What News Outlets do People Have in Mind When They Answer Survey Questions about Trust in “Media?”

Abstract

While ample research on audience trust in the news media uses survey questions that ask respondents about their trust in a generic “news media,” only scant research has investigated what types of news outlets respondents have in mind when answering such questions. These previous investigations originated mostly in the US and resulted in inconsistent findings. To further investigate this question, we use data from a large-scale survey (N = 2,337), collected in Sweden, including both general media trust measures and specific measures about trust in 20 mainstream and nonmainstream news outlets. The results demonstrate that our respondents seemingly averaged across all mainstream sources when they formed their general evaluations of the news media’s trustworthiness.

In the recent years, research on media trust in various disciplines is burgeoning. A search of the Web of Science database demonstrates that while 27 articles in 2001 contained “media trust” in their topic line, the number of parallel articles in 2021 was 1,093. Most studies investigating audience trust in media do so using survey items referring to “the press” or “the news media” in general, without reference to specific news outlets (Ariely, 2015; Engelke, Hase, & Wintterlin, 2019; Fawzi et al., 2021; Hanitzsch, Van Dalen, & Steindl, 2018; Livio & Cohen, 2018; Newman, Fletcher, Schulz, Andi, & Nielsen, 2020), with relatively little research dealing with trust in specific journalists or news content. These generic references may have been straightforward when the news media comprised a limited and homogenous set of outlets. However, in today’s dramatically diversified media landscape, “the media” as a concept has become highly polysemic (Strömbäck et al., 2020, p. 142).

This raises the question of what people have in mind when they are asked about the “news media” in surveys. The question has only received scant research attention, despite the immense popularity of general media trust measures, and despite that the object of the survey questions—the media—is an abstract object “not located in a designated place like Congress” (Ladd, 2011, p. 11). Previous research has used follow-up questions asking respondents what they had in mind when answering media trust questions (Ladd, 2011, pp. 96–106) or drew on focus groups that discussed “the media“ (Tsfati, 2002, pp. 66–68) to assess what people think of when asked about the news media. These studies demonstrated that the respondents generally thought about the news media as an institution, without referring to specific outlets, and that when they did refer to specific outlets, these were typically popular mainstream outlets such as television news. Furthermore, analyses of closed-ended questions asking Israeli respondents to what extent they perceive different outlets as being part of “mainstream media” (Tsfati & Peri, 2006, pp. 176–177) have demonstrated that respondents think of the main television news programs and the most popular newspapers as those outlets that represent the mainstream media most strongly.

Recently, Daniller et al. (2017) used an experimental approach to investigate what comes to people’s minds when answering questions about trust in the press. They found that respondents expressed much greater trust in “the press” when they were asked to consider specific news sources than when they were asked to evaluate a generic news media referent. As an explanation, the authors suggested that people may not think about the most popular mainstream outlets, or average across them, when asked about trust in the media in general, but may be driven by a negative accessibility bias, under which distrusted sources more easily (automatically) come to mind. As Daniller et al. (2017, p. 77) speculate, “Asking respondents about a collective referent such as ‘the media’ generally encourages more negative assessments because negatively-valenced information is typically more accessible than positive information.” Similar to the “I am doing better than we are” bias (Mutz, 1998), which refers to the fact that people tend to rate their own economic conditions as better than others’, their own doctors as better than doctors in general and their own Congressperson as better than Congress. Daniller et al. (2017) argue that people evaluate the media, in general, less favorably compared with the media they consume. They also argue that with the increased diversification of the media landscape over time, there are now many more options of distrusted sources that can come to mind. This may account for the long-term decrease in general media trust over time that has been documented in several longitudinal US studies, such as the General Social Survey.

Some possible reasons for the differing conclusions by Daniller et al. (2017) and earlier studies (e.g., Ladd, 2011) could either be the different methodologies used or the increased diversification of the media landscape that has occurred between the different studies (e.g., the American news site Breitbart did not exist when Ladd collected his data in 2005, and the Huffington Post has just been launched). Still, in contrast to Daniller et al’s (2017) recent findings, a study by Pew Research Center (Shearer & Mitchel, 2021) found broad agreement on which outlets are part of the American mainstream media. The study revealed that 87% of the US respondents perceived ABC News as being a part of the mainstream media, and a majority of Americans considered seven outlets, including the national network and cable news outlets, and three legacy print publications (the New York Times, Wall Street Journal, and the New York Post) to be part of the same mainstream. However, we cannot tell if respondents perceived these outlets as mainstream because they viewed these outlets as popular or because the respondents completing these surveys are simply more likely to use them.

