Exploring Possible Explanations for the Modest Relationship Between News Media Trust and Use

Introduction

A recurring finding in the literature on the relationship between news media trust and news media use is that they are positively associated, but usually in a quite modest or even weak manner (Fletcher and Park 2017; Jackob 2010; Kalogeropoulos et al. 2019; Mourão et al. 2018; Williams 2012; Yuan 2011; for a review, see Strömbäck et al. 2020).

While a weak correlation between news media trust and news use might make sense in low-choice media environments, it is more puzzling in the contemporary high-choice information environment. With an abundance of media options and greater opportunities than ever to avoid distrusted news or replace it with more trusted ones (Strömbäck et al. 2022; Van Aelst et al. 2017), one might anticipate a stronger relationship between news media trust and use. As Park et al. (2025) argue “It is often assumed that distrust in news will turn citizens away from it. However, the relationship between trust and news engagement is not straightforward. Those who distrust news may continue to consume it, and those who have high trust in news may choose to disengage from it.”

Despite the fact that the modest correlation between news media trust and news use has bothered scholars and pundits for decades, empirical research regarding possible explanations for the relationship is limited (Strömbäck et al. 2020). However, recent qualitative studies have made some progress when trying to tackle potential explanations. For example, Panievsky et al.’s (2025) respondents mentioned social and incidental exposure, as well as critical and selective exposure as explanations for attending to untrusted news. Similarly, Štětka and Mihelj’s (2024, pp. 94–112) findings suggest that the low correlation between news media trust and exposure may be explained in part by differences between different types of media trust, especially between generalized media trust and trust in particular news brands. Similarly to Panievsky et al.’s (2025) findings, their results also indicate that while trust is important, it is not always the only or even the primary factor influencing news use. Alongside trust, news use is also influenced by factors such as news habits, unintentional exposure due to contextual factors, and audience desire to access a variety of perspectives (Webster 2014).

Guided by these insights, the purpose of this study is to explore a range of possible explanations for the relatively weak relationship between news media trust and news use. More specifically, we investigate whether the variance in news use explained by media trust increases when shifting the level of analysis to assess trust in specific media outlets and its effects on usage of these corresponding outlets. Drawing on recent qualitative research, we also develop hypotheses and test whether self-reported tendency for selective exposure, news avoidance, News Finds Me perceptions, and conspiracy mindset moderate the association between news media trust and news media use.

Trust in Mainstream Media and Mainstream News Use

Conceptually, audience trust in mainstream news refers to “audience’s perception that relying on mainstream media information is advantageous or at least not detrimental when making decisions” (Tsfati et al. 2025: 17). This applies the definition of trust in general—as the willingness of a trustor to be vulnerable to the actions of a trustee, irrespective of the ability to monitor or control that other party (Mayer et al. 1995)—to the audience’s relationship with news media. The news media cover the world beyond people’s own experiences, and thus it is often extremely complicated for the audience to verify the validity of the information conveyed by mainstream journalists, their assessments, and interpretations (Kohring and Matthes 2007). Given this, when we trust the news media, we allow ourselves to rely on these media to inform our daily decisions (Fawzi et al. 2021).

The expectation that trust in mainstream news media should be linked to mainstream news use is rooted in the idea that people seek reliable information from the news to understand the world and therefore make rather rational choices when selecting their news sources (Tsfati and Cappella 2003). Based on these theoretical premises and the definition of media trust as the expectation of receiving benefits rather than drawbacks from using news media, Tsfati and Cappella (2003) hypothesized and empirically found a positive association between mainstream news media trust and mainstream news exposure. However, the association they detected was surprisingly small. Similar findings have since been found in other studies based on data from different countries (Jackob 2010; Kalogeropoulos et al. 2019; Tsfati 2010) and using diverse measures of media trust and news use (Tsfati and Cappella 2003; Williams 2012).

Several scholars have noted that the modest or low correlation between news media trust and news use is paradoxical (Strömbäck et al. 2020; Swart and Broersma 2022; Szostek 2018). Most often, the fact that people attend to news they do not trust is explained by news consumption being motivated by a variety of factors, many of which may be unrelated to information seeking (Strömbäck et al. 2020: 145; Webster 2014). However, empirical research addressing the question is scant, and most empirical explorations have used qualitative methods (Nelson and Lewis 2023; Štětka and Mihelj 2024; Swart and Marcel Broersma 2022). While the insights brought about by these qualitative explorations are substantial, they offer quite a diverse set of possible explanatory factors, making it difficult to assess the explanatory power of each of these explanations.

