982			5.6	77.7	16.7			22.3	516
1973-1997			5.9	74.3	19.8			25.7	595
1982-1997			6.0	83.7	10.3			16.3	517
Influence others
1973-1982			20.8	65.4	13.7			34.5	518
1973-1997			16.3	65.9	17.8			34.1	596
1982-1997			9.1	70.9	20.0			29.1	516
Sticker or button
1973-1982			19.1	66.9	13.9			33.0	517
1973-1997			15.6	81.4	18.6			34.2	596
1982-1997			9.3	71.5	19.2			28.5	516
Meeting or rally
1973-1982			9.2	76.4	14.3			23.5	518
1973-1997			11.2	67.1	21.6			32.8	596
1982-1997			10.5	71.5	18.0			28.5	516
Campaign work
1973-1982			9.5	79.5	11.0			20.5	518
1973-1997			6.7	73.5	19.8			26.5	596
1982-1997			6.8	75.4	17.8			24.6	516
Contributing US$ to party/candidate
1973-1982			7.7	74.0	18.3			26.0	519
1973-1997			6.7	66.3	27.0			33.7	596
1982-1997			7.0	73.5	19.5			26.5	517
Contact official
1973-1982			10.1	65.6	24.4			34.5	517
1973-1997			6.1	61.4	32.5			38.6	594
1982-1997			11.6	63.6	24.8			36.4	516
Protest
1973-1982			12.7	82.8	4.4			17.1	518
1973-1997			12.4	82.4	5.2			17.6	596
1982-1997			4.8	87.2	7.9			12.7	517
Letter to editor
1973-1982			4.6	90.8	4.6			9.2	519
1973-1997			3.7	88.9	7.4			11.1	596
1982-1997			2.9	90.1	7.0			9.9	517
Community action
1973-1982			10.8	64.0	25.1			35.9	517
1973-1997			11.4	59.4	29.1			40.5	594
1982-1997			15.5	65.3	19.2			34.7	516

Source. Jennings Youth–Parent Socialization Panel Study, 1965-1997 (Waves 2, 3, and 4), Study Number 4037, ICPSR.

Note. The sum of changes may not add up to 100% due to rounding error. Each cell entry represents the percentage change over each year pair for each dependent variable, moving from T₁ to T₂. The “0” column is equal to the percentage of respondents demonstrating no change over the year pair for each behavior and the index. For the individual behaviors, +1 indicates the percentage of respondents moving from not doing the behavior in the first year of the wave pair to doing so in the second, while −1 is the converse. The behavior index, then, is an aggregation of these 10 behavior-change scores within the wave pair, so the positive and negative scores range from −10 (doing every behavior at T₁ and not doing any at T₂) to +10 (doing every behavior at T₂ while not having done any at T₁).

Explaining the variability in participation

Table 4 displays the direction and significance level of the statistically significant coefficients for the variables of interest for the behavior index and each of the 10 separate activities for the three wave pairs (the table of coefficients appears in Online Appendix Table 4). As with Study 1, the headline is the remarkable reliability of the non-findings across all 10 behaviors (some of which are more prone to overreporting or recall errors than others) and the index (which has higher measurement reliability).²⁰ Even with more behavioral variability to explain, and more variability in resources and engagement with which to explain it, neither resources nor engagement is a consistent predictor of micro-level changes in participation. As with Study 1, the resource variables underperform relative to the engagement variables.

Table 4.

Covariates of Political Activities, Jennings Youth–Parent Socialization Study, 1973, 1982, 1997.

Time span	Behavior index			Vote			Influence others			Sticker or button			Meeting or rally			Campaign work			Contributing US$ to party/candidate			Contact official			Protest			Letter to editor			Community action
	1	2	3	1	2	3	1	2	3	1	2	3	1	2	3	1	2	3	1	2	3	1	2	3	1	2	3	1	2	3	1	2	3
Resources
Income		P		P															P	p					n
Education			P												P		P	P		P		p	P		n
Employment						p																											N
Engagement
Interest		P	P						p		P	p		p				p					P	P
Follow politics																		N															p
PID strength	P	P			P		P	P		P	P								P	P	P
Efficacy	P		P	P			p						p		P	p		p
Knowledge			P				p		P						P	N					p

Source. Jennings Youth–Parent Socialization Panel Study, 1965-1997 (Waves 2, 3, 4), Study Number 4037, ICPSR.