Given the inconsistency in previous findings and the fact that they come mostly from the US context, more research on what people have in mind when responding to survey questions about media trust is clearly needed. Given that general media trust measures have become exceedingly popular over the past years, knowledge of what the answers to these questions represent is crucial. In addition, our understanding of what outlets people have in mind when answering questions about trust in media may carry important answers to unsettled theoretical conundrums. For example, it is possible that the answer to the question “why do people watch news they do not trust?” that have puzzled scholars for almost two decades (Strömbäck et al., 2020; Tsfati & Cappella, 2003, 2005) may be simply that typically these studies used vague terms such as “the media” for the trust in media measures which do not line up with exposure measures referring to specific news outlets (Fawzi et al., 2021).

The Current Study: Methodological Approach and Research Question

To understand which outlets people have in mind when answering questions about trust in media, we use data from a large-scale survey that included both general media trust questions and 20 survey items asking about trust in specific mainstream and nonmainstream news outlets. Our approach is that when using the specific trust measures (tapping trust in specific outlets) to predict the general ones (tapping trust in the media in general), stronger regression coefficients represent more weight assigned to trust in the specific outlets in formulating general trust assessments, and vice versa. Regression coefficients as representing weights assigned to specific considerations have been used to answer similar research questions in the context of priming research. Iyengar and Kinder (2010), for example, used regression coefficients as an indication of the weight respondents assigned to specific evaluations of the performance of the president on primed versus nonprimed topics when they formed their general evaluations of the president. In a similar vein, stronger associations between specific trust items and general trust items imply that more weight is given to the specific answer (e.g., trust in the least-trusted outlet) in forming the general media trust evaluations.

Specifically, our research question concerns how different types of (specific) media trust constructs predict general media trust. Previous studies (Ladd, 2011; Tsfati, 2002) argued that trust in popular mainstream outlets will be more influential when people answer general media trust questions, compared to other types of media trust. This approach rests on conceptualizations and theories that view popularity as a main component in both laypeople and expert definitions of mainstream media (Tsfati & Peri, 2006). Cognitively, this approach resembles the representativeness heuristic, the tendency of human estimation to be heavily influenced by mental prototypes (Kahneman & Tversky, 1972), and public opinion research about the bandwagon effect (Schmitt-Beck, 2015).

Alternatively, based on the literature on the negativity bias, Daniller et al. (2017) argue that the most negative impressions of media are more readily accessible when people answer questions about general trust in media, and thus, as a result, the most negative assessment of specific outlets will be more influential than other trust assessments when formulating general trust in media. In contrast, the literature on the accessibility bias (e.g., Kahneman, 2003) leads us to expect that the outlets that people use most often will be more readily available in their memories when they are asked about the media in general. In a different vain, exemplification theory (e.g., Bar-Hillel, 1980; Zillmann, 1999) argues that encounters with unusual and extraordinary examples can stand out when we make general (base rate) assessments, leading to the expectation that trust in the most trusted outlet may perhaps exert a stronger influence on the general trust. Finally, another possibility that Daniller et al. (2017) mention—although not empirically examine—is that people may average across trust in specific news outlets when they answer general questions about media trust. Such a process is similar to some models in the impression formation literature that suggest that people average across specific traits when forming general impressions of others (Kenny, 2004).

Given the conflicting theoretical arguments and inconsistencies in past empirical findings, our research question (RQ1) stipulates: Which of the following—“trust in the outlet one trusts the least,” “trust in the outlet one trusts the most,” “trust in the outlets one uses,” “trust in the most popular outlet in the population,” “one’s average trust in all mainstream news outlets”—will best predict general media trust?

Methods

To examine our research question, we use data from the second wave of a population-based panel survey that we conducted in Sweden with the primary intention to explore why people resist knowledge that enjoys academic and expert consensus.¹ An important factor for predicting knowledge resistance that we were interested in investigating with this panel is media trust, hence, the survey data includes a variety of media trust measures, making it ideal for the purpose of the current exploration. The survey was fielded between February 25 and March 30, 2021 (N = 2,337). Additional details about data collection, including response and attrition rates and a demographic breakdown of the sample, which is largely representative of the Swedish population, are presented in Appendix A.

