Abstract
Online harassment is shown to have democratic consequences by limiting freedom of expression, disrupting public discourse, reinforcing prejudice and hostility, and creating barriers for online participation. However, there has been less attention to how online harassment relates to trust, which is a key component of well-functioning democracies. This study investigates the relationship between experiences with online harassment and different forms of social and institutional (media and governmental) trust. The analyses utilize survey panel data from Denmark and Norway, two high-trust societies. Findings indicate that online harassment, especially group-based harassment, negatively correlates with all forms of trust, with the most significant panel effects observed on governmental trust and trust in social media (Facebook), which are both institutions that regulate speech. The study contributes to the research literature by shedding light on how various types of online harassment influence different aspects of trust.
Introduction
As the online sphere has become a substantial part of the daily lives of most citizens, there is a growing concern about problematic aspects of online communication. Online harassment and hate speech are increasingly visible online and are no longer confined to the dark corners of the Internet (Siegel, 2020). While some of this content stems from organized extremist groups, the majority of online harassment and hate speech appears to be produced by individual Internet users without a clear organizational affiliation (Bliuc et al., 2018). The more severe forms of such content disproportionately target minority groups (Lingiardi et al., 2020; Pew Research Center, 2021), although it is visible to—and affects—a larger audience. As online harassment has become a notable part of the online sphere, it is key to understand how exposure to such content impacts citizens and society at large.
In this article, we are concerned with the potential democratic consequences of online harassment. Online harassment can be said to undermine democratic values such as equality, anti-discrimination, minority rights, and free and respectful exchange of ideas (Waldron, 2012). A growing body of empirical research demonstrates how such content negatively affects support for the democratic system and its institutions (Rekker and van Spanje, 2022; Wichgers et al., 2022), norms for civil interaction (Bilewicz and Soral, 2020), and the exercise of free speech (Fladmoe and Nadim, 2017; Nadim and Fladmoe, 2021). Moreover, online harassment targeting minorities can contribute to reinforce prejudice and hostility (Hsueh et al., 2015; Soral et al., 2018). Consequently, online harassment can challenge central foundations of a democratic society, but the extent to which democratic values are compromised depends on the individual and collective responses of citizens.
Although several studies address the democratic consequences of online harassment, surprisingly little attention has been given to how online harassment can impact on trust (but see Näsi et al., 2015). Trust is central to a well-functioning society, as it fosters social cohesion, facilitates cooperation, encourages civic participation, and enables effective governance (Putnam, 2000). Online harassment, as a negative and unwanted experience, can influence how individuals perceive (the intentions of) others, and thus impact on social trust. It can also shape their perceptions of—and trust in—key democratic institutions, which might be seen as incapable of protecting citizens from harassment. Relying on a two-wave panel survey, carried out in Denmark and Norway in 2020 and 2022, this article examines how experiences with online harassment can impact on citizens’ social and institutional (media and governmental) trust. In other words, we examine the impact of online harassment on the general trust people have in unknown others (social trust), in institutions that facilitate and regulate content of harassment and hate speech (media trust), and in the institutions that design the overarching rules and protect citizens from harm (governmental trust).
More specifically, we pose the following research question: What is the relationship between experiencing online harassment and social and institutional (media and governmental) trust? The longitudinal nature of panel data allows us to analyze how citizens’ levels of trust change between two points in time depending on whether they have had different types of experiences with online harassment. To consider different types—and degrees of severity—of online harassment, we differentiate between group-based and other types of harassment, and between direct and indirect experiences with online harassment.
Denmark and Norway are both characterized as high trust- and consensus-oriented societies (Andersen and Dinesen, 2017; Stenius, 2010). As such, they offer an interesting context for examining competing hypotheses on the relation between online harassment and trust. Initial high levels of trust may reflect a robust societal resilience that can cushion against negative effects of harassment. Conversely, the high levels of trust may also imply a vulnerability, representing a greater potential for loss.
By examining the impact of different types of online harassment experiences on different types of trust, the study allows for a nuanced discussion of how such experiences can influence trust in society, contributing new knowledge about democratic consequences of online harassment. The findings generally indicate that online harassment has a greater effect on institutional trust than on social trust, specifically leading to lower trust in social media, as well as in political and judicial institutions.
