https://orcid.org/0000-0003-2598-2453
Abstract
Interest groups increasingly communicate with the public, yet we know little about how effective they are in shaping opinions. Since interest groups differ from other public communicators, we propose a theory of interest group persuasion. Interest groups typically have a low public profile, and so most people are unlikely to have strong attitudes regarding them. Source-related predispositions, such as credibility assessments, are therefore less relevant in moderating effects of persuasive appeals by interest groups than those of high-profile communicators. We test this argument in multiple large-scale studies. A parallel survey and field experiment (N = 4,659) establishes the persuasive potential of low-profile interest groups in both controlled and realistic settings. An observational study (N = 700) shows that substantial portions of the public are unable to assess interest group credibility. A survey experiment (N = 8,245) demonstrates that credibility assessments moderate the impact of party but not interest group communication.
Governments, partisan actors, and international organizations routinely use communication campaigns to shape public opinion. We increasingly see a new type of actor, interest groups, running these campaigns. They typically have a low public profile but often have considerable backroom influence (Dür & Mateo, 2016). In the context of such “outside tactics” (Kollman, 1998), these actors try to persuade the public as a way to generate additional pressure on decision makers to enact their preferred policies. For communication research, the question is whether established theories allow us to evaluate the persuasive potential of these new communicators correctly.
In this article, we argue that interest groups differ from established communicators in an important way. Unlike communication by established communicators such as political parties, interest groups typically have low public profiles. This raises the possibility that two central tenets in general theories of persuasion do not hold for interest group communication: namely, that source-related predispositions toward these types of actors are available in people’s belief systems and that these predispositions moderate communication effects (Pornpitakpan, 2004; Self, 2009; Wilson & Sherrell, 1993).
We develop this argument in several steps. First, we discuss the contemporary role of interest groups as communicators with the public. Specifically, we identify the strength of their public profile as a crucial characteristic in which interest groups typically differ from sources that have conducted public communication campaigns in the past. To give one example, in 2019, the National Rifle Association (NRA) was mentioned 1,575 times in three leading U.S. newspapers, while the Republican and the Democratic parties were mentioned 39,617 and 45,201 times, respectively.1 Thus, the public profile of even the NRA—one of the most influential and visible interest groups in American politics (Lacombe, 2019)—pales in comparison to that of the two predominant political parties. Consequently, the public can be expected to be much less familiar with the typical interest group than with actors continuously at the center of public discourse.
We argue that this difference has important implications for the effects of persuasive attempts by these actors, especially with respect to the role predispositions play in people’s reactions to their interventions. Indeed, the findings of three empirical studies reported below show that the persuasive effects of communicative interventions by low-profile sources are not moderated by prior attitudes about the source, whereas they do so with respect to high-profile sources such as political parties.
In the first study, we tested whether a communicative intervention by a German business group with a low public profile could increase public support for an advocated economic policy. We tested this in a parallel survey and field experiment with three survey waves (N = 4,659). The relatively low-key intervention increased the salience of the policy addressed, persuaded recipients of the interest group’s position, and increased the number of accessible supporting considerations. These effects were modest in size and decayed over the course of a week. It is notable that we found no evidence that recipients’ prior attitudes toward the interest group moderated the persuasive appeal of the intervention.
In the second preregistered study (N = 700), we examined the availability of credibility assessments for 10 environmental groups and 6 political parties in Germany. We found that more respondents were unable to express credibility assessments for the environmental groups than for the political parties. We also found that respondents who did express credibility assessments held them with less certainty. Furthermore, perceived credibility of environmental groups corresponded only weakly and not systematically with respondents’ party affiliations.
We used a high-powered, preregistered survey experiment (N = 8,245) to test directly whether a moderation of persuasive effects by credibility assessments is weaker in interest group communication than political party communication. As expected, we found that source credibility assessments did not moderate the persuasive appeal of an intervention by a low-profile environmental group in favor of a program to increase energy efficiency. Similarly, for an environmental group with a comparatively high public profile, there was at best weak evidence for a moderating role of source credibility ratings. In contrast, prior attitudes did moderate persuasive effects when we attributed the message to a political party. As theorized, those who considered the party as a credible source on environmental issues changed their attitudes in line with the party’s message, but those who found the party not credible moved away from the advocated position, rendering the intervention counterproductive among these recipients.
We add to communication research by demonstrating the importance of examining closely the dynamics of persuasion by actors not traditionally in the focus of communication scholars. Our findings indicate considerable differences in the degree to which people hold and fall back on explicit credibility assessments for interest groups and political parties. While interest groups and political parties both seem able to shape public opinion, people appear to rely on prior source-related attitudes when evaluating the messages sent by political parties but not interest groups. By examining the effects of interest group communication and their dynamics, we thus gain a deeper understanding of the conditions for predispositions to matter in persuasion. In light of these findings, persuasion research needs to reflect that the influence of source-related predispositions on the effects of communicative interventions varies for different types of actors.
Interest groups as communicators with the public
Interest groups and public communication campaigns
Interest groups are organizations that try to influence governments and the public in line with the interests of their membership, which may be comprised of individuals or of other organizations. Interest groups are distinct from political parties in that they do not seek offices by running candidates themselves but aim to influence candidates and office holders (Berry & Wilcox, 2018, p. 5f.). While certain interest groups have formed exclusive alliances with specific political parties in the past, such ties have been weakening in many Western democracies in recent years, and new types of interest groups have emerged that remain independent of traditional political divisions. As a result, firm and exclusive ties between interest groups and parties have become rare (see Allern, 2010).
Beyond immediate links between interest groups and political parties, there may be deeper links between political ideology and specific groups (Noel, 2013). This is clearly the case for interest groups emerging along deep political cleavages, such as the labor movement and its links to the political left. But even these strong historical links have weakened over time (Kitschelt, 1994; Piazza, 2001). Clear ideological sorting is even less pronounced for interest groups that have emerged only recently. For example, the environmental movement can be linked to the political right as well as the political left, depending on whether environmental groups present environmentalism as anti-technology and emphasize conservationist concerns or frame it as anti-capitalist and anti-establishment (Dunlap et al., 2001; Gray, 1993; Nawrotzki, 2012; Neumayer, 2004; Pilbeam, 2003). Interest groups are thus best thought of as distinct political actors who, in pursuit of specific issues, can have varying political allegiances and who do not necessarily remain fixed in ideological space.
In face of weakened political alliances, interest groups increasingly employ “outside tactics,” directly addressing the public (Kollman, 1998). While many powerful interest groups used to concentrate on engaging politicians and decision makers directly (Dür & Mateo, 2016), they increasingly engage in extended public communication campaigns. Recent examples include campaigns by MARQUES (The Association of European Trade Mark Owners) in support of the “Anti-Counterfeiting Trade Agreement” in 2012 (Dür & Mateo, 2014) and by the Bundesverband der Deutschen Industrie in support of the Transatlantic Trade and Investment Partnership (TTIP) in 2015–2016. Thus, there emerges a new set of actors with considerable resources that engage in public communication campaigns.