Measures

General trust in media was measured using two constructs, representing two very popular approaches to measuring media trust in various studies. First, we used a single-item indicator of general media trust, worded: “Generally speaking, to what extent do you trust information from the news media in Sweden?” with response categories ranging from 1 = “do not trust at all” to 7 = “trust completely.” Second, given that some view trust in media as the belief in the professionalism of the journalistic practice (Liebes, 2000, p. 295), and given that credibility is at the heart of journalistic professionalism (see, e.g., Tsfati & Cappella, 2003, p. 506), we used five items from Gaziano & McGrath’s (1986),News Credibility Scale (see Meyer, 1988). Respondents were asked whether the Swedish news media are “fair,” “unbiased,” “tell the whole story,” “accurate,” and “separate fact and opinion” in their news coverage. The response categories ranged between 1 = “strongly disagree” to 7 = “strongly agree,” and the items loaded on a single factor in an Exploratory Factor Analysis (Cronbach’s α = .93).

To measure trust in specific media outlets, respondents were asked: “Generally speaking, to what extent do you trust information from the following news outlets?” with response categories ranging between 1 = “do not trust at all” to 7 = “trust completely.” The list of outlets included eight mainstream news outlets followed by 12 nonmainstream outlets (including right-wing and left-wing news outlets and a business news outlet); the full list of outlets is presented in Appendix B.² For each respondent, we calculated the average trust in the eight mainstream outlets (trust in mainstream outlets), and the average trust for all outlets that the respondent reported using more than twice a week (when answering the exposure question described below). We also created a variable that, for each respondent, captured the lowest trust s/he gave to any of the news outlets included in the data.

Exposure to the mainstream (and nonmainstream) news outlets was measured using the question “In a typical week, how often do you use the following news media, in their traditional formats or online?” followed by the list of all outlets detailed in Appendix B. The response options were 0 = “never,” 1 = “more seldom,” 2 = “1 day a week,” 3 = “2 days a week,” 4 = “3 days a week,” 5 = “4 days a week,” 6 = “5 days a week,” 7 = “6 days a week,” 8 =  “7 days a week.” Given that the eight mainstream outlets loaded separately from the nonmainstream outlets in an Exploratory Factor Analysis, the eight items measuring exposure to the mainstream news outlets were averaged to create a mainstream news exposure scale, used as a covariate in the analytic models.

Covariates

All analytic models include political ideology, mainstream news exposure, political interest, sex, education, and age, for control. Descriptive statistics and the exact question wording for these variables can be found in Appendix C.

Results

The research question was tested using Ordinary Least Squares regression models predicting the general trust measures using the various specific trust constructs. Because adding all trust-in-specific-outlets constructs as independent variables in the same model resulted in severe multicollinearity (with all tolerance values lower than .43 for the trust measures), the specific trust constructs were entered into separate models one at a time. As the models are exactly the same (i.e., control for the same covariates) except for the specific trust items, and given that all trust items (general and specific alike) were measured on 1–7 scales, the unstandardized regression coefficients and the R-square scores are comparable across models. To facilitate such comparisons, we present 95% confidence intervals (CIs) around each coefficient and assume that lack of overlap between CIs implies significant differences between the coefficients.

We first examined which of the “trust in specific outlets” measures better predict the two general media trust measures. The results, reported in Appendix D, revealed that all coefficients for trust in mainstream outlets are positive and rather robust, and significantly larger than the coefficients for trust in nonmainstream outlets which are mostly nonsignificant and, when significant, negative rather than positive. These findings can be interpreted as suggesting that (1) people assign more weight to mainstream outlets when assessing general media trust compared to nonmainstream media outlets, and (2) that the association between trust in nonmainstream sources and general trust, if anything, is negative (somewhat consistent with Ladd, 2011; Tsfati & Cappella, 2003; Tsfati & Peri, 2006). Of the coefficients for mainstream media, the largest ones are for trust in the two public service television news outlets (Rapport i SVT and Aktuellt i SVT), and these two are significantly stronger predictors of general news media trust and the News Credibility Scale than all other media outlets included in our data. Trust in Rapport i SVT, for example, added roughly 32.1% to the explained variance in the News Credibility Scale, and 40.4% in the single media trust item. For the nonmainstream outlets, respondents’ outlet-specific trust did not contribute to the explained variance in any of the two general media trust measures.