Furthermore, the research literature has documented that emotional and behavioral reactions to online harassment and hate speech vary depending on both the type of harassment and the characteristics of the targeted individual (e.g. Boeckmann and Liew, 2002; Herek et al., 2002; Nadim and Fladmoe, 2021). However, less is known about whether these varied effects of online harassment extend beyond individual responses to influence broader democratic factors, such as trust. Indeed, the study corroborates previous findings by demonstrating that group-based harassment has a stronger negative association with trust compared to other, less severe forms of online harassment.
Understanding online harassment
Much online communication and online debate is characterized by hostility, aggression, and personal attacks (Papacharissi, 2004; Quandt, 2018). Accordingly, the Internet is a potential site for negative interactions, which can have adverse consequences both on an individual and societal level. In the research literature and in public discourse, online harassment is used as an overarching concept that covers a range of behaviors—from less severe name-calling and belittling, to more severe forms of threats and hate speech (e.g. Pew Research Center, 2021). Online harassment can contribute to erode norms for civil interactions, hamper freedom of expression, and endanger social cohesion (Bilewicz and Soral, 2020; Frischlich, 2023; Hsueh et al., 2015; Soral et al., 2018). For instance, studies suggest that exposure to online harassment may push people to withdraw from expressing their views in public and make them more reluctant to make themselves visible, thus affecting the conditions for free speech and civic engagement (Barnidge et al., 2019; Henson et al., 2013; Nadim and Fladmoe, 2021).
Studies of online harassment find that victims are most often targeted because of their (political) opinions or because of their identity characteristics, such as ethnic or religious minority background, sexual orientation, or gender (Lingiardi et al., 2020; Pew Research Center, 2021). While the first type of harassment targets an individual in terms of what they think, harassment targeting identity characteristics can be understood as a group-based harassment as it targets individuals in capacity of the social group they belong to. Such group-based harassment is often understood as hate speech (Boeckmann and Turpin-Petrosino, 2002; Gagliardone et al., 2015; Lawrence et al., 1993), though the term “hate speech” is typically reserved for harassment directed at historically marginalized minority groups, including racial, ethnic, religious, and sexual minorities (Boeckmann and Turpin-Petrosino, 2002; Gagliardone et al., 2015; Lawrence et al., 1993).
A central argument in the hate speech literature is that group-based harassment is more harmful than other types of harassment as it attacks a person’s core identity or “difference” (Chakraborti and Garland, 2015: 13). Thus, it triggers the awareness of the vulnerable and exposed position of the target. In line with this argument, studies find that group-based harassment produces stronger emotional reactions in its recipients than do other forms of harassment (Boeckmann and Liew, 2002), and that even less severe expressions of hostility against minorities can be experienced as traumatic (Herek et al., 2002). Furthermore, group-based online harassment does affect not only the direct victim but also the wider community observing the incident. Knowledge of the experiences of others can incite fear (Gelber and McNamara, 2016; Nadim, 2023; Perry, 2001) and shape people’s perceptions of the trustworthiness of other people and of key democratic institutions and their ability to protect citizens against this type of harm. Thus, it is important to examine not only the effects of being a direct target of online harassment, but also of indirect experiences of observing others becoming victim of harassment.
The idea that group-based harassment is more harmful than other types of harassment, is part of the rationale for legislating against hate speech. Both Denmark and Norway 1 prohibit hate speech based on specific groups characteristics, including among other things ethnicity, nationality, religion, and sexual orientation. In both cases, the prohibition concerns hate speech that is expressed publicly or in a way that can reach an audience beyond the individual recipient. Other types of online harassment can also be unlawful, for instance, under the legislation against threats, or offensive behavior. Although the exact legal boundaries might be unclear to the individual recipient, it is fair to expect that experiences of online harassment are interpreted as negative, unwanted and, in many cases, as an experience that one would wish or expect to be protected against—either by law or by the regulations of online platforms. Consequently, such experiences can potentially impact on both social and institutional trust.
