Incidental news exposure via social media and political participation: Evidence of reciprocal effects

Abstract

The aim of this study is to investigate the causal direction of the relationship between incidental news exposure via social media and political participation. Unlike prior studies, which have relied on cross-sectional data to examine this link, we used two panel data sets to better identify causal relationships. Specifically, we evaluate two unidirectional models (i.e. mobilization and reinforcement) and a reciprocal causal model using both cross-lagged and autoregressive path models. The findings reveal a more complex relationship than most previous studies have suggested. The relationship between incidental news exposure via social media and political participation appears to be reciprocal, with incidental news exposure and political participation indirectly influencing each other through social media use for political purposes. Furthermore, while the relationship between incidental news exposure and political participation is reciprocal, the participation-to-incidental news exposure path exerted a stronger effect than the reverse path in both studies.

Keywords

Incidental news exposure mobilization model comparison political participation reciprocal causality reinforcement

A few today would challenge the basic proposition that social media use and political engagement are linked. After all, many users encounter a virtual torrent of political messaging—including exhortations to engage various different forms of political participation posted by political figures, friends, former high-school classmates, co-workers and others—on a regular basis, and even most who do not are likely (sometimes painfully) aware of the possibility. Such posts may be simultaneously viewed both as invitations or catalysts to participation (in the case of those who see them), as well as acts of political participation in themselves (in the case of those who post them as a form of interaction with others in their network). Turning to more systematic evidence, a meta-analysis of three-dozen research studies examining the relationship between social media use and participation found strong evidence of a basic positive relationship between these two variables, with more than 80% of coefficients in these studies estimated as positive (Boulianne, 2015).

Expanding on this premise, a growing body of communications research has turned to deeper questions about the precise nature of the linkage between social media use and participation. In particular, studies in this area have found footing in the observation that for most users, exposure to information about politics and public affairs in social media is not their primary intention when accessing their social media feeds. In other words, most users encounter such information incidentally, as a byproduct of using social media for other purposes, such as entertainment (Fletcher and Nielsen, 2017; Matsa and Mitchell, 2014). This raises the optimistic possibility that incidental exposure to politics through social media may stimulate users to engage in more active forms of political engagement, such as more directed uses of social media for political purposes and a variety of other forms of political participation. The possibility of such a dynamic is particularly enticing given its similarity to the kind of mobilization provided for many by broadcast media in the low-choice era before the rise of cable and the Internet, which has been shown to be associated with broader participation rates overall, as well as a relatively less polarized electorate—democratic outcomes with high normative value (Prior, 2007). Thus a number of recent studies have specifically focused on possible relationships between incidental exposure to political information via social media and political participation (Heiss and Matthes, 2019; Kim et al., 2013; Lee, 2018; Valeriani and Vaccari, 2016).

There are, however, at least two other plausible scenarios, besides a mobilization effect, that could be consistent with the observed linkage between social media use and political participation. First, causality could simply run the other way, with political activity serving as an antecedent rather than a consequence of incidental exposure. We label this the “reinforcement” scenario. Alternatively, the relationship could be reciprocally causal, including both the influence of incidental exposure to information about politics and public affairs on political participation, as well as the influence of participation on incidental exposure. Either of these two possibilities would offer markedly less sanguine—but certainly no less important—implications regarding the overall influence of social media on core democratic processes. Unfortunately, with a small number of notable exceptions (Heiss and Matthes, 2019; Yamamoto and Morey, 2019), most research in this area suffers from the common limitations of cross-sectional data and is thus unable to support solid inferences about which of the three possible scenarios outlined here receives the most empirical support.

To help advance research and theory development in this area, this study draws on multiple two-wave panel surveys, conducted in two different US election cycles, in order to directly compare competing models reflecting mobilization, reinforcement, or reciprocal causality dynamics at the center of the relationship between incidental exposure to news via social media and political participation. We develop these models in the next section, drawing on existing empirical findings and theoretical arguments to further clarify assumptions about the nature of these relationships, including likely mediation of both key underlying causal influences through purposeful political uses of social media. Including the most recent presidential and midterm or congressional election cycles in the United States, which have arguably been the most social media intensive to date (e.g. Center for Information and Research on Civic Learning and Engagement, 2018; Lee, 2020; Roose, 2018), we believe our data provide an ideal context for exploring causal relationships involving incidental exposure to politics through social media. Using both cross-lagged and autoregressive path models to test the extent to which each set of assumptions finds support in our two panel survey data sets, we find that a model based on reciprocal causality best fits the data, though within that model the causal path from participation to incidental exposure is stronger than the reverse path.

Literature review

Social media use and political participation

Social science scholars have long tried to discover what factors and conditions lead to political engagement (e.g. McLeod et al., 1999; Rolfe, 2012). Communication scholars in particular have focused on the role of media use in fostering political engagement by examining how specific media environments generate new opportunities for individuals to engage with political information and news, thus furthering political engagement (e.g. Prior, 2007; Shah et al., 2005). For instance, scholars have identified the positive role played by newspaper reading (e.g. Eveland and Scheufele, 2000), television news use (e.g. Shah et al., 2005), and Internet news (e.g. Kenski and Stroud, 2006) in political participation. The core argument explaining the positive relationship between news and political participation concerns the fact that news serves as both a source of political information and a prompt for political discussions with others, which consequently promotes politically participatory behaviors (Boulianne, 2011; Shah et al., 2005).

Building on these findings, a number of studies have examined the relationship between political participation and the use of social media to access news (e.g. Dimitrova et al., 2014; Gil de Zúñiga et al., 2014). The majority of studies point to a positive relationship between the two, although there have been some exceptions (see meta-analysis by Boulianne, 2015). In summarizing her meta-analysis of 36 comparable studies, Boulianne (2015) has argued that research by and large supports the proposition that social media use has a positive impact on engagement.