Note. Table entries represent signs (P = positive, N = negative) on coefficients from OLS multiple regression analyses. Empty cells correspond to p values > .10. Capital letters (P or N) indicate p ≤ .05. Italicized lowercase letters (p or n) indicate .05 < p≤ .10. Time spans are indicated below activity type, and are indexed as follows: 1, 1973-1982; 2, 1973-1997; 3, 1982-1997. Coefficient estimates and other model statistics are reported in Online Appendix Table 4. OLS = ordinary least squares.

Despite the relative lack of significant effects, a few patterns do emerge. As with the ANES results, the set of engagement change variables are consistent (albeit not always) predictors of change in the behavior index and attempting to influence others how to vote. This pattern also extends to the additional social/expressive behaviors of displaying a sticker or button, attending a meeting or rally, and, to a lesser extent, volunteering for a campaign. Among the engagement variables, the one that shows the most consistent effects is strength of party identification, which is positively associated with 5 of the 11 behavior-change variables in the 1973-1997 wave pair and 4 of the 11 in the 1973-1982 wave pair. This is quite different from Study 1, where strength of party identification had no positive effects (and three negative effects). In addition, changes in political interest and efficacy are also consistent predictors of social/expressive behaviors. Interestingly, whereas change in following politics was one of the most reliable predictors in Study 1, it had virtually no effect in the Jennings panel.

Summary of Results

To summarize our results across the two studies, we counted the number of statistically significant coefficients (including ones with a significance level of p < .10; see Table 5). The percentage of significant, positive coefficients in the ANES data for the resources/skills, recruitment, and engagement variables was 7%, 22%, and 13%, and the percentage of significant negative coefficients was 7%, 2%, and 3%, respectively. In the Jennings data, the percentage of positive, significant coefficients for resources/skills and engagement was 12% and 20%, respectively, and the percentage of negative, significant coefficients was 3% and 1%. We see no consistent evidence that the antecedents of participation that are at the forefront of resource-based theories explain micro-level ebbs and flows in behavior over time, except for the engagement variables, which did a slightly better job than did resources/skills and recruitment.²¹ These null findings are robust across two different panels, collected over different time periods, with different time spans between waves.

Table 5.

Count of Significant Coefficients for Multiple Regression Models of ANES and Jennings Panel Data.

	ANES	Jennings	ANES	Jennings
	Total positive	Total positive	Total negative	Total negative
Resources/skills	5/70 (7%)	12/99 (12%)	5/70 (7%)	3/99 (3%)
Recruitment	13/60 (22%)		1/60 (2%)
Engagement	12/90 (13%)	33/165 (20%)	3/90 (3%)	2/165 (1%)

Source. ANES 2000, 2002, and 2004 Full Panel Study, Study Number 21500, ICPSR, and the Jennings Youth–Parent Socialization Panel Study, 1965-1997 (Waves 2, 3, and 4), Study Number 4037, ICPSR.

Note. Skills measures are only available for 2002-2004 ANES analyses. Results in our counts are inclusive of all coefficients with significance levels of p < .10. ANES = American National Election Study.

These findings point to what we see as a “boundary condition” of resource-based theories: Resources do a good job of explaining participation differences between individuals, but they are less effective at explaining participation within individuals. As such, extant theories of political participation are, at best, incomplete, insomuch as they cannot explain individual-level variance in political behavior over time.

