Abstract
The effects of religion on political behavior are difficult to study for a number of reasons. One difficulty is that “religion” is not a singular entity and is thus unlikely to have a unidirectional effect on political behavior. Another difficulty is that everyone in a particular place and time might be embedded in the same set of religious practices, such that the counterfactual is difficult to assess. In response to these and other challenges, we suggest opening up the black box of religion in order to examine the influence of its component parts. Specifically, we focus on exposure to sermons. We describe a study about the impact of Christian sermons in sub-Saharan Africa on reactions to inequality. We discuss the approach’s advantages and limitations and discuss how to integrate it with the study of other aspects of religion and how the approach might apply to other domains of political behavior.
Introduction
Religion as a category of analysis and object of study is multifaceted, internally heterogeneous, and lacks coherent boundaries. How then can we assess its possible influence on political behavior? In many ways, the issues that plague the study of religion’s influence on political behavior are a subset of the issues that plague the study of cultural influences on political behavior more generally. Culture encompasses “the meanings and values which arise amongst distinctive social groups and classes. . . and the lived traditions and practices through which those understandings are expressed and in which they are embodied” (Hall, 1980, p. 63). Although scholars are often aware of the difficulties of studying the influence of political culture, it is easy to fall back on arguments characterized by essentialism, assuming implicitly or explicitly that a particular group of people share a fixed set of values and practices (Martin, 2018). However, “religion” is not a singular, unchanging entity and is thus unlikely to have a single, unidirectional effect on political behavior. And yet to investigate the causal influence of any particular religious tradition or experience on political behavior, one has to identify it, and one has to identify instances in which people are subject to it as well as instances that approximate the counterfactual (Paluck & Ricart-Huguet, 2017). Finally, one has to deal with the issue of self-selection: that adherents might self-select into religious practices to which they are already politically predisposed (Margolis, 2018), and thus religious belonging might simply reflect (rather than cause) political behavior.
One way forward is to unbundle religions (and cultural systems more generally) into disaggregated parts. 1 Within any given faith tradition, and across its varied forms around the world, religion typically combines many different elements: ideas about the spiritual and physical worlds, rituals and practices, organizational resources, hierarchies, social networks, social identity categories, social insurance, service delivery and so on (Table 1). These component parts are each likely to influence political behavior in different ways, and they can sometimes be randomly assigned by nature or the researcher, allowing scholars to circumvent concerns about self-selection. For instance, as-if random processes (unexpected death) might alter the leadership of a particular house of worship, and thus its governance or service content (Tunón, 2017). Or, particular components of religion (its sermons, practices, associated social identities) might be brought into the lab and randomly assigned or primed by researchers without stepping too far outside the bounds of what would happen in the real world. Thus, disaggregating religion can provide ways to approximate the counterfactual and to circumvent concerns about self-selection. Disaggregating religion can also allow the researcher to describe and situate the component parts of religion under study in a particular time and place without assuming religion is a monolith.
Examples of Disaggregating Religion.
In this paper, we discuss a way to disaggregate religion that focuses on sermons and their possible influence on political behavior. Exposure to sermons is a regular and integral part of lay religious experience in many faith traditions, including in Christianity, Islam, and Judaism. 2 The content of sermons also varies across individual adherents and over time for any given adherent as clergy craft and deliver different sermons each week. A focus on sermons thus makes clear that religion is not a unified thing and demands that researchers investigate, rather than assume, the multiple possible directions of religious influence on political behavior. Sermons also offer an opportunity to approximate the counterfactual (of what people would have done in the absence of religious influence) while studying a core part of religious experience for many people around the world.
Sermons provide authoritative interpretative maps about how the world works and are thus strong candidates for influencing political attitudes and behavior. The ideas communicated in sermons often grapple with deep and difficult-to-answer questions about how to understand what is going on in the world: Why have certain problems occurred? How seriously should we take these problems, and how serious are they likely to become? What do they mean? What should we do in response? Answers to these questions, especially when delivered by a trusted, authoritative source, like a local pastor, 3 are likely to be sought after and incorporated into people’s everyday thinking. Sermons can thus provide mental maps of how to conceptualize how the world works and therefore how to engage with the world materially and socially. Even when not explicitly political, sermons are likely to incline people toward different strategies of political, economic, and social action. When political problems, and proposed solutions, arise, sermons are likely to condition citizens’ attitudes and behaviors in response. Narrowing in on sermons as ideological frames that influence their listeners is a promising way to understand a specific pathway through which religion shapes political behavior.
Disaggregating religion to focus on sermons acknowledges that the content of religion varies greatly, across traditions and sects, within the same time period and place, and also within the same tradition and sect across time and space (Hurd, 2015; c.f., Huntington, 1997). The approach sets aside official doctrine and does not take the content of locally-delivered sermons for granted. Instead, the approach proposes that we first observe and describe the content of sermons being delivered within a particular time and place, before theorizing and testing for their effects on political behavior. This disaggregated, contextual approach to studying religion means not reflexively using religious denomination as a proxy for, or approximation of, the bundle of beliefs, practices, social identities, organizational structures and resources that might hang together within a single denomination. Before testing for the influence of “Catholicism” or “Pentecostalism” on political behavior, our approach demands that we know something about the particular ideas being disseminated by Catholic and Pentecostal houses of worship in a particular time and place. The approach allows the use of denomination as an explanatory variable for political behavior only if sermon content varies systematically by denomination in that particular time and place.
Indeed, rather than treating religious influence as singular and robust, or as a chronic force on everyone within a particular society from childhood on, the approach allows that exposure to religious content might vary from person to person, or even from week to week, or within the same week for a given person. The ideas presented in sermons might vary within the same house of worship from week to week. People might attend different houses of worship with family or friends, or in an effort to choose the house of worship that appeals to them most. Different world views might be particularly salient to them from 1 week to another, depending on the sermons to which they have recently been exposed. This approach to studying religion may be particularly apt where multiple faith traditions or denominations are practiced side by side with relatively fluid social boundaries between them and among them. There, people are regularly exposed to sermons offering different metaphysical perspectives.
But the disaggregating approach makes certain aspects of religious influence visible while rendering others less visible, as reflected in Table 1. The approach of focusing on sermons zeros in on individual-level religious experience and behavior, and allows for plausible approximations of counterfactuals, but it also pushes collective and institutional aspects of religion to the background. After the spread of COVID-19 around the globe and the subsequent restrictions on public gatherings, focusing on the effects of isolated exposure to religious ideas is obvious in its importance: people around the world are now worshipping in their homes, accessing content online or on the radio or TV, without setting foot in a church and without gathering with others. Understanding the political effects of being exposed to sermon recordings, unbundled from other aspects of religious experience, are urgent now, even as they were important before.
