Increasing Youth Political Engagement with Efficacy Not Obligation: Evidence from a Workshop-Based Experiment in Zambia.

In many places around the world, young voters participate in politics at low rates. What factors might increase youth political participation? We investigate one possibility: exposure to a religious message that emphasizes the possibility of change through faithful action. We argue that this message, which is common in religious groups that attract large numbers of youth around the world, addresses several barriers to political participation by young voting-age adults. Working in collaboration with the major religious coalitions in Zambia, we randomly assigned young adults (18-35 years old) into civic engagement workshops. Identical informational material, based on pre-existing, non-partisan curricula, was presented in each workshop. Workshops then concluded with one of two randomly assigned, pre-recorded Christian motivational messages based on existing religious programming in Zambia. In some workshops, the concluding message emphasized a Christian obligation to work towards the greater good. In other workshops, the message emphasized the power of faith to make change in the world. We found that the power of faith message moved workshop participants to be more willing to participate in protest, to disavow political violence, and to criticize other people who choose not to participate, relative to pre-workshop measures and to an information-only condition. By contrast, the message focused on an obligation to the greater good did not change political participation, resulting in lower willingness to participate in politics than the power of faith message. We discuss implications for youth political participation and the study of religion and politics. Supplementary Information: The online version contains supplementary material available at 10.1007/s11109-022-09794-2.

In many countries, young adults participate in politics at low rates (Solijonov, 2016; Uberoi, 2021; Turcotte, 2015; Holbein & Hillygus, 2020; Resnick & Casale, 2014). This pattern is particularly striking in Sub-Saharan Africa, where young voting-age citizens (18–35 years old) vastly outnumber citizens in older cohorts but are under-represented among those who participate in electoral politics (Adebayo, 2018; Berinsky et al., 2016; YouthMap, 2014). Both because of their numbers and their policy priorities, youth disengagement has serious consequences for the region’s political future (Resnick & Casale, 2014). Hence, many scholars, policymakers and activists argue that the future of democracy “depend[s] critically” on increasing younger citizens’ “active political engagement” (Bratton et al., 2017, pp. 15–16).

What factors might propel more young adults into the political arena?1 In this paper we investigate one possibility: exposure to religious messages that emphasize the potential to make change through faithful action. This power of faith message is characteristic of preaching in Pentecostal and Charismatic Christian communities (Gifford, 2009; Marshall, 2009; McClendon & Riedl, 2019), which are increasingly popular, attracting large numbers of young people, particularly in developing countries with disproportionately large youth populations, termed youth bulges (Freston, 2004; Parsitau & Mwaura, 2010; Lindhardt, 2012; Kalu, 2008; Sperber and Hern 2018; Audette et al., 2020). The power of faith message suggests that those who act—and persist—in strong Christian faith will achieve personal, social and political change, even if such change might seem impossible ex ante. As such, the message directly addresses barriers to participation that commonly affect youth—e.g., youth perceptions that their actions will not make a difference (Berinsky et al., 2016) and the higher likelihood that youth give up in the face of obstacles (Holbein & Hillygus, 2020).2 These barriers are likely to be particularly consequential for political actions that involve higher risks and costs (e.g., protesting and contributing financially to politics), which suggests that a power of faith message might be particularly influential for such actions.

We present the results of a workshop-based field experiment investigating the consequences of the power of faith message on youth political participation in Zambia, an overwhelmingly Christian country in Southern Africa. In Zambia—as in many other young, developing democracies—religious civic education is widespread but rarely studied. In collaboration with Caritas-Zambia (the NGO of the Zambian Council of Catholic Bishops), the Council of Churches in Zambia (the umbrella body for mainline Protestant churches), Pentecostal churches, and Innovations for Poverty Action-Zambia, we ran in-person workshops for young voting age adults (18–35 years old, hereafter termed youth)3 in 2020 before the spread of COVID-19. Workshops disseminated non-partisan civic information from existing curricula. Half of the workshops supplemented this information with Christian messages conveying the power of faith to make change. The other half of the workshops supplemented the curriculum with Christian messages stressing a moral obligation to serve the common good. Both types of messages were based on our partner organizations’ preaching and programming. Youth were randomly assigned to one type of workshop. Then, using surveys and a behavioral measure, we measured participants’ willingness to engage in various forms of political action and their judgments of others’ political actions.

We found that exposure to the power of faith message increased participants’ willingness to engage in various forms of political participation, especially more effortful and costly forms of participation, such as protesting and donating, and to critically evaluate other people who chose not to participate, relative to pre-workshop measures. On some outcome measures, the effect of the power of faith message was particularly pronounced among women, who exhibited lower initial levels of political interest and political participation, and higher initial levels of religiosity. Contrary to some research that emphasizes the mobilizing potential of prosocial religious messages (Oviedo, 2016), the message focused on obligation to the common good did not increase workshop participants’ political participation relative to pre-workshop levels. Additionally, in some workshops the audio equipment required to play religious messages inadvertently failed, creating a civics information-only condition. When we compare measures of political participation between workshops with and without functional audio equipment, we find that the power of faith message increased political participation relative to this information-only condition, but the obligation to the common good message did not.

This study makes several contributions. First, the paper joins a growing literature on youth political participation.4 Bringing new experimental evidence to bear, this paper advances the claim that youth participation is often curtailed by the belief that participation is futile and by inexperience with obstacles to participating (Holbein & Hillygus, 2020). As Campbell (2019) and Finkel and Lim (2020) note, experimental studies of civic engagement and youth participation are still rare, especially in the Global South. We expand the geographic scope of existing literature by conducting the study in Zambia. Second, to our knowledge, our study is the first experiment to examine how religious messages influence the impact of faith-based civic education programs on youth political participation. Faith-based organizations (FBOs) are one of the most common providers of civic education world-wide,5 but political scientists rarely study the consequences of these activities for political participation.6 The study also sheds light on the influence of religious ideas on political participation more generally by examining the consequences of randomized exposure to one of two popular messages in contemporary Christianity (McClendon & Riedl, 2019), each of which conveys an idea that is prevalent across world religions. Finally, the study is an example of a community-collaborative experiment, designed and implemented in partnership with local organizations and a team of researchers from both the country of study and U.S.-based institutions.