It is one thing for interest groups attempting to persuade the public and another to succeed at doing so (Arceneaux & Kolodny, 2009a, 2009b; Broockman & Kalla, 2016; Dewan, Humphreys, & Rubenson, 2014; Dür, 2019; Rogers & Middleton, 2015). While the literature on interest groups discusses the use of outside tactics in issue campaigns, it focuses on portraying the repertoire of tactics employed and does not theorize or test the effects of communication campaigns by interest groups (e.g. Binderkrantz, 2005; 2008; Binderkrantz & Krøyer, 2012; Dür & Mateo, 2016; Hanegraaff et al., 2016; Kollman, 1998). The few studies that have tested communicative interventions by interest groups indicate that they can act as effective communicators and persuade recipients (see Broockman & Kalla, 2016; Dür, 2019).
Persuasive effects come in many forms. Following Miller (1980), communicative interventions can mean downright changing recipients’ minds on a topic (response-changing effects) or just shaping the way they think about a topic, such as by raising its salience or emphasizing the relevance of specific arguments in a debate (response-shaping effects). But any persuasive intervention has to compete for attention in the noisy and competitive information environment of public discourse (Druckman & Lupia, 2016). We must, therefore, test whether communicative interventions by interest groups have persuasive effects not only under optimal conditions of exposure and reception but also under realistic, noisy conditions that are far from optimal from a communicator’s perspective. In general, though, it seems plausible that interest groups, just as any other skilled communicator, have the potential to persuade people:
H1: Communicative interventions by interest groups persuade recipients of the position for which they argue, doing so by (a) increasing a topic’s salience, (b) shifting attitudes toward the position, and/or (c) increasing the availability of arguments raised.
By drawing on previous insights from communication and persuasion research, we can go beyond this baseline hypothesis, however, and propose a more specific argument about the impact of interest group communication on public attitudes and how it might differ from that of other communicators in the political arena, such as political parties.
Low public profiles. The most important characteristic that sets interest groups apart from other political actors is their typically low public profile, that is, their presence in public discourse. The main political parties in a given country, for example, are actors with high public profiles. They participate regularly in public discourse and take sides on the issues of the day. Most interest groups, such as business or employers’ groups, have contrastingly much lower public profiles. Of course, the public profiles of interest groups vary, and there are those that routinely participate in public discourse (such as Greenpeace and the NRA, well-known public-facing interest groups with long-established political positions). But even the public profiles of relatively prominent interest groups are dwarfed in comparison with those of the main political parties and their representatives.2
For example, German national media mentioned the interest group Greenpeace 2,917 times in 2019, as compared with 85,375 and 80,615 mentions of the German parties Sozialdemokratische Partei Deutschlands (SPD) and Christlich Demokratische Union Deutschlands (CDU), respectively. This difference persists once we examine the smaller—but nevertheless politically relevant—German parties such as Freie Demokratische Partei (FDP) or Bündnis 90/Die Grünen. These parties had 26,741 and 23,242 mentions, respectively, in 2019. Thus, according to this measure, the public profile of the German interest group perhaps best known for its aggressive outside tactics is lower than that of these small political parties by a factor of 10.
This difference becomes even more apparent if we focus on the public profile of business groups. For instance, the Arbeitgeberverband Gesamtmetall, a highly influential German employers’ group (Schroeder, 2017), was mentioned in German media 96 times during the same period.3
These differences in public profile are no accident; they are rooted in the different social functions of these actors. Political parties are “full-service” political communicators: they communicate continually with the public and take sides on issues along the entire political spectrum. As a consequence, they are themselves the focus of political coverage. In addition, people encounter political parties and their representatives on various occasions in their daily lives, such as through interactions as a part of local politics. Interest groups, however, communicate selectively, focusing only on issues that are important to them and appear only in the media if those issues are featured. Furthermore, since most people do not follow politics closely, only members of respective issue publics will encounter a given interest group directly or regularly. The public profile of interest groups is, hence, structurally determined and thus unlikely to change fundamentally.
Because the public profile of interest groups is typically low, it is not possible to apply directly one of the core insights communication research has produced when studying interest groups as public communicators. Specifically, we must question the role of predispositions that moderate the effects of communication campaigns. Can people be expected to link an interest group with a political ideology or party reliably, or do they have prior attitudes toward the group that they rely on in reaction to communicative interventions?
Do interest groups provide meaningful source cues? A key insight of communication and persuasion research is that recipients, in evaluating a message, consider not only the content of the message itself—such as the quality of argument or soundness of evidence—but also consider contextual cues (Chaiken & Ledgerwood, 2012; Chen & Chaiken, 1999). Contextual cues, by allowing recipients to assess the credibility of a message, add to or detract from its persuasiveness.
In political communication, party cues are particularly important (Bullock, 2011, 2020; Nicholson, 2011; Popkin, 1991). Studies show the strong moderating influence of party affiliation on the assessment of policy options and candidate evaluations: Party messages influence supporters but are ineffective at changing the minds of people “across the aisle”—or may even drive them further away (Arceneaux, 2008; Arceneaux & Kolodny, 2009a; Bartels, 2002; Berinsky, 2009; Bullock, 2011; Kirkland & Coppock, 2018; Li & Wagner, 2020). Similarly, political partisanship has been shown to influence the perception of content when presented by a politically aligned or opposing information source, such as partisan news media or partisan social media users (Gunther et al., 2017; Lee et al., 2018; Reid, 2012). The importance of partisanship cannot be generalized to all political communication, however, as it hinges on certain conditions that are not necessarily met. Crucially, in settings where research has found partisanship effects, the political affiliations of the sources were clearly identifiable and unambiguous. Where communicators are less clearly aligned along ideological or partisan divides, partisanship is bound to be less relevant (Bullock, 2020, p. 138f).
In the case of interest group communication, it is doubtful that these conditions are met. As described above, interest groups in contemporary Western democracies are typically not linked exclusively or unambiguously with specific ideologies or political parties, and their public profile is, on average, quite low. While in some Western democracies, such as the United States, extreme levels of political polarization lead to constellations in which some interest groups are linked unambiguously with specific parties (Noel, 2013), such ties cannot be taken for granted. If we want to understand the role of source-related predispositions in interest group communication, we must turn to specific source assessments and not general political predispositions, which may not be related to a given interest group in a given context.
Credibility assessment of the information source is the type of source evaluation that seems to have received the most research attention (Hovland, Jannis, & Kelley, 1953; McGuire, 1969, 1985; O’Keefe, 2016; Pornpitakpan, 2004; Self, 2009; Wilson & Sherrell, 1993). A source’s credibility consists of judgments by people regarding whether they should believe its communications (O’Keefe, 2016, p. 188f.). To make such judgments, people are expected to use prior attitudes toward the source as well as situational perceptions to assess its credibility. The positive evaluation that a source is credible should increase the persuasiveness of a message. In turn, a negative credibility assessment of the same message should weaken its persuasive appeal (Chaiken & Ledgerwood, 2012).
This line of reasoning is supported by a long series of empirical studies. Sources recipients deem trustworthy and competent are more persuasive than those deemed untrustworthy and incompetent (Gass & Seiter, 2018; McCroskey & Young, 1981). Arguments presented by sources of high social repute, such as scientific publications or government agencies, tend to be more persuasive than those by sources deemed to be ill informed or predominantly self-interested, such as undergraduates or commercial companies (Eastin, 2001; Gunther & Liebhart, 2006). This makes the credibility assessment, the most likely candidate for moderating the persuasive appeal of messages (Iyengar & Valentino, 2000; Lupia, 2002). It is an explicit assessment of the source in question and is useful even in the absence of additional knowledge about the sender—such as its links to political ideologies or parties.