Nevertheless, a disaggregating approach to studying religion need not ignore other aspects of religious experience entirely. We may want to know whether the effects of sermons on political behavior are similar when sermons are listened to in isolation and when sermons are bundled together with other aspects of religious experience. Does a sermon influence political behavior to a greater or lesser degree, or in a different direction, when the ideas in that sermon are shared through a social network rather than absorbed alone? Or when the sermon is accompanied by institutional rules regulating church members’ behavior? Mixed-method research designs can help researchers explore these questions. For instance, when studying the effects of sermons, combining laboratory experiments with focus groups or ethnography could integrate the study of both individual and social aspects of religion and help us understand how different components of religion interact in order to influence political attitudes and behavior.
Below, after first providing more details about the general approach as well as about a specific research design that used descriptive analysis of sermons in combination with a laboratory experiment to study sermons’ effect on political attitudes toward inequality, we then suggest alternative research designs that might integrate other aspects of religion. In particular, we offer suggestions for how to combine laboratory or survey experiments with analyses of observational survey or administrative data, with focus groups, with ethnographic work, and with comparative historical analyses of cases. We then suggest how such research designs could straightforwardly be applied to better understand sermons’ effects on a range of individual political attitudes and behaviors. Many domains of political action demand that people take a stance on how the world works now and is likely to work in the future. For instance, evaluations of climate change policies, or responses to public health crises, reactions to racial inequality, or mobilization against democratic backsliding, require that people have a world view about how serious such problems are, or are likely to become, and about what those problems’ likely causes are. Metaphysical content communicated in sermons can provide an authoritative, interpretive map of the world and thus influence adherents’ positions on these issues as well.
Disaggregating Religion and Isolating the Effects of Sermons
In Table 2, we provide an example sequence that one might follow in order to disaggregate religion and focus on the effects of sermon content on political behavior. One advantage of the approach, as we have mentioned, is that the approach does not make assumptions about how doctrine is interpreted in a particular time and place and does not treat religious content as fixed. A first step, then, is to describe the sermon content circulating in a particular context. In our study (results of which we present in the next section), we accomplished this descriptive task by constructing a sampling frame of churches in Nairobi from Google Maps and randomly sampling from the three largest categories of African churches—Pentecostal, 4 Mainline Protestant and Catholic. 5 We used Google Maps because there was no existing repository of sermons in Nairobi, and because not all churches in Nairobi are formally registered. In places with existing sermon repositories, 6 or where registration lists are comprehensive, researchers could use those sources as a starting point instead. Researchers could also use a random walk procedure to sample churches. One advantage of the random walk procedure over our use of Google maps would be that the random walk might pick up religious associations that do not have their own dedicated buildings (e.g., religious associations that worship in schools or assembly halls or other borrowed space on the sabbath). Google maps tends to miss such churches. A disadvantage of the random walk method is that it could be more resource intensive, because enumerators would have to canvas widely during worship service times (in order to capture churches without dedicated structures) and then arrange to go back to observe sermon content. Google maps’ coverage of Nairobi is detailed, accurate and comprehensive enough that we decided to use it as our starting point there. In places with less comprehensive or outdated coverage by Google maps, researchers might choose to use a random walk procedure or registration lists instead.
The Empirical Approach in Practice.
Our sampling method captured a wide range of churches in Nairobi (megachurches and tiny churches, long-standing and newer churches, churches serving rich/upper-class congregations and churches serving very poor congregations). Along with a team of Kenyan research assistants, we then visited each church in order to observe Sunday worship services, to interview the head religious leader (pastor or priest) and to collect and transcribe sermon texts from the randomly sampled churches. Some of the churches in our sample had websites with recordings of their sermons or dedicated YouTube sites. For these churches, we could collect sermons at many points in time. 7 For other churches, we could collect sermon content only by observing the services ourselves, which limited the number of sermons we could observe from these churches but ensured that those churches (typically smaller or serving poorer congregations) were still included in the study. We attempted to visit each church at least twice: once during a non-election period and once during the lead-up to the 2017 general elections so that we would have at least two observations for each. We were not able to randomly assign the observer (ourselves or one of our research assistants) to the services, but we noticed very little difference in sermon content depending on whether a white researcher or a black researcher visited the service, or depending on whether we collected the sermon from an online archive or in person. Researchers conducting future studies could randomize observers to services to investigate more systematically whether observer identity and presence affects content. Interviewing the head religious leader allowed us to obtain consent from people whose work product we were studying and also allowed us to collect additional meta-data about the houses of worship (e.g., self-identified denomination, ties with other churches, characteristics of the congregation) that we could use to understand whether certain types of sermon content were associated with particular types of churches. Because the online archives and in-person sermons did not demonstrate wildly different themes, we think religious leaders did not substantially change their sermons in response to knowing that they were part of a study, but this is a research design element that future researchers could also investigate. After we finished collecting sermons from churches in these ways, we analyzed the corpus of sermon transcriptions and worship service notes through hand coding and computer-assisted text analysis. 8
Another advantage of the approach is that exposure to sermons can be varied orthogonally to other individual and contextual factors in order to draw inferences about the causal effect of sermons on political behavior. In our specific study, we thus followed up our descriptive analysis of sermon content by creating two treatments for a laboratory experiment that drew on the contemporary, real-world sermons we had observed. The texts of the treatments are in the appendix. We drew phrases for the two religious treatments directly from actual sermons, keeping some phrases the same in each message (only if they were phrases that, based on our descriptive survey, could reasonably be found in any type of church we observed) and varying other phrases to reflect the main thematic differences we found across the churches. We created two religious treatment recordings because we found that our sermons fell into two broad camps. We had each of the messages translated into Swahili and back-translated into English, because we found in our descriptive survey that churches in Nairobi regularly deliver sermons in Swahili or in English, or in a mix of both. 9 A male research assistant who is a life-long resident of Nairobi and has an authoritative sounding voice then recorded all audio messages with the same cadence and tone. 10 Together, these steps ensured that the treatments differed only in their content, rather than in the identity, personality or style of the speaker. We constructed the treatments in this way so as to be able to test whether the content of the ideas alone can influence people’s political attitudes and behaviors. However, researchers could also use this general approach to vary the identity of the speakers (as communicated by accent, for instance), or to vary the style of delivery, in order to explore how identity and emotion interact with ideational content to influence political attitudes and behavior. Our specific study used audio recordings because of the technological set-up of the lab we used, but other studies using this general approach could use video or text. 11
