Introduction
According to conventional wisdom representative democracies require high levels of public trust in political institutions (cf. Crozier et al., Reference Crozier, Huntington and Watanuki1975; Pharr et al., Reference Pharr, Putnam and Dalton2000; Dalton, Reference Dalton2004). Low political trust is argued to affect electoral behavior, driving citizens away from mainstream politics by either abandoning the democratic process or voicing discontent by supporting anti-establishment parties (Norris, Reference Norris2002; Mair, Reference Mair2013). This link between political distrust and two specific forms of electoral behavior – non-voting (Bélanger & Nadeau, Reference Bélanger and Nadeau2005; Grönlund & Setälä, Reference Grönlund and Setälä2007; Hadjar & Beck, Reference Hadjar and Beck2010; Hooghe & Marien, Reference Hooghe and Marien2013) and voting for anti-establishment parties, particularly those on the radical right (Miller & Listhaug, Reference Miller and Listhaug1990; Mudde, Reference Mudde2007; Hobolt & Tilley, Reference Hobolt and Tilley2016) – has been well-documented. Whereas the former undermines the principle of equal representation (Lijphart, Reference Lijphart1997), the latter often challenges democratic norms and fundamental principles of liberal democracy (Učeň, Reference Učeň2007; Rovira Kaltwasser & Taggart, Reference Rovira Kaltwasser and Taggart2016). As such, political distrust may carry both direct and indirect negative consequences for democratic stability.
Despite an abundance of supporting empirical evidence, the precise nature of the relationship between political distrust and electoral behavior remains elusive. While most existing studies are based on the central distinction between low and high levels of trust (e.g., Miller & Listhaug, Reference Miller and Listhaug1990; Norris, Reference Norris2002; Grönlund & Setälä, Reference Grönlund and Setälä2007; Ivarsflaten, Reference Ivarsflaten2008; Hooghe & Marien, Reference Hooghe and Marien2013; Van Kessel, Reference Van Kessel2015), its dynamic component is underdeveloped. Yet, democracy is argued to be undermined not by low trust as such, but rather by persistent, long-term distrust that remains unaffected by the objective performance of institutions (Levi, Reference Levi, Braithwaite and Levi1998; Norris, Reference Norris2022). This suggests we need to focus not only on levels of trust, but also on its long-term variability (cf. Norris, Reference Norris2022) and its short-term variation (cf. Voogd et al., Reference Voogd, van der Meer and van der Brug2019). Integrating the distinction between levels, variability, and temporal changes in political trust is important as declining or low levels of trust by themselves should not be the primary concern for democracy. In fact, the resilience of representative democracy has been argued to rely on vigilant and alert citizens (Levi, Reference Levi, Braithwaite and Levi1998; Lenard, Reference Lenard2008). Variable low trust resulting from skepticism can therefore be likened to a form of “constructive criticism” that stimulates the accountability of incumbents in a way that stable distrust does not (Devine et al., Reference Devine, Gaskell, Jennings and Stoker2020). Rather, stable distrust is often posed to be the significant threat (Citrin & Luks, Reference Citrin, Luks, Hibbing and Theiss-Morse2001; Van der Meer & Zmerli, Reference Van der Meer, Zmerli, Zmerli and van der Meer2017; Norris, Reference Norris2022).
While the theoretical assumptions of existing empirical work often do involve an aspect of trust stability or change (e.g., Citrin & Luks, Reference Citrin, Luks, Hibbing and Theiss-Morse2001; Fieschi & Heywood, Reference Fieschi and Heywood2004; Denemark & Bowler, Reference Denemark and Bowler2002; Lubbers et al., Reference Lubbers, Gijsberts and Scheepers2002), this dynamic component is rarely tested explicitly. One reason why studies have been unable to distinguish systematically between the three components of political trust (level, variability, and change) is their reliance on cross-sectional data or very short-ranged panel data. Hence, to gain a deeper understanding of how political distrust influences electoral behavior, a multidimensional approach to the operationalization and theorization of political trust is necessary.
This paper provides a systematic test of the influence of individuals’ levels, variability, and temporal variation in political trust on both vote abstention and voting for anti-establishment parties. This requires data on individuals’ general level of political trust and the variability of their political trust over time, as well as within-person changes from year to year. Because direct measures of dispositional and evaluative forms of (dis)trust have not yet been developed, only long-term panel data allow us to identify different types of (dis)trusters. For this purpose, we rely on all fifteen available waves of the Dutch LISS panel data (Scherpenzeel & Das, Reference Scherpenzeel, Das, Das, Ester and Kaczmirek2010), collected between 2008 and 2023 (N = 6455 individuals). Unlike typical panel studies primarily designed to identify causal relationships, we leverage the panel structure in two complementary ways. First, we use the repeated observations to develop a typology of trust types based on individuals’ average level and temporal variability in trust over an extended period of time. Second, we exploit the within-person variation in trust to examine how changes in political trust relate to changes in electoral preferences over time. The multinomial logistic random effects within-between (REWB) model (Bell et al., Reference Bell, Fairbrother and Jones2019) enables us to differentiate between the effect of trust levels, variability, and changes.