To the best of our knowledge, the only moderator that has been tested quantitatively thus far is need for cognition (Tsfati and Cappella 2005), documenting a stronger association among individuals with lower cognitive needs, likely because those with a high need for cognition appreciate thinking about political issues from different vantage points, making it valuable to engage with mainstream news media even when they do not trust them. However, the interaction added only 2 percent to the explained variance in mainstream news exposure (p. 261), suggesting that there might be other explanations for the modest correlation between news media trust and use.

The key question is what are these other explanations? Here, we will investigate two types of explanations identified in recent qualitative research on why people consume news they do not trust (see Figure 1). We start with the habitual patterns with which people navigate contemporary high-choice media environments (outlet specificity, selective exposure, news avoidance, and News Finds Me perceptions)—and then move to the individual cognitive predisposition conspiracy mindset. These explanations will be investigated through quantitative analyses of a large-scale panel survey dataset.

OPEN IN VIEWER

Figure 1.

Conceptual model.

The Habitual Patterns With Which People Navigate the Information Environment: Outlet Specificity, Selective Exposure, News Finds Me Perceptions, and News Avoidance

When asked to explain their use of mainstream news media despite their mistrust in the same, some people clarify that they do trust the specific news outlets they read and watch even if they do not trust mainstream news media in general. As one of the respondents in Paniyevsky et al.’s study (2024) responded: “I am exposed only to news on specific credible channels.” Along the same lines, Štětka and Mihelj (2024) argue that inconsistency in levels of analysis may be a likely explanation for the modest association between news media trust and exposure. Therefore, they argue that it would be beneficial to distinguish “between different levels of media trust, including not just generalized media trust, but also trust in individual news brands and different types of media” (p. 112).

This suggests that the modest correlation between mainstream news media trust and use stems from respondents’ uncertainty regarding the target of trust and use in the survey questions they answer. Similarly, Toff (2022) argues that audiences are increasingly “selectively trusting,” discerning between specific brands. This suggests that a modest correlation between generalized news media trust and use might stem from the fact that people consume specific brands they trust while harboring skepticism toward the broader media. Therefore, lower-level measures of news use and trust (e.g., trust in a specific outlet and use of that outlet) should yield stronger correlations and larger explanatory power. We therefore hypothesize that (H1) the variance in news use explained by media trust will be higher when using outlet-specific measures of use and trust, compared to when using general trust and use measures.

In the previous qualitative studies that have asked respondents why they follow news media they do not trust, some respondents mentioned attending to ideological nonmainstream outlets. For example, one of the respondents in the Panievsky et al study wrote “I only watch Channel 14; All the rest are ‘mind engineers’.” (Channel 14 is an ideological television news channel in Israel). Attending to news outlets congruent with one’s ideological leanings was also uncovered in the Štětka and Mihelj (2024, pp. 70–75) and Nelson and Lewis (2023) studies and mentioned as a strategy to avoid untrusted news, or to corroborate and balance the information provided by mainstream media. This explanation aligns with extensive research on political selective exposure (for a recent review, see Sude and Knobloch-Westerwick 2022). If some audiences turn to politically alternative ideological channels due to their mistrust in mainstream media, in order to (at least partially) replace these mainstream channels, then for those with a tendency for ideological selective exposure, the association between mainstream media trust and mainstream news use should be stronger and their mistrust in media accompanied with lower exposure to mainstream channels. Based on this, we hypothesize that (H2) the tendency for ideological selective exposure should moderate the association between trust in mainstream news media and mainstream news use such that the higher the tendency for ideological selective exposure, the stronger the association.

Several studies have shown that news media use can at least partly be explained by factors beyond individuals’ own choices (see also Webster 2014). For example, Štětka and Mihelj (2024) mentioned contextual restrictions linked to family, household, and working context as a primary explanation for people’s news exposure in spite of their low trust (pp. 97–98). One of their respondents, an 18-year-old Hungarian, said that he “‘sadly’ had to watch the pro-government MTV while visiting his grandparents” (p. 97). In Panievsky et al.’s (2025) study, some respondents argued that mainstream news are everywhere, so it is impossible to avoid them: “It’s unavoidable – it pops up on Facebook and Instagram although I don’t follow news websites,” or “When I listen to the radio there is suddenly breaking news” (p. 13).