In creating treatments in this way, we erred on the side of realism rather than on the side of conceptual purity. We could have written treatments from scratch in order to control fully how concepts and ideas varied across conditions. But in the real world, sermons are bundles of ideas and topics. By drawing on actual sermons, our treatments mimicked the kinds of bundles of ideas we observed in our context and thus ensured our treatments would resonate with study participants and would be viewed as realistic. For instance (as we discuss in more detail below), in Nairobi we observed sermons that bundled the idea of institutional and collective responsibility for earthly problems with a view that people may not be able to do much to solve those problems (other than find ways to cope). We observed other sermons that bundled the idea of individual responsibility for earthly problems with a view that through faith alone people can make a material difference in the world. If we had erred on the side of conceptual purity, we might have separated these ideas from one another when creating the treatment recordings for the lab. We might have created treatments that combined the idea of institutional and collective responsibility with a view that people can make a material difference in the world, for instance. But we did not actually observe this combination in our context, and so we refrained from introducing it in the lab. In making this design decision, we could be sure not only that treatments would resonate but also that we were estimating quantities that are of real-world importance: the difference in behavior between exposure to one circulating set of ideas and another. We also avoided introducing study respondents to new or foreign bundles of religious ideas, which might have been viewed as a form of proselytizing and thus ethically problematic (Nielsen, 2015). 12
As a third step, we created secular comparison recordings. We removed any religious phrasing explicitly evoking God, Jesus, the Bible or the spiritual world from the religious audio recording texts and replaced them with non-religious language in order to create secularly-worded texts on the same general topics. Each study participant was assigned to listen to one of the four treatment messages (either one of the two sermons, or one of the two secular equivalents). When studying the influence of sermons on political attitudes and behavior, an important question is what the appropriate comparison conditions are. What is the counterfactual of being exposed to sermon content? One counterfactual is exposure to different sermon content. In our study, this counterfactual is captured in the comparison between the two religious treatments, which are in fact examples of two principal types of sermons to which study participants might be exposed in Nairobi. A second counterfactual is exposure to messages that are not religious in nature. This counterfactual is captured in our specific study with the comparison between religious and secular recordings. In our study, this counterfactual makes sense because there are in fact secular equivalents of the themes being communicated in sermons. The bundle of ideas that emphasizes individual responsibility for earthly problems and says that people can make a material difference in the world often appears secular self-help books. The bundle of ideas that emphasizes institutional and collective responsibility for earthly problems and laments that people may not be able to do much to solve those problems often appears in punditry and on talk shows. In fact, we shared the texts of our four treatment conditions with local informants before running the lab experiments, and they identified each treatment condition as plausibly heard in Nairobi and identified churches as sources of the religious treatments and books and radio programs as sources of the secular treatments.
Our study thus employed a research design that examined realistic and theoretically interesting comparisons between different primes, rather than between a prime and a pure control (Chong & Druckman, 2007). In other studies examining the influence of sermons, we have used secular placebo treatments on totally different topics as the comparison prime (McClendon & Riedl, 2019). For instance, when studying sermons’ influence on political participation, we used a weather report designed to lift listeners’ moods in the same way that listening to a certain type of sermon might (but using weather forecasts rather than religious ideas), in order to test whether the ideas themselves, separate from the mood induced, changed behavior. That design also employed a comparison of exposure to sermon content with exposure to a secular message that respondents might realistically encounter in their everyday lives. A third counterfactual that researchers could employ when studying the influence of sermons is exposure to no message at all (a “pure control”). In our studies, we did not use a pure control because participants assigned not to listen to a recording could have been easily identified by lab staff and by other study participants, and we wanted to ensure that lab staff treated all participants the same and that all participants had the same experience in and outside of the lab, apart from the sermon content they heard. But if researchers were to conduct an online study of sermons, or could otherwise administer treatments to study participants without alerting other study participants or research staff to treatment assignment, then a pure control could approximate the state of not hearing a sermon.
As a fourth step, we worked with the Busara Center for Behavioral Economics, 13 a standing laboratory in Nairobi, to run the lab experiment. Using a lab environment ensured a high degree of control over the experience of the study participants during the study, further ensuring that we could isolate the influence of sermon content on behavior. 14 During all laboratory sessions, participants sat in cubicles, each with a touchscreen computer and an audio headset, which was used to play the treatment messages, and without other sources of distraction. Instructions for all tasks were delivered in Swahili by laboratory staff. The language on the computer screens was English. The Busara Center has found that this combination of English screens and Swahili oral instructions maximizes participant comprehension in the Nairobi lab. As a result of all of these procedures we could ensure that all participants had the same access to content and the same experience during the study apart from the differences in content that they heard in their assigned treatment message. If all participants have equal access to and facility with requisite technology, researchers could ensure similar levels of control with an online, mobile phone, Whatsapp or other social media platform, though it would be more difficult to make sure that respondents are participating in the study in similar environments and without distractions. Listening to the sermons in the lab is similar to listening to sermons in church pews in the sense that in both environments accessing various forms of distractions is generally discouraged.