Our argument is structured around three main points. First, building on trust levels and variability, we theorize and develop a typology of theoretically relevant and empirically distinct groups of stable/variable political (dis)trusters. Second, among the three trust components, static trust levels emerge as the strongest predictor of electoral behavior. Stable and variable distrusters exhibit no substantial differences in their likelihood to support anti-establishment parties or to abstain from voting. Finally, there is an independent effect of declining trust that drives citizens away from established politics, irrespective of their static level and variability of trust.
Theory
A conceptualization of political trust
Political trust captures the extent to which citizens believe political actors and institutions will operate in accordance with one’s own interests and expectations in the absence of constant scrutiny (Citrin & Stoker, Reference Citrin and Stoker2018). Insofar as political trust captures a diffuse form of support, it can be likened to a buffer of positive sentiments toward the political system which allows citizens to give political actors “the benefit of the doubt” when faced with unfavorable institutional outcomes (Easton, Reference Easton1965; Citrin & Luks, Reference Citrin, Luks, Hibbing and Theiss-Morse2001). Yet, it also captures elements of specific support, as it tends to respond to political events and (changes in) institutional performance (Hetherington, Reference Hetherington1998; Bowler & Karp, Reference Bowler and Karp2004; Van der Meer, Reference Van der Meer and Uslaner2018).
The effect of political trust levels on electoral behavior
Political distrust has been widely recognized in the literature for its electoral implications, typically manifesting in two specific behaviors. First, distrusting citizens are systematically less likely to turn out in elections than their trusting counterparts (Citrin & Stoker, Reference Citrin and Stoker2018). This alienating influence of distrust has been a longstanding concern among political scientists (Finifter, Reference Finifter1970; Crozier et al., Reference Crozier, Huntington and Watanuki1975; Mair, Reference Mair2013). Low trust leads citizens to believe that their vote will be futile, discouraging them from turning out in elections. Ample evidence has linked lower levels of political trust with a greater propensity to abstain from voting (Bélanger & Nadeau, Reference Bélanger and Nadeau2005; Grönlund & Setälä, Reference Grönlund and Setälä2007; Hadjar & Beck, Reference Hadjar and Beck2010; Hooghe & Marien, Reference Hooghe and Marien2013). Building on this literature, our first hypothesis is:
Hypothesis 1: Individuals with low levels of political trust are more likely to abstain from voting than individuals with high levels of political trust
At the same time, political distrust is linked to support for parties that seek to destabilize the established political order (Citrin & Luks, Reference Citrin, Luks, Hibbing and Theiss-Morse2001; Bélanger & Nadeau, Reference Bélanger and Nadeau2005). According to critical citizenship theory, distrust can motivate civic engagement, including protest voting (Norris, Reference Norris2002; Rosanvallon, Reference Rosanvallon2008). When mainstream parties fail to address citizens’ concerns, distrusting voters may abandon them in favor of anti-establishment alternatives, both to penalize established parties and to express their desire for change (Norris, Reference Norris2002; Hobolt & Tilley, Reference Hobolt and Tilley2016).Footnote 1 The observation that citizens with low political trust are more likely to opt for anti-establishment parties, particularly on the radical right, has been a consistent finding in the empirical literature (e.g., Denemark & Bowler, Reference Denemark and Bowler2002; Lubbers et al., Reference Lubbers, Gijsberts and Scheepers2002; Bélanger & Nadeau, Reference Bélanger and Nadeau2005; Mudde, Reference Mudde2007; Hooghe et al., Reference Hooghe, Marien and Pauwels2011; Hobolt & Tilley, Reference Hobolt and Tilley2016). Our second hypothesis tests this presumed relationship between low political trust and anti-establishment voting:
Hypothesis 2: Individuals with low levels of political trust are more likely to vote for anti-establishment parties than individuals with high levels of political trust
The effect of political trust variability on electoral behavior
However, the empirical literature fails to explain which aspects of distrust drive these relationships due to an almost exclusive focus on cross-sectional trust levels (see also Citrin & Luks, Reference Citrin, Luks, Hibbing and Theiss-Morse2001). This is remarkable, given longstanding theoretical debates suggesting that deep-rooted forms of political distrust may have different behavioral implications than more transient types (Easton, Reference Easton1965; Levi & Stoker, Reference Levi and Stoker2000; Fieschi & Heywood, Reference Fieschi and Heywood2004; Citrin & Stoker, Reference Citrin and Stoker2018). Yet, rigorous empirical tests of this claim remain scarce.