But how convincing is it that credibility assessments matter for interest group communication? This question has not been examined empirically; most studies addressing source credibility focus on sources that are either well-known themselves or belong to groups about which survey respondents hold pre-formed opinions (such as the medical profession and scientific institutions). Studies on the role of credibility assessment in cases where people have little information about the source in question are missing from the literature (see, e.g., the list of studies in Pornpitakpan, 2004 or Wilson & Sherrell, 1993).
Crucially, there are theoretical considerations that cast doubt on the notion that credibility assessments play an important role in interest group communication. Key among them is that the preconditions for the heuristic in question to become relevant are not met (Chaiken & Ledgerwood, 2012). At the most basic level, if recipients are unfamiliar with an interest group, they may simply not have a credibility assessment to draw on. In such a situation, the particular heuristic of processing the persuasive message is not available and cannot add to or detract from a message’s persuasiveness. This is exactly what we expect to be the case when people encounter political messages from interest groups, due to their low public profile. And even if recipients are able to make a credibility assessment of an interest group on the fly,4 it is unlikely to have the same impact on the processing of the message as crystallized assessments.
This argument can be recast and generalized by drawing on the concept of attitude certainty, that is, the “subjective sense of conviction or validity about one’s attitude” (Gross, Holtz, & Miller, 1995, p. 215): The more certain recipients are about their credibility assessments of the source, the more relevant those assessments will be in evaluating the message. If credibility is assessed to be uncertain, this particular heuristic will play less of a role in adding or detracting from the message’s persuasive appeal.
Finally, it seems unlikely that individuals can rely on partisan affiliations when constructing credibility assessments of an interest group. As noted above, this is because most interest groups do not exhibit clear links to a single political party or political ideology.
In sum, neither sender-specific credibility assessments nor party affiliation—or similar general political predispositions—are likely to matter much for processing persuasive messages of interest groups. These considerations lead to the following hypotheses:
H2a: Respondents are less likely to express credibility assessments for interest groups than for political parties.
H2b: Respondents express less certainty about credibility assessments of interest groups than about credibility assessments of political parties.
H2c: Credibility assessments of interest groups are not linked systematically to party identifications.
H3: Credibility assessments will influence the persuasive effects of party communication stronger than interest group communication.
While these hypotheses treat interest groups as a homogenous category, we have pointed out above that some of these groups have higher public profiles than others. The proposed differences between parties and interest groups should be particularly pronounced when comparing political parties with the large number of interest groups with low public profiles. It is less clear what to expect when comparing political parties to the handful of relatively prominent interest groups with high public profiles. Two scenarios seem possible. One is that the public profile of interest groups such as Greenpeace or the NRA could be sufficiently high for people to form meaningful credibility assessments (i.e. assessments held with certainty), which might in turn serve as available heuristics. Accordingly, in the case of interest groups with high public profiles, credibility assessments might add to or detract from persuasive appeals in the same way as they do for political parties. But given the vast difference between the public profiles of even the most visible interest groups and the least visible major parties, as shown above, another outcome is just as likely: it may be that even interest groups with a comparatively high public profile do not achieve the degree of visibility necessary for the average citizen to develop crystallized credibility assessments. If true, we should expect credibility assessments of interest groups with high public profiles to have little or no impact on the processing of their messages. Consequently, we consider it an open question whether the patterns expected in H2a, H2b, H2c, and H3 also hold for the small set of interest groups with relatively high public profiles.
Study 1
Data and methods
We used a combined field and survey experiment design to examine the effects of a communicative intervention by an interest group with a comparatively low public profile in a competitive and noisy communication environment. In order to provide a realistic context for our experiment, we collaborated with the German business group Arbeitgeberverband Gesamtmetall (Federation of German Employers’ Associations in the Metal and Electrical Engineering Industries), which represents the interests of German companies in those two sectors. Given the importance of these industries to the national economy, Gesamtmetall can be considered one of the most politically influential employers’ groups in Germany (Schroeder, 2017). However, its influence does not translate into prominence in public discourse.5
In collaboration with the group, we developed an information mailing reflecting its actual policy position. The mailing presented three arguments in support of the Transatlantic Trade and Investment Partnership (Jungherr, Mader, Schoen, & Wuttke, 2018). In addition, the letter included various statements in favor of the ratification of the agreement. It was sent in an envelope clearly identifying the sender and its support for TTIP (see Appendix S2 for treatments).
The experiment was conducted in three waves between 21 April and 22 June 2016 with members of a large German online panel Payback, and was executed on our behalf by the survey company Infratest dimap. To ensure correspondence between the respondents and the German online population, we used quotas blocking access to the survey for respondents with specific demographic characteristics once their share had reached levels comparable with their share of the German online population. The realized sample size was N = 7,826. We randomly assigned participants into three groups. The field experiment group (N = 2,606) received the information letter by mail, the survey experiment group (N = 2,609) was exposed to the mailing embedded in the questionnaire, and the control group (N = 2,611) received no treatment (see Appendices S3–S5 for details).
Wave 1 of the survey (N = 7,826), which ran before treatment administration (21 April–11 May 2016), measured attitudes toward TTIP, demographic characteristics, and predispositions expected to moderate the effects of the planned treatment. Nineteen days after the close of wave 1, the information mailing was sent to the field experiment group (30 May). In wave 2 (N = 6,171), which began six days later (6 June–13 June), respondents in the survey experiment group were shown the same letter on screen that respondents in the field experiment group had received by mail. To measure effect decay, we conducted a final survey wave one week later (N = 4,659) (20 June–22 June). Waves 2 and 3 included items to measure TTIP attitudes, considerations about the agreement, and subjective attitude certainty (see Appendix S6 for the questionnaire; see Appendix S7 for indicators). For the analyses presented in the article, we rely only on answers by respondents who participated in all three waves and reported an attitude toward TTIP in wave 1 (N = 4,659).6
The comparison of the treatment’s effects under field and survey conditions allows complementary insights into its effects on respondents. Under field conditions, we can assess the persuasive strength of interest group communication over time in competitive and noisy communication environments. The survey condition promises a detailed view of the strength of persuasive effects under close-to-ideal measurement conditions, given controlled exposure and measurement directly after exposure. This allows the identification of small or heterogeneous effect patterns. Because treatment compliance and time of exposure are known, it is also much easier to calculate a realistic decay rate of persuasive effects. We thus capitalize on the strengths of both experimental modes while offsetting their weaknesses (Harrison & List, 2004), whereas previous studies focused on the comparability and relative advantages of each experimental mode (Barabas & Jerit, 2010; Findley, Laney, Nielson, & Sharman, 2017; Jerit, Barabas, & Clifford, 2013).
Results
To address the first hypothesis, we estimated the main treatment effects in a series of Ordinary Least Squares (OLS) regressions. For each of the three panel survey waves, we regressed the three dependent variables (TTIP SALIENCE, TTIP ATTITUDE, and NUMBER OF PRO-TTIP ARGUMENTS) on dummy variables indicating a respondent’s treatment assignment and, to increase efficiency, a pretreatment measure of the dependent variable (Gerber & Green, 2012).7Figures 1, 2, and 3 show the results. Specifically, we report differences in predicted means of the dependent variables between the control, field, and survey groups, respectively. Coefficients in wave 1 show that the experiment groups did not differ significantly before the treatments were administered.8
Treatment effects on TTIP salience. Notes: The figure reports differences between the control group and the treatment groups; bars indicate 95% confidence intervals. The results are based on linear regressions of TTIP salience on dummy variables indicating group membership. The range of the salience measure is 0–1. To increase the efficiency of the treatment estimate, a pretreatment measure of the outcome variable from wave 1 is included as covariate in the regression.