Despite the popularity of research concerning social media use for political purposes (or for news-related purposes) and political participation, much less attention has been paid to how incidental news exposure via social media is related to political participation. Active or purposeful news consumption is not the only way that people consume news by means of social media (Bode, 2016; Fletcher and Nielsen, 2017; ^{Oeldorf-Hirsch, 2018)}. Indeed, most social media users encounter news incidentally rather than searching for it purposefully (Matsa and Mitchell, 2014). However, despite the prevalence of incidental news exposure relative to purposeful news seeking via social media, the majority of research efforts in this regard have been directed toward exploring how purposeful news consumption via social media specifically affects individuals’ political activities. Relatively few studies have explored whether incidental news exposure through social media is related to political participation (for notable exceptions, see Heiss and Matthes, 2019; Valeriani and Vaccari, 2016). Yet, although the research in this area is notably sparse, it does form a foundation upon which to build.

Kim et al. (2013) found that incidental news exposure online was positively related to a citizen’s political participation, and this positive linkage was found to be stronger for those who less frequently used the Internet for entertainment. Valeriani and Vaccari (2016) also identified a positive correlation between incidental news exposure via social media and online political participation across a variety of settings, including Germany, Italy, and the United Kingdom.

And yet, not all studies in this area have found evidence for a positive effect of incidental news exposure on political participation. For example, drawing on cross-sectional data from the candlelight protests in South Korea, Lee (2018) explored how purposeful and incidental news exposure via Facebook individually influence participation in protests; results suggested that incidental news exposure did not lead to further political action. Heiss and Matthes (2019), using panel data, also found no evidence of a stimulation effect of incidental exposure on high-effort digital participation; in fact, their analysis suggests a negative effect of incidental exposure on high-effort digital participation, especially for those with low levels of political interest.

These conflicting findings could stem from the methodological limitations of the relevant studies. In fact, most of these studies share the limitation that they were unable to document a causal relationship between incidental news exposure and political participation due to a reliance on cross-sectional survey data. In other words, cross-sectional data do not allow for the determination of whether incidental exposure to news via social media mobilizes political action (i.e. stimulation), or whether those already active in politics are more likely to be incidentally exposed to news via social media (i.e. reinforcement), or whether they influence each other (i.e. reciprocal causality). However, our panel data allow us to compare alternative models and to determine which model best explains the relationship between incidental news exposure and political participation.

Theoretical models

The mobilization model: incidental exposure stimulates political participation via political social media use

Theoretically, incidental news exposure via social media may serve to mobilize political action. Political communication scholars have long suggested that accessing news is often the first step toward political engagement (e.g. Newman, 1991). However, because news consumption behaviors often demand substantial effort, time, and commitment (often referred to simply as “costs”), the costs associated with news consumption often serve as an initial barrier to people’s political engagement—even if one has some interest in politics. Yet, the incidental nature of news exposure via social media sites may lower this barrier by providing social media users with opportunities to engage in a broad range of low-cost political activities (e.g. clicking “Like” or making comments), which may serve as a gateway to further offline political participation (e.g. Gil de Zúñiga et al., 2014).

This idea that softer or easier forms of political participation online may provide a pathway toward offline participation is often referred to as the “spillover effect” (e.g. Bode et al., 2014; Cantijoch et al., 2016). Cantijoch et al. (2016) suggest that the pathways from online activities to offline political engagement often work in this fashion, such that “individuals take a gradual ‘step-up’ the ladder of participation, migrating from low intensity activities to marginally more active versions” (p. 38). This idea of a spillover effect has largely been supported in the social media context (e.g. Bode et al., 2014; Gil de Zúñiga et al., 2014).

Another theoretical explanation for how incidental exposure can stimulate further offline participation, consistent with the spillover effect just described, has been articulated by Knoll et al. (2020). Taking a psychological approach, their work suggests that even incidental exposure to political content can stimulate further political action by insidiously embedding key messages in a viewer’s subconscious. This incidental exposure to political content can in turn influence one’s cognitions (e.g. goals), regardless of whether said content was consciously or systematically processed. These implicitly processed messages may be activated later if one is again exposed to similar political information, which may induce more intentional participatory behavior (Knoll et al., 2020).

Previous discussions have implied potential pathways of incidental exposure stimulating political participation; however, this stimulation pathway does not necessarily suggest that news exposure per se would directly lead to political participation (Bimber, 2001). Most of the prevailing literature suggests that the effect of news consumption on political participation tends to be positive, but such effects are mostly indirect. According to the communication mediation model (e.g. Shah et al., 2005), informational media use indirectly promotes political participation by stimulating interpersonal political communication, such as participating in political discussions or politically expressive activities, whether online or offline.

In this sense, social media is an ideal platform where individuals can first express their political views (Bode et al., 2014; Gil de Zúñiga et al., 2014). It can include posting links to a news story, posting one’s own thoughts about politics/social issues, encouraging others to take action on political/social issues, and even “liking” posts on political issues (Lee and Xenos, 2019). Such expressive activities on social media may serve as important pathways to offline political participation, because expression involves a self-reflective process of consuming information (Pingree, 2007). In other words, by engaging in politically expressive activities, individuals move from observer to participant (Gil de Zúñiga et al., 2014). This suggests that incidental exposure to news via social media likely stimulates offline political participation, and that this relationship is mediated by political uses of social media.