Online harassment and trust
Although several studies have addressed democratic consequences of online harassment, few have studied its impact on trust (Näsi et al., 2015). Trust is generally considered a key component of a well-functioning democracy, promoting social cohesion between citizens (Putnam, 2000) and maintaining institutional stability and legitimacy (van der Meer and Zmerli, 2017). Trust is both relational (A trusts B to do X) and situational (affected by the actions of B in a given situation) (Dalton, 2004; Hardin, 2000). To facilitate democratic participation and political consensus, citizens need to have at least some minimum level of trust in the government, the information provided by the media and other actors facilitating public debate, and in their fellow citizens (Warren, 2018). Conceptually, it is common to distinguish between horizontal social trust between individuals and groups, and the vertical trust citizens have in institutions and authorities (Eek and Rothstein, 2005; Weipert-Fenner et al., 2024). Although they might be interrelated empirically, we expect that different theoretical mechanisms explain the relationship between online harassment and different forms of trust.
Social trust
Social trust denotes various forms of trust between individuals or groups, from trust in known others (particularized social trust) to trust in unknown others (generalized social trust) (Uslaner, 2002; Weipert-Fenner et al., 2024). In this article, we are interested in generalized social trust, which in democratic theory is emphasized as a form of social capital (Putnam, 2000) that contributes to reduce dilemmas of collective action (Sønderskov, 2009). Trustful individuals expect others to behave socially desirable, and act accordingly, which reduces transaction costs in economic affairs (Alesina and La Ferrara, 2002), and make societies better at avoiding “social traps,” such as the tragedy of the commons (Goetzmann, 2021; Ostrom, 1998).
The research literature on social capital emphasizes interactions with known others and involvement in formal networks of civic engagement as key in forming generalized social trust because this builds familiarity, reinforces social norms of reciprocity and accountability, and connects diverse groups, fostering solidarity and trust in broader society (Putnam, 2000). However, Freitag and Traunmüller (2009) argue that interactions with strangers are even more important for generalized social trust than interactions with known others, as such interactions contribute to reduce prejudice and stereotypes, and reinforce the belief that most people are trustworthy regardless of familiarity. Conversely, negative interactions with unknown others can strengthen the perception that most people are not to be trusted. Exposure to online harassment constitutes a negative encounter with others and can be experienced as evidence that not everyone acts with good intentions, thereby weakening generalized social trust.
Empirical studies on online harassment and social trust are limited. However, a Finnish study of young people found that exposure to online “hate material” correlated negatively with both particularized and generalized trust toward people in different contexts of life (Näsi et al., 2015). Relatedly, Salmi et al. (2007) found that crime victimization in general is associated with lower levels of social trust.
Thus, as online harassment typically represents a negative experience with unknown others, we understand it as a situational factor that can lead A to put less trust in people in general, and propose the following hypothesis:
H1. Experiences of online harassment negatively impact social trust
Institutional trust
Institutional trust refers to the trust individuals place in institutional actors (Weipert-Fenner et al., 2024). In a democratic perspective, institutional trust is important as a mechanism to reduce societal complexity (Luhmann, 1979). Individuals lack knowledge and overview of important aspects of society, and by placing trust in the information given by institutions, individuals can make informed choices (Siegrist and Cvetkovich, 2000). Whether individuals trust specific institutions, or not, depend on multiple factors, including perceptions of institutions’ competence, integrity, and benevolence (Mayer et al., 1995). Thus, to trust an institution, citizens must conceive it to be competent in its “modus operandi” and perceive it as honest, transparent, and well-intentioned. For instance, studies have shown that trust in political institutions increases when citizens perceive that governments are performing according to expectations, both in terms of their ability to achieve stated goals and whether these goals result in preferred outcomes (van der Meer, 2018).
Given that individuals expect institutions to protect them from negative experiences, being exposed to online harassment can be conceived as evidence of a failure of institutions to do what is expected from them (cf. van der Meer, 2018). This can affect the general levels of trust individuals have in both media and governmental institutions. Through ethical and editorial rules, traditional and social media institutions have the regulatory power to decide what is published and to regulate the presence of online harassment. Governmental institutions, on their part, design and enforce the overarching judicial framework regulating unlawful content through, for instance, hate speech legislation. Thus, exposure to online harassment can affect the general trust in the platforms that convey such messages, as well as in the political and judicial authorities that design and enforce the legislation that regulates online harassment. Experiencing online harassment could be perceived as an indication that media and governmental institutions are unable to protect citizens from such harm, and can consequently shape people’s general sense of trust in these institutions.
In summary, we are concerned with how experiences with online harassment affect people’s general levels of institutional trust, and propose the following hypotheses, distinguishing between trust in traditional and social media (media trust), and between political and judicial trust (governmental trust):
H2a. Experiences of online harassment negatively impact trust in social media.