Recent conceptual work distinguishes between two types of distrust based on its temporality: variable versus stable distrust (Lenard, Reference Lenard2008; Bertsou, Reference Bertsou2019; Jennings et al., Reference Jennings, Stoker, Valgarðsson, Devine and Gaskell2021; Norris, Reference Norris2022; Korvela & Vento, Reference Korvela and Vento2023).Footnote 2 The first type, variable distrust, often emerges from political skepticism. Skepticism entails a vigilant stance toward authorities, where citizens critically scrutinize the behavior and performance of institutions. It generally targets current officeholders rather than the political system as a whole (Lenard, Reference Lenard2008; Rosanvallon, Reference Rosanvallon2008). As such, it tends to fluctuate over time in response to political developments (Fieschi & Heywood, Reference Fieschi and Heywood2004). However, while skeptical distrust is inherently variable, not all fluctuations stem from skepticism; some may result from non-attitudes and are essentially noise. We return to this in the results section.
Second, distrust can also be stable (Norris, Reference Norris2022). Stable distrusters hold a ‘settled belief that the other is untrustworthy’ (Citrin & Stoker, Reference Citrin and Stoker2018, p. 50), often stemming from a cynical, dispositional pessimistic outlook on politics. This type of distrust remains largely unresponsive to the behavior and performance of political institutions (Lenard, Reference Lenard2008; Citrin & Stoker, Reference Citrin and Stoker2018). As a result, stable distrust can often be “misplaced” in the face of objectively trustworthy institutions, while variable distrust is more adaptive and contextually appropriate.
While both stable and variable distrust indicate political dissatisfaction, they differ in how they are expected to shape electoral behavior. The dominant argument in the literature reads that deep-seated and stable distrust leads to disengagement from the democratic process (Hooghe et al., Reference Hooghe, Marien and Pauwels2011; Opdycke et al., Reference Opdycke, Segura and Vasquez2013; Dassonneville et al., Reference Dassonneville, Blais and Dejaeghere2015). Stable distrusters typically perceive the entire system as fundamentally flawed and therefore lack the motivation to “throw the rascals out.” Their cynical outlook makes all political options appear equally untrustworthy, and as a result, they expect no meaningful political change from voting, leading to electoral abstention (Butzlaff & Messinger-Zimmer, Reference Butzlaff and Messinger-Zimmer2020). While others contend that entrenched distrust may instead reflect a willingness to engage under low expectations (Citrin & Luks, Reference Citrin, Luks, Hibbing and Theiss-Morse2001; Bromley & Curtice, Reference Bromley and Curtice2004; Fieschi & Heywood, Reference Fieschi and Heywood2004), this view remains less prominent in the literature.
By contrast, variable distrust is associated with higher levels of political engagement (Citrin & Stoker, Reference Citrin and Stoker2018). Since it arises from dissatisfaction with specific actors or outcomes rather than a wholesale rejection of the system, variable distrusters are motivated to act on their concerns. They believe the system can be improved by holding incumbents accountable, and this belief makes them more likely to participate – particularly when dissatisfied – in an effort to influence political outcomes (Fieschi & Heywood, Reference Fieschi and Heywood2004). Rather than disengaging, they channel their discontent through electoral alternatives that directly challenge the status quo (Norris, Reference Norris1999; Bergh, Reference Bergh2004).
Crucially, however, this engagement does not necessarily translate into support for anti-establishment parties. The political choices of variable distrusters depend on the alternatives available. When credible options exist apart from illiberal parties, they may direct their votes accordingly (Bélanger & Nadeau, Reference Bélanger and Nadeau2005; Duch & Stevenson, Reference Duch and Stevenson2008). Yet when established (opposition) parties fail to offer meaningful alternatives, skeptical voters may instead support anti-establishment parties that more forcefully challenge the political consensus and democratic norms (Grant, Reference Grant2021; Van de Wardt & Rooduijn, Reference Van de Wardt and Rooduijn2023). In light of the growing ideological convergence among mainstream parties in many Western European democracies, such illiberal anti-establishment voting may become increasingly common among politically engaged but dissatisfied citizens (Spoon & Klüver, Reference Spoon and Klüver2019). In sum, we expect stable distrust to predominantly manifest as alienation from the political system, while variable distrust is more likely to result in critical engagement with the system. Therefore, we hypothesize that these two groups will differ in both their tendency to abstain from voting and their support for anti-establishment parties:Footnote 3
Hypothesis 3: Stable distrusters are more likely to abstain from voting than variable distrusters
Hypothesis 4: Variable distrusters are more likely to vote for anti-establishment parties than stable distrusters
The effect of temporal variation in trust on electoral behavior