As a fifth step, we invited Christians, of varying primary denominational affiliations, to participate in the study. We sought recruitment of Christians into the study, because boundaries between Christian denominations are soft in Nairobi and in many parts of Sub-Saharan Africa (Gifford, 1994; Marshall, 2009; Ngong, 2014). Churches certainly try to demand commitment from congregants by asking for regular tithes (McCauley, 2013), but Christians still have discretion over which churches to attend and can switch between them, or attend ones other than their primary house of worship with family or friends. Incidental exposure—via street preachers or radio messages or billboards—is also common. One study, Wafula (2003), found that, in a representative sample of adults in Nairobi, approximately 30% of respondents reported having moved between Christian denominations at least once (p. 49). In our lab experiment, we asked study participants, “In the last year, how many different churches have you attended, if any?” They could answer “none,” “one,” “two” or “more than two.” Twenty-nine percent said that they had attended at least two. Twelve percent reported attending more than two different churches in the last year. These are reports of purposeful attendance, where “attended” may have been interpreted as more commitment than an occasional visit to a worship service. Casual visits to services at houses of worship other than their own may be even more prevalent. Certainly, the share of Nairobi residents incidentally exposed to messages from houses of worship and preachers other than their own is likely much higher (England, 2005). We have ourselves been exposed to sermons from itinerant preachers on buses and the streets of Nairobi, and Christian radio programming is common. Wafula (2003) notes that it is also not uncommon in Nairobi for different members of the same family to attend different houses of worship (pp. 4–5), so Nairobi Christians may be exposed to messages from other denominations even within their own homes. Random assignment of Christians (of any denomination) to different Christian sermons is not an implausible phenomenon in this context. One can imagine a Christian being assigned to any of the treatment conditions, even if they are not necessarily in practice. 15
Note that we used the lab environment to randomly assign sermon content to participants so that we could circumvent the concern that people have some choice over the house of worship they attend, and thus over the sermon content to which they are most frequently exposed. (In the absence of random assignment, any differences in behavior among people exposed to different sermon content might instead have had to do with differences in their pre-existing political proclivities that led them both to expose themselves to that content and to engage in that political behavior.) Nevertheless, researchers could also use this general approach and the lab environment to study the drivers of selection into different types of sermon content. After having described the sermons being delivered in a given context, researchers could employ participants’ decisions to listen to a particular type of sermon as a main outcome measure. DeBenedictis-Kessner et al. (2019) have shown how lab and survey experiments can combine free choice over treatments with forced choice in order to estimate treatment effects among those who would have self-selected into exposure. DeBenedictis-Kessner et al. (2019)’s focus was media influence, but the idea is easily extended to sermons. If applied to our approach, one could combine free choice of sermon content with forced assignment to sermon content to estimate the degree of influence of sermons on political behavior among those who are in practice most likely to choose to listen to those sermons.
As a sixth step, we used behavioral tasks in and outside of the lab to collect outcomes of interest. After exposure to their assigned treatment recording, participants in the lab study engaged in dictator games to measure prosocial behavior and revealed preferences over inequality. We describe these in more detail below. Attaching monetary stakes to the tasks ensured that participants would take the decisions seriously and not simply seek to give us the answers we wanted, or satisfice, when answering attitudinal questions. In another study (McClendon & Riedl, 2015), we partnered with a non-religious, non-government organization that was running a political text message campaign in order to measure political participation outside of the lab. During the lab experiment, we told study participants about the organization and the text message campaign. We gave them information about how to take part. But we did not ask them to decide whether to participate or not while in the lab. We ended the lab experiment and then, in collaboration with the NGO, we observed who participated in the text message campaign in the couple of weeks following the lab experiment, matching cell phone numbers to treatment assignment. Text message campaigns happen to be one common form of political participation in Nairobi, and are increasingly common in various parts of the world, but researchers could use whatever measures capture common forms of political behavior in their research context. For instance, one could imagine petition signing or showing up to political organization meetings, attending a rally, or participating in a WhatsApp group, as measures of political participation in other studies. Researchers can use either supervised or unsupervised outcome measures when studying the influence of sermons on political attitudes and behavior; they could conduct a follow-up survey, soliciting political attitudes. Researchers could also measure outcomes long after the administration of treatments in order to gauge the longevity of sermons’ influence. Periodically in the weeks or days following the administration of treatments, researchers could also randomly assign some participants to receive reminders of the treatment messages they were assigned to in order to estimate the decay rate of sermons’ influence. The general approach opens up lots of possibilities for exploring the influence of this aspect of religion on many forms of political attitudes and behavior—both in the short-term and the long-term.
In sum, the approach of disaggregating religion to focus on the political influence of sermons is general and flexible enough to be used in other contexts and for the study of other types of outcomes. The basic idea is to take real-world sermon content, to isolate that content from other aspects of religious experience, to leverage random assignment of that content in order to draw causal inferences, and to measure outcomes that realistically capture political attitudes and behavior that might be influenced by the content. The approach can be applied to studying the influence of religions other than Christianity, if those religions have sermons as a core component. Indeed, the approach can be used anywhere where there is variation in sermon content, provided researchers first observe and describe, rather than assume, the sermon content circulating in that time and place. The lab environment lends itself well to the approach, but online, mobile phone and other platforms might also be able to accomplish the basic goals of the approach, depending on feasibility and on the realism and degree of control they provide in a given context. Choice over sermon content can be incorporated, as can examination of the duration of sermon influence. The approach is general enough that it could be combined with measures of a range of political attitudes and behavior, depending on what is appropriate and realistic in the study context.
An Example Study: Sermon Content and Redistribution
In an example of this general approach, we conducted a study of the influence of Christian sermons in Nairobi to shed light on a long-standing area of political interest: inter-personal redistribution. The guiding logic is based on the contention that sermons are a prevalent and authoritative source of guidance for grappling with deep metaphysical questions. Sermons often discuss causal attribution for earthly problems and discuss what human beings can and should do to address those problems. Such problems include economic inequality, so we thought it possible that exposure to sermons might have implications for people’s reactions to inequality.
Although the descriptive analysis of sermons being delivered across churches in Nairobi demonstrates that sermon content was not explicitly political, we found that sermons did display themes that could plausibly shape the way people reacted to economic inequality. Clergy-constituent communication during worship services was very rarely on openly political topics, even when we visited the churches in the weeks directly before the 2017 general elections. Rarely if ever did preachers mentioning particular parties, candidates or one’s role as a citizen explicitly. They did not give explicit political directives. 16 Nevertheless, there were other striking differences in sermon content that were potentially relevant. Most starkly, we found that sermons (1) attributed responsibility for earthly problems (poverty, inequality, suffering) to very different sources and (2) made different characterizations of the nature of human agency in enacting earthly change. On the first dimension, Pentecostal sermons tended to locate the source of earthly problems primarily inside the individual, 17 whereas Catholic/Mainline sermons tended to locate the source of earthly problems more in factors external to the individual—in the relationships between people and in community structures and institutions. 18 On the second dimension (portrayals of the possibilities for creating change in this world), the Pentecostal sermons were vehemently optimistic about the power of human agency, enabled by strong faith: they promised material change if people would strengthen their faith, think positively and act on that faith. 19 Others have called this the prosperity gospel or “gospel of success” Gifford (2009). The Mainline Protestant and Catholic sermons, by contrast, encouraged people to try to better the world but promised no high likelihood of success or possibility of an end to earthly problems. 20 Notice that these differences in content are not the same differences that one might have observed when surveying Mainline Protestant and Catholic, or Pentecostal churches, in other places and at other points in history, 21 so conducting the descriptive exercise was important.