H2b. Experiences of online harassment negatively impact trust in traditional media.
H3a. Experiences of online harassment negatively impact trust in political institutions.
H3b. Experiences of online harassment negatively impact trust in judicial institutions.
Group-based harassment and group differences
In addition to the general relationship between online harassment and various forms of trust, we are interested in the significance of the type of online harassment one is exposed to. As the research literature suggests that group-based harassment is more harmful than other types of harassment (e.g. Chakraborti and Garland, 2015), we expect that the effects of online harassment on trust are stronger when individuals are exposed to group-based harassment compared to other types of harassment. Group-based harassment is likely to be understood as a particularly negative experience with (unknown) others as it is an attack on a person’s core identity. Furthermore, group-based online harassment represents an example of how media and political institutions have failed to protect citizens from harm that can border on being unlawful hate speech. Consequently, we pose the following hypothesis:
H4. The impact of online harassment on trust is stronger among individuals who have been subject to group-based harassment compared to individuals who have been subject to other forms of harassment.
Finally, we expect there to be group differences in the extent to which online harassment impacts on trust. While a detailed theorization of group differences is beyond the scope of this article and no specific hypotheses are proposed, we outline a few expectations informed by previous empirical studies. Studies have shown that women react more strongly than men to experiences of online harassment (Nadim and Fladmoe, 2021), and that women are more likely than men to support stronger regulations on harassing content (Wollebæk et al., 2022). In sum, this leads us to expect a stronger effect of online harassment on trust among women than among men. Regarding age, studies have shown that young adults are much more exposed than older adults to online harassment (Pew Research Center, 2021), and that they generally express higher levels of stress, anxiety, and depression (Varma et al., 2021). Combined, these findings lead us to expect that online harassment has a stronger impact on younger individuals than on older ones.
Data and methods
We rely on survey panel data, carried out by means of a web-questionnaire in Denmark and Norway in December 2020 (t1) and May-June 2022 (t2). The data originates from a larger research project (Far right politics online and societal resilience [FREXO]) that studied the two countries, and by including both, we achieve increased statistical power, allowing for more nuanced analysis. We do not expect the main relations to vary between the countries, but take account of potential differences in the statistical models.
The survey sample was drawn from Kantar’s pre-established online panel and was stratified to resemble the adult online 2 population in the two countries. 3 Recruitment to Kantar’s online panel is based on random sampling and invitation only (not self-recruitment). About 4000 respondents participated in wave 1 (~2000 in each country), of which about half also participated in wave 2. Removing respondents with missing values on any variable used in the analyses leaves us with a panel sample of 1657 respondents, 748 from Denmark and 909 from Norway. Attrition rates from wave 1 to wave 2 were highest among women and the youngest respondents (see Table 1). Given that young adults are more likely than the oldest age cohorts to be targets of online harassment (see Table 3), the effects estimated in the analyses might be conservative. We therefore also estimate effects in different gender and age groups (reported in Supplementary Material, Table A3).
Descriptive statistics—original sample (wave 1) and panel sample (wave 1–wave 2).
Dependent variables: social, media, and governmental trust
We measure social trust by means of the standard generalized trust item “Most people can be trusted vs can’t be careful enough.” We also utilize a second item asking about general trust in people living in the country (“Generally, how much do you trust different groups of people?—People living in Denmark/Norway”). Both questions had response scales ranging from 1 (can’t be careful enough/no trust at all) to 7 (most people can be trusted/complete trust).
To measure media trust, we distinguish between trust in social and traditional media. Trust in social media is measured through a single item on trust in Facebook (“How much or little trust do you have in . . . Facebook” 4 ). Ideally, we would like to measure several social media platforms, but trust in Facebook was the only item included in the survey. This limits generalizability to other platforms; however, Facebook was by far the most commonly used platform among the respondents across gender and age groups. 5 Trust in traditional media is measured through a battery of items on trust in the national broadcaster (DR, NRK), the main tabloid newspaper (BT, VG), and the main broadsheet newspaper (Politiken, Aftenposten) in Denmark and Norway, respectively (“How much or little trust do you have in the following news sources?” [list of specific outlets]). Trust levels between these three types of outlets are more strongly correlated in Norway (Cronbach’s alpha t1 = .82) than in Denmark (Cronbach’s alpha t1 = .63), primarily because trust in the leading tabloid newspaper is significantly higher in Norway (mean score t1 = 4.3, 1–7 scale) than in Denmark (mean score t1 = 3.5). The index measuring trust in traditional media was constructed by combining mean scores on the single items.