The third component of political trust concerns fluctuations in trust within individuals over time (Voogd et al., Reference Voogd, van der Meer and van der Brug2019; Ouattara & Van der Meer, Reference Ouattara and Steenvoorden2023). The temporal dimension of trust is not only relevant for classifying (static) types of citizens, but also for understanding how shifts in trust may causally influence political behavior. While existing work often assumes that trust has a causal effect on electoral behavior, it typically only examines variation between individuals. Yet, establishing a causal relationship warrants an investigation of whether temporal changes in trust instigate shifts in electoral behavior. Similar to the potential impact of trust variability, the short-term effects of changing trust on political behavior remain insufficiently examined. The limited existing research indicates that temporary declines in trust are associated with a heightened likelihood of intended vote switching (Voogd et al., Reference Voogd, van der Meer and van der Brug2019), diminished support for representative democracy, and increased support for direct democratic alternatives (Ouattara & Van der Meer, Reference Ouattara and van der Meer2023). It is plausible to expect comparable effects on more general electoral behavior: as individuals experience a drop in trust over time, they may become more inclined to abstain from voting or to support anti-establishment parties (Hetherington, Reference Hetherington1999). Our last two hypotheses therefore read as follows:
Hypothesis 5: When individuals’ political trust decreases, their likelihood of abstention increases
Hypothesis 6: When individuals’ political trust decreases, their likelihood of anti-establishment voting increases
Data and method
Data
In the absence of direct survey measures, testing the effects of different types of trust based on general trust levels and variability, as well as the effect of temporal changes simultaneously requires panel data over a long time span. All waves should include measures of political trust and intended electoral behavior. The Dutch LISS panel (Longitudinal Internet studies for the Social Sciences) dataset forms a robust basis for this purpose for several reasons. First, the dataset covers fifteen annual waves, covering political trust and intended voting behavior across a span of sixteen years (2008–2023) (Scherpenzeel & Das, Reference Scherpenzeel, Das, Das, Ester and Kaczmirek2010).Footnote 4 This extensive temporal span marks a significant improvement over existing cross-sectional studies (e.g., Lubbers et al., Reference Lubbers, Gijsberts and Scheepers2002; Grönlund & Setälä, Reference Grönlund and Setälä2007; Hooghe & Marien, Reference Hooghe and Marien2013) and surpasses the few longitudinal studies that rely on fewer panel waves (e.g., Rooduijn et al., Reference Rooduijn, van der Brug and de Lange2016). Such a long observation window is crucial for both purposes of this study: constructing a robust typology of political trust based on long-term patterns of stability and fluctuations, and examining the effect of within-person changes in trust on electoral preferences. Second, the dataset includes trust assessments set against a backdrop of changing institutional arrangements, covering five government periods: Balkenende IV (2007–2010), Rutte I (2010–2012), Rutte II (2012–2017), Rutte III (2017–2022), and Rutte IV (2022–2024). With parliamentary elections in 2006, 2010, 2012, 2017, and 2021, the dataset enables us to isolate long-term distrusters from those dissatisfied with one specific administration.
Case selection
The Dutch political landscape between 2008 and 2023 offers a compelling substantive case to test our hypotheses. First, the Netherlands has an open party system facilitated by a highly proportional electoral system (PR) and no electoral threshold (Pellikaan et al., Reference Pellikaan, de Lange and van der Meer2018). This allows for the relatively easy emergence of new parties, including those with an anti-establishment message. Second, the country also has a well-defined political establishment: with one short-lived exception, the same group of parties alternated in government coalitions since the late 1970s. Third, the absence of compulsory voting makes abstention a viable option for Dutch citizens, although electoral turnout in parliamentary elections remains relatively high, fluctuating around 80 percent. Fourth, Dutch trust patterns mirror those of other Western European democracies (Torcal, Reference Torcal, Zmerli and van der Meer2017), with approximately 50 percent of citizens reporting moderate to high trust in parliament and government from 2008 to 2019 (Den Ridder et al., Reference Ridder, Josje and van den Broek2023). Although trust in political institutions surged in the spring of 2020 in response to the onset of the COVID-19 pandemic (Van der Meer et al., Reference Van der Meer, Steenvoorden and Ouattara2023), it had returned to its previous levels by the spring of 2021.Footnote 5 Lastly, and importantly, the period under study was marked by considerable political turbulence, including multiple government collapses, economic crises, major political scandals, and the COVID-19 pandemic. These events produced visible fluctuations in aggregated political trust levels (see online Appendix A2), providing a dynamic political context that allows for a meaningful differentiation between stable and variable (dis)trusters.
Sample and operationalization
The LISS panel is a probability-based sample and contains 7500 Dutch individuals aged 16 and above, distributed across 5000 households. We restrict our analysis to respondents who have participated in at least three panel waves and are eligible to vote. This results in an analytical sample of 6455 individuals, with an average of 6.34 observations per respondent. Yearly sample sizes vary between 1287 (wave 15 in 2023) and 4051 respondents (wave 6 in 2013).
Independent variables
We compare different types of trust based along three components: level, variability, and change. The dataset offers four indicators of political trust per survey wave: trust in government, trust in parliament, trust in political parties, and trust in politicians. Scores on these items range from 0 (no confidence) to 10 (full confidence). Cronbach’s alpha falls between 0.92 (wave 4 in 2011) and 0.96 (wave 15 in 2023).