Our research question (RQ1) asked which of the following: trust in the outlet one trusts the least, trust in the outlet one trusts the most, trust in the outlets one uses the most, trust in the most popular outlet (Rapport i SVT—the one most frequently used in Sweden according to our and other data), and one’s average trust in all (eight) mainstream news outlets, will better predict general media trust. Table 1 presents the coefficients for the prediction of the general trust measures by the above constructs.³ The findings demonstrate that the largest coefficients were for the average trust in mainstream outlets (e.g., when predicting the News Credibility Scale: b = .764 [CI = .727, .802]).⁴ The findings also show that average trust in the mainstream news outlets was a significantly stronger predictor of the two general trust measures, than were trust in the least trusted outlet (b = .399 [CI = .364, .433]), trust in the most trusted outlet (b = .659 [CI = .619, .699]), trust in the outlets the respondents used the most (e.g., for used mainstream outlets: b = .628 [CI = .586, .666]) and trust in Rapport i SVT—the most popular news outlet in Sweden.⁵

Table 1.

Models Predicting the News Credibility Scale and the Single Item Measure of Media Trust, Using Outlet-based Media Trust Measures

Predictors	News Credibility Scale	Single-item Measure of media trust
Average trust in used media	.098a [.079, .117] R2 = .171 ΔR2 = .041	.101a [.083, .118] R2 = .188 ΔR2 = .054
Average trust in used mainstream media	.628bdf [.586, .666] R2 = .408 ΔR2 = .288	.624bd [.591, .657] R2 = .485 ΔR2 = .361
Trust in the least-trusted outlet	.399c [.364, .433] R2 = .298 ΔR2 = .166	.376c [.345, .406] R2 = .320 ΔR2 = .178
Trust in the most-trusted outlet	.659db [.619, .699] R2 = .409 ΔR2 = .277	.685db [.651, .718] R2 = .502 ΔR2 = .360
Average trust in eight mainstream outlets	.764e [.727, .802] R2 = .498 ΔR2 = .366	.760ef [.728, .791] R2 = .577 ΔR2 = .435
Trust in the most popular outlet	.607fb [.574, .649] R2 = .454 ΔR2 = .321	.621fbe [.594, .649] R2 = .547 ΔR2 = .404
N	2,203	2,209

Note. Entries are unstandardized beta coefficients with 95% confidence intervals (CIs) in square brackets, model R-squared (R²) and delta R-squared (ΔR²; change in R-squared when adding the focal media trust variable into a model that included only the covariates). In all models, political interest, left-right ideology, mainstream news use, age, education, and gender are included as covariates for control. All coefficients are statistically significant at the p < .0001 level. Two coefficients are significantly different from each other if their CIs do not overlap. Pairwise tests based on the Hotelling–Williams test for the equality of two dependent partial correlations are reported in Appendix E. Results were almost identical. For each dependent variable, coefficients not sharing subscript letters are significantly different from each other according to this test.

Given processes of media diversification and political polarization, and the fact that exposure to news media in contemporary democracies around the world is shaped by political ideology (Strömbäck et al., 2020), it makes sense to ask if political ideology affects which outlets people have in mind when answering survey questions about the media. To investigate partisan differences in general media trust using the specific trust constructs, we ran the same models reported in Table 1 separately for right-wing and left-wing respondents, while adding models predicting general trust using trust in right-wing and left-wing outlets. Results, reported in Appendix G, show that overall, average trust in mainstream media remains a strong (and in three out of four cases, the strongest) predictor of the general media trust indicators. However, right wingers assigned more weight to the least trusted outlet, and to the average of eight mainstream media outlets, when responding to both indicators of general trust. They also assigned more weight to the mainstream outlets they used when responding to the single-item general trust measure. Further interestingly, for both general trust indicators, the coefficients for nonmainstream media (both conservative and liberal) were negative and significant for the right wingers, while parallel coefficients did not differ from zero for the left wingers.

Discussion

Different social psychological phenomena—the representativeness heuristic (Kahneman & Tversky, 1972), the accessibility heuristic (Kahneman, 2003), the negativity bias (Mutz, 1998), exemplification (Zillmann, 1999), and the interpersonal perception Weighted Average Model (Kenny, 2004)—point to different cognitive strategies that people may employ when assessing general constructs (such as “the media”) from impressions of specific ingredients that compose the general construct (specific news outlets). Taken together, our findings shed light on these processes in several ways. First, the fact that both measures of general media trust were more strongly associated with trust in mainstream, in particular popular, outlets (also see Ladd, 2011; Shearer & Mitchel, 2021; Tsfati, 2002), combined with the null and/or negative associations with trust in nonmainstream outlets, suggest that, at least in the current Swedish context, a representativeness heuristic in which prototypical examples define the more general and abstract construct is at play. Second, rather than placing more weight on the least trusted media, as suggested by research on the negativity bias and by Daniller et al. (2017), our Swedish respondents seem to have averaged across all mainstream sources when forming their general evaluations of the news media’s trustworthiness. This is in line with research on the weighted average model of interpersonal impression (Kenny, 2004).