The reinforcement model: political participation stimulates incidental exposure via political social media use

Although prior studies have suggested that incidental exposure to news via social media can mobilize further political action (e.g. Kim et al., 2013; Valeriani and Vaccari, 2016), it is also relatively easy to argue the reverse—that those who are already active in politics are more likely to engage in expressive political behaviors and be incidentally exposed to news within social media. The basic premise for this argument can be traced back to the “normalization thesis” proposed by Margolis and Resnick (2000). The core of the normalization thesis holds that digital media platforms simply help people to do what they were already doing prior to the digital media era. In other words, the classic normalization thesis suggests that those who are already interested in, or who already actively participate in, politics will use new forms of media to more easily engage in politics. Consistent with this, it is also widely assumed that those who are not already interested in politics will likely craft their media environment in such a way that they can easily opt out of political matters (Prior, 2007).

Social media sites, as a distinct form of digital media, enable users to actively control their news exposure and to selectively filter information so that they can shape the nature of the content they wish to consume (e.g. Pariser, 2011). In this sense, those who are already active in politics (i.e. those who regularly participate in political matters) can craft their social media environment in such a way that political content is more readily visible, which means that they will be more likely to be exposed to political content, incidentally or otherwise. If this is the case, then the positive association between incidental exposure to news and political participation may simply indicate a phenomenon whereby those who are active in politics deliberately curate their social media environment so that they are regularly exposed to political content. This scenario does not necessarily imply that those who are uninterested in politics will not be incidentally exposed to news/political content on social media at all. As data from the Pew Research Center indicate, incidental news exposure should occur regardless of political interest at least to some extent (Matsa and Mitchell, 2014). Yet, rather than such exposure mobilizing further politically participatory action, those who already regularly participate in politics may deliberately shape their social media environment so as to be more easily (incidentally) exposed to news and/or political content.

These discussions imply the theoretical possibility that those who participate in politics more are, as a result, more likely to engage in political uses of social media and, in so doing, be incidentally exposed to political content. In other words, it is reasonable to assume that those who already actively participate in politics likely deliberately shape their social media environment in the first place, and in doing so make themselves more easily (incidentally) exposed to political content. This suggests that politically participatory behaviors may prompt individuals to use social media for political purposes, which would in turn increase the likelihood that those individuals are incidentally exposed to news via social media.

A reciprocal causal model

The aforementioned theoretical arguments propose unidirectional causal models around the relationship between incidental news exposure and political participation. However, there is no reason to assume that such relationships are mutually exclusive. Indeed, media scholars have demonstrated that relationships between media use and its effects are often reciprocal (Boulianne, 2011; Slater, 2007). This reciprocal relationship between the two variables is especially plausible when the nature of social media is considered. In social media, while it is true that a considerable amount of news content is encountered through incidental exposure, in practice, such exposure may not be completely incidental, since the news content may be shown due to the workings of an algorithm based on the consumer’s past behaviors (Thorson and Wells, 2016). In other words, news encountered through incidental exposure may lead one to engage in further political activities, and the increased participation in turn may render individuals more likely to use social media for political purposes, which could further increase the chances of their being incidentally exposed to news. This feedback path suggests the possibility of a reciprocal relationship.

Thus, due to the need to address the complex cause-and-effect nature of the relationship between incidental news exposure and (the resultant) political participation, it is important to consider a reciprocal causal model in this study.

One advantage of using panel data is that such data offer the possibility to test hypotheses concerning reciprocal causality and to compare a reciprocal causal model with the unidirectional causal models (for details, see Eveland et al., 2005). Based on these theoretical arguments, we seek to assess which of these three models is best supported by the data (RQ1).

Method

Sample

We relied on two survey datasets to explore our research question. In Study 1, survey data were collected during the 2016 US presidential election by YouGov. The sampling frame was constructed through stratified sampling that resembles the US population in terms of gender, age, race, education, party identification, ideology, and political interest. The first wave data were collected between 20 September and 27 September 2016 (N = 937). The second wave data were collected between 18 November and 28 November 2016 (N = 750; 80.04% retention rate). All variables used in this article were measured at both Waves 1 and 2 (except demographics).

In Study 2, survey data were collected during the 2018 US midterm election by Survey Sampling International (SSI). The sample closely mirrored census data on key dimensions, such as gender and age. The first wave data were collected between 26 September and 30 September 2018 (N = 1555). The second wave data were collected between 7 November and 13 November 2018 (N = 824; 53% retention rate). All variables except demographics were measured at both waves.

Measures

Incidental news exposure via social media

To measure incidental news exposure via social media, we follow the same approach applied in numerous prior investigations concerning this issue (e.g. Kim et al., 2013; Tewksbury et al., 2001; Valeriani and Vaccari, 2016). Specifically, respondents were asked the following question:

When you use social networks/social media platforms (e.g. Facebook, Twitter, YouTube, etc.),¹ how often do you come across news and information on current events, public issues, or politics when you may have been going online for a purpose other than to get the news?

In Study 1, response options ranged from never (1) to always/very often (4) (Study 1 W¹: M = 2.61, SD = 0.90; Study 1 W²: M = 2.73, SD = .86). In Study 2, to better capture variability in participant responses, we used a 10-point scale (1 = never, 10 = all the time) (Study 2 W¹: M = 5.60, SD = 3.21; Study 2 W²: M = 4.86, SD = 3.31).

Political participation

According to Verba and Nie’s (1972) classic definition, political participation refers to “those activities by private citizens that . . . aim at influencing the government, either by affecting the choice of government personnel or by affecting the choices made by government personnel” (p. 2). Following the majority of the literature in which online and offline participation have been found to be (related but) distinctive concepts (e.g. Hoffman, 2012; Vissers and Stolle, 2014), we have focused on the offline dimension of political participation. In Study 1, respondents were asked to report whether they have the following: (1) worked or volunteered in a community project, (2) worked or volunteered for a nonprofit group such a hobby club, environmental group, or ethnic association, (3) raised money for a charity or ran/walked/biked for charity, (4) worked or volunteered for political groups or candidates, (5) contacted an elected leader in any way, (6) written to a newspaper or other news organization, (7) worked on or volunteered with an election campaign, (8) tried to persuade others how to vote in an election, (9) wore or displayed a badge, sticker, or sign related to a political or social cause, (10) deliberately bought (or avoided buying) a product for political, ethical, or environmental reasons, (11) circulated or signed a petition, (12) attended a demonstration or rally, and (13) discussed politics with friends or family. Respondents were asked to choose either “yes” (coded as 1) or “no” (coded as 0) for each of these activities. Then, these items were summed to form an additive index ranging from 0 to 13 (Study 1 W¹: Cronbach’s α = .83, M = 2.95, SD = 2.85; Study 1 W²: Cronbach’s α = .83, M = 2.82, SD = 2.77).