Such comments by respondents are linked to two interrelated concepts, namely “News Finds Me” and “news avoidance.” Both concepts are related to the fact that in contemporary media environments, individuals feel surrounded and overwhelmed by news on social media and elsewhere and report various forms of incidental exposure to news online (Boczkowski et al. 2018; Toff et al. 2024). As a result, some believe they no longer need to actively seek out news to stay informed about public affairs. As Gil de Zúñiga and Cheng (2024) explain, News Finds Me (NFM) perceptions include perceptions that they will encounter news and as a result will be informed, combined with the perception that they do not have to actively seek news. Individuals who feel so encapsulated by news should, like respondents in the studies by Panievsky et al. (2025) and Štětka and Mihelj (2024), probably report that they expose themselves to news media they do not trust. Thus, we hypothesize that (H3) NFM perceptions should moderate the association between news media trust and use such that the stronger the NFM perceptions, the weaker the association.

Turning to news avoidance, a bourgeoning literature indicates that different types of news avoidance have very different antecedents, are highly context dependent, dynamic in nature, and do not always coincide with low news exposure (see De Bruin et al. 2024). Our focus here is on one aspect, namely self-reported, intentional news avoidance, that is characterized by “distaste for news that drives individuals to deliberately steer clear of news exposure” (Skovsgaard and Andersen 2022: 1099). If people indicate to actively avoid the news overall, theoretically they are likely to do this also with trusted sources, making the connection between trust and use less strong. In addition, as active news avoidance is negatively correlated with media trust (Toff and Kalogeropoulos 2020), it seems likely that some mistrusting individuals are actually actively trying to avoid mainstream news. As Panievsky et al. (2025) note, some of their respondents “expressed a strong sense of helplessness, describing their experience as feeling trapped in an inescapable news environment. For example: ‘It’s everywhere!’, or ‘You cannot escape the news, although I really try!’” (p. 13) (see also Toff et al. 2024). These respondents were in other words reporting that they are failed news avoiders. This implies that some news avoiders avoid even trusted news sources, while others try to avoid news but run into untrusted sources. Both of these possibilities lead to the hypothesis that (H4) intentional news avoidance should moderate the association between news media trust and use such that the higher the intentional news avoidance, the weaker the association.

Cognitive Predispositions: Conspiracy Mindset

The power of trust in shaping news use rests not only on habitual exposure patterns but also on individual predispositions. One such potential moderator is conspiracy mindset. Conceptually, conspiracy mindset is distinct from beliefs in specific conspiracy theories and refers to a “general propensity to suspect that conspiracies are at play” (Imhoff et al. 2022: 392). As such, it might help explain why some individuals—all else being equal—are more likely than others to explain events or circumstances by means of conspiracies. In Schwarzenegger’s interview study (2023) with German users of alternative media, some thus describe themselves as “awake” and able to detect “the lies” that the “ruling class”—including mainstream news media—are trying to push. Similarly, Noppari et al. (2019) found in their interview study a link between mistrust against mainstream news media and a susceptibility to conspiracy thinking. In addition, several quantitative studies have found linkages between media use and conspiracy mindset (e.g., Strömbäck et al. 2023; Valenzuela et al. 2024). This leads us to expect that a conspiracy mindset leads to a more determining role of trust when it comes to news media use and hypothesize that (H5) conspiracy mindset should moderate the association between news media trust and use such that the higher the conspiratorial mentality, the stronger the association.

Conceptualizing News Use as a Type of News Exposure

Before testing our hypotheses, a conceptualization of our main dependent variable is in order. Following Slater (2004), news exposure can be understood as the extent to which audiences encounter news as a class of media content. According to Potter (2008), exposure differs from contact (i.e., physical proximity to a message) in that it includes concentration on the message and its processing (“attention”). However, different “exposure states” allow for varying forms of processing, ranging from automatic/unconscious processing to self-reflective states where the audience consciously processes the message. The current technological context makes exposure even more complex, varying “from active, deep, and focused to passive, habitual, ephemeral, unconscious, short and superficial” (Neijens et al. 2024: 12).

In this study, we rely on a standard retrospective self-reported measure in which respondents were asked to report their exposure to a list of news outlets during a typical week. Such self-reported measures are widely used in research examining the association between news media trust and news exposure (Jackob 2010; Tsfati and Cappella 2003; Williams 2012). Given the popularity of this measure in existing research, this approach enables us to explain and compare our results with past findings.

It is important to note, however, the limitations of retrospective self-reported measures (Neijens et al. 2024, Ch. 4), such that they tend to capture intentional behavior more effectively than incidental encounters. This is because self-reports rely on respondents’ ability to recall messages, and messages that are processed without intent or with low attention often fail to leave a memory trace. Therefore, while some scholars argue that certain unintentional encounters with media do leave such a trace (Neijens et al. 2024)—especially if these are habitual encounters (e.g., Štětka and Mihelj’s 2024 interviewee watching a pro-government news channel when visiting his grandparents)—and might therefore be covered by retrospective exposure measures—our measure will likely be biased toward intentional use. Hence, we use the term “news use” instead of “news exposure” to emphasize that we refer to this more intentional and attentive form of exposure.