Discussion

Taken as a whole, extant political participation theories argue that resources (both SES and institutional skills), mobilization efforts by elites, and political engagement (such as political interest and efficacy) are the root causes of participation. In weighing the relative importance of these three categories of variables, the authors of the most comprehensive theory to date, the CVM (Verba et al., 1995), argue for the primacy of resources and skills (see also Schlozman et al., 2012). However, the considerable amount of evidence that has accumulated in support of the proposal that resources/skills, recruitment, and political engagement are all correlated with political participation has been primarily cross-sectional. Not only do cross-sectional data provide limited empirical leverage with regard to causality—to wit, one of the reasons Verba et al. (1995) argue for the primacy of resources is that it is easier to establish, via logic, that resources cause participation and not the reverse—but studying the antecedents solely at the cross-sectional level obscures the empirical fact that participation in the United States is more accurately characterized by instability, rather than by stability, at the micro-level.

We theorize that resources and the like may not be able to predict micro-level variations in behavior because such variables change much more slowly over time than does participation. In the first empirical test of the explanatory power of these variables using two panel studies, we find that changes in resources, recruitment, and political engagement are not consistently and/or meaningfully related to micro-level changes in participation. As such, resource-based models like the CVM, as comprehensive models of political participation, are at best incomplete.

Providing convincing evidence in support of a null hypothesis is tricky business, to be sure. We consider three alternatives: (a) Our data lack the statistical leverage necessary to adequately reject the null hypotheses being tested here; (b) the variability in the dependent variables is an artifact of random measurement error, thus any attempt to model the observed (random) variability would fail, regardless of the plausibility of the theory; and (c) respondent reports of political behaviors are subject to systematic overreporting due to social desirability, which would suppress individual-level variability, making it difficult to detect significant effects on that reduced variability.

We find the lack of statistical leverage argument unpersuasive for two reasons. First, the number of cases available for analysis is quite sizable in both studies (between 616 and 792 in Study 1 and 509 and 596 in Study 2), and, to give even a greater chance of detecting significant effects, we relaxed the threshold for rejecting the null hypothesis by using 90% confidence intervals. Second, and perhaps more convincingly, as a further check, we used the ANES and Jennings data to examine whether resources/skills, recruitment, and political engagement were significant predictors of participation at the cross-sectional level (as they should be, given the preponderance of the extant evidence in this regard).²² The cross-sectional findings largely replicate past research; we have no reason to suspect that we would need any more statistical power for the change models.

The cross-sectional findings also reduce our concerns about random measurement error. Specifically, if the dependent variables from two of the most credible and oft-utilized panel data sources available to our discipline were plagued with random measurement error (e.g., due to respondents misremembering whether they engaged in a political behavior or not), we would not have been able to use these data to replicate the findings of resource-based theories like the CVM at the cross-sectional level. Moreover, we would expect that if misremembering is an issue, it would be more of a problem for the Jennings data (for which the waves are at least 9 years apart)—but both the cross-sectional and time-series analyses yield similar results across the ANES and Jennings panels. Although we cannot completely rule out random measurement error as an alternative explanation for our null findings, we are relatively confident that the individual measures that constitute the change variables are high enough in reliability to mitigate concerns about artificially attenuated relationships.

Finally, our results are remarkably consistent across both datasets, as well as across individual behaviors (that vary in their susceptibility to overreporting) and the behavior indices (which have higher reliability). They also replicate when respondents who we suspect are the most likely to be engaging in socially desirable responding are dropped from the analyses. These consistencies reduce our concerns that systematic measurement error, and the concomitant decrease in over-time variability in behaviors, is appreciably affecting our ability to detect significant effects.²³

Future Research

If not the traditionally examined, resource-based antecedents, then what could explain the micro-level instability in participation over time? One possibility is that ebbs and flows in participation are due to short-term (and possibly idiosyncratic) campaign or context effects. Some campaigns “activate” people more than others, either because of the specific characteristics of the candidates or because the particular election context increases the salience of politics (e.g., an economic downturn, the country is at war, etc.). Although this is certainly part of the story, we would have expected some of these campaign effects to have been picked up more consistently by one or more of the political engagement variables in our models.

Instead, we propose a modification of resource-based theories that places individual-level motivation²⁴ at the forefront, paving the way for a better understanding of the variation in micro-level participation over time as well as the causes and consequences of participation at the cross-sectional level. Specifically, we assert that resource-based theories overemphasize the ability dimension of the decision to participate in politics at the expense of the motivation dimension. As such, we know a lot about who participates, but not as much about why they participate.