The above findings were rather consistent across political ideologies (see Appendix G) and different levels of incidental news exposure (see Appendix F). Of course, given that conservatives and liberals use different language to understand and describe the world (Holtgraves & Bray, 2022), perceive the news media differently (Coe et al., 2008; Feldman, 2011), and are surrounded by different media ecologies in which the label “the media” may be used differently (Meeks, 2020), the left-wing and right-wing respondents differed in the weight they assigned to the specific trust constructs when they evaluated the media in general. Interestingly, stronger associations between general media trust on the one hand and average trust in mainstream outlets and average trust in used media on the other were observed for the right-wingers compared with the left-wingers. These seemingly contradictory findings, in addition to the finding that larger parts of the variance in general trust were explained for right-wingers, suggest that general trust in media is simply lower among right-wingers. This, in turn, suggests that the different trust constructs have similar values and play a similar, and stronger role in shaping media trust for these right-wing individuals.

We also found that while for both right-wingers and the sample as a whole, trust in both the right-wing and left-wing nonmainstream outlets were negatively associated with general media trust, this was not the case among left-wingers (for them, the association was null). These negative associations are in line with the fact that use of nonmainstream media is inversely related with media trust (e.g., Tsfati & Cappella, 2003). Right-wingers’ trust in ideological outlets is higher than their trust in mainstream media, and the more they trust nonmainstream sources, the more they distrust mainstream sources. Perhaps because both left-wing and right-wing nonmainstream outlets take a very critical stance against mainstream news sources and tend to attack these mainstream sources (Tsfati & Peri, 2006, p. 169), they earn the trust of audiences alienated from mainstream media. Another possibility is that the causal mechanism is reversed and exposure to these criticisms fosters mistrust of mainstream news.

Somewhat in line with Daniller et al’.s arguments (2017), literature on how people assess social conditions and institutions in general, and the “I am doing better than we are” bias (Mutz, 1998), trust in the mainstream media that one uses (M = 5.30, SD = 1.26) was on average higher than general media trust (for the single item measure M = 4.93, SD = 1.22; For the NCS M = 4.01, SD = 1.36; Repeated measures ANOVA F(5, 2130) = 779.47, p < .001). It is noteworthy that this was true in particular for trust in the used mainstream outlets, and that including nonmainstream outlets in the “trust in used media” measure lowered the influence of this trust construct (M = 5.10, SD = 2.82; see Appendix H for full results).

Interestingly, across the board, there were only minor differences in the results between the News Credibility Scale and a general single-item media trust measure. This is noteworthy because single-item measures can be employed more easily and efficiently in longitudinal designs and thus provide an over-time view of public opinion towards media. Had projects such as the GSS or the WVS used a more nuanced measurement (such as the outlet-specific approach utilized here), such over-time comparisons may not have been possible. Our exploration thus points out the benefits of the simple one-item question.

The findings carry implications for our understanding of past findings on the weak association between media trust and news exposure. Particularly, some have suggested (Fawzi et al., 2021) that a mismatch between the object of trust (the “media”) and the news outlets that are included in the “news exposure” questions may account for the weak associations found in previous studies. The current findings suggest that people have in mind more or less the same mainstream outlets that are included in the media exposure questions when they are asked about trust in “media.” If generalizable to other contexts, these findings weaken, if not refute, the “mismatch in question object” explanation. Instead, some other explanations (e.g., trust-unrelated motivations for exposure; see Tsfati & Cappella, 2005) must underlie the weak association between news media trust and news exposure.

The current finding that people seem to average across mainstream sources when they assess general trust in media does not only carry theoretical and empirical implications; it can also be seen as normatively desirable from a democratic perspective. As trust in media and trust in democracy are intertwined (Ariely, 2015; Tsfati & Cohen, 2005, it would be regrettable if citizens’ assessment of this important democratic institution was based on a single bad apple, or on distrust of fringe sources with an ideological positioning that runs in contrast to the recipient’s point of view. If the results from the current study would generalize to other countries with lower levels of trust in media than Sweden, this would also suggest that lower scores on general media trust is not a survey artifact, but that trust in mainstream media is in fact lower.