Although the measurement used for Study 1—that is, summing the binary variables to obtain a composite score—represents a standard approach to measuring political participation (e.g. Eveland and Scheufele, 2000; Xenos et al., 2014), the binary response format does not allow respondents to report their intensity or frequency of usage. To better capture variability in participant responses, in Study 2, respondents were asked on a 10-point scale (1 = never, 10 = all the time) to indicate how often during the past 12 months they had participated in any of the following activities: “signed a petition,” “boycotted or bought certain products for political, ethical or environmental reasons,” “participated in any political rallies,” “attended a public meeting dealing with political or social issues,” “posted a political sign, banner, button, or bumper sticker,” “taken part in concerts or a fundraising event with a political cause,” and “contacted a politician or public official.” Responses to each statement were added into a single index (seven items averaged scale; Study 2 W¹ Cronbach’s α = .95, M = 3.16, SD = 2.66; Study 2 W² Cronbach’s α = .94, M = 2.31, SD = 2.17).

Political social media use

In Study 1, respondents were asked to report whether they have ever used Facebook or other social networking tools for any of the following purposes in the past month: (1) posting links to political stories or articles for others to read, (2) posting your own thoughts or comments on politics or social issues, (3) encouraging other people to take action on a political or social issue that is important to you, (4) encouraging other people to vote, (5) re-posting content related to politics or social issues that was originally posted by someone else, and (6) “Liking” or promoting material related to political or social issues that others have posted. Respondents were asked to choose either “yes” (coded as 1) or “no” (coded as 0) for each of these activities. The six items were summed to form an additive index ranging from 0 to 6 (Study 1 W¹: Cronbach’s α = .86, M = 2.36, SD = 2.20; Study 1 W²: Cronbach’s α = .87, M = 2.58, SD = 2.28).

To address the same measurement issue noted in the political participation variable in Study 1, in Study 2, we adopted methods from Gil de Zúñiga et al. (2014), asking respondents to answer questions on a 10-point scale (1 = never, 10 = all the time) to indicate how often during the past 12 months they had engaged in any of the following activities: (1) posting personal experiences related to politics or campaigning; (2) friending or following a political advocate or politician; (3) posting or sharing thoughts about politics; (4) posting or sharing photos, videos, or audio files about politics; (5) forwarding someone else’s political commentary to other people; and (6) reading posts about politics. The six items were summed to form an additive index (Study 2 W¹: Cronbach’s α = .96, M = 3.99, SD = 2.96; Study 2 W²: Cronbach’s α = .96, M = 3.23, SD = 2.76).

Control variables

We included a variety of control variables including political interest, news attention, and a series of demographic variables including age, education, gender, race, and household income. Political interest was measured by asking respondents how interested they were in politics on a 5-point scale ranging from “not at all interested” (1) to “extremely interested” (5), (Study 1 W¹: M = 3.18, SD = 1.01; Study 2 W¹: M = 3.19, SD = 2.33). To measure attention to news, respondents were asked on a 4-point scale (1 = not at all, 4 = A great deal) to report how much attention they pay to (1) political news, including news about the presidential election and other elections happening this fall, (2) news about their community, (3) national news, and (4) news about international affairs. Responses to these items were averaged to create a composite score (Study 1 W¹: Cronbach’s α = .83, M = 3.03, SD = 0.75; Study 1 W²: Cronbach’s α = .85, M = 3.07, SD = 0.75). To better capture variability in participant responses, in Study 2, respondents were asked on a 10-point scale (1 = no attention at all, 10 = very close attention) how much attention they pay to news about politics and public affairs from the following media types: newspaper, TV, radio, Internet (except social media), and social media.

Finally, demographic variables include age (Study 1: M = 46.81, SD = 16.57; Study 2: M = 46.81, SD = 16.57), gender (Study 1: 48.4% females; Study 2: 50.5% females), race (Study 1: 66.0% White; Study 2: 65.0% White), education (operationalized as highest level of education that they have completed; Study 1: Mdn = some college; Study 2: 2-year college), and total annual household income (Study 1: Mdn = US$40,000–US$49,999; Study 2: Mdn = US$60,000–US$69,999).

Analytical procedure

To test the causal direction between incidental news exposure and political participation, a path model that relates the variables concerning incidental news exposure to the political participation index was estimated with the “lavaan” package in R. One important advantage of panel data when running path analysis is that it gives us substantial leverage in inferring causation when compared to cross-sectional data. Because the cross-sectional data provide only correlational, rather than causal, information about the relationship, cross-sectional data do not allow researchers to compare model fits across three alternative models—namely, the mobilization model, the reinforcement model, and the reciprocal causal model. Even though two-wave panel data are not the ideal type of dataset to test the reciprocal causal model—indeed, multiple waves of panel data, which provide a more robust and longitudinal dataset, would better serve this purpose—it nonetheless allows us to examine how the variables change over time, including possible reciprocal relationships.