Below, we situate the study in literatures on religion and youth political participation and present our hypotheses. We then discuss the Zambian context, provide an overview of study methods and data, and present the results. We conclude with discussion of the theoretical, empirical, and policy implications of our findings and prospects for future research.

The Power of Faith Message and Political Participation

A significant literature suggests that religion affects political participation. In many parts of the world, religious people tend to be more politically active than non-religious people.7 In Sub-Saharan Africa, specifically, observational evidence reveals a robust correlation between religious and civic engagement (Dowd & Sarkissian, 2017). This association may be attributed to factors such as skills gained at church (Campbell, 2004), religious identity (McCauley, 2017), social networks (Lewis et al., 2013), or patron-client relationships (McCauley, 2013).8

There is also reason to believe that the content of religious ideas can motivate political participation. In particular, a growing body of research on religious messaging argues that religious communication stressing the power of faithful action can influence political participation by altering individuals’ expectations that their actions will succeed (McClendon & Riedl, 2015, 2019; Haynes, 2017; Marshall, 2009). In Kenya, McClendon and Riedl (2019) analyzed the text of Christian sermons and found that Pentecostal church sermons routinely delivered messages that suggested faithful action can lead to change. Moreover, exposure to this power of faith message increased political participation among both Pentecostals and other types of Christians. In other words, power of faith messages increased political participation across denominations. These findings accord with evidence that self-efficacy is a critical ingredient in individuals’ decisions to engage in collective action (Lieberman & Zhou, 2020; Niemi et al., 1991), particularly among youth (Holbein & Hillygus, 2020). These findings are also consistent with evidence that when people believe their actions will make a difference, they are less deterred by the costs of joining with others (Croke et al., 2016). Self-efficacy and grit are particularly consequential for collective actions that are costly or require significant effort (Niemi et al., 1991). For political acts that are more habitual and less costly, grit is often unnecessary, as individuals face fewer obstacles and may undertake routine actions even if they doubt their impact. A power of faith message may therefore be particularly motivating with respect to costlier and more effortful actions, such as protesting or donating to a political cause.

A power of faith message may also be particularly motivating for young adults. Scholars focused on youth participation (Adebayo, 2018; Berinsky et al., 2016; Holbein & Hillygus, 2020) have recently identified several especially important barriers to youth participation, including lack of grit and political efficacy. Holbein and Hillygus (2020) find that lower political participation among younger cohorts stems partly from a lack of experience overcoming obstacles to participation (due to their relatively recent eligibility to vote) and because of disillusionment about the potential for their individual actions to make change (often due to dominance of the political arena by older cohorts). Accordingly, in the more secular setting of that study, the authors found that measures promoting grit, persistence, and self-efficacy increased youth political participation (see also, Holbein et al., 2020, ). We contend that in highly religious contexts, messages about the power of faith, which often encourage belief in the power of action and the importance of persistence despite apparent challenges, may have similar consequences.9 To our knowledge, this question has not yet been explored.

Of course, there are other ways in which religious messages might increase political participation, apart from altering individuals’ expectations that their actions will succeed. A common theme in research on religious messaging is that most major religions impart prosocial values, which stress individuals’ moral obligation to contribute to the common good, even when this requires personal sacrifices (Preston et al., 2013). Norenzayan (2013) argues that religions that impart the idea of a moral obligation to others and imply that omnipotent “big gods” monitor prosocial behavior are uniquely capable of facilitating large-scale cooperation and collective action. Indeed, simply mentioning God can induce adherents of major religions to behave more prosocially (Shariff & Norenzayan, 2007; Warner et al., 2015). Studies of religious communities find that people adhere more durably to communal obligations framed as sacred forms of sacrifice (Sosis, 2000; Skitka et al., 2005). And, while it may seem counterintuitive that messages encouraging sacrifice would ever persuade rational individuals, Bénabou and Tirole (2011) stress that religious individuals often invest time and energy in sacrifice as a way to signal virtue to themselves. Additionally, action that involves sacrifice for others can bind people to moral communities (Bulbulia, 2004; Graham & Haidt, 2010), and impart a strong sense of belonging. Religious ideas can thus encourage individuals to overcome barriers to collective action by altering perceptions of costs and benefits associated with joining in (e.g., belonging).

By not only priming a religious mindset and its possible attendant prosocial commitments, but also emphasizing that faithful action can make change even in adverse settings, a power of faith message may increase youth political participation even over and above other religious messages. In fact, in highly religious environments, prosocial religious values—which scholars have argued are basic elements of most major religions—may be chronically available (Djupe & Smith, 2019).10 Thus, in a highly religious setting, religious messaging stressing prosocial obligations might not necessarily increase youth participation as much as a power of faith message, because the latter primes a less chronically available idea that directly addresses a prominent barrier to youth participation.