Our measure of change in political trust is relatively straightforward. Consistent with literature showing that within-person political trust dynamics should be understood as temporal fluctuations around a long-term baseline (Devine & Valgarðsson, Reference Devine and Valgarðsson2024; La Roi et al., Reference La Roi, van Alebeek and van der Meer2025), we operationalize changes in trust as the deviation of respondents’ level of political trust at time t from their overall mean level of political trust.Footnote 6 This allows us to empirically disentangle the dynamic (within-person) trait of change from the two static (person) traits of levels and variability. Since this scale is centered around the person mean and therefore symmetrical, we multiplied these deviations by −1 so that a 1-point increase on the scale indicates a 1-point decrease in trust.
Next, we combine levels and variability to categorize groups of respondents. Respondents’ mean level of political trust across all waves indicates whether they generally tend to be more trusting or distrusting. To calculate respondents’ variability of political trust, we take the average absolute deviation from each wave-specific level of political trust to the static level of trust. Consequently, the larger (smaller) respondents’ average absolute deviation, the more variable (stable) their political trust. For individuals with more survey waves, the number of observations helps ‘averaging out’ fluctuations in trust scores, resulting in higher stability. To account for this, we correct the variability measure by dividing it by the number of survey waves respondents participated in.
From these two measures, we construct our trust typology. Because the classification of cases is to a certain extent arbitrary, our primary strategy is to achieve a balance between internal validity and group size. After excluding 118 influential outliers on the variability measure (i.e., cases with a variability score at least 3 standard deviations (SD) away from the mean), observations with a trust level and variability of at least 0.50 SD away from both means are assigned to one of the four main categories: stable trusters (1), variable trusters (2), stable distrusters (3), or variable distrusters (4). Cases within 0.50 SD of either of the two means are placed in the middle category (5), which functions as a residual group for respondents whose average trust level and variability fall too close to the sample means to allow for meaningful classification. As such, this group does not represent a clear, theoretically driven category. To assess the sensitivity of our results to the classification criteria, we also estimate our models using a distance of 0.25 SD. A distance of 1 SD will be used as a robustness test, as it results in groups with fewer than 30 cases.
Figure 1 visualizes the distribution of our data across both levels and variability of trust. Applying a threshold of 0.50 SD from both means, we identify 1364 stable trusters (21.1% of cases), 242 variable trusters (3.7%), 290 stable distrusters (4.5%), 662 variable distrusters (10.3%), and 3897 observations in the middle category (60.4%). Reducing the threshold to 0.25 SD, the groups expand to include 2027 stable trusters (31.4%), 559 variable trusters (8.7%), 650 stable distrusters (10.1%), 999 variable distrusters (15.5%), and 2220 cases in the middle category (34.3%).Footnote 7 The distribution of trust levels reveals an asymmetry between the two groups: trusters predominantly score between 6 and 8 on the 0–10 scale, while distrusters tend to score between 0 and 4. This asymmetry underscores the contrast in the intensity of trust and distrust, with distrusters exhibiting a broader range of low scores compared to the more concentrated trust ratings of trusters.Footnote 8
Figure 1. A typology of political trust by level and variability.
Note: N = 6445. Solid lines indicate the mean level and variability of trust. Categorization based on 0.50 SD includes dark data points only, 0.25 SD includes both light and dark data points.
Dependent variable: vote intention
For our dependent variable, we use an item that asked respondents for which party they would vote if parliamentary elections were held today. We categorized responses into four groups: established parties (1), anti-establishment parties (2), other parties (3), and abstention (4). The first category includes five traditional parties that held government responsibility before and during the 2008–2023 period: the Labor Party (PvdA), Christian Democratic Appeal (CDA), People’s Party for Freedom and Democracy (VVD), Christian Union (CU), and Democrats 66 (D66).
Anti-establishment parties are identified using the Chapel Hill Expert Survey (CHES) anti-elite and anti-establishment indicators from 2014 and 2019, selecting parties scoring 5 or higher on a scale of 0–10 (Jolly et al., Reference Jolly, Bakker, Hooghe, Marks, Polk, Rovny, Steenbergen and Anna Vachudova2022). The anti-establishment status of parties unavailable in the CHES data is supported by existing literature (Lucardie, Reference Lucardie2010) and The PopuList (Rooduijn et al., Reference Rooduijn, Pirro, Halikiopoulou, Froio, van Kessel, de Lange, Mudde and Taggart2024). This group includes the Socialist Party (SP), Party for the Animals (PvdD), BIJ1, Party for Freedom (PVV)Footnote 9 , Forum for Democracy (FvD), JA21, Farmer-Citizen Movement (BBB), Proud of the Netherlands (TON), 50Plus, and DENK.
Other parties such as GreenLeft (GL), the Reformed Political Party (SGP), and Volt Netherlands (Volt) have never been part of a governing coalition but lack a clear anti-establishment stance. Note that the number of anti-establishment and other parties might differ between survey waves due to party emergence and dissolution. A full classification of the parties included per wave can be found in Appendix A6.