Treatment effects on TTIP attitude. Notes: The figure reports differences between the control group and the treatment groups; bars indicate 95% confidence intervals. The results are based on linear regressions of TTIP attitude on dummy variables indicating group membership. The range of the attitude measure is 0–1. To increase the efficiency of the treatment estimate, a pretreatment measure of the outcome variable from wave 1 is included as covariate in the regression.
Treatment effects on number of pro-TTIP considerations. Notes: The figure reports differences between the control group and the treatment groups; bars indicate 95% confidence intervals. The results are based on linear regressions of the number of pro-TTIP arguments on dummy variables indicating group membership. One unit on the considerations measure represents one pro-TTIP consideration mentioned. To increase the efficiency of the treatment estimate, a pretreatment measure of the outcome variable from wave 1 is included as covariate in the regression.
In comparing treatment effects in the field and the survey conditions, note that these experiments followed somewhat different logics. In the survey condition, treatment compliance was ensured by design because respondents were exposed to the treatment as part of the survey itself.9 In the field condition, we look at people we intended to treat with the mailing. In the unobtrusive conditions of a field experiment, however, it is likely for a variety of reasons that some participants assigned to receive the treatment did not read the letter. Comparing the effects in the field and the survey conditions thus means comparing treatment effects in two different groups: one in which all subjects saw the information letter at least briefly (survey) and a group in which only a fraction did (field). In addition, the effect measured in wave 2 has likely already decayed in the field condition due to a time gap of unknown length between exposure and measurement, whereas measurement immediately followed treatment exposure in the survey condition. We should, therefore, examine the survey condition to assess the treatment’s effect strength under ideal exposure and measurement conditions. To gain an impression of how the treatment fares under realistic exposure conditions, we should turn to the field condition while considering the caveats described.
According to Hypothesis 1, the treatment should increase the salience of TTIP, support for TTIP, and the number of pro-TTIP arguments of which respondents are aware. As Figures 1, 2, and 3 report, the findings from both experimental modes support this, but to different degrees. Figure 1 shows that the treatment had noticeable and somewhat durable effects on respondents in the survey experiment condition. Respondents who had read the letter embedded in the survey considered the TTIP issue to be slightly more important than respondents in the control group (treatment effect on the dependent variable scaled from 0–1: 0.049 [95% CI 0.035–0.062], Cohen’s d: 0.23). The effect is still visible at about half its original size when attitudes were measured a week later in wave 3 (0.029 [0.016–0.042]).10 In the field condition, there is no comparable effect dynamic—although TTIP salience is higher in the field group than in the control group both in wave 2 (treatment effect: 0.013 [0.000–0.0268], Cohen’s d: 0.03) and wave 3 (0.019 [0.005–0.031]), the numerical effects are small and somewhat unreliable.
The results for TTIP support, the second dependent variable, are more consistent across experimental modes. Figure 2 shows that in wave 2, respondents assigned to read the letter showed significantly higher support for TTIP than respondents in the control group. As expected, the effect is again substantially larger in the survey (treatment effect: 0.119 [0.104–0.133], Cohen’s d: 0.43) than in the field condition (0.031 [0.017–0.045], Cohen’s d: 0.12). Both effects decay but remain statistically significant in wave 3 (effect in survey: 0.050 [0.037–0.064], effect in field: 0.015 [0.001–0.029]).
Recent developments in experimental research have raised awareness in the field that coefficients and p-values do not necessarily indicate the practical relevance of effects (Calin-Jageman & Cumming, 2019; Flora, 2020; Pek & Flora, 2018). Yet, this is important in assessing whether interest groups can influence public opinion to a meaningful degree.
Since the literature has not converged on a universal assessment of effect sizes, suggesting instead a number of different approaches (Lakens, 2013), we apply three intuitive approaches to assess effect sizes on TTIP support—which may arguably be considered the most important dependent variable in our study. We begin by looking at the percentage of overlap in the distributions of the outcome variable in the control and the experimental groups (Calin-Jageman, 2018). In the survey experiment, we observe an overlap of 84% in the distributions of TTIP attitudes between the control and the experimental group in wave 2. This distribution translates into a 63% probability that a member of the survey treatment group supported TTIP more strongly than a person from the control group, indicating a substantial treatment effect. In comparison, the effect in the field condition is much smaller: the attitude distributions of the field and the control groups overlap by 95% (wave 2), which means that the probability of a randomly chosen individual from the field group having a more favorable opinion on TTIP than a randomly chosen individual from the control group is only 54% (i.e., only slightly higher than chance).
Dichotomizing the outcome variable provides another useful metric to get a better understanding of an effect size’s substantive significance. In wave 2, 32% of the respondents in the control group showed TTIP attitude scores larger than 0.5 (i.e., had positive attitudes toward TTIP). In the survey group, the corresponding percentage is 53%, suggesting that the survey treatment transformed aggregate opinion from clear opposition to slim support. In the field experiment setting, the share of TTIP supporters was 38% in wave 2, indicating that mailing the letter resulted in a difference of a mere 6 percentage points compared to the control.
Categorizing respondents into supporters or opponents of TTIP further allows us to estimate the practical significance of the persuasion effort by the interest group from a cost–benefit perspective. The “numbers needed for change” effect size (Gruijters & Peters, 2019) illustrates that treating seven participants with the survey experiment stimulus would persuade one additional participant that the free trade agreement merits support, compared to the control group. In the field, the interest group can expect to generate one additional supporter for every 23 letters it sends out.
As noted above, we should consider that not all participants we assigned to the field condition actually received and processed the treatment. Indeed, only 51% of these respondents reported to have read the letter. Naturally, the treatment’s effect on attitude change will be larger among compliers than among the entire group of individuals assigned to the field treatment. Although self-reported recalls of treatment compliance are likely subject to measurement bias, they allow us to estimate, at least coarsely, the effects on the treated using two-stage least squares (Gerber & Green 2012). Accordingly, the effect on an individual who reported to have opened and read the letter was 0.06 (p < .001) in wave 2 (as opposed to .03 when considering all members of the treatment group). Hence, in a natural environment, the interest group’s letter had a small but noticeable effect on those individuals who reported to have read it. In practical terms, this effect means that the interest group would generate one additional supporter of TTIP for every 13 people who read the letter.
In combination, all three assessments show that the strength of the treatment effect on changing opinions was limited but clearly identifiable and potentially practically relevant during the course of a public information campaign.
Moving on, the treatment’s effect on the number of accessible considerations in favor of TTIP resembles the effects on TTIP salience (see Figure 3). In wave 2, respondents in the survey experiment could recall, on average, 0.2 (95% CI 0.145–0.255, Cohen’s d: 0.23) additional arguments in favor of TTIP. A smaller but significant effect was also visible in the field experiment condition (0.066 [0.010–0.121], Cohen’s d: 0.08). This initial effect decayed between waves 2 and 3 to roughly half its size, and remained statistically significant only among respondents in the survey condition (0.091 [0.040–0.142]).11
Effects were substantial in the survey experiment condition and small but not negligible in the field experiment condition with respect to all three dependent variables in wave 2. In wave 3, these effects had substantially decayed. In sum, we find support for Hypothesis 1. These results are consistent with the view that interest group interventions have an effect on public opinion, but suggest that the success of a single intervention (at least in the form of an information letter) is short lived.