If we begin to think about the decision to participate (or not) as an iterative process, then the basic questions change. Not only are the “Have I been asked?” and “Do I have the resources?” questions important, but so, too, are questions such as “Do I want to participate?” and “Did I get what I wanted out of the last time I participated?” Specifically, we suggest that motivation (Why do people participate?), in conjunction with resources, may be the key to understanding variations in micro-level participation (see, for example, Han, 2009; Harder & Krosnick, 2008; J. M. Miller, 2013; Rogers, Fox, & Gerber, 2012; Sinclair, 2012).

Theories and empirical research from a variety of disciplines all converge on a threefold typology of motives likely to lead to political participation: instrumental, social/identity, and expressive (see, for example, Chong, 1991; Clark & Wilson, 1961; Clary et al., 1998; Klandermans, 2003; Salisbury, 1969; Teske, 1997; Verba et al., 1995). Instrumental motives are conceptualized as the desire to obtain a tangible reward (e.g., Olson, 1965). Social/identity motives include the desire to make or maintain friends and the desire to express one’s group membership and to be a “good group member” (J. M. Miller & Rahn, 2002). Expressive motives are defined as the desire to express one’s attitudes, values, or beliefs (e.g., Katz, 1960).

Measuring Motivation

One argument against incorporating motivation into theories of political participation is that the construct is too difficult to measure reliably and validly (Verba et al., 1995). This argument privileges an approach to science that places ease of measurement as a primary determinant of theoretical emphasis. In contrast, the psychology literature provides instruction regarding how to measure motives. Clary et al. (1998) have developed and validated a scale of motives for volunteerism (the Volunteer Functions Inventory [VFI]) using five questions to measure each of six motives for volunteering (value-expressive, instrumental, social, knowledge, ego-defensive, and self-esteem enhancement) similar to the three-part classification of motives described above. Sheagley, Miller, and Snyder (2010) adapt the VFI to measure motives for political participation and provide evidence for the predictive validity of their measure.

The Effects of Motivation on Participation

Previous survey and experimental research has shown that psychological motives derived from the VFI predict many types of participation above and beyond the resource-based models at the cross-sectional level. For example, using a cross-sectional survey with a representative sample of U.S. adults, J. M. Miller (2013) found that the value-expressive, self-interest, and social motives are correlated with engaging in political behaviors in theoretically expected ways, while controlling for the components of the CVM. Using an experimental design, J. M. Miller (2013) also provides evidence that these motives are causally related to willingness to engage in specific political behaviors in the near future.

Social/Identity Motives and Voter Turnout

A growing literature points to the social and identity motives as important determinants of voter turnout. For example, Rogers et al. (2012) synthesize their program of research involving GOTV field experiments spanning almost 15 years (beginning with Gerber and Green’s, 2000, seminal work) to propose a motivational model of turnout that hinges on two motives: (a) the desire to maintain feelings of affiliation and belonging, and (b) identity expression (Gerber & Rogers, 2009). The authors point to mounting field experimental evidence (e.g., Gerber & Green, 2000; Gerber et al., 2008; Green & Gerber, 2008; Nickerson, 2008; see also Bedolla & Michelson, 2012) to argue that the decision to vote is not a static one, but is influenced by social and self-perception factors that occur prior to and after the act of voting itself. Recent field experimental research on social/political networks also converges on the notion that voting is much more of a socially motivated act than SES models presume it to be (Sinclair, 2012; see also Rolfe, 2012).

Moving Beyond the Cross-Sectional Level

We propose that changes in motives over time may be particularly effective (more so than resources/skills, recruitment, and political engagement) at explaining variability in participation over time. Most importantly for any theory of participation that aims to explain both cross-sectional and over-time variability, motives are more malleable than resources. As people’s motives change, so might the participation avenues they choose or their overall participation levels. And motives can be changed or strengthened by participation itself, paving the way for theories that incorporate feedback effects (see, for example, Hibbing & Theiss-Morse, 2002; Madsen, 1987; Stukas, Snyder, & Clary, 1999).