We hypothesized that the differences in metaphysical content we observed could have implications for listeners’ attitudes and action around inequality and redistribution. Ex ante, we had hypotheses running in both directions. 22 On the one hand, we thought that the Pentecostal sermons’ attribution of responsibility for earthly problems to the individual, combined with Pentecostal sermons’ emphasis on individual agency (through strong faith), might result in listeners being more proactive in redistributing their own wealth. That is, we thought it possible that people exposed to Pentecostal sermons would feel more empowered to do something about inequality and to act more prosocially as a result. 23 Inspired by the notion that embodying faith can change one’s material circumstances, those assigned to the Pentecostal-like sermon might have felt more capable of being generous with others. Pentecostal sermons’ attribution of responsibility for earthly problems to the individual might also make listeners feel that it is their responsibility to do something to fix the problem of unequal distributions of wealth, whereas the Mainline/Catholic sermons’ acknowledgment of larger cultural and collective responsibility might have absolved listeners of having to take action on their own to solve the issue.
On the other hand, Mainline Protestant and Catholic churches are historically associated with more extensive social welfare provision (Scheve & Stasavage, 2006; Seay, 2013), and so if listeners respond to a sermon by emulating its associated church’s activities, we thought it also possible that those exposed to Mainline/Catholic sermons would behave more prosocially than those exposed to Pentecostal sermons. The Pentecostal sermons’ attribution of responsibility for earthly problems to the individual could in fact run against prosociality and lead listeners to blame those with less for their lot, leaving them less inclined to be generous with others. This hypothesis aligns with a broader literature which suggests that individualistic societies are less likely to support redistribution than collectivist ones (Gorodnichenko & Roland, 2012), based largely on the organizational and systemic aspects of social interaction. It has been argued that members of collectivist societies are more inclined toward interventionist attitudes such as the belief that “government should take responsibility to ensure that everyone is provided for” (Pitlik & Rode, 2017). We thus used the experimental design to adjudicate between two competing hypotheses about the influence of Pentecostal, versus Mainline/Catholic, sermons on inter-personal redistribution.
Finally, there is a strand of research in the study of psychology and religion that suggests that religion makes people more generous with members of their in-group, even though it does not always lead people to be more generous with everyone. (See Preston and Ritter [2013] for a review.) The idea is that religion may prime people to feel more approval of and obligation toward those who share their same values and beliefs but at the same time feel more righteous toward and disapproving of those who share other values and beliefs. If this were true, we might expect both religious treatments in our study to lead to greater generosity compared to the secular treatments if study participants know that the recipient of their generosity is part of their in-group.
One hundred thirty-one Christians, of varying denominations, were assigned to one of the four treatment recordings (two religious, two secular) and then asked to play dictator games to measure their level of generosity with others. They played an anonymous and a non-anonymous version of the dictator game. In the anonymous game, participants were randomly and anonymously paired with another participant in the lab and then given 100 Kenyan shillings (approximately one US dollar, or a third of a daily wage in Kenya). They had to choose whether to keep all 100KES or to give some or all of it to their anonymous partner. (They also received money from their randomly assigned partner—if that person shared with them—but no player found out how much she earned until the end of the experiment.) In the “known” dictator game, participants were given another 100KES. They were then told that they could pick someone they knew and have any part or all of that 100KES transferred to that person by mobile phone (MPesa). This is a real and common form of remittance in Nairobi. Because subjects could pick the recipient, we considered this an opportunity to give to someone belonging to one’s in-group. 24 There was no deception in any of these studies. Any amount designated was actually given to the anonymous partner (though not until the end of the experiment) and sent to the chosen person via the phone number provided for her by the participant.
Sermon exposure in the lab affected how individuals played these redistributive tasks. Figure 1 presents histograms of the amount kept (out of 100 Kenyan shillings) in the anonymous dictator game and the known recipient dictator game, by treatment condition. In general, allocations shifted to the left across all treatment conditions when we moved from the anonymous game to the known-recipient game. In other words, not surprisingly, all participants—regardless of the experimental condition to which they were assigned—behaved more generously with someone they knew than with an anonymous partner. Nevertheless, despite this commonality, behavior in both games looks quite different depending on the content of the sermon to which study participants were exposed.
Histograms of amounts shared in dictator games.
Study participants assigned to listen to the Mainline/Catholic sermon behaved more selfishly in these tasks, despite general associations of the Catholic and Mainline Protestant churches with social welfare provision. Study participants assigned to listen to the Mainline/Catholic sermon exhibited the highest rate of keeping all of the money for themselves (mean of 86.50KES kept in the anonymous game, mean of 80.30KES kept in the known recipient game). The distribution of amount kept is also more negatively skewed in this condition than in others. 25
By contrast, people assigned to listen to the Pentecostal sermon behaved much more generously than those in the Mainline/Catholic condition, particularly in the anonymous dictator game. The mean of the amount kept in the anonymous dictator game by those exposed to the Pentecostal sermon was only 73KES, which is 9KES less than the average amount kept in the other conditions 26 and 13.5KES less than the average amount kept in the Mainline/Catholic condition. 27
Table 3 shows one version of results using simple OLS regressions. 28 Assignment to the Mainline/Catholic sermon increased the amount that participants kept for themselves, rather than share with an anonymous other player, compared to all other conditions. By contrast, those assigned to hear the Pentecostal sermon were more likely to behave generously in both games than those in the Mainline/Catholic condition. As can be seen by looking at the modes in Figure 3, those in the Pentecostal condition were more likely, for instance, to give up to 50% of their allocation (act as an “equal divider”) than those in comparison conditions, where as people assigned to the Mainline/Catholic condition were less likely to act generously enough so as to divide their allocation equally with their assigned partner.
Treatment Effects on Behavior in Dictator Games (OLS).
Comparison condition is the Secularized Pentecostal message.
p-value < .10. Standard errors in parentheses.