With this advantage of panel data, we developed six models that examine the relationship between incidental news exposure and political participation. These models are two unidirectional models (i.e. the mobilization and reinforcement model) and a reciprocal causal model, with each specified in cross-lagged and autoregressive forms. In all the models, gender, age, education, income, political interest, and news attention were included as exogenous variables that affect all Wave 2 outcome variables. Each of these methods (i.e. cross-lagged and autoregressive path analysis) has significant advantages (see Shah et al., 2005). For example, the cross-lagged path analysis enables researchers to assess how an independent variable in Wave 1 can be related to the mediator and the dependent variable in Wave 2, offering a clearer picture of how these variables are related over time. For autoregressive path models, we assess how Wave 2 measures are related, while each Wave 2 measure is regressed on its corresponding Wave 1 measure. In all models, we include a structural path between Wave 1 and Wave 2 incidental news exposure, between Wave 1 and Wave 2 political social media use, and between Wave 1 political participation and Wave 2 political participation. These paths control for prior levels of the outcome, making other paths to these outcome variables interpretable as predicting “change” in the outcome variable—and enabling researchers to explain unexplained variance in Wave 2 variables while accounting for variable stability over time.

To compare the relative fit of the competing models, we used several fit indices: (1) the Akaike information criterion (AIC), (2) the Bayesian information criterion (BIC), (3) the root mean squared error of approximation (RMSEA), and (4) the ratio of the chi-square statistic to the degrees of freedom for the model (χ²/df). AIC and BIC impose penalties on models that include more structural paths and are therefore less parsimonious. Yet, AIC and BIC are sensitive to sample size. Thus, we also used RMSEA and χ²/df as measures of relative fit. Lower values of AIC, BIC, RMSEA, and χ²/df indicate better model fit. Following Eveland et al.’s (2005) approach, we did not include widely used fit indicators when assessing the global fit of the structural/path model (such as comparative fit index (CFI), Tucker–Lewis index (TLI), and goodness of fit index (GFI)), because these indices tend to be used as measures of absolute fit, and are not necessarily well suited for determining optimal fit across competing models.²

Results

We compared three different theoretical models to identify the best fitting model, controlling for news attention, political interest, and demographic variables (RQ1). To ensure methodological rigor, we used both cross-lagged path analyses and autoregressive path analyses, and also relied on two different datasets (Study 1—2016 US presidential election data and Study 2—2018 US midterm election data). The results were consistent across different types of path analyses across two studies. In both Study 1 and Study 2, the data are most consistent with a reciprocal causal model; that is, both incidental news exposure and political participation indirectly influence each other through social media use for political purposes. In addition, while the relationship between incidental news exposure and political participation appears to be reciprocal, the participation-to-incidental news exposure path exerted a stronger effect than the reverse path in both studies. More specific findings for each study are presented in the following sections.

Study 1

In terms of the global fit of the model, the chi-square tests along with other fit indices suggest that the reciprocal causality model fits the data better than both models of unidirectional causality models. This pattern was consistent across both the cross-lagged and autoregressive path models.

The results for the three different cross-lagged path models (i.e. the mobilization model, reinforcement model, and reciprocal causal model) are presented in Figures 1 to 3. Fit indicators suggest that the reciprocal causal model showed a substantially better fit to the data compared to both the mobilization model and the reinforcement model (see Table 1). The autoregressive path models showed the same pattern as the cross-lagged path models; that is, the reciprocal causal model fits the data better than other two unidirectional causal models (see Table 1). The results for the three different autoregressive path models (i.e. the mobilization model, reinforcement model, and reciprocal causal model) are presented in Figures 4 to 6.

Figure 1.

Cross-lagged mobilization model (Study 1).

Figure 2.

Cross-lagged reinforcement model (Study 1).

Figure 3.

Cross-lagged reciprocal causal model (Study 1).

Table 1.

Model comparisons (Study 1).

Model comparisons (cross-lagged)	AIC	BIC	RMSEA	χ²/df
1. Incidental → Participation lagged	5482.718	5619.570	0.141	11.618
2. Participation → Incidental lagged	5449.264	5593.520	0.088	5.105
3. Reciprocal lagged	5445.912	5591.318	0.059	2.833
Best fitting model	Model 3	Model 3	Model 3	Model 3
Model comparisons (autoregressive)
1. Incidental → Participation autoregressive	5479.945	5603.968	0.107	7.100
2. Participation → Incidental autoregressive	5451.609	5575.632	0.062	3.052
3. Reciprocal autoregressive	5447.409	5579.985	0.055	2.633
Best fitting model	Model 3	Model 2³	Model 3	Model 3

AIC: Akaike information criterion; BIC: Bayesian information criterion; RMSEA: root mean squared error of approximation.

Note. For, AIC, BIC, RMSEA, and χ²/df, lower values indicate better fit.

Figure 4.

Autoregressive mobilization model (Study 1).

Figure 5.

Autoregressive reinforcement model (Study 1).

Figure 6.

Autoregressive reciprocal causal model (Study 1).

In addition, as with the cross-lagged path model, we also found that the relationship between incidental news exposure and political participation appears to be reciprocal. However, the participation-to-incidental news exposure path (cross-lagged path model: point estimate = 0.01, p = .001; autoregressive path model: point estimate = 0.01, p = .002) represents a more consistent significant relationship than the reverse path (cross-lagged path model: point estimate = 0.03, p > .05; autoregressive path model: point estimate = −0.02, p > .05) in both the cross-lagged and autoregressive path models.

Study 2

The patterns found in Study 2 are very similar to the patterns found in Study 1. As with Study 1, the reciprocal causality model fits the data better than both of the unidirectional causality models—a pattern that was consistent across both the cross-lagged and autoregressive path models (see Table 2). The results for all the different path models (i.e. the mobilization model, reinforcement model, and reciprocal causal model) are presented in Figures 7 to 12. In addition, consistent with the findings of Study 1, we found that the participation-to-incidental news exposure path (cross-lagged path model: point estimate = 0.03, p = .001; autoregressive model: point estimate = 0.07, p < .001) was more pronounced than the reverse path in our other two models (cross-lagged path model: point estimate = 0.01, p = .008; autoregressive path model: point estimate = 0.01, p > .05).