Context

Sweden is commonly classified as a democratic corporatist media system (Hallin and Mancini 2004), characterized by high levels of journalistic professionalism, politically independent news media, and high levels of press freedom (Syvertsen et al. 2014). It is also characterized by relatively high levels of general news media trust compared to other democratic countries (Newman et al. 2023; Strömbäck 2021) and relatively high “selective trust” (trust in certain brands and not in others; Toff 2022). Importantly, all major national media are nonpartisan with research showing no patterns of politically partisan political news coverage (e.g., Johansson and Strömbäck 2023). That said, studies have shown that Swedes tend to believe that the news media are left-leaning (Oleskog Tryggvason 2021), which plausibly can be explained by delegitimizing attacks predominantly from the political right (Strömbäck and Åkerlund 2025) and the hostile media phenomena (Vallone et al. 1985). Importantly, there are also recent studies showing evidence for selective exposure in Sweden (Dahlgren et al. 2019). A more detailed review of the Swedish background can be found in Appendix D in the Supplemental file.

Methods

To test our hypotheses, we rely on secondary analyses of a four-wave panel dataset collected annually between 2020 and 2024. The four-wave panel was drawn as a stratified subsample from a larger standing survey panel of Swedish residents aged 18 and older, administered by the SOM Institute, a survey research infrastructure at the University of Gothenburg, Sweden. Our analytic sample consists of 3,542 respondents who participated in all four waves. The subsample includes both probability- and nonprobability-recruited respondents and was prestratified to resemble the Swedish population in terms of education, age, and sex.

For the main analyses reported below, we pooled the data such that each respondent had separate records for each wave. While news use and media trust were asked in all waves of the study, the only moderators that were asked in all waves of the study were the tendency for selective exposure and news avoidance. NFM was measured only in Waves 1 and 2, and conspiracy mindset was asked in Waves 2, 3, and 4. In each of the models reported, we use all available data for each of the moderators. This resulted in N (number of observations) ranging between N-within = 6,710, N-between = 3,669 (conspiracy mindset model), and N-within = 10,902, N-between = 3,777 (selective exposure model).

Appendix A in the Supplemental file contains details on data collection, interview dates, and response and attrition rates, while the demographic breakdown is available in Supplemental Appendix B, where descriptive statistics for all variables are available. In all multiple-item scales, we used the mean of the items to calculate the scale. Exploratory factor analyses are reported in the appendices.

Results

To begin with, we tested whether the general measure of mainstream news media trust (news credibility trust index) predicted later news media use (as measured by the index of the eight specific mainstream news outlets, see methods section). In line with previous studies, the results show that the association between general mainstream media trust and mainstream news use, controlling for all covariates, was significant in each of the panel waves for both trust indicators (news credibility index and the single item of trust). That was true even when using a more stringent test, in longitudinal models, predicting later news use using previous trust when controlling for previous news use (see results in Table 1, Model M1; See also Supplemental Appendix G). As the results in Table 1 demonstrate, when fitting a random-effects GLS regression model on the pooled data, controlling for covariates and previous use, media trust in previous waves significantly affected later use (b = .071, se = .007, p < .001). These results were replicated when running the same random-effects models using the single-item indicator of media trust (b = .007, se = .001, p < .001; see Table 1, Model M1a). However, consistent with previous research, the effects were quite small and added only slightly (less than 1%) to the explained variance in exposure. To assess potential nonlinear effects, we included a quadratic specification of lagged trust variable (i.e., trust and trust²). The results show that the effect of media trust on subsequent news use is indeed somewhat nonlinear. Increases in trust are associated with higher news use at lower levels, but the marginal effect diminishes and approaches zero near the upper end of the scale (approximately trust ≈ 6 on a 0–7 scale), indicating diminishing returns. However, the improvement in model fit compared to a linear specification is negligible (R² = .573 vs .572). These results are reported in model M2.

OPEN IN VIEWER

Table 1.

News Use by Trust in News Media: General- and Outlet-Specific Relationships.