To gain a first intuition of whether persuasive appeals by interest groups with low public profiles are moderated by recipients’ prior attitudes toward a source, we re-estimated the regression models reported above, adding interaction terms between the treatment dummies and the respondents’ prior attitudes toward the sender. Specifically, as a moderating variable we rely on attitudes toward the role of employers’ groups in politics in general.12
Table 113 shows little evidence of any substantial interaction effect of prior attitudes toward the employers’ group influence in five of the six analyses in this setup. The one deviant finding is that attitudes toward employers’ groups seemingly moderated the number of pro-TTIP arguments mentioned by respondents in the field condition (p = .018). This effect did not replicate in the survey condition (p = .270), however, and disappeared once we controlled for multiple comparisons (Bonferroni-corrected p-value of interaction with field-treatment: .108), casting doubt on the notion that prior attitudes played a systematic role in moderating the effects of the interest group communication.14
Conditionality of Treatments Effects Depending on Prior Attitudes on Employers’ Group Influence (study 1: wave 2)
| Salience | Attitude | No. of pro-TTIP arguments | |||
|---|---|---|---|---|---|
| Field | Survey | Field | Survey | Field | Survey |
0.00 (0.03) | –0.01 (0.03) | 0.04 (0.03) | 0.03 (0.03) | 0.25* (0.11) | 0.11 (0.11) |
| Salience | Attitude | No. of pro-TTIP arguments | |||
|---|---|---|---|---|---|
| Field | Survey | Field | Survey | Field | Survey |
0.00 (0.03) | –0.01 (0.03) | 0.04 (0.03) | 0.03 (0.03) | 0.25* (0.11) | 0.11 (0.11) |
Notes: Reported are interaction effects between the moderator and the experimental condition mentioned in the respective column headers. Results are based on linear regressions and standard errors are reported in parentheses. Dependent variables were measured in wave 2, while moderators were measured in wave 1. Full regression tables are provided in Appendix S10.
p < .05.
Conditionality of Treatments Effects Depending on Prior Attitudes on Employers’ Group Influence (study 1: wave 2)
| Salience | Attitude | No. of pro-TTIP arguments | |||
|---|---|---|---|---|---|
| Field | Survey | Field | Survey | Field | Survey |
0.00 (0.03) | –0.01 (0.03) | 0.04 (0.03) | 0.03 (0.03) | 0.25* (0.11) | 0.11 (0.11) |
| Salience | Attitude | No. of pro-TTIP arguments | |||
|---|---|---|---|---|---|
| Field | Survey | Field | Survey | Field | Survey |
0.00 (0.03) | –0.01 (0.03) | 0.04 (0.03) | 0.03 (0.03) | 0.25* (0.11) | 0.11 (0.11) |
Notes: Reported are interaction effects between the moderator and the experimental condition mentioned in the respective column headers. Results are based on linear regressions and standard errors are reported in parentheses. Dependent variables were measured in wave 2, while moderators were measured in wave 1. Full regression tables are provided in Appendix S10.
p < .05.
In sum, the findings are in line with our argument about the persuasive appeal of interest groups with low public salience. However, we have not yet presented systematic evidence to support our claim that interest groups have a relatively low public profile, nor have we directly tested the effects of messages from different senders. We turn to these questions next.
Study 2
Data and methods
We now assess our claim that interest groups have a relatively low public profile (Hypotheses 2a–c).15 To assess the public profiles of political actors, we conducted a survey to measure the availability of credibility assessments and associated attitude certainty of 10 environmental groups (registered with the German federal government) and 6 political parties. We queried how credible survey respondents thought a given group or party was when communicating on environmental issues (see Appendix S15 for the questionnaire; see Appendix S16 for indicators). Interviews with 700 members of the German online panel Respondi were conducted between 31 March and 6 April 2020. To ensure that this sample corresponded to the German online population, we used quotas blocking access to the survey for respondents with specific demographic characteristics once their share had reached levels comparable with their share of the target population.
To identify a set of interest groups with comparatively high and low public profiles, we began with a list of all environmental groups registered with the federal government that are membership organizations with a national focus.16 To this initial set, we added groups based on communication with a professional working in environmental advocacy. This left us with 108 environmental groups, for whom we infer public profiles based on their mentions in German media in 2019.17 For our set of 10 environmental groups, we selected the 5 with the highest number of media mentions during 2019 and added another 5 that were randomly selected. To identify a set of political parties, we relied on representation in the German federal parliament as the selection criterion. Thus, we are able to test the availability of credibility assessments and the certainty with which they are held for five environmental groups with comparatively high profiles, five with comparatively low profiles, and six political parties.
Results
The distributions of credibility assessments support Hypothesis 2a, which states that respondents are less likely to express credibility assessments (be they positive or negative) for interest groups than for political parties (see Table 2). Roughly 9 of 10 respondents were ready to provide credibility assessments of political parties. Greenpeace was the only interest group to reach a similar level. For all other interest groups, respondents indicated much more frequently that they were unable to assess the groups’ credibility. Our preregistered, formal test also shows that the aggregated index of unavailable credibility assessments of interest groups (mean: 0.50, 95% CI 0.47–0.53) was much higher than credibility assessments of political parties (mean: 0.10, 95% CI 0.08–0.12). In short, considerably fewer respondents could form credibility assessments for interest groups than for political parties. At the same time, the higher variation in available credibility assessments of interest groups compared to political parties also encourages the above-mentioned differentiation between low-profile and (relatively) high-profile groups.
Characteristics of Interest Groups and Parties (study 2)
| Group | Media mentions, 2019 | Credibility assessment available (in %) | Explained variance by party identification | No. of observations |
|---|---|---|---|---|
| Interest group, high profile | ||||
| Greenpeace | 11,167 | 88 | 0.09 | 535 |
| Deutsche Umwelthilfe | 5,660 | 75 | 0.06 | 449 |
| Naturschutzbund Deutschland (NABU) | 1,909 | 79 | 0.07 | 476 |
| Bund für Umwelt und Naturschutz Deutschland (BUND) | 1,405 | 79 | 0.08 | 478 |
| Germanwatch | 1,212 | 47 | 0.01 | 280 |
| Interest group, low profile | ||||
| NaturFreunde Deutschlands | 135 | 52 | 0.05 | 313 |
| Deutscher Naturschutzring | 128 | 53 | 0.04 | 321 |
| Bundesdeutscher Arbeitskreis für Umweltbewusstes Management (B.A.U.M.) | 16 | 49 | 0.02 | 300 |
| MUNA e. V. Mensch, Umwelt-, Natur- und Artenschutz | 2 | 45 | 0.02 | 276 |
| Naturschutzforum Deutschland e. V. | 0 | 50 | 0.04 | 302 |
| Party | ||||
| CDU/CSU | – | 91 | 0.24 | 555 |
| SPD | – | 90 | 0.27 | 548 |
| AfD | – | 89 | 0.37 | 542 |
| FDP | – | 88 | 0.12 | 537 |
| Die LINKE | – | 88 | 0.25 | 537 |
| Bündnis 90/Die Grünen | – | 92 | 0.28 | 558 |
| Group | Media mentions, 2019 | Credibility assessment available (in %) | Explained variance by party identification | No. of observations |
|---|---|---|---|---|
| Interest group, high profile | ||||
| Greenpeace | 11,167 | 88 | 0.09 | 535 |
| Deutsche Umwelthilfe | 5,660 | 75 | 0.06 | 449 |
| Naturschutzbund Deutschland (NABU) | 1,909 | 79 | 0.07 | 476 |
| Bund für Umwelt und Naturschutz Deutschland (BUND) | 1,405 | 79 | 0.08 | 478 |
| Germanwatch | 1,212 | 47 | 0.01 | 280 |
| Interest group, low profile | ||||
| NaturFreunde Deutschlands | 135 | 52 | 0.05 | 313 |
| Deutscher Naturschutzring | 128 | 53 | 0.04 | 321 |
| Bundesdeutscher Arbeitskreis für Umweltbewusstes Management (B.A.U.M.) | 16 | 49 | 0.02 | 300 |
| MUNA e. V. Mensch, Umwelt-, Natur- und Artenschutz | 2 | 45 | 0.02 | 276 |
| Naturschutzforum Deutschland e. V. | 0 | 50 | 0.04 | 302 |
| Party | ||||
| CDU/CSU | – | 91 | 0.24 | 555 |
| SPD | – | 90 | 0.27 | 548 |
| AfD | – | 89 | 0.37 | 542 |
| FDP | – | 88 | 0.12 | 537 |
| Die LINKE | – | 88 | 0.25 | 537 |
| Bündnis 90/Die Grünen | – | 92 | 0.28 | 558 |
Notes: The columns “Explained variance by party identification” and “No. of observations” report summary statistics of linear regression analyses of the organizations’ perceived credibility with respect to environmental issues as the dependent variables regressed on the respondent’s identification with a political party (measured in six binary variables).