Control variables
We control for various sociodemographic and political respondent characteristics. Our first set of control variables includes sex, age, level of education and household income. To ensure that the electoral effects of trust variability are not driven by panel participation rates – as is the case with panel conditioning effects (Sturgis et al., Reference Sturgis, Allum, Brunton-Smith and Lynn2009) – we control for both the number of panel waves and the number of government periods in which respondents participated. Given that political trust is shaped by ideological proximity to governing parties (Hetherington & Rudolph, Reference Hetherington and Rudolph2015; Citrin & Stoker, Reference Citrin and Stoker2018), and since Dutch governments between 2008 and 2023 have consistently been centrist, we control for respondents’ squared left-right self-placement to account for the potential confounding effect of ideological distance to the governing coalition on both trust and intended electoral behavior. Lastly, we include wave fixed effects to adjust for unobserved, time-specific factors – such as economic crises or major political scandals – that could simultaneously affect political trust and electoral preferences. Detailed operationalizations and descriptive statistics are provided in Appendix A7–A8.
Method
This study aims to simultaneously test the effects of static and dynamic aspects of political trust. To this end, we employ the REWB method (Bell et al., Reference Bell, Fairbrother and Jones2019). Unlike conventional two-level random effects models, REWB models allow for the testing of the impact of the same panel variable at both levels concurrently. This is done by separating the between-person variation from the within-person variation. At the between-level, the method relies on aggregated person scores of trust level and variability (i.e., the typology). At the within-level, original time-varying values are centered around the person mean (i.e., our measure of trust change). The models include random effects for the intercept. Because the dependent variable consists of four nominal categories, we apply multinomial logistic regression.
While our longitudinal study provides a stronger test of the link between political trust and electoral behavior, caution is needed in making causal claims. The assertion that the causal relationship between discontent and anti-establishment voting is unidirectional has been challenged by scholars who demonstrate that exposure to anti-establishment rhetoric can exacerbate existing distrust, contributing to a so-called “spiral of distrust” (e.g., Rooduijn et al., Reference Rooduijn, van der Brug and de Lange2016). Unfortunately, we are unable to control for (changes in) exposure to anti-establishment messages.
Results
Static effects of political trust
The static REWB multinomial logistic regression results (Figure 2; full models in Appendix B) reveal that individuals with low trust are more likely to abstain from voting and favor anti-establishment parties compared to those with high trust. Stable distrusters are significantly more likely to support an anti-establishment party and abstain from voting compared to both stable and variable trusters (p < 0.05). In Figure 2, the coefficients for anti-establishment voting are statistically significant and negative for both categories of trusters. This indicates that stable distrusters are more likely to choose an anti-establishment party than variable trusters, and even more so than stable trusters. These coefficient patterns are also visible for abstention. The probabilities in Figure 3 confirm that distrusters (Pr = 0.56) are more likely to support anti-establishment parties and abstain (Pr = 0.13) than trusters (Pr = 0.15 and Pr = 0.03, respectively).Footnote 10 These findings support both H1 and H2.
Figure 2. Between-effect of political trust on intended electoral behavior (95% CI).
Note: Coefficients represent log-odds. Coefficients derived from all Model 3 specifications in Appendix B1–B12. Controlled for panel participation, government period participation, survey wave, sex, age, education, income, and ideological self-placement.
Figure 3. Probability of electoral outcomes per trust group (full model) (95% CI).
Note: Controlled for panel participation, government period participation, survey wave, sex, age, education, income, and ideological self-placement.
We now shift our focus to the main contribution of this paper: trust variability. Our second pair of hypotheses proposed that stable distrusters are more likely to abstain from voting than variable distrusters (H3), while the latter are more inclined to vote for anti-establishment parties (H4). However, our data provide no support for these predictions. Starting with support for anti-establishment parties, despite the positive and statistically significant coefficient for variable distrusters in Figure 2a, the difference is negligible in magnitude (b = 0.49) and inconsistent across different trust categorizations (i.e., 0.50 SD) and reference outcomes (i.e., Figure 2d). The probabilities in Figure 3 show that stable and variable distrusters are also equally as likely to report an intention to vote for an anti-establishment party: probabilities are 0.56 and 0.55, respectively. This leads us to conclude that H4 is not supported. Surprisingly, variable distrusters (Pr = 0.15) are significantly more likely to choose abstention than stable distrusters (Pr = 0.11), which contradicts H3. Yet, the effect size is again too small to be interpreted as substantial (b = 0.62), and the significance of the result appears to be contingent on the categorization criterion of the trust groups, as well as the reference outcome of the model (i.e., Figure 2e).
When we compare the likelihood of abstention of the two groups directly to their likelihood of support for anti-establishment parties (Figure 2f), we observe no significant difference. Our analysis thus offers rather weak evidence that stable and variable distrusters differ in their tendency toward electoral abstention and support for anti-establishment parties. They do, however, differ significantly in these aspects from those with high political trust. Consistent with the traditional view (e.g., Finifter, Reference Finifter1970; Mair, Reference Mair2013) and despite more recent theories of critical citizenship (e.g., Norris, Reference Norris1999), political distrust pushes citizens away from status quo politics, regardless of whether this distrust is stable or more variable over time. Yet, distrust does not necessarily alienate citizens from electoral politics altogether.