Predictors = media trust	DVs: Media exposure	N
General trust-use relationship (DV = General mainstream news use)
M1: General news media trust (news credibility index)	.071*** [0.056, 0.085] R2 = .572ΔR2 = .004	11,007
M1a: Robustness test: General media trust—(single item)M2: Nonlinear model	.007*** [0.002, 0.013] R2 = .565ΔR2 = .003Trust: .220*** [0.154, 0.285]Trust2: −.019*** [−0.028, −0.011] R2 = .573	11,13711,007
Outlet specific trust–use relationship(DVs = Outlet specific news use)
M3: Aftonbladet	.109*** [0.077, 0.142] R2 = .716ΔR2 = .001	6,837
M4: Expressen	.002 [−0.000, 0.005] R2 = .689ΔR2 = .000	6,917
M5: Dagens Nyheter	.002 # [−0.000, 0.004] R2 = .705ΔR2=.000	6,803
M6: SvenskaDagbladet	.002* [0.000, 0.003] R2 = .615ΔR2 = .000	6,717
M7: Rapport I SVT	.006*** [0.002, 0.010] R2 = .642ΔR2 = .005	7,277
M8: Aktuellt I SVT	.000 [−0.004, 0.005] R2 = .535ΔR2 = .000	7,159
M9: Nyheterna I TV4	.003* [0.001, 0.006] R2 = .690ΔR2 = .000	7,042
M10: Ekonyheterna iSveriges Radio	.008*** [0.004, 0.011] R2 = .553ΔR2 = .001	7,011

Note. Results are obtained using random-effects GLS regression models, predicting use of news outlets using trust in the outlet in previous wave, and controlling for previous exposure and all covariates (left-right ideology, political interest, age, education, and gender). Entries are unstandardized regression coefficients with 95% confidence intervals 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).

#

p < .10, *p < .05, **p < .01, ***p < .001. Two coefficients are significantly different if their confidence intervals do not overlap.

H1 predicted that the variance explained by media trust in news use will be higher when using outlet-specific measures of exposure and trust, compared to when using generalized trust and use measures. To test this hypothesis, we first estimated a series of eight random-effects GLS regression models predicting use of eight specific mainstream news outlets (all the individual outlets represented in the index used for testing generalized exposure), using previous trust in the same specific eight news outlets, while controlling for previous use and covariates. We then compared the results from these models in terms of coefficient sizes and explained variance to the previous model that predicted general mainstream news use using the general media trust index. The results from all outlet-specific models are presented in Table 1 (models M3–M10). In the table, two coefficients are significantly different when their confidence intervals do not overlap.

As we can see in Table 1, in contrast to the expectation of H1, the coefficient for general media trust predicting general news use was significantly larger compared to the coefficient for trust in a single outlet predicting exposure to that same outlet. The only result that deviated from this pattern was for Aftonbladet (tabloid newspaper), for which there was an overlap in confidence intervals, indicating no significant differences in coefficients and variance explained (see Model M4, Table 1). This finding is in contrast to H1, which predicted larger coefficients for trust in specific outlets. Concurrently, the variances explained in news use were small and comparable in all models predicting outlet-specific use (models M3–M10, Table 1), similarly to the models, predicting general media use (Model M1, Table 1).

Furthermore, looking at the coefficient for the alternative, single-item measure of news media trust predicting general mainstream news media use (robustness test, see Model M2 in Table 1), it was not significantly different than any of the coefficients for specific trust predicting specific outlet use, with the exception of the coefficient for Expressen (Model M4, Table 1). The only result in Table 1 consistent with H1 was for this outlet, in the case of the robustness test. As predicted, the effect of trust in Aftonbladet on subsequent use of this outlet was stronger than the effect of single-item trust in mainstream media on mainstream news use (b = .109, se = .016, p < .001 for Aftonbladet compared to b = .007, se = .002, p < .001 for mainstream news media in general). All in all, and in contrast to the prediction of H1, we conclude that switching the level of analysis to the outlet level compared to the general mainstream news media level did not improve the prediction of news media use by news media trust (H1 not supported).

Our remaining hypotheses focused on four different possible individual-level moderators in the association between mainstream media trust and use. To test for these, we ran a series of GLS random-effects regression models (with respondents nested in waves) predicting consequent news use by levels of media trust, levels of the moderators, and the interaction terms for media trust and the moderator, in the preceding panel wave. In all models, we included previous levels of mainstream news media use and the demographic and political covariates for control. Models 1 through 4 in Table 2 report results for each moderator separately. The different sample sizes reflect that some moderators were measured in all panel waves, while others were included only in certain waves. In Model 5, we present the results when entering all moderators and interaction terms to the model simultaneously. As the only wave that included all moderators was Wave 2, this model predicts Wave 3 news exposure using Wave 2 media trust, moderators, and interaction terms.

OPEN IN VIEWER

Table 2.

Mainstream News Use Predicted by Lagged Media Trust, Lagged Moderators, and the Trust by Moderator Interaction.