Some research on religion has explored the possibility that religion writ large might lead people to behave more prosocially (Ben-Nun Bloom et al., 2015; Norenzayan, 2013; Sachs, 2010; Shariff & Norenzayan, 2007; Warner et al., 2015), or might change people’s level of support for redistribution (Scheve & Stasavage, 2006), or might lead them to be more generous with members of their in-group (Preston & Ritter, 2013). We should question whether religion writ large is always the right focus. Religion may have very different consequences for prosociality and reactions to inequality depending on the content of religious ideas communicated to adherents. People exposed to religious recordings were not more generous with known recipients than people exposed to secular recordings. And study participants exposed to the Pentecostal sermon behaved differently than study participants exposed to the Mainline/Catholic sermon in the anonymous dictator game.
Indeed, the results most strongly support our hypothesis that people exposed to Pentecostal sermons would feel more empowered to do something about inequality and to act more prosocially as a result. Those sermons point to the individual and his or her strength of faith as responsible for earthly problems while at the same time promising that by embodying strong faith individuals can make a difference in their own lives and in society at large. This message may lead people to be more proactive in redistributing their own wealth. Listeners may feel that it is their responsibility to do something to fix the problem of unequal distributions of wealth, and less able to sit back and rely on the actions of others. The act of giving might itself be an act that embodies faith: even if I give up this income, I believe that things will still work out for me. More generally, the results support the notion that exposure to the ideas in sermons can influence politically-relevant behavior.
This was a small lab study, and the results could therefore be driven by imbalances across treatment groups. Perhaps the group assigned to hear the Pentecostal sermon happened to be very different in type (e.g., more educated, of a particular ethnic group) than those assigned to hear the Mainline/Catholic sermon, even though individuals were randomized into treatment conditions. In Appendix B.1–B.3, we provide several tables showing balance on the observable attributes the lab collected (age, level of education, sex, marital status, asset ownership, number of children, ethnic group membership, and denominational affiliation). There is little statistical indication that the experimental conditions were populated by wildly different types of people, but because the numbers are small we still note that the Pentecostal message condition did have a few more women, Pentecostal identifiers, people without secondary education and a few more members of the Kisii tribe than other conditions did. We therefore also show the results of Table 3 in Appendix Table B.4 when including controls for observable attributes. The results are not driven by imbalances on these attributes. 29
Moreover, studying the effects of sermons on political behavior need not be done in the lab alone or in the lab at all. The inferences drawn from lab or survey experimental studies of sermons’ effects on political behavior can be complemented by inferences drawn from other methods. Mixing methods has the advantage of compensating for the weaknesses in each method when used alone, lending greater confidence to the findings. Mixing methods might also allow for the effects of sermons on political behavior to be integrated into a deeper understanding of the effects of religion as a bundle on political behavior. We discuss these possibilities in the next section.
Mixing Methods in Order to Improve Inferences and Re-Bundle Religion
Although our approach has many advantages, there are inevitably aspects of religion rendered less visible when we employ this approach to disaggregating religion (see Table 1). Social networks, social identity, resources, rituals, skills, organization, and structures of authority are all also part of religious experience and practice but are rendered less visible when sermon content becomes the focus of study. Our approach also has implications for levels of analysis: we examined the influence of sermons on individual political behavior, rather than on social interaction among church members or large-scale collective action.
In addition, there are weaknesses in the above approach, as there are trade-offs in any research design. One weakness is the resource-intensity of the lab set-up, which typically means relatively smaller samples. In our separate study of sermons’ effects on political participation (McClendon & Riedl, 2019), we were able to increase the sample size to over 1,000 study participants, giving us more confidence in the inferences from that study; however, a large lab study is not always feasible. Moving from the lab to an online experimental platform, using mobile phone, social media or other internet technology, may be feasible in some contexts and avoid the overhead and staffing costs required for a study in a standing or mobile laboratory, thus allowing for much larger samples. 30 But experimental methods can also be effectively combined with observational methods to improve upon the weakness of either when used alone.
How might a mixed methods approach harness the advantages of disaggregating religion to focus on sermons while at the same time improving inferences and reintegrating sermons into the fabric of church life and organization in order to deepen our study of religious influence? Here, we provide an overview of additional methodologies that might improve and deepen insights from a study of sermons’ influence on individual political behavior, and also move to other levels of analysis (Table 4).
Mixed Methods to Improve the Research Design.
In our specific study, when we collected sermon data, we were also able to collect observations on the organizational resources, networks, social service provision, and geographic settings of each church, as well as socio-economic class markings, congregation size, language, and other features of the congregation. This kind of information, gleaned from interviews with church leaders as well as from participant observation in worship services, can shed light on the ways the metaphysical ideas conveyed in the sermons overlap and intersect with other aspects of religious practice and experience. For instance, we observed that the programming and social service provision of Pentecostal churches seemed to underscore (rather than contradict) the messages in their sermons: these churches often offered financial and psychological counseling services that complemented the message that individual and internal (psychological) transformation was the route to material and social betterment. We therefore expected that the effects of the sermons heard in the lab would, if anything, be amplified when heard in their institutional homes, where social programming served to underscore their ideas. In addition, our observations of mid-week services, which often re-iterated the content of the previous Sunday’s sermon, led us to hypothesize that sermons might work as primes in the lab. They did not seem to convey ideas that were chronically available to adherents but instead to convey ideas that needed to be primed and re-primed and that this was one goal mid-week and weekly services were designed to achieve.
These points call attention to the fact that sequencing may be important in a mixed methods research design. Beginning with descriptive and participant observation work in religious associations can not only improve interpretation of experimental results after the fact; it can also inform and improve the experimental design (Dunning, 2012; Paluck, 2010;). This participant observation work can also generate hypotheses about possible interaction effects between sermons and other institutional, material and social features of religion, which could inform follow up studies.
In order to re-bundle sermons with social aspects of religion, lab or survey experiments randomly assigning sermon exposure can also be combined with focus groups. The advantage of combining experiments with focus groups is that it allows identification of treatment effects (from the experiments) to be coupled with congregants’ expressions of and discussions about their own interpretation of and reaction to the sermons (from the focus groups). These interpretations and reactions can deepen the researchers’ understanding of the nature of the treatments used in the lab. If conducted with congregants of the same house of worship, focus groups can also allow for social interaction among congregants and allow researchers to observe whether and how the meaning of sermons is socially constructed through congregant discussion.