Additionally, power of faith messages may inspire peaceful political participation more than other religious messages. Religious communication has long been used both to foment and to quell political violence, so there is no necessary relationship between religious messaging per se and peaceful political action.11 Yet, power of faith messages suggest that even mundane action, when done in faith, is likely to succeed because of divine assistance. Primed with this idea, listeners to the power of faith message may be more likely to view their aims as achievable and, in particular, achievable without the use of violence. By extension, they should be less likely to condone political violence or to participate in violence. This expectation is consistent with the evidence that greater self-efficacy—i.e., a belief that one’s actions will succeed—correlates with peaceful political participation in democracies (Condon & Holleque, 2013), and that political efficacy moderates support for political violence in post-conflict settings, such as Northern Ireland and Nepal (Dyrstad & Hillesund, 2020). Dyrstad and Hillesund (2020) find that people with higher political efficacy are less likely to participate in and support political violence, because they believe other channels of participation can succeed.12 Religious messages concerning prosocial obligations might also reduce approval of violence insofar as prosocial concerns might reduce inclinations to harm others,13 but if prosocial ideas are chronically available, the power of faith message may increase peaceful participation and disapproval of violence over and above these more prevalent messages. In other words, by introducing an idea that is less chronically available but that directly addresses barriers to participation, the power of faith message may have greater influence over participation and attitudes toward violence than a religious obligation message.

Many studies of political participation, and of the effects of civic education on youth participation, point to the prevalence of compensatory effects. That is, when an intervention seeks to mitigate barriers to participation, its effects (if any) tend to be largest among individuals who faced the highest initial barriers or exhibited the lowest initial rates of participation ex ante (Campbell et al., 2016; Finkel et al., 2012; Neundorf et al., 2016; Gainous & Martens, 2012). Similar compensatory patterns may follow exposure to a power of faith message, and they may be particularly powerful if individuals with the lowest initial levels of participation also find religious messaging particularly appealing. In sub-Saharan Africa, these differences tend to play out across genders: Men consistently report greater willingness than women to discuss politics, join with others to address problems, attend community meetings, and contact political leaders (Logan & Bratton, 2006; Hern, 2018). At the same time, women around the world, including in Sub-Saharan Africa, tend to be highly religious (Pew Research Center, 2016, p. 54). Further, the region’s gender gap in political participation also appears to be impacted by age, with relatively young women exhibiting the lowest levels of participation (Lekalake & Gyimah-Boadi, 2016). These patterns suggest that any effects of the power of faith message on peaceful political participation are likely to be more robust and/or large among young women.

Bringing these insights together, we investigate the following hypotheses:

Participants assigned to the power of faith message will express greater willingness to participate in politics relative to (a) pre-workshop levels; (b) participants exposed only to civics information; and (c) participants exposed to a religious obligation to the common good message.

Participants assigned to the power of faith message will be less likely to view the use of political violence as necessary compared to (a) pre-workshop levels; (b) participants exposed only to civics information; and (c) participants exposed to a religious obligation to the common good message.

Subgroups of listeners with lower initial levels of political participation (e.g., women) will exhibit larger gains in willingness to participate in peaceful actions in response to the power of faith message.

The Zambian Context

Zambia is an analytically appropriate site for this study for several reasons. First, it has a significant “youth bulge” and low youth turnout,14 and is thus a place where our research question is salient and pressing. Our evidence confirms that young Zambian women’s political participation was lower than their male peers’. Second, faith-based programming is central to civic education in Zambia, where major FBO’s (including our partner organizations) provide civic education opportunities in all of Zambia’s provinces. Third, Zambians are highly religious and overwhelmingly Christian.15 In this context, it was unlikely that our study would expose participants to content perceived as foreign, jarring or likely to provoke inter-religious tensions. (We discuss additional ethical considerations in Online Appendix B.) Last but not least, there was strong local demand for this research in Zambia: Local Zambian churches and civil society leaders, including our partner organizations, wanted to understand the impact of their civic education programming on youth, including religious messages that are sometimes included in this programming.

Research Design

To test our hypotheses, we designed a community collaborative study to understand the influence of Christian messages on peaceful political participation. Common in such fields as public health and education, the defining features of this approach are: (1) an enduring commitment to work cooperatively with community partners as equals in the process of research design and implementation; and (2) the pursuit of mutual benefits for community partners and researchers alike.16 This study stemmed from one of the PI’s long-term conversations with Zambian religious and community leaders while conducting fieldwork in Zambia between 2011 and 2016. The author team (composed of one Zambian researcher and two US-based researchers) designed a study with Caritas–Zambia and the Protestant Council of Churches in Zambia (CCZ) to evaluate shared research questions, subsequently inviting the Evangelical Fellowship of Zambia (EFZ) and the Christian Churches Monitoring Group (CCMG) to become study partners, too. The study was approved by the research ethics review committee at the University of Zambia and the IRBs of the two affiliated US universities. For additional ethical considerations, see Online Appendix B.

Sample and Recruitment

We conducted the study in Lusaka, Zambia from February to early March 2020, before the spread of COVID-19. Two hundred ninety one Zambian youth (between the ages of 18 and 35 years old) participated. As noted above, this definition of youth is borrowed from several Zambian and international agencies. Our community partners and Zambian PI also confirmed the appropriateness of this definition of youth in the Zambian context, where youth is commonly understood as those under 35 years of age or as those who, regardless of age, are unmarried and without children. During recruitment, we screened only for age and not for marital status or number of children. Nevertheless, as we report below, the sample skewed relatively young and childless, as expected.

We adopted a two-pronged approach to recruitment, with roughly 39% of study participants randomly recruited from public streets and markets where young people were known to congregate. Enumerators stopped every fifth young-looking person (stratifying by gender) who passed them in the public marketplace, and invited that person to volunteer to join the study. The other 61% of our sample were recruited directly from Sunday worship services at Catholic, mainline Protestant, and Pentecostal congregations, also stratifying by gender. Our partner organizations helped connect us to the leaders of these churches, from whom we obtained advanced permission to recruit on Sundays. After the service, enumerators waited outside the church and handed out tickets at random to youth exiting the church. All workshops were held in nearby classrooms, 88% of which were rented from local public schools.17 The goal of this recruitment strategy was to ensure a mix of regular and potentially less regular church attendees. Online Appendix Table C.8 shows that our main results in the paper did not differ substantively across the subsamples of public street-market-recruited participants and church-recruited participants.