Finally, to measure governmental trust we distinguish between political and judicial trust, which both are indices consisting of two single items (“How much or little trust do you have in . . .” [list of institutions]). The political trust index was constructed by combining mean values on trust in the government and trust in the parliament (Cronbach’s alpha t1 = .85 [DK], .90 [NO]), and the judicial trust index was constructed based on mean scores on trust in the police and in the courts (Cronbach’s alpha t1 = 0.83 [DK], 0.85, [NO]).
Independent variables: experiences with different types of online harassment
The main independent variable is experiences with online harassment, which is a categorical variable distinguishing between different types of experiences, both in terms of the severity of the message and in terms of what the harassment targets. In the survey, respondents were first asked whether they themselves, or someone they knew, had been exposed to unpleasant comments, hateful comments, or violent threats online. Those responding “yes” were asked whether these comments were targeting one or more group-based characteristics (gender, sexual orientation, religion, ethnicity, disability, skin color, nationality, gender identity) or non-group-based characteristics (appearance, occupation, political opinion, other, don’t know). Respondents could choose all characteristics that applied.
We generated a single variable with four values, from the most- to least severe experiences: those who had experienced what they perceived as (1) harassment (hateful comments or violent threats) directed at one or more group-based characteristics (group-based harassment), (2) harassment (hateful comments or violent threats) directed at other (non-group-based) characteristics (other harassment), (3) those who had experienced unpleasant comments online (unpleasant comments), and finally, (4) those who had not themselves been targeted, but who knew someone who had experienced online harassment (friends targeted). Thus, respondents with the value “1” could theoretically have experiences with all types of harassment, while those with the value “4” would exclusively have experiences with friends being targeted.
Table 2 sums up the share of respondents in each wave reporting that they had experienced the various forms of online harassment. The table shows that only a small share of the respondents had experience with group-based online harassment (t1: 2.3%, t2: 3.6%). The most targeted group characteristic in each wave was gender (0.9/1.6%), while the most common non-group characteristic was political opinion (2.9/3.6%).
Experiences with online harassment—panel sample (percent).
Less severe forms of online harassment were more common, with 5% (t1) and 6.5% (t2) reporting other forms of harassment and more than 7% reporting unpleasant comments. More than 18% reported that they knew someone who had been targeted (but had not been targeted themselves). About two-thirds of the respondents reported that they had no experience—neither personal nor through friends—with online harassment. While not directly comparable due to different question wordings, these numbers suggest that online harassment is less widespread in Denmark and Norway compared to the United States, where data from Pew Research Center show that, in 2021, about 40% had experienced “any type of online harassment,” and as many as 25% had experienced a “severe” form of online harassment (Pew Research Center, 2021).
Analytical strategy
The analytical strategy is threefold. First, we explore descriptive statistics on the relationship between experiences with online harassment, various forms of trust, and gender and age. Second, we analyze the cross-sectional relationship between experiences with online harassment and various forms of trust. Third, we estimate panel fixed-effects regression models, analyzing whether changes in experiences with online harassment correlate with changes in trust (Rabe-Hesketh and Skrondal, 2008). To explore differential effects, we repeat the models over gender and age. Unfortunately, we did not have information on the respondents’ immigrant background. Since our panel data are limited to two waves over a 1.5-year period and we lack information on prior experiences, the panel analysis does not definitively establish the causal relationship between online harassment and trust. However, when combined with cross-sectional analyses, it nevertheless represents a meaningful step forward from correlation to causality.
Results
Descriptive analysis
Table 3 shows the share of different groups of respondents who had experienced different forms of online harassment. Gender differences in experiences of group-based harassment are small, and both genders report higher levels at t2 than t1. Men are more likely than women to have experienced other forms of harassment, and this share increased from t1 (5.8%) to t2 (8.1%). Regarding less severe experiences (unpleasant comments and having friends targeted), there is a decline among men and an increase among women from t1 to t2. Respondents aged 18–49 are more likely than older respondents to be targets of group-based and other forms of harassment. Across all age groups, there is an increase from t1 to t2, with the largest increase being the youngest adults’ (18–34 years) experiences with other forms of harassment. The younger cohorts were are also more likely than older respondents to have friends that have been targeted, which could reflect that young adults spend more of their social life online. Combining all types of online harassment measured in the survey, about half of the young adults report having some kind of experience with online harassment, either personally or through friends. This contrasts the oldest age cohort (70+), where the equivalent number is 26–27%. Finally, respondents in Norway are more likely to report experiences of group-based harassment, while respondents in Denmark are more likely to report other harassment. In both countries, there was an increase in both forms from t1 to t2.