	Model 1: Selective exposure	Model 2: News find me	Model 3: News avoidance	Model 4: Conspiracy mindset	Model 5:All
Mainstream news uset−1	.678 (.008)***	.604 (.010)***	.678 (.008)***	.537 (.010)***	.644 (.014)***
Media trustt−1	.047 (.021)*	.018 (.023)	.116 (.023)***	−.016 (.027)	.090 (.080)
Selective exposuret−1	−.064 (.026)**				−.105 (.050)*
News find met−1		−.086 (.028)**			−.072 (.043)
News avoidancet−1			.044 (.015)**		.070 (.031)*
Conspiracy mindset				−.106 (.028)***	−.006 (.038)
Media trust*selective exposuret−1	.007 (.006)				.009 (.012)
Media trust*NFMt−1		.021 (.006)**			.020 (.010)
Media trust*News avoidancet−1			−.008 (.004) *		−.020 (.008)**
Media trust * Conspiracy mindsett−1				.023 (.006)***	.006 (.009)
Ideology	.015 (.004)***	.014 (.006)*	.016 (.004)***	.023 (.006)***	.011 (.008)
Political interest	.083 (.015)***	.092 (.020)***	.086 (.015)***	.165 (.021)***	.138 (.029)***
Sex	−.038 (.020)	−.049 (.028)	−.035 (.020)	−.038 (.031)	−.034 (.040)
Education	−.000 (.005)	.009 (.007)	.000 (.005)	−.009 (.008)	.005 (.010)
Age	.165 (.008)***	.201 (.010)***	.165 (.008)***	.209 (.011)***	.144 (.015)***
Constant	.155 (.117)	.313 (.142)*	−.307 (.114)**	.432 (.168)*	.266 (.362)
R2 (within)	.067	.180	.067	.259
R2 (between)	.874	.780	.873	.771
R2 (overall)	.570	.560	.569	.561	.546
Observations	10,841	7,249	10,828	6,699	3,100
N	3,769	3,701	3,770	3,667	3,100

Note. The results in models 1–4 are based on random-effects GLS regression models with respondent ID as the grouping variable, predicting mainstream news use using lagged media trust, lagged moderators, and the trust*moderator interaction. Entries are unstandardized regression coefficients with standard errors in parentheses. The results in Model 5 are based on regular OLS regression and do not contain within and between R², as it is based on a single survey-wave (i.e., no panel structure). In all models, left-right ideology, political interest, age, education, and gender are included as covariates for control.

*

p < .05. **p < .01. ***p < .001.

H2 predicted that the tendency for ideological selective exposure should moderate the association between trust in mainstream news media and mainstream news media use. As both Models 1 and 5 in Table 2 demonstrate, the interaction between news media trust and the tendency for selective exposure was not statistically significant. H2 is thus not supported.

H3 predicted an interaction between NFM (News Finds Me) perceptions and news media trust such that the association between trust and use would be weaker for those high on NFM. This interaction was significant in both Model 2 (b = .021, se = .006, p < .01) and marginally significant in Model 5 (b = .020, se = .010, p = .051). The effect runs contrary to the expected direction, with higher levels of NFM yielding stronger effects in trust (H3 not supported).

Our fourth hypothesis, H4, predicted an interaction between news avoidance and media trust in predicting news media use, such that weaker associations are expected for those ranking themselves high as news avoiders. The analyses revealed a significant interaction in the anticipated direction, both when tested separately (Model 3, Table 2: b = –.008, se = .004, p < .05) and simultaneously (Model 5, Table 2: b = –.020, se = .008, p < .01). The interpretation of the interaction is presented in Figure 2. As this figure shows, and in line with H4, for individuals with low intentions to avoid the news, there was a stronger association between trust in mainstream news media and mainstream news use, whereas this connection was weaker among those who identified as frequent news avoiders. Hence, H4 is supported.

OPEN IN VIEWER

Figure 2.

Differential impact of media trust on media use based on different levels of news avoidance.

Our fifth, and final, hypothesis (H5) predicted that conspiracy mindset would moderate the association between mainstream news media trust and use, expecting stronger associations for those high on conspiracy mindset. As predicted by this hypothesis, the conspiracy mindset by media trust interaction was significant when the moderator was tested in isolation (Model 4: b = .023, se = .006, p < .001). However, this effect did not hold when tested under control for all the other moderators and their by media trust two-way interaction terms (Model 5: b = .006, se = .009; p > .05), yielding mixed support for our final hypothesis.