The study took place in three urban compounds (townships) in Lusaka: Bauleni, Chawama, and Mtendere. We chose Lusaka as the site for this study for its high concentration of different churches (Catholic, Protestant, and Pentecostal) in close geographic proximity. We purposefully selected these compounds for their relatively low-income populations (which make them closer to the average Zambian’s socioeconomic status), and because they were known to be politically competitive and diverse areas. As a result, although the University of Zambia Research Ethics Committee did not allow us to ask respondents about partisanship, we believed that our workshops would not be perceived as partisan in nature (i.e., we did not target outright party strongholds). We discuss questions about generalizability at the end of the paper.

Intervention

Upon arrival at the workshop, study participants were randomly assigned to one of two identical classrooms. Study participants completed a brief pre-workshop (baseline) survey with an enumerator before participating in the workshop. Subsequently, they listened to and discussed approximately 1.5 h of information about the process of becoming an engaged citizen and voter, citizens’ rights and duties, and the electoral process. This information was identical across workshops, explicitly non-partisan, and was drawn directly from curricular materials previously developed by our partner organizations (see Online Appendix A.3 for the full content of the informational curriculum). Local research staff were trained to deliver this information, answer questions, and facilitate workshops in similar ways. Sixteen workshops were conducted, including 8 assigned to play the obligation message and 8 assigned to play the power of faith message. On average, there were 17–18 participants in a workshop.

At the end of every workshop, facilitators played one of the two pre-recorded Christian messages (“treatments”) through a speaker to workshop participants. Together with our community partners, we developed these messages from their member churches’ preaching and existing materials. One message emphasized the power of faith to make change, while the other emphasized Christians’ moral obligation to the common good. Messages were of comparable length and were recorded by the same man, allowing us to hold the identity and style of the speaker constant and to vary only the content of the audio messages. (See Appendices A.1 and A.2 for the full text of each message.) By comparing outcomes following the power of faith message to outcomes following the religious obligation message, we are able to test whether the power of faith’s explicit focus on efficacy, grit and change increased participation over and above religious and prosocial messaging per se.

Specifically, the obligation to the common good message presents Jesus’ sacrifice for humanity as a model for thinking beyond one’s own immediate needs and wants, and for taking action in service of the greater good. This message approximated messages from (non-Charismatic) Catholic organizations.18 A key passage from this message states: The obligation message thus uses the example of Jesus’ life to emphasize prosocial values (i.e., “personal sacrifices for the greater good,” “commitment to the common good”). Prior research has found that these values are often primed by even general references to major religions (Norenzayan, 2013; Shariff & Norenzayan, 2007; Shariff et al., 2014; Preston et al., 2013).

[T]he dignity of every human person, the respect of human rights, commitment to the common good ...these are central to the teachings of Jesus Christ and important values of all Christians. ...Christians must make personal sacrifices for the greater good, just as Jesus made the ultimate sacrifice for us. We may be tempted to get up in the morning and say that participating in politics is too hard. It takes time and commitment to be politically engaged — resources that we may not feel we have, as we try to make ends meet and try to get through the struggles of daily life. But Jesus knew the meaning of sacrifice. And for the sake of the morals that uphold our democratic system, we must be willing to do the same.

The power of faith message makes religious references, thus potentially also priming prosocial values. Importantly, however, the power of faith message also emphasizes that God works through the actions of the faithful to make change in the world. This message approximates messages commonly delivered in Pentecostal and Charismatic sermons but is not unique to them (McClendon & Riedl, 2019; Dowd & Sarkissian, 2017). A key passage states: Both messages acknowledge that participating in politics can be daunting. Both exhort listeners to engage in peaceful political action despite associated personal costs in time and energy. Both use Jesus as a model and underscore the importance of acting to protect the “morals that uphold our democratic system.” The messages differ only in the reasons offered to motivate participation: The obligation message makes prosocial values explicit, whereas the power of faith message primes prosocial values implicitly and then introduces the idea that strong faith can bolster an individual’s ability to make positive change through peaceful political participation. “With strength of faith anything is possible,” it says. We “must declare [our] faith and embody it every day.” “We must declare and show our faith and know that whatever we have declared, it has already been established.” This last line is protoypical of power of faith messages and implies not that action is unnecessary, but rather that when things are done in the faith, they will be achieved, as if they were already established by God. The power of faith message thus places particular emphasis on the efficacy of faithful action. Comparing these two messages allows us to assess the effect of the specific content of the power of faith message on youth political participation, separate from the effect of priming religiosity or religious values per se.

You must know that with strength of faith anything is possible. ...Christians must declare their faith and embody it every day, just as Jesus would have us do. We may be tempted to get up in the morning and say that participating in politics is too hard. It takes time and commitment to be politically engaged — resources that we may not feel we have, as we try to make ends meet and try to get through the struggles of daily life. But these are the ways of thinking of people of little faith. And for the sake of the morals that uphold our democratic system, we must declare and show our faith and know that whatever we have declared, it has already been established.

We chose to compare these two prevalent religious messages for several reasons. First, against a background of extremely high religiosity and prevalent church-based civic education programs in Zambia, the contrast between these religious conditions was the most relevant to our local partners, who wanted to understand which religious ideas most effectively motivated youth. Indeed, in settings saturated with religious content, the more interesting research questions may be about the consequences of different types of religious messages, rather than questions about the consequences of religious versus secular messages, on political behavior (McClendon & Riedl, 2015, 2019). Therefore, without the resources to run a larger study, we maximized statistical power by focusing on the contrast between these two conditions. Yet, as we detail below, our team experienced some inadvertent implementation mishaps, such that in a small sub-set of workshops neither religious message was played successfully. In those workshops, participants received only (non-religious) civics information. Below, we demonstrate that these mishaps were not significantly correlated with potential outcomes and compare workshops with each of the religious messages to these "no-audio" information-only workshops.