Experiences of online harassment in different groups—panel sample (percent).
Table 4 shows mean trust levels across the same groups. Respondents express high levels of social trust (generalized trust, trust most people) and governmental trust (political and judicial trust), as well as high levels of trust in traditional media. Trust in Facebook is much lower. With a few exceptions, gender and country differences are small, but older respondents have higher levels of social and political trust compared to younger respondents. Women report somewhat higher levels of political trust than men, but across both genders there were a small decline from t1 to t2.
Social, media and, and institutional trust in different groups—panel sample (mean scores [scales 1–7]).
Cross-sectional analysis
Figure 1 displays results from several ordinary least squares (OLS) regressions of different forms of trust, explained by online harassment at t1 and t2, controlling for gender, age, education, and country (see Table A1, Supplementary Material, for complete regression tables). The reference category (vertical lines) is “no experience” with online harassment, with deviations to the left of this line indicating negative relationships between harassment and trust. Overall, these models show a clear pattern in which more severe forms of harassment are more strongly and negatively related to trust than less severe forms, at both t1 and t2. Experiences with group-based harassment are negatively related to all forms of trust. Experiences with other harassment are negatively related to trust in traditional media, political trust and judicial trust. Experiences with unpleasant comments are negatively related to traditional media at t1 and political trust at t2, while indirect experiences with online harassment (knowing someone who has been targeted) are generally unrelated to trust.
Relationship between online harassment and trust. Plots from OLS regressions.
Country-specific analyses show similar patterns; however, in Denmark, the relationships tend to be weaker at t2 than at t1 (see Table A1).
In sum, the cross-sectional analysis gives indicative support for H1–3 (experiences of online harassment are negatively related to social, media, and governmental trust), and H4 (these relationships are stronger among individuals who have been subject to group-based harassment). The causal pattern is, however, unclear. Online harassment may reduce trust, but it could also be that less trustful individuals are more likely to report (whether accurately or not) being victims of online harassment. To reduce problems of selection bias and move closer to identifying a causal relationship between experiences of online harassment and the various forms of trust, the next section presents results from panel fixed effects models where changes in harassment experiences from t1 to t2 are correlated with changes in trust levels, controlling for unobserved factors.
Panel fixed effects analysis
The results of the fixed effects estimations are displayed in Figure 2 (Table A2, Supplementary Material, reports full regression tables). The effects are weaker in these models than in the standard OLS models, suggesting that selection bias is part of the explanation for the negative relationships between online harassment and trust. In other words, less trustful individuals appear to be more likely to report experiences of online harassment than more trustful individuals. Nevertheless, the fixed effects estimations also suggest that experiences with online harassment can (further) reduce trust. The main effect of experiences with group-based harassment are reduced political trust (b = −324, p = 0.070), judicial trust (b = −0.319, p = 0.018), and trust in Facebook (b = −300, p = 0.061). Less severe experiences with online harassment (other harassment, unpleasant comments, friends targeted) are not significantly related to trust in the panel analyses.
Relationship between online harassment and trust. Plots from panel fixed effects models (wave 1–wave 2).
Overall, this analysis suggests that the most robust finding is that online harassment reduces governmental trust, supporting H3a and H3b. Media trust is also impacted, but the main effect is reduced trust in social media, supporting H2a. Furthermore, H4 (group-based harassment has stronger effects than other types of harassment) is supported for both measures of governmental trust and for trust in social media. Experiences with online harassment do not impact social trust nor trust in traditional media, rejecting H1 and H2b.
Country-specific analyses suggest that the panel effects are virtually similar across the two countries for judicial trust and trust in Facebook, while the effect on political trust is stronger in Norway (b = −0.445, p = 0.095) than in Denmark (b = −0.186, p = 0.422) (see Table A2).