Moreover, psychological motives such as the ones described above are pre-political, similar to resources and institutional skills. Motives such as wanting to spend time with friends, or wanting to express one’s identity, or wanting to obtain tangible rewards, are not tautological with participation (as are, arguably, political engagement and recruitment)—they can be met through engaging in political or non-political activities. As such, motives share the quality that Verba et al. (1995) argue is the reason why resources should be at the center of any model of participation—they are clearly distinct from, and are therefore theoretically interpretable as antecedents of, political behavior.

Given that the costs to participation in the political process in the United States can be considerable, combined with evidence of steep declines in citizen participation, the motivational question is even more imperative. As Gamson (1968) has argued, there are many reasons why an individual with a wealth of resources may choose not to participate. For example, “He may care very little about the outcome of most issues and thus has no motivation for influence despite his ability” (p. 96). Furthermore, McGraw (2000) argues, “There has been very little consideration of the impact of various motivations and goals . . . our understanding of political judgment and choice is doomed to be incomplete until we incorporate motivational parameters” (p. 820).

To test our theory that ebbs and flows in participation over time are explained by changes in psychological motives, we need time-series data (preferably with embedded experiments) that have good measures of psychological motives along with measures of the CVM and specific behaviors. Unfortunately, these data do not exist. We therefore echo McGraw’s concern and call for further research into the motivational bases of political participation.²⁵

Coda

For years, academics and commentators alike have decried not only the low levels of political participation in the United States, but more so the representational biases inherent in that participation—those who participate are more likely than not to be white, male, highly educated, and of higher SES. Models that focus primarily on resources or recruitment certainly can help explain why these representational biases exist—because people who participate are not only more likely to have the necessary resources to act, but they are also more likely to be asked to participate.

However, such models can do less in the way of contributing possible prescriptions to alleviate these biases at the individual level. As our findings indicate, resources, skills, and the like are not significant predictors of micro-level change in participation. Moreover, they are relatively fixed constructs; they cannot be easily increased to amplify the participation levels of traditionally underrepresented groups (in fact, the American Political Science Association Task Force on Inequality and American Democracy (2004) recommends changes at the macro- or institutional level, rather than the individual level). In contrast to resources, motives are dynamic; they can be learned, changed, and strengthened. For example, as reviewed above, social interactions and social networks have the potential to increase political behavior by connecting participation to one’s motivation to engage in behaviors that provide social benefits (or help one avoid social sanctions). Other motives may also be similarly changed, learned, or strengthened. The very fact that motives are dynamic means that a motivational model that also incorporates resources will be better able to explain ebbs and flows in participation over time than resource-based models alone (and, we suspect, why citizens become active in the first place). Motivational models can also provide guidance to programs (such as GOTV campaigns) aimed at increasing minority representation through changing or strengthening motives (e.g., Bedolla & Michelson, 2012).

We look forward to future research that takes its cue from the growing body of work on the antecedents of the social motive for political participation to examine how other motives are constructed and expressed politically. Moreover, we suggest that future research put the dynamic nature of motivation (and the feedback effects between motivation and participation) front and center of both cross-sectional and time-series explanations of individual-level participation.

Footnotes

Acknowledgements

We would like to thank the anonymous reviewers and Editor Brian Gaines for a helpful review process. The authors would also like to acknowledge Ben Ansell, Bud Duvall, Christopher Federico, Paul Goren, and Howie Lavine for their advice and early stage feedback, as well as Sarah Treul for her contributions to an earlier version of this research and Matt Motta for research assistance. Any errors are solely the responsibility of the authors.

Authors’ Note

A previous version of this paper was presented at the Midwest Political Science Association Conference, Chicago, IL, April, 2012.

Declaration of Conflicting Interests

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

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

Supplementary Material

Online appendix is available on the American Politics Research website at .

Notes

Author Biographies

Joanne M. Miller is associate professor of political science at the University of Minnesota. She specializes in Political Psychology, Public Opinion, and Political Behavior. Her research interests include the impact of psychological motivations on political participation and the causes and consequences of conspiratorial thinking. Miller can be reached at jomiller@umn.edu.