If it turned out that the interpretation of sermons changes substantially during social interaction and with communal discussion, then researchers could pursue an experimental design that differs from the one we pursued: they might randomly assign sermon content across listening groups, rather than across individuals. This design could look similar to the one used by Paluck and Green (2009) to study media influence in Rwanda. That is, sermon exposure would be assigned to groups rather than to individuals; the group would be allowed to listen to the treatment together (thus generating common knowledge about the sermon message) and to discuss it. Researchers could then measure political behavior at both the individual level and the group level. Researchers could survey individuals, or offer them individual opportunities to make political decisions, as we did in our studies. Or researchers could provide the groups with opportunities to act collectively toward some political end and use group-level measures as outcome variables.
To check that the effects identified in the lab are not an artifact of isolating sermon content from other aspects of religious experience, or a product of small sample sizes, lab experiments randomly assigning sermon exposure could also be combined with observational survey data, or other observational individual-level data. If sermon content correlates with denomination (as it did in our study), and evidence suggests that this correlation extends more broadly across geographic breadth and longitudinal duration of the survey (given expected variation in sermon content across time and space), and the observational survey data captures denominational affiliation and measures political behavior, then the experimental and observational data can be analyzed in tandem. The experimental data can provide insight into whether sermon content in isolation can affect political behavior. Researchers can then use denominational affiliation and frequency of practice in the observational survey data as a proxy for regular, real-world exposure to a particular type of sermon content as experienced in combination with other social and institutional aspect of worship. If the results from the lab experiment and the observational survey data are consistent with one another, then sermon content has a direct causal effect on political behavior and that effect is probably not specific to the lab environment or to the sermon being delivered in isolation from other aspects of religious practice. In addition, if the survey or administrative data contain the day of the week on which a political attitude was reported or the political behavior took place, researchers can look at how differences in behavior by denomination vary by temporal proximity to sermon exposure. For instance, if differences in political attitudes and behavior between Pentecostals and Catholics in a given setting are most stark early in the week (on Sundays, Mondays, Tuesdays) but converge later in the week as time passes after Sunday sermons, then it may be the case that sermons influence political attitudes and behavior only for a short period. In other words, this method can be used to further explore the duration of sermon influence or to validate experimental results about the duration of sermon influence but with a larger sample.
Just as focus groups can shift focus from the individual level of analysis to the group level of analysis, incorporating comparative historical analysis can also shift research focus from individual behavior to group action. Having identified sermons’ causal effect on individual behavior in the lab or through an online study, researchers could employ comparative case studies to investigate whether individual-level behavior aggregates to group-level behavior in ways that are consistent with the experimental results (Lieberman, 2015). Doing so allows researchers to explore the interaction of religious ideas with the institutional structures and power hierarchies in which they are composed and delivered. Indeed, researchers could choose cases that ensure variation in institutional structures and power relations within churches, and between church leaders and the state, in order to explore whether the institutional features and power position of religious associations moderate the effects of sermons on individual and group behavior. In some contexts, where sermons are not written down and archived, researchers will have difficulty recovering the past content of sermons, but in contexts with sermon repositories, researchers might also be able to use comparative historical analysis to examine whether religious leaders alter the content of sermons as the institutional features or power position of their church change.
Last, although our general approach identifies sermons’ influence on short-term and relatively isolated forms of political behavior, the approach could be combined with ethnographic observations of study participants over a period of time, after exposure to treatment. Such observations need not be limited to any one setting; researchers could learn more about members’ interactions in their communities, families, and workplaces, accompanying them as they engage in political activities or do not. Combining experimental research with ethnography in this way could broaden the researchers’ focus from short exposure to sermons and one-shot measures of political behavior to longer term political activities and selecting into further religious practice (and additional sermon exposure). The combination of these methods would allow researchers to re-bundle sermons with more sustained aspects of church life and religious practice and to observe the ways people do or do not bring the ideas communicated in sermons to their decisions about more sustained forms of social and political activity. As others (Paluck, 2010) have suggested, experimental and ethnographic methods could fruitfully be combined to deepen and contextualize the inferences we draw.
The above may not be an exhaustive litany of methods that could be combined with our general approach. But the methods we have listed could each be integrated with a descriptive and experimental study of the influence of sermons on political behavior, and doing so would both leverage the advantages of the disaggregation approach and also deepen and improve the inferences drawn from it.
Extending the Study of Sermons’ Consequences into Other Domains of Politics
Having outlined some of the practical ways researchers could use a focus on sermons to study religious influence on political attitudes and behavior, we outline some of the domains of politics in which that influence is likely to be important. Many domains of political action demand that people take a stance on how the world works and an individual’s role in it. Sermons are prevalent sources of guidance in shaping this stance, and thus are likely to influence political attitudes and behavior in these domains. Our studies have provided some evidence that exposure to sermon content can influence responses to inter-personal inequality and to opportunities to participate in political action, but there are many other political areas of sermon influence that are worthy of exploration and to which our general approach could be applied.
Climate Change
Deciding whether to support or oppose climate change-reducing policies requires a view about whether climate change is happening, about whether it is, or is likely to become, severe enough to warrant the costs involved to combat it, and about what the causes of climate change are. By providing mental maps about cause and effect, sermons can influence citizens’ attitudes and actions on this issue. Some religious leaders may discuss climate change policies explicitly in their communication with congregants (Djupe & Olson, 2010), but these explicit communications and political directives may be unusual (ibid) and sermons that discuss non-political questions (e.g., how much of a role does human activity play in shaping the physical world, is there a future in which apocalyptic change will occur) might still be expected to shape attitudes toward climate change policies.
Public Health Crises
As the COVID-19 global crisis has made ever more clear, when novel and poorly understood public health crises arise, political leaders and citizens have to take a stance on the severity of the crisis, its likely causes, and its likely trajectory, in order to devise a response. In responding, both political leaders and citizens may be influenced by sermons they have heard. Previous studies of religious responses to HIV/AIDS found variation across houses of worship in how clergy attributed blame for the disease and its spread, and in how they interpreted adherents’ responsibilities and appropriate response (Trinitapoli & Weinreb, 2012). In the wake of a public health crisis, researchers may find that sermons offer different interpretations of the divine’s role in the crisis, of individual versus collective responsibility for disease origins and spread, as well as of possible cures (or lack thereof). Even if sermons do not explicitly mention the disease or offer prescriptions for a response, sermons’ discussions of apocalyptic scenarios and of novel threats to community and personal security could influence listeners’ attitudes and behaviors in response to a public health crisis.