Measurement

At the start and at the end of the workshop, participants were asked several questions designed to elicit their willingness to engage in peaceful political participation. The end of workshop survey asked the following questions: “There are various actions that people sometimes take as citizens. We are interested in whether you would do these things if offered the opportunity in the near future. For instance, would you attend a peaceful protest or demonstration march?...Would you contact a government official?” “If elections were held tomorrow, would you vote?” and “How interested would you say you are in politics? [very, somewhat, not very, not at all]”19 These questions measure political interest and willingness to take peaceful political action. By asking them in our pre-workshop survey, as well, we are able to control for pre-workshop attitudes in our main analyses and thus to analyze within-respondent changes. Additionally, the post-workshop survey asked, “Would you attend a protest where violence by political cadres was likely to break out?” Together, these measures capture participants’ willingness to engage in diverse political actions ranging from the more effortful or costly (participating in a protest) to the more common and less costly (contacting a government official, voting).20

While these questions provide useful measures of self-reported willingness to engage in different kinds of political participation, we also included a more costly behavioral measure. Specifically, following every post-workshop survey, study participants were given the opportunity to donate any portion of their compensation to the Christian Churches Monitoring Group (CCMG) “to facilitate election observation in the future.” Participants had learned earlier that the non-partisan CCMG works for peaceful and fair elections. This donation measure therefore allowed study participants to demonstrate their level of support for peaceful political participation in a relatively costly way. Additionally, we view this particular donation as an act of peaceful political participation, as it directed support to an organization seeking to prevent electoral violence. Study participants were assured that they were under no obligation to donate (indeed, many did not) and all donations were allocated to the CCMG, as promised.

Furthermore, because respondents may not be willing to report the full extent to which they would engage in certain political behaviors, we included a vignette question that asked participants to evaluate a sympathetic person who chose not to vote. The vignette read, We then asked (1) “In your view, is this person’s decision [fully justified, mostly justified, mostly unjustifed, or totally unjustified]?” (2) “In your view, is this person a good citizen?” and (3) “In your view, is this person a good Christian?” We used responses to these questions to estimate the effect of treatments on participants’ views about the appropriateness of (not) participating in peaceful political action. We chose a trade-off between familial obligations and voting because it is familiar and common in Zambian society, at large. It is also a trade-off that is likely to be viewed by many Zambians as legitimate and thus as not easily dismissed.21 We thus interpret willingness to criticize this person as a meaningful indicator of a participant’s valuation of the act of voting.

Now I want to read you a brief description about someone’s actions and then ask for your opinion. Specifically, I want to tell you about a Zambian parent who is very busy taking care of their children and household while also struggling to make ends meet. This parent strongly believes that, if elected, one of the politicians running for office will lead a corrupt and exploitative government. They are optimistic that the other candidate will lead a more effective, accountable, and purpose-driven government. However, this parent also believes that a strong majority will turn out to vote for the candidate that they prefer, so they decide to stay home on election day rather than going out to vote.

Finally, the post-workshop survey asked participants about the use of political violence: “Which of the following statements is closest to your view: ‘The use of violence is never justified in Zambian politics today.’ OR ‘In this country, it is sometimes necessary to use violence in support of a just cause.”’ Participants could agree with either statement strongly or weakly, or they could agree with neither.

Results

We present summary statistics in Online Appendix Table C.1. By design, the sample was young, with an average age of 24 years. Three-quarters of participants were unmarried and more than half were childless. The vast majority reported attending church at least once a week. Even among participants recruited from public markets, the majority (58%) reported attending church weekly or daily. Having also recruited from Catholic, mainline Protestant (UCZ), and Pentecostal churches, the largest denominational group in our sample was Catholic (35%), with another 31% belonging to UCZ, and 22% belonging to Pentecostal churches. The remainder belonged to other churches except for one Muslim who volunteered to participate in the study. Online Appendix Table C.2 shows no detectable differences in observable characteristics of participants across the two types of workshops.

We find that exposure to the power of faith message boosted reported willingness to participate in peaceful protests, relative to exposure to the obligation to serve the common good message. Willingness to engage in many of the other, less effortful types of participation (willingness to contact a government official, interest in politics) also moved in the same direction as willingness to protest but did not reach statistical significance. The power of faith message did not have a detectable treatment effect on donations in the full sample but did among women, as we discuss below. Table 1 shows results for inclinations to participate among the full sample. We control for pre-workshop levels of willingness to engage in each behavior and cluster standard errors by workshop session, the level at which treatment was assigned. Figure 1 underscores that, although pre-workshop willingness to participate was high, the power of faith message increased it. Before workshops, 87% of participants reported willingness to protest; exposure to the power of faith message corresponded with an increase to 94%. By contrast, exposure to the obligation message did not significantly affect willingness to protest.

Table 1

Effects of power of faith message (relative to obligation to common good message) on political intentions

	Would protest	Would vote	Would contact Gov’t	Very interested in politics	Donated
Power of Faith Message	0.072**	0.009	0.031	0.068	0.045
	(0.026)	(0.042)	(0.048)	(0.050)	(0.076)
Pre-workshop Level	0.312***	0.567***	0.508***	0.558***	–
	(0.082)	(0.077)	(0.079)	(0.047)	–
Respondents	288	288	288	288	279
Workshops	16	16	16	16	16

***p < 0.01, **p < 0.05, *p < 0.10; standard errors clustered by workshop session

Fig. 1

Direction of treatment effect on willingness to protest

Online Appendix Table C.4 affirms that results in Table 1 are consistent with results from a difference in difference estimation, which examines within-subject changes from pre- to post-workshop survey responses. Given the relatively small number of workshops, one might also worry that results are driven by an outlier workshop. Yet, Appendix Fig. C.2 shows that results are consistent when any individual workshop is excluded from the analysis.