Kyle L. Saunders is associate professor of political science at Colorado State University. Saunders specializes in American Politics, Political Behavior, and Political Parties. His current research interests include activism and political participation, as well as the study of various forms of ideologically influenced misinformation. Saunders can be reached at kyle.saunders@colostate.edu.

References

Aldrich

J. H.

Montgomery

J. M.

Wood

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

American Political Science Association Task Force on Inequality and Democracy. (2004). American democracy in an age of rising inequality. Washington, DC: American Political Science Association.

Ansolabehere

Hersh

(2011). Who really votes? In Sniderman

P. M.

Highton

(Eds.), Facing the challenge of democracy: Explorations in the analysis of public opinion and political participation (pp. 267-291). Princeton, NJ: Princeton University Press.

Ansolabehere

Hersh

(2012). Validation: What big data reveal about survey misreporting and the real electorate. Political Analysis, 20, 437-459.

Bedolla

L. G.

Michelson

M. R.

(2012). Mobilizing inclusion: Transforming the electorate through get-out-the-vote campaigns. New Haven, CT: Yale University Press.

Berelson

B. R.

Lazarsfeld

P. F.

McPhee

W. N.

(1954). Voting: A study of opinion formation in a presidential campaign. Chicago, IL: University of Chicago Press.

Berent

M. K.

Krosnick

J. A.

Lupia

(2011). The quality of government records and over-estimation of registration and turnout in surveys: Lessons from the 2008 ANES Panel Study’s registration and turnout validation exercises (American National Election Studies Working Paper No. nes012554. Retrieved from http://www.electionstudies.org/resources/papers/nes012554.pdf

Berinsky

A. J.

Lavine

(2012). Self-monitoring and political attitudes. In Aldrich

J. H.

McGraw

K. M.

(Eds.), Improving public opinion surveys: Interdisciplinary innovation and the American national election studies (pp. 27-45). Princeton, NJ: Princeton University Press.

Bernstein

Chadha

Montjoy

(2001). Overreporting voting: Why it happens and why it matters. Public Opinion Quarterly, 65, 22-44.

10.

Campbell

A. L.

(2002). Self-interest, social security, and the distinctive participation patterns of senior citizens. American Political Science Review, 96, 565-574.

11.

Campbell

A. L.

(2003). How policies make citizens: Senior political activism and the American welfare state. Princeton, NJ: Princeton University Press.

12.

Campbell

Converse

P. E.

Miller

W. E.

Stokes

D. E.

(1960). The American voter. New York, NY: Wiley.

13.

Chong

(1991). Collective action and the Civil Rights Movement. Chicago, IL: University of Chicago Press.

14.

Clark

P. B.

Wilson

J. Q.

(1961). Incentive systems: A theory of organizations. Administrative Science Quarterly, 6, 129-166.

15.

Clary

E. G.

Snyder

Ridge

R. D.

Copeland

Stukas

A. A.

Haugen

Miene

(1998). Understanding and assessing the motivations of volunteers: A functional approach. Journal of Personality and Social Psychology, 74, 1516-1530.

16.

Cravens

M. D.

(2013). The role of habit in voting: Making and breaking habitual voters (Unpublished doctoral dissertation). University of Minnesota, Minneapolis.

17.

Dahl

(1961). Who governs? Democracy and power in an American city. New Haven, CT: Yale University Press.

18.

Davenport

T. C.

Gerber

A. S.

Green

D. P.

Larimer

C. W.

Mann

C. B.

Panagopoulos

(2010). The enduring effects of social pressure: Tracking campaign experiments over a series of elections. Political Behavior, 32, 423-430.

19.

Enos

R. D.

Hersh

E. D.

(2015). Party activists as campaign advertisers: The ground campaign as a principal-agent problem. American Political Science Review, 109, 252-278.

20.

Finkel

S. E.

Opp

K.-D.

(1991). Party identification and participation in collective action. Journal of Politics, 53, 339-371.