Table 2.

Model comparisons (Study 2).

Model comparisons (cross-lagged)	AIC	BIC	RMSEA	χ²/df
1. Incidental → Participation lagged	9345.521	9495.508	0.143	17.352
2. Participation → Incidental lagged	9311.441	9461.428	0.099	8.832
3. Reciprocal lagged	9287.527	9446.889	0.058	3.707
Best fitting model	Model 3	Model 3	Model 3	Model 3
Model comparisons (autoregressive)
1. Incidental → Participation autoregressive	9372.622	9508.548	0.130	14.644
2. Participation → Incidental autoregressive	9327.236	9463.162	0.094	8.160
3. Reciprocal autoregressive	9290.855	9436.155	0.054	3.348
Best fitting model	Model 3	Model 3	Model 3	Model 3

AIC: Akaike information criterion; BIC: Bayesian information criterion; RMSEA: root mean squared error of approximation.

Figure 7.

Cross-lagged mobilization model (Study 2).

Figure 8.

Cross-lagged reinforcement model (Study 2).

Figure 9.

Cross-lagged reciprocal causal model (Study 2).

Figure 10.

Autoregressive mobilization model (Study 2).

Figure 11.

Autoregressive reinforcement model (Study 2).

Figure 12.

Autoregressive reciprocal causal model (Study 2).

Discussion

The overall aim of this study was to investigate the causal direction of the relationship between incidental news exposure via social media and political participation, which has rarely been attempted before, most likely because most data do not allow researchers to untangle causal relationships. Unlike prior studies, which relied on cross-sectional data to examine this link, our study used two different panel data sets to better identify and test competing models specifying both unidirectional and reciprocal causal relationships. Specifically, we employed a model comparison approach to identify the best fitting model across two unidirectional models (i.e. mobilization and reinforcement) and a reciprocal causal model. In addition, we used both cross-lagged and autoregressive path analyses across the two different datasets (i.e. Study 1 and Study 2) not only to draw better causal inferences about the dynamic relationships among our variables of interest but also to test the generalizability of our findings in multiple election contexts.

Our results reveal that reciprocal causal models fit the data better than models that specify unidirectional causality (in either direction) between incidental news exposure and political participation (what we refer to here as the mobilization and reinforcement models). This is consistent with the proposition that the relationship is, in fact, reciprocal, including (1) the influence of incidental news exposure on political participation (through political social media use) and (2) the influence of political participation on incidental news exposure (also through political social media use). Yet, it is important to note here that these results do not suggest that incidental news exposure and political participation influence each other to an equal degree. In fact, a closer look at the findings suggests that the participation-to-exposure path consistently exerts a stronger effect than the reverse path. These results are in line with a number of other recent studies, suggesting that while it is true that a considerable amount of news content is encountered through incidental exposure, such exposure is not completely incidental. Rather, algorithms likely influence content exposure based on the consumer’s past behaviors (e.g. Thorson and Wells, 2016).

These findings have important normative implications for democratic societies. Political participation is widely considered as a cornerstone of healthy democratic systems (Verba et al., 1995). Yet, despite the normative importance of political engagement, not everyone actually participates in politics. Unfortunately, those with more resources tend to participate in politics to a greater degree. Furthermore, the resources (or factors) traditionally known to influence political participation, including education, income, political interest, and political socialization (Verba et al., 1995), are relatively stable, and hence, cannot be improved overnight. By contrast, different forms of social media use may be promoted relatively easily (if they need to be promoted at all), which has undoubtedly drawn many researchers to explorations of the potential mobilizing effects of this form of media use. One implication of our findings is that while incidental news exposure can serve as a pathway to further political action, such mobilizing potential should not be overstated, as those who are already active in politics are more likely to be exposed to such news content in the first place. This finding is also consistent with previous research by Lee (2019), which found that social media effectively mobilized citizens to engage in protests, but that this mobilizing effect was more pronounced for those with higher level of political interest.

Indeed, our results show that potential mobilizing effects are just a part of the whole picture, and a smaller one at that. Our findings show that although incidental news exposure via social media does appear to encourage political participation to a certain degree, the opposite causal relationship—that is, those who regularly participate in politics are more likely to be incidentally exposed to news via social media—is stronger. We believe it is likely that this more complicated picture of reciprocal causality has contributed to a lack of clarity in the research literature on these questions to the extent that they have typically been explored in ways that obscure these relationships. In other words, because of the greater relative strength of the pathway from participation to exposure, it is only by essentially controlling for this dynamic that we can develop a clear sense of what we would consider a nontrivial mobilizing effect.

Despite the strengths of our study, there are some methodological and analytical limitations. The first such limitation relates to how we measured the key variable of this study, incidental news exposure via social media. Although it is a common practice for capturing this variable (e.g. Tewksbury et al., 2001; Valeriani and Vaccari, 2016), using a self-reported measure of incidental news exposure via social media has several notable limitations. First, scholars have previously noted that people’s self-reports regarding their exposure to news are often inaccurate, producing either over- or under-estimates of exposure (González-Bailón and Xenos, 2020; Prior, 2009). These patterns raise important questions concerning the overall clarity of measurements that rely on individuals’ ability to accurately remember and then differentiate the news content they encounter purposefully or incidentally. Second, we used a “global” measure of incidental news exposure via social media platforms, rather than asking the degree of incidental exposure separately across different social media platforms. However, different social media platforms have different characteristics (Treem and Leonardi, 2013), which may yield different effects on one’s news exposure patterns. Third, this single item measure does not fully capture the complex nature of incidental news exposure. Matthes et al. (2020) suggest that operationalizing incidental news in this manner may be problematic, insofar as it does not distinguish “passive scanning of information deemed as irrelevant” (“first level incidental exposure”) versus “the intentional processing of incidentally encountered information” (“second level incidental exposure”). In other words, respondents sometimes encounter news and process that information more thoroughly (e.g. if the content is appraised as relevant to one’s goal), while others may encounter news and merely glimpse at it. It is important to note that these limitations mainly serve to weaken the “signal” provided by such self-reports, relative to “noise” introduced by the above mentioned factors, however, which would serve to make it more difficult to identify patterns such as those reported here. Nonetheless, to address these limitations, future research can use (1) more multifaceted survey items to tap into incidental news exposure, and/or (2) additional methods to reduce the biases attendant with self-reporting, such as the “forced mobile experience sampling” pioneered by Karnowski et al. (2017) to capture in situ exposure.