Finally, some of the effects are unequally distributed across social groups (see Table A3 in the Supplementary Material for repeated panel fixed effects models across gender and age groups). While the negative panel effect of group-based harassment on judicial trust is nearly identical for men and women, the effects of group-based harassment on political trust and trust in Facebook are much stronger among women than among men. Women are also more likely than men to express lower levels of political trust and trust in Facebook after experiencing less severe forms of online harassment.
We also expected that young adults would react more strongly than older adults in terms of reduced trust, but the analyses suggest that differential effects related to age are often non-linear. Young adults (aged 18–34) are most likely to express reduced trust in Facebook, but older age cohorts are more likely to express reduced governmental trust. Regarding Facebook, the youngest cohort constitutes the most “digitally native” in the sample, that is, they have grown up surrounded by digital technology (Evans and Robertson, 2020), and one may speculate if this could lead them to hold higher standards to social media platforms (Schoenebeck et al., 2021). However, very few in the older age cohorts had experienced the most severe forms of online harassment (see Table 3), and the effect differences between young and old adults must be interpreted with caution.
Discussion and conclusion
A growing body of research addresses how online harassment can have democratic consequences, suggesting that online harassment undermines democratic values by limiting freedom of expression, disrupting public discourse, reinforcing prejudice and hostility, and creating barriers for online participation (e.g. Bilewicz and Soral, 2020; Hsueh et al., 2015; Nadim and Fladmoe, 2021; Soral et al., 2018). However, there has been less attention to how exposure to online harassment might impact on citizens’ trust in other people and institutions. Trust is generally considered a key component of a well-functioning democracy. A basic level of trust in the government, the media, and other actors that facilitate public debate, as well as in fellow citizens, is essential for a well-functioning democracy (Warren, 2018). Thus, examining the impact of online harassment on trust contributes important knowledge on the potential democratic consequences of online harassment.
Relying on survey panel data from Denmark and Norway, this article has examined how experiences with online harassment are associated with trust in unknown others (generalized social trust), trust in platforms that facilitate and regulate content of harassment (media trust), and trust in the institutions that design the overarching rules and protect citizens from harm (governmental trust). Furthermore, as online harassment can encompass a range of different experiences, of varying severity, we distinguished between (1) harassment and violent threats targeting group characteristics, (2) harassment and violent threats targeting other (non-group) characteristics, (3) unpleasant comments, and (4) having friends who have been targets of online harassment (without having been targeted oneself).
We hypothesized that experiences of online harassment would negatively impact (H1) social trust, trust in (H2a) social and (H2b) traditional media, and trust in (H3a) political and (H3b) judicial institutions. Furthermore, we hypothesized (H4) that the effects would be strongest among individuals who had been subject to group-based harassment. We found support for H2a, H3a, H3b, and H4, while H1 and H2b were not supported. Although online harassment involves a negative experience with other people, it does not appear to diminish generalized social trust. Furthermore, online harassment did not impact on trust in traditional media. In both instances, the cross-sectional analyses suggested negative relationships, which turned out to be insignificant in the panel analyses.
Thus, the main conclusions from the analyses are that, first, in line with the key argument in the hate speech literature (e.g. Chakraborti and Garland, 2015), harassment targeting someone’s group identity appears to have more severe consequences than harassment targeting someone as an individual. Second, experiences with online harassment have a stronger impact on institutional trust than on social trust. Both governmental and social media actors are in various ways responsible for regulating and moderating online content. Online harassment could be perceived as an example of institutional failure, in that these actors have failed to protect individuals from online harm, either through content moderation (in the case of Facebook) or by failing to design and enforce overarching judicial frameworks that effectively regulate harmful content, for instance, through hate speech regulation (in the case of governmental institutions).
Regarding trust in social media, the panel effects were limited to women and young adults (<34 years). This corroborates previous studies finding that women react more strongly to online harassment (Fladmoe and Nadim, 2017; Nadim and Fladmoe, 2021) and have lower acceptance of harassing speech than men (Wollebæk et al., 2022). Young adults constitutes the most “digitally native” in the sample, that is, they have grown up surrounded by digital technology (Evans and Robertson, 2020), and one may speculate if this could lead them to hold higher standards to social media platforms (Schoenebeck et al., 2021). Interpreted in line with the hate speech literature, online harassment may also trigger a stronger sense of vulnerability among women and young adults, accentuating their historically inferior position in society (Perry, 2001). This may reinforce their perception that social media institutions, such as Facebook, are not performing according to expectations in regulating online harassment and protecting citizens from such negative experiences—which in turn can affect their general trust in social media.