Dynamic effect of political trust
Next, we turn to the impact of dynamic changes in trust. H5 and H6 anticipated that short-term decreases in political trust are associated with increases in individuals’ likelihood of both abstention and anti-establishment voting. Figure 4 shows that declining political trust has an independent and significant effect in all models except those in Figure 4f. In Figure 4a, a temporary drop in political trust below the person mean is related to an increase in the probability that an individual opts for an anti-establishment party instead of an established party. Figure 4c shows a comparable effect for abstention. Declining trust does not seem to have a larger impact on either support for anti-establishment parties or abstention (Figure 4f).
Figure 4. Within-effect of political trust on intended electoral behavior (95% CI).
Note: Coefficients represent log-odds. Coefficients derived from all Model 3 specifications in Appendix B1–B12. Controlled for panel participation, government period participation, survey wave, sex, age, education, income, and ideological self-placement.
Much like static distrust, a decline in trust moves individuals away from the status quo. However, unlike static distrust, this decrease is not exclusively linked to anti-establishment voting and abstention. Instead, it is also associated with greater support for mainstream opposition parties over established parties (see Figure 4b). These results align with previous research demonstrating a short-term effect of shifts in political trust on electoral volatility (Voogd et al., Reference Voogd, van der Meer and van der Brug2019) and support for various forms of democracy (Ouattara & Van der Meer, Reference Ouattara and van der Meer2023).
Robustness tests
To validate whether these findings hold across different model specifications and categorization thresholds, we perform several robustness tests. First, we re-estimate our models incorporating trust levels and variability as continuous variables rather than predefined categories (see Appendix C1–C6). The results of this test reinforce our earlier findings. In addition, we employed several alternative classification criteria to create the trust typology: a threshold of 1 SD, groups based on quartiles, and groups based on fixed ranges for trust level (Appendix C7–C9). Of these, only the 1 SD approach fails to echo our main findings, yet its statistical power is low given the extremely small group sizes.
Second, to explore whether trust variability reflects only quantitative differences or also qualitative differences between groups of citizens, we test whether declining trust has a different effect for each group (Appendix C10). We find no evidence that trust changes affect stable and variable distrusters differently, suggesting that stable distrust may not be as deeply entrenched or behaviorally constraining as often assumed. However, it should be noted that these null interaction effects may also stem from limited statistical power, as stable distrusters have, by definition, experienced fewer changes in trust than their variable counterparts.
Third, to verify whether our typology meaningfully distinguishes stable from evaluative (dis)trusters, we reconstructed it on a subset of respondents who experienced both politically calm and turbulent periods. Here, we focus on the politically most dynamic decade in our data (2013–2022), which covers three government periods (Rutte II, Rutte III, and Rutte IV) and the COVID-19 pandemic. We select respondents who participated in at least one COVID wave (2021/2022), at least one politically turbulent wave (2016/2017), and at least two less turbulent waves (2013, 2014, 2018, 2019, or 2020). The results align closely with the main findings (Appendix C11).
Fourth, to test if our main results hold across domestic political and institutional changes, we assess whether our findings hold across government periods (see Appendix C12–C21).Footnote 11 The impact of trust levels on anti-establishment voting and abstention is robust across all periods. The effect of trust variability is somewhat less robust over time. From Balkenende IV to Rutte III, differences between stable and variable distrusters are minimal. Under Rutte IV, however, stable distrusters are more likely to abstain, while variable distrusters lean toward anti-establishment voting. This suggests that the effect of trust variability is contingent upon external factors, such as national crises or an increased availability of anti-establishment parties.
Finally, we assessed to what extent our trust variability measure is sensitive to non-attitudinal noise. To that purpose, we excluded respondents who gave the exact same trust score on all four trust items in at least eight survey waves (i.e., straightliners), or who show high variability over time on ideological items that supposedly tap into more principled stances (see Appendix C22).Footnote 12 This does not alter our findings. Additionally, we ran our models separately for respondents with low and high political interest (see Appendix C23–C24). The findings for low-interest respondents are consistent with the main analysis. In summary, our conclusions from the main analysis are unlikely to be an outcome of these types of non-attitudes.
Conclusion
The conventional narrative linking political trust to electoral behavior argues that low trust in political institutions undermines democratic stability by encouraging citizens to either vote for anti-establishment parties or withdraw from the electoral process. However, the nature of this relationship remains elusive, largely due to an overemphasis on cross-sectional trust levels. This paper simultaneously tested the effects of three components of trust (i.e., trust levels, trust variability, and declining trust) on both electoral participation and vote choice.