Racial and Ethnic Relations
Some sermons might preach explicitly about the superiority of certain groups over others, but even absent any explicit racial or ethnic ideology, sermons could influence support for policies designed to affect inter-group relations or disparities simply by offering mental maps about how the world operates. Debates about policies that would reduce between-group inequality, or ensure that groups are equally treated under the law, or compensate for racial injustices of the past, often hinge on how people attribute responsibility, for inequality and discrimination (Iyengar, 1989). Citizens who perceive unequal outcomes across groups as the product of aggregated individual choices and of personal responsibility may be less likely to support or push for sweeping institutional reform or new laws to reduce the privileges of advantaged groups. Citizens who, by contrast, perceive unequal outcomes across groups as the product of collective action and institutional responsibility may be more likely to turn to collective and institutional solutions. Our study found that sermons in Nairobi offered one or the other of these world views. Sermons elsewhere in the world are likely to do the same and might therefore be expected to influence people’s political attitudes and behaviors around racial and ethnic relations.
Democracy, Autocracy, and Citizenship
To the extent that sermons offer differing perspectives on the role of the individual in the public sphere, they may also shape people’s understanding of their roles as citizens, and their ability or obligation to exercise exit, voice, or loyalty (Hirschman, 1970). For instance, where sermons emphasize individual transformation, followers may be more attuned to discussions of rights, liberties, and freedoms, and less inclined toward collective action and civic responsibility. They may seek to exit a regime with which they disagree. Exercising voice as part of anti-regime mobilization may be more likely among those primed to understand the world as a product of both individual and collective responsibilities. Finally, sermons that emphasize divine intervention in the selection of leaders might cultivate regime loyalty among listeners. Regime preference could be shaped by sermon guidance as well. Sermons extolling absolute authority in leaders might incline listeners toward greater tolerance of increased executive power.
Conclusion
We hope that this paper can serve as a call for political scientists to pay more attention to the ideational content of religion, and to leverage the advantages of disaggregating religion into its component parts and of focusing on the influence of sermons in order to study religious influence in a tractable way. We have offered practical guidelines for doing so, ones that can be extended to a variety of outcomes as well as to a broader array of ideational content and cultural systems to be explored. We have also described a specific study that we conducted in Nairobi, Kenya in order to illustrate an example of the approach. This study examined the influence of contemporary Christian sermons on inter-personal distribution, and found that exposure to sermons emphasizing material change through individual acts of faith (most typical of Pentecostal churches in Nairobi) led to more generous behavior than exposure to sermons emphasizing collective and institutional responsibility for earthly problems (most typical of Mainline Protestant and Catholic churches in Nairobi). Beyond the particular results of the study is the research design that it employs—one that combines descriptive analysis of sermon content with a lab experiment, and the potential for extensions with a mixed methods, sequential approach. This approach can be used in many different contexts in order to identify religious influence in a host of domains of political behavior, including attitudes and behavior around climate change, public health crises, inter-group relations and discrimination, political participation, perceptions of regime and citizenship and more.
The advantages of following the approach outlined in this paper are several. The approach allows us to draw causal inferences that are much more difficult to obtain when religion is left unbundled. The approach also prevents us from treating religion as if it were one static, coherent “thing.” Contemporary Christian traditions in sub-Saharan Africa exhibit much variation in sermon content and help to illustrate the general point that exposure to different religious worldviews can influence citizens’ political behavior, but the content of these sermons, though not unique to contemporary African churches, is also not representative of religious teachings throughout time and space. Our approach takes the local-ness of religious content and experience seriously while also offering a way to disaggregate religion (and culture) that could be more widely applied. Finally, the approach provides a way to study religious influence that is not hindered by individuals’ choices over the degree and location of their religious practice. Our specific study used the lab environment to circumvent self-selection into religious ideas. The approach can also be adapted to study directly the drivers of self-selection into religious ideas.
The approach is not without its tradeoffs, but we have also outlined several ways mixed methods strategies could re-integrate what is made less visible by the approach. For instance, we have described how focus groups, comparative historical analysis, analyses of survey and administrative data, and ethnographic methods might be combined with a lab or online experiment to improve the inferences from the experiment, to integrate individual- and group/institutional-levels of analysis, and to re-bundle sermons with other aspects of religious experience, practice, and organization. These steps are practical and could be applied in a wide array of contexts and domains of political behavior to deepen our understanding of religious influence on politics.
Research Data
mcclendonriedl_inequalitystudy_demographicdata – Research Data for Using Sermons to Study Religions’ Influence on Political Behavior
Research Data, mcclendonriedl_inequalitystudy_demographicdata for Using Sermons to Study Religions’ Influence on Political Behavior by Gwyneth McClendon and Rachel Beatty Riedl in Comparative Political Studies
Research Data
mcclendonriedl_usingsermons_replication – Research Data for Using Sermons to Study Religions’ Influence on Political Behavior
Research Data, mcclendonriedl_usingsermons_replication for Using Sermons to Study Religions’ Influence on Political Behavior by Gwyneth McClendon and Rachel Beatty Riedl in Comparative Political Studies
Research Data
_mcclendonriedl_inequalitystudy_ztreedata – Research Data for Using Sermons to Study Religions’ Influence on Political Behavior
Research Data, _mcclendonriedl_inequalitystudy_ztreedata for Using Sermons to Study Religions’ Influence on Political Behavior by Gwyneth McClendon and Rachel Beatty Riedl in Comparative Political Studies
Footnotes
Appendix
Acknowledgements
We are grateful to Jeremy Menchik and Tim Longman and participants in a conference hosted by Boston University’s Institute on Culture, Religion and World Affairs (CURA) entitled “Finding Religion" (April 5-6, 2019), as well as to anonymous reviewers, for comments. We are most grateful to Mwongela Kamencu, Julie Santella, Alice Xu, Eddine Bouyahi, Malitt Ishmael, Nelson Ngige, Brenda Ochieng, May Koko, Esther Kerubo, Amalia Bersin, Jack Furness, Kennedy Mmasi, Elvira Salgado, Adelina Pak, Michelle Ki, and Mert Salur for invaluable research assistance. We thank the Busara Center for Behavioral Economics in Nairobi, and the British Institute for Eastern Africa, for help facilitating the research. The Eric Mindich Foundations of Human Behavior Initiative at Harvard University and Northwestern University’s Equality, Development and Globalization Studies provided generous funding for the research described in this paper.
Declaration of Conflicting Interests
The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The authors received no financial support for the research, authorship, and/or publication of this article.
Notes
Author Biographies
References
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.