Additional Analyses

Lastly, given the significant patterns of interactions for three of the four moderators, we tested for the possibility of higher-order interactions. However, we found no significant three-way interactions. Analyses that split the dependent variable into news use on television, tabloid, and quality newspapers/radio reveals that also here, the news avoidance interaction term is negative in all instances, though only significant for television. Other results are relatively consistent as well (see Supplemental Appendix F).

Discussion

The low correlation between news media trust and exposure has bothered scholars and pundits for years. In this study, we tested several possible theoretical explanations drawn from recent qualitative explorations (Nelson and Lewis 2023; Panievsky et al. 2025; Štětka and Mihelj 2024; Swart and Broersma 2022), using quantitative analysis of panel-survey data. Our results inform the literature by negating most of these explanations. First, the possibility that the low correlations between mainstream news media trust and use stem from the generalized measures of news use and trust did not receive support. Our finding that outlet-specific trust does not outperform generalized trust in predicting news media use seems to stand in contrast to Toff’s (2022) emphasis on the importance of brand-level trust. However, Toff notes that in developed markets with strong public service media, generalized trust often reflects the high status of leading brands. Therefore, for Swedish respondents, “generalized trust” might already be a proxy for trust in their primary, high quality, sources. That is, our respondents perhaps simply differentiate between media evaluations as consumers and as “good citizens.” Their trust in high-quality outlets may not be reflected in their consumption patterns that satisfy needs other than information.

Second, our longitudinal analysis revealed a nonlinear relationship between media trust and subsequent news consumption. Theoretically, this suggests that while overcoming cynicism or deep mistrust is a crucial prerequisite for engaging with mainstream media, exceptionally high trust does not yield a linear surge in usage. It is very likely that at the higher end of the trust scale, factors other than trust (such as news habits, situational encounters with media and informational needs) shape news use. This finding suggests that the characteristically modest linear trust-use association found in prior research may, in part, be an artifact of fitting strictly linear models to data that actually exhibit a nonlinear pattern.

Third, selective exposure did not significantly interact with media trust in shaping mainstream news media use. In other words, contrary to expectations, we did not observe stronger associations for those who engaged in likeminded selective exposure. Fourth, respondents’ NFM perceptions did significantly interact with media trust, but the interaction patterns were in the opposite direction from those hypothesized: Stronger, and not weaker, associations were observed for those ranking themselves high on the NFM scale. It seems that the experience of information abundance that may characterize high-NFM respondents necessitates more, not less, trust-based selectivity. That is, contrary to what they may feel, those high on NFM are faced with so much news that “finds them” that they have to be selective, and this selectivity is guided, among other things, by media trust.

Fifth, H5 concerning conspiracy mindset received partial support. As predicted, the association between news media trust and mainstream news media use was stronger for those high on this trait-like mentality scale. Among this group, only 11.1 percent ranked themselves high (>5 on a 1–7 scale) on media trust. This means that the stronger association is driven mostly by differences between those with low and medium levels of media trust. The stronger correlation in this group means that the high conspiracy mindset drives respondents to avoid distrusted mainstream news media even more than in the other groups. It is noteworthy that, in a way, the stronger correlation in the high-conspiracy-mindset group makes this group seem “rational in being irrational” in the sense that they demonstrate more consistent behavior (a trust—use link). Their (seemingly irrational) tendency to explain and understand the world using conspiracies, could hence also be viewed as a more organized and consistent mindset (Lukić et al. 2019), that is, a belief that secret plots by powerful actors explain a variety of significant social and political events (Douglas et al. 2019). From that perspective, it is no wonder that such a mindset is also characterized by consistency between one’s news media diet and trust toward mainstream media. It is noteworthy, however, that in the simultaneous model, controlling for all other moderators and interactions, the interaction between conspiracy mindset and news media trust was not statistically significant. Because it is difficult to know whether the inclusion of all moderators in a single model overcontrols in a way that takes important variance from a given moderator, or if there is simply no direct effect, this relationship should be explored further in future research.

The only hypothesis receiving full support was H4, which concerned news avoidance. As predicted, the association between news media trust and mainstream news media use was stronger for those reporting low intentions to avoid the news and weaker for those ranking themselves high as news avoiders. This weaker associations can be logically accounted for by two processes— “failing news avoiders” (individuals who try hard to avoid news but still encounter it quite often either incidentally at workplaces or public spaces or due to routine exposure by someone close) —or “successful news avoiders” (trusting news avoiders who succeed in avoiding trusted news). The former (failing news avoiders) explanation seems to be the more probable one, given that the correlation between news avoidance and mainstream news trust was found to be significant and negative in previous explorations and in our data (in Wave 1, r = –.107; in Wave 2, r = –.135; in Wave 3, r = –.149; in Wave 4, r = –.118; in all waves p < .001) and given that it aligns with findings from qualitative explorations in which respondents reported trying to avoid mainstream news but failing.