Characteristics of Interest Groups and Parties (study 2)
| Group | Media mentions, 2019 | Credibility assessment available (in %) | Explained variance by party identification | No. of observations |
|---|---|---|---|---|
| Interest group, high profile | ||||
| Greenpeace | 11,167 | 88 | 0.09 | 535 |
| Deutsche Umwelthilfe | 5,660 | 75 | 0.06 | 449 |
| Naturschutzbund Deutschland (NABU) | 1,909 | 79 | 0.07 | 476 |
| Bund für Umwelt und Naturschutz Deutschland (BUND) | 1,405 | 79 | 0.08 | 478 |
| Germanwatch | 1,212 | 47 | 0.01 | 280 |
| Interest group, low profile | ||||
| NaturFreunde Deutschlands | 135 | 52 | 0.05 | 313 |
| Deutscher Naturschutzring | 128 | 53 | 0.04 | 321 |
| Bundesdeutscher Arbeitskreis für Umweltbewusstes Management (B.A.U.M.) | 16 | 49 | 0.02 | 300 |
| MUNA e. V. Mensch, Umwelt-, Natur- und Artenschutz | 2 | 45 | 0.02 | 276 |
| Naturschutzforum Deutschland e. V. | 0 | 50 | 0.04 | 302 |
| Party | ||||
| CDU/CSU | – | 91 | 0.24 | 555 |
| SPD | – | 90 | 0.27 | 548 |
| AfD | – | 89 | 0.37 | 542 |
| FDP | – | 88 | 0.12 | 537 |
| Die LINKE | – | 88 | 0.25 | 537 |
| Bündnis 90/Die Grünen | – | 92 | 0.28 | 558 |
| Group | Media mentions, 2019 | Credibility assessment available (in %) | Explained variance by party identification | No. of observations |
|---|---|---|---|---|
| Interest group, high profile | ||||
| Greenpeace | 11,167 | 88 | 0.09 | 535 |
| Deutsche Umwelthilfe | 5,660 | 75 | 0.06 | 449 |
| Naturschutzbund Deutschland (NABU) | 1,909 | 79 | 0.07 | 476 |
| Bund für Umwelt und Naturschutz Deutschland (BUND) | 1,405 | 79 | 0.08 | 478 |
| Germanwatch | 1,212 | 47 | 0.01 | 280 |
| Interest group, low profile | ||||
| NaturFreunde Deutschlands | 135 | 52 | 0.05 | 313 |
| Deutscher Naturschutzring | 128 | 53 | 0.04 | 321 |
| Bundesdeutscher Arbeitskreis für Umweltbewusstes Management (B.A.U.M.) | 16 | 49 | 0.02 | 300 |
| MUNA e. V. Mensch, Umwelt-, Natur- und Artenschutz | 2 | 45 | 0.02 | 276 |
| Naturschutzforum Deutschland e. V. | 0 | 50 | 0.04 | 302 |
| Party | ||||
| CDU/CSU | – | 91 | 0.24 | 555 |
| SPD | – | 90 | 0.27 | 548 |
| AfD | – | 89 | 0.37 | 542 |
| FDP | – | 88 | 0.12 | 537 |
| Die LINKE | – | 88 | 0.25 | 537 |
| Bündnis 90/Die Grünen | – | 92 | 0.28 | 558 |
Notes: The columns “Explained variance by party identification” and “No. of observations” report summary statistics of linear regression analyses of the organizations’ perceived credibility with respect to environmental issues as the dependent variables regressed on the respondent’s identification with a political party (measured in six binary variables).
Examining the certainty with which credibility assessments were held among the subset of respondents that voiced credibility assessments for a particular group, the findings again conformed with our expectation: respondents were less certain about credibility assessments of interest groups than about credibility assessments of political parties (Hypothesis 2b). The formal, preregistered test shows that an aggregated index of certainty of credibility assessments regarding interest groups (mean: 0.52 [0.50–0.54]) was considerably lower than certainty regarding political parties (mean: 0.60 [0.57–0.62]). Overall, familiarity with political parties is widespread among respondents, but only a fraction knows interest groups well enough to express credibility assessments and, even when they do, they express these assessments with considerable uncertainty.18
Finally, credibility assessments are only weakly related to party affiliations. Column 4 in Table 2 reports how well the respondents’ self-reported attachment to German political parties (measured in six dummy variables) explains variance in credibility assessments of interest groups and political parties. We see that party identification has exceptionally low explanatory power R2 to predict credibility assessments of interest groups. As expected, party identification is more closely related to how credible parties are perceived to be on a specific political issue but, interestingly, the association is far from perfect. Thus, in accordance with our characterization of interest groups, respondents’ credibility assessments do not cluster along partisan lines.19
In sum, the findings support our expectations. The lower public profile of interest groups is reflected in a higher absence of source credibility assessments in the general public. Those who do report such assessments tend to be less certain about the assessment of interest groups than of political parties, and credibility assessments of interest groups in general tend to be weakly connected with party affiliations. All this bolsters the expectation that source-related predispositions should play a smaller role in moderating interest group communication compared to communicators with high public profiles. This is true in the basic sense that predispositions that a person does not hold cannot influence their attitudes or behavior. Furthermore, it stands to reason that attitudes that do exist but are held with lower certainty may not influence the processing of a given message. We turn to this expectation next.
Study 3
Data and methods
We designed this study to contrast the role of credibility assessments in processing communicative interventions from different types of political actors—namely, political parties and interest groups with high and low public profiles (Hypothesis 3).20 We conducted a large-scale survey experiment among members of the German online panel Respondi between 28 April and 13 May 2020 (N = 8,245). The large sample size was deliberate to allow for reliable identification of effect moderation and is based on the results of a preregistered power analysis (see Appendix S19). To ensure correspondence with the German online population, we used quotas blocking access to the survey for respondents with specific demographic characteristics once their share had reached levels comparable with their share of the German online population.