Our analyses indicate that self-selection biases are significant, as individuals with initial low levels of trust were more likely to report having been targets of online harassment than those with higher levels of trust. Regarding social trust, our theoretical argument was that negative situational experiences online would lead to reduced trust in unknown others. Our findings instead suggest that individuals who lack social trust are more likely to find themselves in online spaces where harassment is present or that they are more inclined to interpret online content as harassment. One possible explanation, which links low social trust with low trust in traditional media, is that these individuals are more likely to consume online content from hyperpartisan, alternative media sources (Andersen et al., 2023; de León et al., 2024), which are characterized by hostile media criticism and anti-democratic positions that promote distrust (Ihlebæk et al., 2022).
The overall picture from our findings is that individual experiences with online harassment is a relatively limited phenomenon in Denmark and Norway, as less than 1 out of 10 respondents had been targets of the most severe forms of harassment. Still, we found that almost half of the young adults in the sample had some form of experience with online harassment, either personally or through friends. This underlines that some groups are particularly exposed to online harassment, especially digital natives who spend a substantial part of their social lives online. Given that young adults continue to live much of their lives online as they grow older, this means that the democratic consequences of online harassment could become more severe in the future.
This study has some important limitations, which should motivate further research. First, the panel data only span two waves over a relatively short time and do not contain information on respondents’ prior experiences or frequency of exposure, limiting the possibility to draw strong causal conclusions. Future research should aim to cover a longer time-period with more survey waves, and include even more nuanced items of experiences with online harassment. Second, our sample represents the general population, and we lack information on minority characteristics, which is unfortunate as we know that minority groups are more likely than majority groups to be targets of online harassment. Thus, to fully capture the consequences of online harassment across different groups, future studies should include larger samples of minorities. Third, we have studied a broad conception of trust. More targeted measures of the perception of the performance of institutions in protecting citizens from online harassment can yield further knowledge about the mechanisms through which online harassment impacts on trust. Fourth, our sample only includes two, rather similar, countries. To reach a more general conclusion on the relationship between online harassment and trust, we need data from more diverse contexts. Fifth, we used a single survey item of trust in Facebook as a proxy for trust in social media, and we do not know whether our findings travel to other social media platforms. The advantage of measuring trust in Facebook is that it was the most used social media platform across the groups studied, but we cannot rule out that online harassment would affect trust in other social media platforms differently.
Still, we believe that Scandinavia, as a high trust- and consensus-oriented region (Andersen and Dinesen, 2017; Stenius, 2010), offers a “tough test” for examining the impact of online harassment on trust. While its high-trust context may serve as a buffer against the negative effects of harassment, it could also create a unique vulnerability, as there is more to lose when trust levels are high. Understanding how online harassment affects various forms of trust is therefore essential, not only for recognizing its potential to challenge core democratic values but also for addressing its broader consequences in an increasingly digital world.
Supplemental Material
sj-docx-1-nms-10.1177_14614448251401170 – Supplemental material for The impact of online harassment on trust: Survey evidence from Denmark and Norway
Supplemental material, sj-docx-1-nms-10.1177_14614448251401170 for The impact of online harassment on trust: Survey evidence from Denmark and Norway by Audun Fladmoe and Marjan Nadim in New Media & Society
Footnotes
Acknowledgements
Previous versions of this paper have been presented at the “Politics Seminar” at the Institute for social research and at the Nordic Political Science Congress (Bergen, 25–28 June 2024). We gratefully appreciate all comments from participants at these events.
Funding
The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This research was funded by the Research Council of Norway, grant no. 295985.
Ethical approval and informed consent statements
Anonymized data were retrieved from a commercial web-panel, where respondents have provided informed consent to receive this type of survey. Consent is regulated in a separate agreement between the survey company (Kantar) and the respondents, in accordance with rules set by The Norwegian Data Protection Authority (DPA), which secures informed consent from the individual respondents, as well as the possibility to withdraw information after completing the survey. The data file was anonymized by Kantar.
Data availability statement
Data will be made publicly available by the end of the project (2025), through Sikt—Norwegian Agency for Shared Services in Education and Research.
Supplemental material
Supplemental material for this article is available online.