21.

Fowler

J. H.

(2006). Habitual voting and behavioral turnout. Journal of Politics, 68, 335-344.

22.

Fowler

J. H.

Dawes

C. T.

(2008). Two genes predict voter turnout. Journal of Politics, 70, 579-594.

23.

Gamson

W. A.

(1968). Power and discontent. Homewood, IL: The Dorsey Press.

24.

Gerber

A. S.

Green

D. P.

(2000). The effects of canvassing, telephone calls, and direct mail on voter turnout: A field experiment. American Political Science Review, 94, 653-663.

25.

Gerber

A. S.

Green

D. P.

Larimer

C. W.

(2008). Social pressure and voter turnout: Evidence from a large-scale field experiment. American Political Science Review, 102, 33-48.

26.

Gerber

A. S.

Green

D. P.

Shachar

(2003). Voting may be habit-forming: Evidence from a randomized field experiment. American Journal of Political Science, 47, 540-550.

27.

Gerber

A. S.

Huber

G. A.

Doherty

Dowling

C. M.

(2012). Assessing the stability of psychological and political survey measures. American Politics Research, 41, 54-75.

28.

Gerber

A. S.

Huber

G. A.

Doherty

Dowling

C. M.

Raso

S. E.

(2011). Personality traits and participation in the political process. Journal of Politics, 74, 692-706.

29.

Gerber

A. S.

Rogers

(2009). Descriptive social norms and motivation to vote: Everybody’s voting and so should you. Journal of Politics, 71, 178-191.

30.

Green

D. P.

Gerber

A. S.

(2008). Get out the vote: How to increase voter turnout (2nd ed.). Washington, DC: Brookings Institution Press.

31.

Han

(2009). Moved to action: Motivation, participation and inequality in American politics. Stanford, CA: Stanford University Press.

32.

Harder

Krosnick

J. A.

(2008). Why do people vote? A psychological analysis of the causes of voter turnout. Journal of Social Issues, 64, 525-549.

33.

Hibbing

J. R.

Theiss-Morse

(2002). Stealth democracy: Americans’ beliefs about how government should work. Cambridge, UK: Cambridge University Press.

34.

Highton

Wolfinger

R. E.

(2001). The first seven years of the political lifestyle. American Journal of Political Science, 45, 202-209.

35.

Holbrook

A. L.

Krosnick

J. A.

(2013). A new question sequence to measure voter turnout in telephone surveys: Results of an experiment in the 2006 ANES Pilot Study. Public Opinion Quarterly, 77, 106-123.

36.

James

(1890). The principles of psychology. New York, NY: Dover.

37.

Johnson

P. E.

(1998). Interest group recruiting: Finding members and keeping them. In Cigler

A. J.

Loomis

B. A.

(Eds.), Interest group politics (5th ed., pp. 35-62). Washington, DC: CQ Press.

38.

Katz

(1960). The functional approach to the study of attitudes. Public Opinion Quarterly, 24, 163-204.

39.

Klandermans

(2003). Collective political action. In Sears

D. O.

Huddy

Jervis

(Eds.), Oxford handbook of political psychology (pp. 670-709). Oxford, UK: Oxford University Press.

40.

Lee

(2002). Mobilizing public opinion: Black Insurgency and racial attitudes in the Civil Rights Era. Chicago, IL: University of Chicago Press.

41.

Madsen

(1987). Political self-efficacy tested. American Political Science Review, 81, 571-582.

42.

McGraw

K. M.

(2000). Contributions of the cognitive approach to political psychology. Political Psychology, 21, 805-832.

43.

Milbrath

L. W.

Goel

M. L.

(1977). Political participation: How and why do people get involved in politics? Chicago, IL: Rand McNally.

44.

Miller

J. M.

(2013). The motivational underpinnings of political participation. In La Raja

R. J.

(Ed.), New directions in American politics (pp. 209-232). New York, NY: Routledge.

45.

Miller

J. M.

Rahn

(2002, May 23-24). Identity-based feelings, beliefs, and actions: How being influences doing. Paper presented at the Conference on Social Connectedness & Public Activism, Center for Political Studies, Harvard University, Cambridge, MA.