We designed a treatment in support of an environmental policy targeted at reducing energy waste resulting from inefficient thermal isolation in buildings. Respondents assigned to different treatment groups were shown the treatment without source attribution (N = 1,630), attribution to a high-profile interest group (N = 1,688), a low-profile interest group (N = 1,587), and a political party (N = 1,674). A control group was exposed to no treatment (N = 1,166).21
In selecting interest groups for our treatments, we built on findings from our previous study, in which we tested the availability of credibility assessments for different interest groups. Based on these results, we chose Greenpeace, the group with the highest familiarity among respondents, to stand in as an example for a high-profile group, and we chose Deutscher Naturschutzring, one of those with lower familiarity among our respondents, as an example of a group with a low public profile. For the party source, we selected the German Christian Democrats (CDU/CSU), which is the majority partner in Germany’s governing coalition and thus has overseen deep reforms as part of Germany’s turn toward renewable energy. At the same time, it is not a party traditionally associated with pro-environment policies.
Results
While our goal was to test the moderating role of credibility assessments for persuasion effects, we begin with a discussion of the average effects of the persuasive intervention. We did not find a consistent positive effect of our treatment (means with 95% CIs control: 0.447 [0.440–0.456]; no source: 0.441 [0.432–0.450]; low profile IG: 0.429 [0.420–0.438]; high profile IG: 0.440 [0.431–0.448]; party: 0.430 [0.421–0.438]) (see Appendices S20.2 and S20.3 for treatment materials). This deviates from pre-test results, in which a plain text-version of the treatment had elicited small but significant positive effects (means with 95% CIs: control: 0.482 [0.463–0.501]; no source: 0.519 [0.499–0.538]) (see Appendix S20.1 for treatment materials).
Pre-test and main test, therefore, leave us with contradictory evidence regarding Hypothesis 1, with only the pre-test weakly supporting H1. This could be due to a change in treatment design. The main test treatment relied more heavily on graphics than the pre-test treatment and may have been less persuasive than intended. In combination with the evidence presented in Study 1, this apparently shows that the persuasive appeal of communicative treatments by interest groups cannot be taken for granted but also depends on content and context.22
Our interest here, though, does not concern the main effects of the treatment but the role of perceived source credibility in moderating the effects of political communication. To examine the role of source credibility, we tested whether the level of attitude change in subgroups with positive and negative prior credibility assessments of selected sources differed when the same message was attributed to a political party (CDU/CSU), an interest group with a comparatively high public profile (Greenpeace), and an interest group with a comparatively low public profile (Deutscher Naturschutzring).
The data support Hypothesis 3. Given the same persuasive treatment, respondents relied on prior credibility assessments when the source was a political party but not when it was an interest group.23 As Figure 4 shows, we found no significant moderating effects of source credibility assessments either for the interest group with the highest public profile (Greenpeace) or for an interest group with a comparatively low public profile (Deutscher Naturschutzring). In contrast, among those who considered the party (CDU/CSU) a credible communicator on environmental issues, respondents who received the treatment message from CSU/CSU (N = 303) changed their position on the issue in the party’s direction by 0.039 (95% CI 0.011–0.068; Cohen’s d: 0.23) points on a scale from 0 to 1, compared to respondents with favorable attitudes toward CDU/CSU who received the identical message without any sender cues (N = 317). Among those who considered CDU/CSU not to be a credible communicator on environmental issues, the treatment even made respondents less supportive of the position advocated by the CDU/CSU: among respondents in the party treatment group with unfavorable prior attitudes toward CDU/CSU (N = 1,327), support for the issue dropped by –0.022 scale points (95% CI –0.036 to –0.008; Cohen’s d: –0.12) compared to the no source group (N = 1,231). The result of this preregistered analysis of the difference between treatment effects across subgroups is statistically significant (p < .001), lending support to the proposition that credibility assessments moderate the persuasiveness of communicative interventions by political parties.
Differential treatment effects conditional on type of communicator. Notes: The figure shows coefficients from linear regressions that denote treatment effects within subgroups of individuals who reported high or low prior credibility assessments (split at scale midpoint) toward the respective sender of the message.
In sum, the findings of Study 3 are in line with the broader argument that predispositions such as credibility assessments play a minor role in the effects interest group messages have on public opinion because people are less familiar with these actors. This sets interest group communication apart from party communication, where predispositions matter more because people can fall back on previously held assessments of these high-profile sources.
Discussion and conclusion
As interest groups increasingly run communication campaigns to shape public opinion, understanding their role as public communicators becomes important. This article takes a significant step in that direction. It contributes to the emerging literature on interest groups as communicators with the public (Broockman & Kalla, 2016; Dür 2019) and advances political communication research in general by adding to our understanding of the characteristics and effect patterns associated with a new type of communicator (Arceneaux & Kolodny, 2009a, 2009b; Dewan et al., 2014; Rogers & Middleton, 2015). We have shown that interest groups, like other skilled communicators, have the potential to shape public opinion (Study 1, Study 3 pre-test). That being said, communicative interventions by interest groups are not always successful in persuading respondents (Study 3 main test). We have also presented evidence that interest groups typically have a low public profile, and that this characteristic has implications for how these actors are perceived and how their messages are received by the public. People make credibility assessments of interest groups less frequently than they assess political parties, and when they do, they are less certain (Study 2). Correspondingly, we found that credibility assessments do not moderate effects of communicative interventions by interest groups as they do for those by political parties (Study 3).
These findings have important implications for the impact of interest group information campaigns. As many recipients have no or inconsequential prior attitudes toward these low-profile sources, interest groups may neither hope for a boost in nor need fear a penalty against persuasiveness from these attitudes. In this respect, their interventions appear to differ from those of actors with higher public profiles, such as political parties and government representatives. The absence of credibility moderation does not imply, of course, that interest groups are generally more effective communicators than political parties. However, whereas the persuasiveness of party communication is confined to segments of the population that are open to messages from these parties and may even have detrimental effects on others, many interest groups have the potential to address the entire public successfully.
We showed that recipients draw little information from the source cues of low-profile senders. Therefore, the impact of a persuasive appeal depends on the strength of the message itself. This finding contributes to theorizing on the role of source credibility in persuasion more generally. Prior research on the topic typically assumes (implicitly) that people hold meaningful credibility assessments regarding the sender in question (see, e.g., the studies listed by Pornpitakpan, 2004; Wilson & Sherrell, 1993). We have argued, in contrast, that the effects of source credibility assessments depend on the availability and certainty of attitudes toward specific actors, and we have shown that these vary. If people do not hold meaningful credibility assessments of a sender, these assessments cannot influence the processing of the message. Our study thus also contributes to the important body of research examining the role of source-related predispositions in persuasion (O’Keefe, 2015; Stiff & Mongeau, 2016).
Naturally, our analysis has some limitations, which means that our conclusions are to some extent preliminary and call for additional research. To start, the evidence presented here stems from only one national context, so we cannot say for certain whether the findings—and implications—hold in other contexts. At a very general level, the German communicative context is similar to that of other Western democracies. Zooming in, however, one could expect select interest groups in politically more polarized contexts—such as the NRA in the United States—to have exclusive links to political parties, to be linked unambiguously with political ideologies, or even to have public profiles comparable to those of political parties, and hence also to have to contend with effect moderation. This would be consistent with our general theoretical argument, which hinges on the strength of the public profile communicators have. Accordingly, key tasks for future research remain to identify the threshold at which the public profile of an interest group (or any other public communicator) is sufficiently high to trigger source-based effect moderation and to analyze the effects of different constellations in the alignment of interest groups with political parties and interest groups.