As with any exploration, the current project is not free of limitations. First, the article was concerned with how people assess a general construct based on assessments of specific ingredients or subcomponents of that construct. While this is a relevant question for public opinion research at large (as public opinion scholars frequently ask respondents to assess general constructs such as the health system, the school system, the Congress), it is unclear whether and to what extent the present empirical findings and subsequent theoretical insights are transferable to assessments of constructs other than “the media.” Second, our data come from Sweden, a relatively media-trusting society, and a context in which media diversification is less extreme than in other countries such as the US (see Andersson, 2018). Contrary to many other countries (but see Hanitzsch et al., 2018), media trust in Sweden is not declining. In the World Values Survey, the number of Swedish respondents expressing “a great deal” or “quite a lot” of confidence in the press has increased from 26.7% to 39.7% between the first (1981–1984) and the most recent (2017–2022) waves. Third, the analyses consist of indirect tests based on survey weights. Of course, no one knows exactly what comes into respondents’ minds when they answer questions about the media. While our findings complement previous studies that have used more direct survey questions (e.g., Ladd, 2011), we must acknowledge that the reverse process (in which people rank the specific mainstream outlets based on their more general evaluation of the media) may be taking place. Even if this is the case, this still implies that it is mainstream—and not ideological—outlets that respondents consider as “the media.” This is the present investigation’s contribution to our knowledge of respondents’ understanding of survey questions concerning “the media.”

Biographical Notes

Yariv Tsfati (PhD, 2001, University of Pennsylvania) is a professor at the Department of Communication, University of Haifa. He studies political communication with a focus on audience perceptions of news media and their implications.

Jesper Strömbäck is a professor in Journalism and Political Communication at the Department of Journalism, Media and Communication, University of Gothenburg. His research focuses on political communication, political journalism, and media effects.

Elina Lindgren (PhD) is a researcher in Political Communication at the Department of Journalism, Media and Communication, University of Gothenburg. Her main research interests are in political communication, public opinion, and survey designs.

Hajo Boomgaarden is a professor for Empirical Social Science Methods at the Department of Communication at the University of Vienna. His research focus is on various aspects of political communication and on computational methods for communication science.

Rens Vliegenthart is a professor of Media and Society at and scientific director of the Amsterdam School of Communication Research (ASCoR) at the University of Amsterdam. His research focuses on the role of media in society, and specifically, the way media and politics interact.

Footnotes

1

Wave 1 of the survey included only the general trust measures and only the specific trust in the eight mainstream news outlets. All analyses based on these limited number of Wave 1 items resulted in the exact same patterns reported in the tables and text in the article and serve as robustness checks. For the sake of parsimony, they are reported in Appendix I.

2

We considered Dagens Industri—a business-oriented newspaper—as a nonmainstream outlet because in other countries both audiences and scholars consider the financial press as remote from the mainstream news media.

3

Interestingly, trust in both right-wing and left-wing nonmainstream outlets was negatively associated with both indicators of general media trust [for the News Credibility Scale, the effect of trust in right-wing media was b = −.002; SE = .001, p = .019; the effect of trust in left-wing media was b = −.002; SE = .001, p = .06; for the single-item measure, the effect of trust in right-wing media was b = −.001; SE = .001, p = .039; the effect of trust in left-wing media was b = −.002; SE = .001, p = .08].

4

Note that after employing Bonferroni correction, the difference between average trust in eight mainstream outlets and trust in the most popular outlet becomes insignificant, as noted by the subscripts in Table 1.

5

As pointed out by a reviewer, trust assessments (both specific and general) can be based on incidental news exposure, which our current analyses do not account for. To address this concern, we re-ran the analytical models separately for individuals that scored high and low on a News Finds Me construct (an indicator of incidental exposure). These analyses (see Appendix F) revealed similar results for individuals scoring high and low on the News Finds Me scale. Particularly, in both cases, average trust in mainstream media was the trust measure most strongly associated with both indicators of general trust.

Funding

This work is part of the research project “Knowledge Resistance: Causes, Consequences, Cures,” funded by Riksbankens Jubileumsfond for the Advancement of the Humanities and Social Sciences [grant number M18-0310:1].

Conflict of Interest

The authors have no conflict of interest do declare.