46.

Miller

W. E.

(1980). Disinterest, disaffection, and participation in presidential elections. Political Behavior, 2, 7-32.

47.

Mondak

J. J.

(2010). Personality and the foundations of political behavior. New York, NY: Cambridge University Press.

48.

Nickerson

D. T.

(2008). Is voting contagious? Evidence from two field experiments. American Political Science Review, 102, 49-57.

49.

Olson

(1965). The logic of collective action. Cambridge, MA: Harvard University Press.

50.

Pierce

J. C.

Lovrich

N. P.

(1982). Survey measurement of political participation: Selective effects of recall in petition signing. Social Science Quarterly, 63, 164-171.

51.

Pittman

T. S.

(1998). Motivation. In Gilbert

D. T.

Fiske

S. T.

Lindzey

(Eds.), The handbook of social psychology (4th ed., Vol. 1, pp. 549-590). Boston, MA: McGraw-Hill.

52.

Prior

(2010). You’ve either got it or you don’t? The stability of political interest over the life cycle. Journal of Politics, 72, 747-766.

53.

Rogers

Fox

C. R.

Gerber

A. S.

(2012). Rethinking why people vote: Voting as dynamic social expression. In Shafir

(Ed.), Behavioral foundations of policy (pp. 91-107). Princeton, NJ: Princeton University Press.

54.

Rolfe

(2012). Voter turnout: A social theory of political participation. New York, NY: Cambridge University Press.

55.

Rosenstone

S. J.

Hansen

J. M.

(1993). Mobilization, participation, and democracy in America. New York, NY: MacMillan.

56.

Salisbury

R. H.

(1969). An exchange theory of interest groups. Midwest Journal of Political Science, 13, 1-32.

57.

Schlozman

K. L.

Tierney

J. T.

(1986). Organized interests and American democracy. New York, NY: Harper & Row.

58.

Schlozman

K. L.

Verba

Brady

H. E.

(2012). The unheavenly chorus: Unequal political voice and the broken promise of American democracy. Princeton, NJ: Princeton University Press.

59.

Sheagley

Miller

J. M.

Snyder

(2010, January). Adapting the Volunteer Functions Inventory to measure motives for political participation. Poster presented at the annual meeting of the Society for Personality & Social Psychology, Las Vegas, NV.

60.

Silver

B. D.

Anderson

B. A.

Abramson

P. R.

(1986). Who overreports voting? American Political Science Review, 80, 613-624.

61.

Sinclair

(2012). The social citizen: Peer networks and political behavior. Chicago, IL: University of Chicago Press.

62.

Snyder

(1974). Self-monitoring of expressive behavior. Journal of Personality and Social Psychology, 30, 526-537.

63.

Snyder

(1979). Self-monitoring processes. Advances in Experimental Social Psychology, 12, 85-128.

64.

Snyder

Gangested

(1986). On the nature of self-monitoring: Matters of assessment, matters of validity. Journal of Personality and Social Psychology, 51, 125-139.

65.

Spector

P. B.

(1992). Summated rating scale construction: An introduction. Newbury Park, CA: Sage.

66.

Stukas

A. A.

Snyder

Clary

E. G.

(1999). The effects of “mandatory volunteerism” on intentions to volunteer. Psychological Science, 10, 60-65.

67.

Teske

(1997). Political activists in America: The identity construction model of political participation. Cambridge, UK: Cambridge University Press.

68.

Vavreck

(2007). The exaggerated effects of advertising on turnout: The dangers of self-reports! Quarterly Journal of Political Science, 2, 325-343.

69.

Verba

Nie

N. H.

(1972). Participation in America: Political democracy and social equality. New York, NY: Harper & Row.

70.

Verba

Schlozman

K. L.

Brady

H. E.

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

71.

Volgy

T. J.

Schwarz

J. E.

(1984). Misreporting and vicarious political participation at the local level. Public Opinion Quarterly, 48, 757-765.

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.

0.00 MB

0.10 MB