We also confirm that, prior to the workshops, female participants reported lower political interest and willingness to participate in politics, but higher religiosity, than men (see Online Appendix Fig. C.1). These gender gaps are consistent with national patterns among young people in Zambia and many other countries. Additionally, Table 2 shows that treatment effects of the power of faith message were larger, or consistent with being larger, for women than for men on many outcome measures, as expected. This pattern is most apparent in donations, our behavioral measure of peaceful participation (last column). Among women, willingness to donate increased greatly in response to the power of faith message. Only 28% of women in the obligation to the common good workshops donated to CCMG, compared to 47.6% of women in the power of faith workshops. Men exhibited higher donation rates than women in the obligation workshops but were unmoved by the power of faith message. The power of faith message thus had a compensating effect by decreasing the gender gap in donations. The coefficients on the interaction terms in other columns (i.e., willingness to protest, vote or contact, and interest in politics) are all positive though not statistically significant. These results are consistent with the idea that on self-reported measures, too, the power of faith message had a greater influence on women.

Table 2

Effects of the power of faith message (relative to obligation to common good message) on political intentions, by gender

	Would protest	Would vote	Would contact Gov’t	Very interested in politics	Donated
Power of faith message	0.058*	0.001	0.015	0.030	− 0.047
	(0.035)	(0.053)	(0.059)	(0.066)	(0.088)
Pre-workshop level	0.312***	0.567***	0.505***	0.556***	–
	(0.082)	(0.076)	(0.079)	(0.051)	–
Female	− 0.011	− 0.014	− 0.035	− 0.073	− 0.183**
	(0.045)	(0.038)	(0.062)	(0.050)	(0.084)
Female*Power of faith	0.031	0.017	0.034	0.084	0.200*
	(0.051)	(0.056)	(0.098)	(0.076)	(0.116)
Respondents	288	288	288	288	279
Workshops	16	16	16	16	16
Mean before workshop	0.873	0.880	0.825	0.416	0.423 (power tr)

***p < 0.01, **p < 0.05, *p < 0.10; standard errors clustered by workshop session

The power of faith message also moved respondents to be more critical of the person in the vignette who chose not to vote. Compared to respondents in the obligation workshops, participants in power of faith workshops were much more likely to judge the hypothetical person as a “bad Christian" and a “bad citizen," and to say that the person’s actions were totally unjustified (see Table 3).22

Table 3

Effects of power of faith message (relative to obligation to common good message) on evaluations of non-participating person in the vignette

	Bad Christian	Bad citizen	Action totally unjustified
Power of faith message	0.064(*)	0.084**	0.096
	(0.040)	(0.037)	(0.076)
Respondents	288	288	288
Workshops	16	16	16
Mean in obligation condition	0.74	0.85	0.56

***p < 0.01, **p < 0.05, *p < 0.10; standard errors clustered by workshop session. Stars in parentheses are marginally significant

This increased inclination to participate peacefully in politics (and to criticize others for not doing so) was accompanied by a heightened disapproval of violence. Exposure to the power of faith message increased the rate at which participants viewed violence as never justified and decreased the rate at which they saw violence as sometimes justified, relative to exposure to the obligation message. Exposure to the power of faith message also decreased the rate at which respondents reported willingness to take part in a violent demonstration, relative to exposure to the obligation message (Table 4). About 13% of participants exposed to the obligation message said they would be willing to participate in a violent protest, compared to only 8% of those exposed to the power of faith message. These findings are consistent with the power of faith message’s increasing participants’ proclivity toward peaceful political participation.

Table 4

Effects of power of faith message (relative to obligation to common good message) on attitudes toward violence

	Violence never justified	Violence sometimes necessary	Would join violent protest
Power of faith message	0.101(*)	− 0.052	− 0.050*
	(0.060)	(0.033)	(0.028)
Respondents	288	288	288
Workshops	16	16	16
Mean in Obligation Condition	0.64	0.21	0.13

***p < 0.01, **p < 0.05, *p < 0.10; standard errors clustered by workshop session. Stars in parentheses are marginally significant

As explained above, we designed the study to compare two religious messages rather than a religious message to a secular message, or pure control condition, for several reasons, including a theoretical and practical interest in teasing apart any influence of the power of faith message from a common religious message or standard religious prime. However, the lack of an information-only condition makes it difficult to assess whether the information component of the workshops may have lowered willingness to participate in politics and the obligation to the common good message then raised that willingness back up to pre-workshop levels.

To assess the effect of the power of faith and obligation messages relative to a control condition, we take advantage of an unplanned implementation failure during the study. In four of the workshops—two that were supposed to play the power of faith message and two that were supposed to play the obligation message—audio equipment failed. The 69 respondents in these four workshops could not hear either the power of faith or obligation message. Because the informational portion of all workshops was delivered live by facilitators, delivery of civics information was unaffected by the technology failure. Thus, the four no-audio workshops were effectively civics information only workshops. This implementation failure was caused solely by unanticipated equipment failures and was not related to attributes of the settings or participants.