Similarly, public profiles of interest groups might change over time. Constant campaign activity could increase the public profiles of interest groups so that people develop meaningful credibility assessments, and consequently trigger the moderation of persuasive effects. We expect such evolutions to be rare, however, given the different functions interest groups and political parties serve in modern societies. Interest groups tend to communicate on only one or a few issues they care about, while political parties communicate continually on multiple issues on the agenda. Even factoring out the nontrivial question of financing prolonged communication campaigns, for interest groups to develop public profiles comparable to that of political parties would require either that they begin to communicate on the full spectrum of issues on the agenda or that the agenda be reduced, for long periods, to the few issues about which they care. That being said, the dynamics of the public profile of interest groups and their implications for source credibility assessments at the public level is an important topic for future research.
While we covered two different issue domains in our empirical analysis, we did not study systematically how different features of a given issue domain might strengthen or weaken source-based moderation effects. For example, topics with high ego-involvement might motivate recipients to engage with the content of a message—irrespective of prior attitudes they may have toward the sender. Conversely, with respect to highly politicized topics, people may act on even vague suspicions about the political alignment of a low-profile sender. Future research should explore these possibilities systematically in order to explore the generalizability of our findings across issue domains.
Future research could also examine the cumulative effects of repeated messages. Our studies tested only the effects of isolated interest group interventions. Actual communication campaigns, in contrast, often involve numerous interventions, and it would be too easy to assume that repeated interventions have simple additive effects. Initially successful persuasive messages are likely to exhibit decreasing marginal effects, as support cannot increase indefinitely. These effect ceilings determine the total effects communication campaigns can accumulate. Research on repeated exposure to the persuasive interventions will allow important insights into the persuasive power of interest group communication.
We have argued that sources with low public profiles do not have to worry about source-related predispositions weakening the persuasive impact of their messages, nor can they rely on source-related predispositions bolstering those messages. While our argument and empirical tests focused on interest groups, this distinction of effect patterns based on varying strengths of public profiles is likely to be broadly applicable. It also likely holds for communicative interventions by other low-profile actors who, given the affordances of digital communication environments, have come to feature very prominently in contemporary public discourse (Jungherr, Rivero, & Gayo-Avello, 2020). One prominent example of this can be found in the wide variety of “alternative” sources challenging scientific and governmental bodies with respect to epidemiological diagnoses and policies in the response to public health crises. If communicative interventions by these low-profile actors follow the same dynamics identified here, high-profile sources such as governmental organizations face a considerable challenge in reaching doubters while low-profile sources such as alternative information providers on digital channels can count on broad reach across populations.
Low-profile information sources of many different sorts are a consistent feature of contemporary communication environments. As we have shown in this article, communication research must account for their specific characteristics to understand their persuasive appeal and influence in contemporary public opinion formation.
Supporting Information
Additional Supporting Information—such as questionnaires, treatment materials, and supporting analyses—may be found in the online Appendices available in the online version of this article.
Footnotes
Data from Media Cloud (https://mediacloud.org) for search terms “National Rifle Association,” “Republican Party,” and “Democratic Party” in sources New York Times, USA Today, and Washington Post for mentions between 1 January and 31 December 2019.
Party systems feature parties that are marginal and have a very low public profile. Our argument on the communicative situation of low-profile interest groups is likely also to apply to low-profile political parties. To keep things simple, however, we ignore this variation in public profiles among political parties in the remainder of the article.
Data from Media Cloud for search terms “Greenpeace,” “Arbeitgeberverband Gesamtmetall,” “CDU,” “SPD,” “Die Grünen,” and “FDP” in source set “German Media, National” for mentions between 1 January and 31 December 2019.
Several cues could allow recipients to form ad-hoc assessments. For instance, the name or the logo of an interest group might provide information recipients can use to evaluate the source.
In a CATI survey representative of the German electorate conducted for us by the company Infratest dimap, only 10.8% of respondents indicated that they felt very or somewhat familiar with the political position of Gesamtmetall. At the same time, the share of respondents who reported familiarity with a non-existing, fictional employers’ group was only slightly lower (6.6%). In comparison, a larger share of respondents, 26%, felt very or somewhat familiar with the political positions of Greenpeace, and 19.2% claimed to be very or somewhat familiar with the political positions of Germany’s metal worker union (see Appendix S1 for a detailed description of the survey and distributions of answers). This indicates a low public profile for Gesamtmetall.
See Appendix S11 for replication with the full sample.
Regression tables are provided in Appendix S10.
To enable the use of our findings in subsequent third-party meta-analyses, we report standardized coefficients in Appendix S12.
More technically, the coefficients for the field condition in the figures are ITT (intent to treat) effects; the coefficients for the survey condition are ATE (average treatment effects).
All p-values are from two-tailed tests.
A close analysis of the arguments listed by respondents suggests that the increase in the aggregated measure is indeed driven by the specific arguments raised in the letter.
Credibility assessments of Gesamtmetall are, unfortunately, not available in this survey.
The full regression tables are provided in Appendix S10.
This finding also holds if we assume non-linear interaction effects (see Appendix S13).
The hypotheses, survey design, and analysis syntax of Study 2 were preregistered (except for Hypothesis 2c, which was introduced following reviewer comments). See Appendix S14 for details of the preregistration and deviations from the plan.
Data from Media Cloud for mentions of interest group names between 1 January and 31 December 2019 in media source set “German Media, National.”
See Appendix S17 for detailed findings.
See Appendix S25.1 for full documentation of the underlying regression analyses and underlying distributions.
The hypotheses, survey design, and analysis syntax of Study 3 were preregistered. See Appendix S18 for details of the preregistration and deviations from the plan.
See Appendices S20–S24 for details.
As a corollary of the unexpected main effects of the treatments, the data also do not lend support for our expectation that the informational text should have nearly identical effects when attributed to the low-profile interest group and when attributed to no source. The preregistered equivalence tests yielded inconclusive results and thus did not indicate with sufficient certainty that the differences between these two treatment groups were below the pre-specified smallest effect size of interest.
It is important to note that these results are the same if we test party affiliation (see Appendix S25) or political ideology (see Appendix S26) as moderators of treatment effects rather than source credibility. This further supports our argument that party identification or political ideology do not have to be relevant predispositions in assessing the persuasive power of interest groups.
Acknowledgments
We thank Antonin Finkelnburg and Ralf Güldenzopf in particular for their support in realizing this project. We also thank Kevin Collins, Matthias v. der Malsburg, and Peter Selb for helpful comments on the experimental design of our studies, Frederik Aust, Aaron Caldwell, and Daniël Lakens for helpful comments on the power calculation for our third study, and Tim Andler, Scott Cooper, and Lukas Isermann for assistance in the preparation of the manuscript. The complete data set, preregistration, and replication materials are available in Jungherr et al. (2020).
Funding
This research received funding from the Konrad-Adenauer-Stiftung (KAS), the Arbeitgeberverband Gesamtmetall, and the University of Konstanz’s Young Scholar Fund (YSF). These partners had no role in the analyses underlying this article and imposed no restrictions in the presentation of the findings.
Conflicts of Interests
The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