In addition, another limitation lies in the two-wave structure of both of our datasets. To be sure, the use of an analysis based on two-wave panel data meets minimum requirements for us to estimate causal relationships between incidental news exposure and political participation in a more robust manner than the use of cross-sectional data would. However, at the same time it still has limitations in terms of the clarity offered regarding the causal processes when compared to multiple-wave panel data or experimental research. We believe the findings reported here, however, provide important evidence that such studies would be worthwhile.

These limitations notwithstanding, this study makes a number of significant contributions to the literature. Although a growing number of studies have recently explored the relationship between incidental news exposure and political participation, the majority of these studies have relied on cross-sectional data to examine such relationships (e.g. Kim et al., 2013; Lee, 2018; Valeriani and Vaccari, 2016). Thus, they were only able to demonstrate correlation with the observed relationships. However, our panel research design offers us substantial leverage with regard to inferring causation, because it enables us to assess how the variables of our interest are related over time. In addition, our findings are even more robust, since we adopted (1) two different types of path analyses (i.e. cross-lagged and autoregressive path analyses), of which each approach has its own unique strengths in drawing causal inferences; and (2) two datasets from two different types of US elections, the use of which enhances the generalizability of our findings.

Drawing on these methodological advantages, our findings suggest that the relationship between incidental news exposure and political participation appears to be reciprocal, meaning that the two constructs appear to influence each other indirectly through political social media use. This finding reveals a more complex relationship than previously suggested, and has implications for the democratic process. Notably, the dynamic of mobilization via incidental exposure receives the least empirical support, relative to models emphasizing reinforcement and reciprocal causality. In particular, this finding suggests that with respect to political participation, there is reason to temper excitement over the potential for incidental exposure to political information via social media to have broad-based mobilizing effects similar to those associated with the broadcast era. Despite the allure of phenomena that go against the grain of well-known dynamics within an overall abundance of media choice, future research in this area should carefully attend to the complexities suggested here regarding the interplay of purposeful and incidental exposure with processes of democratic participation.

Footnotes

Funding

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

ORCID iD

Sangwon Lee

Notes

Author biographies

Sangwon Lee is an assistant professor in the Department of Communication Studies at New Mexico State University. His research focuses on how social media use changes the ways in which people learn about, engage with, and form opinions on political and social issues. His work has appeared in publications such as Mass Communication and Society, Computers in Human Behavior, and Political Behavior among others.

Michael A. Xenos is a professor in the Department of Life Sciences Communication, and Professor of Communication Science in the Department of Communication Arts, at the University of Wisconsin-Madison. His research and teaching interests are centered on the effects of new media on political participation, public deliberation, and public engagement with controversial science issues.

References

Bimber

(2001) Information and political engagement in America: the search for effects of information technology at the individual level. Political Research Quarterly 54(1): 53–67.

Bode

(2016) Political news in the news feed: learning politics from social media. Mass Communication and Society 19(1): 24–48.

Bode

Vraga

Borah

, et al. (2014) A new space for political behavior: political social networking and its democratic consequences. Journal of Computer-Mediated Communication 19(3): 414–429.

Boulianne

(2011) Stimulating or reinforcing political interest: using panel data to examine reciprocal effects between news media and political interest. Political Communication 28(2): 147–162.

Boulianne

(2015) Social media use and participation: a meta-analysis of current research. Information, Communication & Society 18(5): 524–538.

Cantijoch

Cutts

Gibson

(2016) Moving slowly up the ladder of political engagement: a “Spill-over” model of Internet participation. The British Journal of Politics and International Relations 18(1): 26–48.

Center for Information and Research on Civic Learning and Engagement (2018) Five takeaways on social media and the youth vote in 2018. Available at: https://circle.tufts.edu/latest-research/five-takeaways-social-media-and-youth-vote-2018

Dimitrova

Shehata

Strömbäck

, et al. (2014) The effects of digital media on political knowledge and participation in election campaigns. Communication Research 41(1): 95–118.

Eveland

Hayes

Shah

, et al. (2005) Understanding the relationship between communication and political knowledge: a model comparison approach using panel data. Political Communication 22(4): 423–446.

10.

Eveland

Scheufele

(2000) Connecting news media use with gaps in knowledge and participation. Political Communication 17(3): 215–237.

11.

Fletcher

Nielsen

(2017) Are people incidentally exposed to news on social media? A comparative analysis. New Media & Society 20(7): 2450–2468.

12.

Gil de Zúñiga

Molyneux

Zheng

(2014) Social media, political expression, and political participation: panel analysis of lagged and concurrent relationships. Journal of Communication 64(4): 612–634.

13.

González-Bailón

Xenos

(2020) Surveys underestimate online news exposure: a comparison of self-reported and observational data in nine countries. Available at: http://ssrn.com/abstract=3522774

14.