Accordingly, Online Appendix Table C.3 shows that on most observables (sex, education, age, marital status, religiosity, etc.) the no-audio workshop respondents were indistinguishable from respondents in other workshops. Importantly, participants in no-audio and audio workshops are indistinguishable on pre-workshop measures of political interest and willingness to participate in politics, which suggests that implementation mishaps were unrelated to potential outcomes. The only exceptions on observables were imbalances on church denomination and, to a lesser extent, language preference. The technological failures happened to occur on weekends when enumerators were recruiting from Catholic worship services, so the participants in the no-audio workshops were disproportionately Catholic compared to participants in the religious messages workshops, though there were mainline Protestants, Pentecostals, and Catholics in all three categories of workshops. Of course, this imbalance appears between no-audio workshops and obligation message workshops as well as between no-audio workshops and power of faith workshops and it is only between the latter types that we observe treatment effects. Additionally, Online Appendix Table C.6 shows that Catholics were influenced by the power of faith message, not just Pentecostals or mainline Protestants. There was also a slight imbalance between no-audio workshops and power of faith workshops in language preference. No-audio workshops included slightly more English speakers relative to Bemba speakers, and power of faith workshops included slightly more Bemba speakers relative to English speakers. However, all materials and surveys were provided in preferred languages, so this slight imbalance is unlikely to have affected comprehension. We also control for these variables in models below.

In Table 5, we present results from an analysis that treats study participants assigned to the four no-audio workshops as “information-only” participants and compares their outcomes at endline to the outcomes of participants assigned to the two types of treatment workshops. These models control for denomination and language to account for any detected imbalances across the types of workshops. We find that, relative to an information-only civics workshop, the power of faith message increased willingness to participate peacefully in politics, whereas the obligation to the common good message had little effect. Specifically, exposure to the power of faith message increased the share of participants willing to participate in protest by almost 6 percentage points relative to information-only workshops. There is no detectable difference in willingness to take part in protest between participants in the obligation to the common good workshops and information-only workshops. Likewise, the power of faith message strengthened the view that violence is never justified and that the person described in the vignette (who does not vote) is a “bad citizen.” The obligation message, by contrast, made no difference relative to information-only workshops. The power of faith message also decreased willingness to participate in violent protest by 8 percentage points, whereas the effect of the obligation message was nil and null relative to no-audio sessions.

Table 5

Effect of power of faith message on willingness to participate, relative to no-audio workshops (reference category)

	Willing to protest	Violence never justified	Willing to violent protest	Bad citizen
Power of faith message	0.057**	0.096**	− 0.082**	0.077
	(0.027)	(0.042)	(0.036)	(0.052)
Obligation to Common Good	− 0.030	− 0.033	− 0.000	− 0.009
	(0.037)	(0.072)	(0.038)	(0.061)
Pre-workshop Level	0.301***	–	–	–
	(0.082)
Respondents	285	285	285	285
Workshops	16	16	16	16
Controls	Yes	Yes	Yes	Yes

***p < 0.01, **p < 0.05, *p < 0.10; standard errors clustered by workshop session. To account for imbalances on observables, controls include English, Bemba, UCZ and Catholic. Nyanja and Pentecostal are reference categories

Overall, our results indicate that exposure to the power of faith message in face-to-face workshops moved youth to be more willing to participate, more critical of others who do not participate, and more disapproving of, and inclined to avoid, violence.

Discussion

Our findings are consistent with our hypotheses that, compared to exposure to an obligation to the common good message, the power of faith message would increase youth willingness to participate in political protest, to donate to an organization dedicated to promoting peaceful electoral participation (among women), to criticize others who do not participate, and to disavow the use of political violence. These results persist when we approximate for an information-only condition using workshops in which, inadvertently, neither religious message was delivered.

These results are consistent with observational findings, including from Zambia itself, that followers of Pentecostal churches—where the power of faith message is a prominent component of preaching—often exhibit higher levels of political participation than their counterparts from other Christian denominations (Sperber & Hern, 2018). As noted earlier, Christianity has grown at a remarkable pace in Sub-Saharan Africa in recent decades. Along with this growth has come the diversification of Christian messages relevant to political action. Alongside the Catholic and mainline Protestant Churches, which tend to emphasize congregants’ moral obligation to serve the public good, Pentecostal and Charismatic leaders commonly motivate their growing ranks (including large numbers of youth) not just by underscoring moral obligations but also by stressing that faithful action will bring change (Gifford, 2009; Marshall, 2009; McClendon & Riedl, 2019). This paper sheds light on the possible influence that a power of faith message has over and above other, more traditional prosocial Christian messages, which may already be chronically available in highly religious contexts like Zambia.

Our results also indicate that exposure to civics information alone (with no motivational messages) did not alter participants’ willingness to participate in politics, relative to pre-workshop levels. These results are consistent with other studies of (non-religious) civic education in other contexts, which have found that civics information alone often has little effect on political participation, even when it increases political knowledge (Campbell et al., 2016; Berinsky et al., 2016; Green et al., 2011). Information alone thus appears unlikely to resolve important barriers to political participation, at least among some citizens.

We argued that the power of faith message might be particularly mobilizing among youth because it addresses two particularly relevant barriers to participation: the belief that individual political action is futile, and hesitancy in the face of obstacles to participation. Neither an obligation to the common good message, nor civic information (e.g. about voter rights and procedures) alone, addresses these obstacles directly. By contrast, the power of faith message communicates explicitly that faithful action can bring about change and that persistence is divinely sanctioned and supported.

Nevertheless, one might posit that other processes could be driving the observed treatment effects. For instance, perhaps the power of faith message was more familiar to study participants than the obligation to the common good message and thus more easily integrated into their thinking about subsequent political questions. By design, our sample included members of mainline Protestant (UCZ), Catholic and Pentecostal churches, as well as born again and not born again respondents.23 Yet, when we examined whether treatment effects were driven by Pentecostal and born again respondents, who might be more familiar with the power of faith message, we found no evidence that they were. Indeed, Table 6 shows the estimated treatment effects of the two Christian messages among only those participants who do not identify as Pentecostal or born again. All treatment effects move in the same direction as in the pooled sample and most treatment effects of the power of faith message are large and statistically significant despite the smaller size of these subsamples.