Some of the evidence, especially the lack of support for H2 and the contradictory patterns for H3, stand in contrast to previous qualitative research. Several explanations can be offered to these differing findings. First, while this is not detailed in all previous qualitative investigations, the different explanations were offered by only some respondents (e.g., only 23.2 percent of respondents mentioned selective exposure in the Panievsky et al. 2025 study, and only 21.6 percent mentioned “forced exposure” which is akin to the NFM mechanism detailed above). While our study has sufficient power to detect even small interactive effects, the fact that some of the mechanisms explained the low correlation between trust and exposure for only part of the sample may make these mechanisms less noticeable in quantitative research.

Second, our moderators might only partly overlap with the concepts raised by participants in the qualitative studies. For example, the tendency for ideologically congruent selective exposure taps trying to avoid incongruent information, but such attempts may differ from successful selective exposure. Prior research shows that self-reported tendency toward congenial information seeking and behavioral selectivity tap strongly related, but not identical phenomena (e.g., Vraga and Tully 2020). In the Swedish context, where behavioral avoidance is relatively low (Sweden is characterized by comparatively extensive and overlapping use of traditional news outlets, Newman et al. 2023; Oleskog Tryggvason 2021), self-reported tendencies may therefore be reflective of motivational orientations toward selectivity rather than actual exposure patterns. In other words, it may be the case that our measures of the moderators do not fully capture the exact processes described by respondents in the qualitative studies or are otherwise plagued by measurement error.

A third possibility is that responses in the qualitative studies reflect how respondents explain their own paradoxical behavior to themselves, which may differ from how they report their news media use in a longer survey questionnaire containing many other variables. If this is the case, responses to the selective exposure and NFM survey items in this study may better reflect individuals’ actual media behaviors. On the other hand, the present results might be influenced by survey-related biases, for example, the selective exposure items might have provided an opportunity for respondents to indicate negative attitudes toward the media, a form of expressive responding.

While obtaining consistent findings across quantitative and qualitative methodologies is desirable and often seen as an ideal, the literature contains many examples where different methods produce contradictory results (Gallivan 1997; Wagner et al. 2012). In the present case of research examining why people appear to consume news they do not trust, the different studies were conducted in different countries with distinct media systems (the qualitative studies in the US, Hungary, the Czech Republic, Poland, Serbia and Israel; the current quantitative study in Sweden). They were also carried out at different times and used different terminologies. Stronger triangulation, and methodological integration, would be possible if these approaches were applied in the same context and time period, and, ideally, to the same respondents. As always, much work remains for future research.

This exploration is not free of limitations. First, previous work shows that the association between news media trust and use is partly reciprocal (Aharoni et al. 2024; Tsfati et al. 2025). The lagged models used here do not fully capture these asymmetrical reinforcing spirals dynamics and assume a causal direction from trust to news use. Second, the use of secondary data limits the analysis, in that we cannot make sure that the measures of the moderators correspond exactly with the responses provided by participants in the qualitative studies. Third, we were not able to test all potential explanations that came up in the previous qualitative studies. For example, some respondents in the Panievsky et al. (2025) study said that they consume news they do not trust given that they need the information provided by the news media. Current conceptualizations do not offer a concept for such a “no-choice exposure,” and there was no measure capturing this in the panel survey data we relied on. Again, this and additional explanations must be tested in future research.

Despite these limitations, the current study advances our understanding of potential factors underlying the low correlation between mainstream news media trust and news media use. Importantly, it negates the hitherto plausible “different levels of analysis” explanation by showing that the correlation between trust in specific outlets and use of these outlets is no stronger than the correlation between the generalized trust and use constructs used in extant literature. This could be seen as somewhat positive news, in the sense that it reassures that the weak correlations seen when using the simpler generalized measures do not seem to be an artifact of these measures being too simple. In other words, this study does not indicate that future studies should switch to the more demanding level of analysis, including a range of specific media outlets, which would be time consuming for survey respondents and potentially affect the quality of the data. The significant role of news avoidance and the partial support for the conspiracy mindset as a moderating variable, furthermore, adds important insights regarding the low correlation between news media trust and exposure that has puzzled scholars for years. Ultimately, scientific knowledge progresses through the accumulation of minute insights that jointly form a broader understanding. From this perspective, this study represents another step in the ongoing journey to understand how audiences navigate a world defined by information abundance and institutional mistrust.

Supplemental Material