Heiss

Matthes

(2019) Does incidental exposure on social media equalize or reinforce participatory gaps? Evidence from a panel study. New Media & Society 21(11–12): 2463–2482.

15.

Hoffman

(2012) Participation or communication? An explication of political activity in the Internet age. Journal of Information Technology & Politics 9(3): 217–233.

16.

Bentler

(1999) Cutoff criteria for fit indexes in covariance structure analysis: conventional criteria versus new alternatives. Structural Equation Modeling: A Multidisciplinary Journal 6(1): 1–55.

17.

Kaplan

Haenlein

(2010) Users of the world, unite! The challenges and opportunities of social media. Business Horizons 53(1): 59–68.

18.

Karnowski

Kümpel

Leonhard

, et al. (2017) From incidental news exposure to news engagement. How perceptions of the news post and news usage patterns influence engagement with news articles encountered on Facebook. Computers in Human Behavior 76: 42–50.

19.

Kenski

Stroud

(2006) Connections between Internet use and political efficacy, knowledge, and participation. Journal of Broadcasting & Electronic Media 50(2): 173–192.

20.

Kim

Chen

Gil de Zúñiga

(2013) Stumbling upon news on the Internet: effects of incidental news exposure and relative entertainment use on political engagement. Computers in Human Behavior 29(6): 2607–2614.

21.

Knoll

Matthes

Heiss

(2020) The social media political participation model: a goal systems theory perspective. Convergence: The International Journal of Research into New Media Technologies 26(1): 135–156.

22.

Lee

(2018) The role of social media in protest participation: the case of candlelight vigils in South Korea. International Journal of Communication 12: 1523–1540.

23.

Lee

(2019) Connecting social media use with gaps in knowledge and participation in a protest context: the case of candle light vigil in South Korea. Asian Journal of Communication 29: 111–127.

24.

Lee

(2020) Probing the mechanisms through which social media erodes political knowledge: the role of the news-finds-me perception. Mass Communication and Society. Epub ahead of print 15 September. DOI: 10.1080/15205436.2020.1821381.

25.

Lee

Xenos

(2019) Social distraction? Social media use and political knowledge in two U.S. Presidential elections. Computers in Human Behavior 90: 18–25.

26.

Margolis

Resnick

(2000) Politics as Usual: The Cyberspace “Revolution.” New York: SAGE.

27.

Matsa

Mitchell

(2014) 8 key takeaways about social media and news. Pew Research Center’s Journalism Project. Available at: https://www.journalism.org/2014/03/26/8-key-takeaways-about-social-media-and-news/

28.

Matthes

Nanz

Stubenvoll

, et al. (2020) Processing news on social media. The political incidental news exposure model (PINE). Journalism. Epub ahead of print 16 May. DOI: 10.1177/1464884920915371.

29.

McLeod

Scheufele

Moy

(1999) Community, communication, and participation: the role of mass media and interpersonal discussion in local political participation. Political Communication 16(3): 315–336.

30.

Newman

(1991) The Future of the Mass Audience. Cambridge: Cambridge University Press.

31.

Oeldorf-Hirsch

(2018) The role of engagement in learning from active and incidental news exposure on social media. Mass Communication and Society 21(2): 225–247.

32.

Pariser

(2011) The Filter Bubble: How the New Personalized Web Is Changing What We Read and How We Think. New York: Penguin Books.

33.

Pingree

(2007) How messages affect their senders: a more general model of message effects and implications for deliberation. Communication Theory 17(4): 439–461.

34.

Prior

(2007) Post-Broadcast Democracy: How Media Choice Increases Inequality in Political Involvement and Polarizes Elections. New York: Cambridge University Press.

35.

Prior

(2009) The immensely inflated news audience: assessing bias in self-reported news exposure. Public Opinion Quarterly 73(1): 130–143.

36.

Rolfe

(2012) Voter Turnout: A Social Theory of Political Participation. New York: Cambridge University Press.

37.

Roose

(2018) We asked for examples of election misinformation. You delivered. The New York Times, 4 November. Available at: https://oklahoman.com/article/feed/6945449/we-asked-for-examples-of-election-misinformation-you-delivered

38.

Shah

Cho

Eveland

, et al. (2005) Information and expression in a digital age. Communication Research 32(5): 531–565.

39.

Slater

(2007) Reinforcing spirals: the mutual influence of media selectivity and media effects and their impact on individual behavior and social identity. Communication Theory 17(3): 281–303.

40.

Tewksbury

Weaver

Maddex

(2001) Accidentally informed: incidental news exposure on the world wide web. Journalism & Mass Communication Quarterly 78(3): 533–554.

41.

Thorson

Wells

(2016) Curated flows: a framework for mapping media exposure in the digital age. Communication Theory 26(3): 309–328.

42.

Treem

Leonardi

(2013) Social media use in organizations: exploring the affordances of visibility, editability, persistence, and association. Annals of the International Communication Association 36: 143–189.

43.

Valeriani

Vaccari

(2016) Accidental exposure to politics on social media as online participation equalizer in Germany, Italy, and the United Kingdom. New Media & Society 18(9): 1857–1874.

44.

Verba

Nie

(1972) Participation in America: Political Democracy and Social Equality. New York: Harper & Row.

45.

Verba

Schlozman

Brady

(1995) Voice and Equality. Cambridge, MA: Harvard University Press.

46.

Vissers

Stolle

(2014) The Internet and new modes of political participation: online versus offline participation. Information, Communication & Society 17(8): 937–955.

47.

Xenos

Vromen

Loader

(2014) The great equalizer? Patterns of social media use and youth political engagement in three advanced democracies. Information, Communication & Society 17(2): 151–167.

48.

Yamamoto

Morey

(2019) Incidental news exposure on social media: a campaign communication mediation approach. Social Media + Society. Epub ahead of print 16 May. DOI: 10.1177/2056305119843619.