Table 6

Results among non-pentecostals and non-born-again respondents, with no-audio sessions as reference category

	Willing to protest (Not Pentecostal)	Willing to protest (Not Born Again)	Violence never justified (Not Pentecostal)	Violence never justified(Not Born Again)
Power of Faith Message	0.053*	0.183**	0.195***	0.394**
	(0.028)	(0.064)	(0.056)	(0.168)
Obligation to Common Good	− 0.064	− 0.140	0.031	0.245
	(0.049)	(0.118)	(0.086)	(0.145)
Pre-workshop Level	0.335***	0.265	–	–
	(0.106)	(0.199)
Respondents	183	40	183	40
Workshops	16	13	16	13
Controls	Yes	Yes	Yes	Yes

***p < 0.01, **p < 0.05, *p < 0.10; standard errors clustered by workshop session. Controls are English and Bemba (with Nyanja as the reference category), to account for imbalances in observables. Respondents are already subsetted by denomination

Another possibility is that the power of faith message is a particularly positive message and was therefore more mobilizing. Perhaps the obligation to the common good message by contrast was depressing. We do not have measures of affect to test this possibility directly, but we do have a question about how rosy a view of democracy in Zambia the respondent has. Respondents were asked, “In your view, is the quality of democracy in Zambia today much better, better, the same, worse or much worse compared to ten years ago?” If the power of faith message improved the mood (Bassi, 2019) of study participants, one might expect participants in the power of faith condition to have responded more positively to this question. However, the results in Table C.7 in the Online Appendix do not support this conclusion. Participants in the power of faith workshop were no more optimistic about the quality of democracy than participants in the obligation to the common good message, and the differences with the no-audio condition are not statistically significant.

Building on this study, future research should examine additional potential mechanisms. For instance, listening to the power of faith message may have led people to expect that others would also be moved by the treatment, thus performing a coordinating function, whereas the commitment to the common good message did not. Another possibility is that the power of faith message mobilized a sense of national identity, which moved listeners to want to engage in politics after hearing the importance of doing so for the nation. There is no obvious, stable alignment between a particular denomination and a particular contemporary political party in Zambia at the moment.24 However, former President Frederick Chiluba (1991–2000) of the Movement for Multiparty Democracy used Pentecostal ideas to define Zambian national identity as a “Christian nation.” Most prominently, in 1996, President Chiluba officially declared Zambia to be a “Christian Nation” with a toothless provision in the constitution. The power of faith message therefore might have primed a sense of national identity and thereby increased the perceived importance of participating in politics. Future research should test these possible alternative mechanisms (positive affect, coordination, national identity) in this or other settings.

Future research could also investigate the scope of this study’s findings. First, our sample included young adults in select lower-income neighborhoods of urban and peri-urban Lusaka, Zambia. We do not know whether our findings extend to youth residing in rural areas or wealthy neighborhoods. Future studies could evaluate the generalizability of our findings to such settings. Second, our study involved a one day workshop intervention with measurements taken pre- and post-workshop. Future research could probe the consequences of repeated interventions and explore the extent to which treatment effects detected in our study persist over time. Indeed, McClendon and Riedl (2019) suggested that the influence of religious messages on political participation may last for only a few days or a week unless repeated and reinforced. The duration of the treatment effects in this study should be the subject of future research. Third, this study was limited to an overwhelmingly Christian country and to political participation outside of a general election cycle. Others have noted comparable messaging within Islam to the power of faith message examined here, so future research could examine whether our findings travel to predominantly Muslim contexts or to electoral periods.

Finally, it is notable that, despite the low levels of youth political participation in Zambia and many other countries around the world, the pre-workshop levels of willingness to participate politically among our study participants were quite high: before the workshops, 87% of people in the obligation to the common good condition said they would be willing to participate in protest, 88% said they would vote, 82% said they would contact a government official. These high rates could be due to the low cost of self-reporting these behaviors in surveys or could be due to the nature of the sample: The study participants, though invited on a random basis, had to choose to show up to civics education workshops in order to be part of the study, so they may be individuals who were already more inclined to participate in politics. It is important to point out that civic education programs like the ones we studied here typically rely on volunteer participants, so our study speaks to the effects of such programs on the participants they are actually likely to attract. Furthermore, we suspect that the high pre-workshop levels of willingness to participate actually made our study a tough test for the intervention messages because of the possibility of ceiling effects. Remarkably, the power of faith message raised willingness to participate even when willingness was already relatively high. Finally, the power of faith message also moved measures of participation that were more costly and not already high: Donations in the obligation to the common good workshop were lower (46% of men and only 28% of women donated), but donations moved in response to the power of faith message, particularly among women, who reported lower willingness to participate in the pre-workshop survey. Nevertheless, future research could investigate whether our findings would extend to study participants recruited through different means and to other behavioral measures of participation. We hope that this study’s community collaborative approach, recruitment strategy, interventions and findings lay a foundation for future research.

Conclusion

To our knowledge, this is the first study to examine how Christian messages influence youth political participation. The dearth of research on Christian civic engagement programming has meant that scholars and civil society leaders have lacked insight into the impacts of these widespread programs. Enlisting experimental and community-collaborative methods, this study found that, among 18-35 year olds in Lusaka, Zambia, exposure to Christian messages focused on the power of faith to make change increased individuals’ willingness to participate in peaceful political action, to disavow political violence, and to critically evaluate others who chose not to vote, compared to religious messages focused on the moral obligation to contribute to the common good. These findings extend insights from studies of the influence of different types of sermons (McClendon & Riedl, 2019), accord with arguments in the youth politics literature that stress the importance of self-efficacy and perseverance in promoting youth political participation (Holbein & Hillygus, 2020; Holbein et al., 2020), and challenge theories that posit moral obligation as the primary mechanism linking religious practice and political participation (Oviedo, 2016; Skitka et al., 2005). Future research should explore the generalizability and scope of this study’s findings.

Below is the link to the electronic supplementary material.

Supplementary material 1 (PDF 12824 kb)