Drawing on fifteen waves of LISS panel data (2008–2023) and employing a multinomial logistic REWB modeling technique, our findings reveal that a multilayered treatment of political trust is in order. First, among the three studied components of trust, (low) trust levels remain the strongest predictor of both support for anti-establishment parties and electoral abstention. Second, stable and variable distrusters show only minimal differences in their inclination toward anti-establishment voting and abstention. Third, short-term declines in political trust increase the chances of anti-establishment voting and abstention, independent of individuals’ overall trust levels and variability.
These findings have important implications. First, political distrust, whether persistent or fluctuating, seems to drive citizens away from the political status quo. In this regard, the traditional emphasis on trust levels has been well justified. However, this does not mean that distrust will inevitably lead to alienation from democracy (cf. Finifter, Reference Finifter1970; Mair, Reference Mair2013). Both stable and variable distrusters remain predominantly involved in electoral politics, typically feeling drawn toward parties that challenge the established order. This challenges the assumption that it is blind rather than evaluative distrust that forms a threat to democratic stability (cf. Levi, Reference Levi, Braithwaite and Levi1998; Citrin & Stoker, Reference Citrin and Stoker2018). Although vigilant citizenship is held up as the democratic ideal, once it turns into distrust, its electoral outcomes are barely distinguishable from those of stable distrust. Nor can we infer from macro-level electoral trends whether increasing support for anti-establishment parties reflects a rise in dispositional or skeptical distrust (cf. Norris, Reference Norris2022). Yet, if democracies have indeed witnessed a transition from blind trust to skeptical citizenship (Norris, Reference Norris1999), our findings may explain why many established political parties struggle to retain a stable voter base. This points to potential challenges for party systems and democratic institutions in responding to citizens’ changing trust orientations.
Second, the categorization of political distrusters based on trust levels and variability offers a valuable framework for future studies within the field. Zooming in on the characteristics, motivations and behaviors of these groups will enhance our comprehension of the complex role of political trust in contemporary democratic dynamics. For instance, the typology can be employed to explore whether the difference between stable and variable distrusters is more substantial in less institutionalized forms of participation than voting in elections (Ouattara & Steenvoorden, Reference Ouattara and Steenvoorden2023), such as joining protests. Moreover, contrary to the trust literature, which has focused solely on different types of distrust, our analysis suggests that distinguishing between types of trust can offer valuable insights into the diverse implications of high political trust for democracy (Norris, Reference Norris2022; Hadarics, Reference Hadarics2024).
Third, to fully grasp how political distrust shapes behavior, it is insufficient to focus solely on static differences between citizens. The simultaneous existence of an independent effect of declining trust suggests that there are at least two separate pathways through which political trust is connected to behavior. The latter is potentially causal in nature, while the former is likely to remain correlational at most. This underscores the need for further research to disentangle the distinct mechanisms driving between- and within-person trust effects (see Voogd et al., Reference Voogd, van der Meer and van der Brug2019; Ouattara & Van der Meer, Reference Ouattara and Steenvoorden2023). While short-term shifts in political trust are generally modest (Devine & Valgarðsson, Reference Devine and Valgarðsson2024), they remain significant enough to influence political behavior and should not be disregarded.
A few concluding remarks on the limitations of this study are in order. First, our typology rests on the assumption that stable distrust particularly captures a dispositional element of trust, while variable distrust captures more evaluative, performance-based orientations. Although this provides a pragmatic approach under data constraints, it remains an indirect proxy. Future research should examine whether this distinction reliably maps onto the psychological foundations of dispositional and evaluative trust, ideally by incorporating direct measures of these constructs. Similarly, while this paper drew on theories of political cynicism and skeptical citizenship, we were unable to measure these attitudes directly. Addressing these limitations would contribute to a more precise understanding of the complex relationship between distrust and democratic engagement.
Second, while the openness and proportionality of the Dutch political and institutional context make it an ideal case for isolating the multidimensional nature of political trust and its effects on electoral behavior, these same features may limit the generalizability of our findings. In two-party systems, where voters have fewer alternatives and switching behavior is more constrained (Bartolini & Mair, Reference Bartolini and Mair1990), or in party systems with few viable anti-establishment parties, the relationship between political trust, its variability, and electoral choices may manifest differently. Similarly, institutional arrangements that delay or even fully suppress the breakthrough of anti-establishment parties might obscure the effect of trust levels and variability on electoral behavior. Subsequent research should explore the extent to which these findings generalize to more closed and majoritarian political contexts. Moving toward a framework that acknowledges the multifaceted nature of political trust is crucial for advancing our understanding of its consequences for the functioning of contemporary democracies.
Supplementary material
To view supplementary material for this article, please visit https://doi.org/10.1017/S1755773925100210.
Acknowledgements
The authors want to thank the Challenges to Democratic Representation programme group at the University of Amsterdam and the Representation and Trust group at The Netherlands Institute for Social Research for their valuable comments on earlier drafts of this paper.
Funding statement
This study was supported by the European Research Council Consolidator Grant 2021 101045653 to Tom van der Meer.
Competing interests
The authors declare that there are no competing interests to disclose.
Data availability
The data that support the findings of this study are available at www.dataarchive.lissdata.nl. Replication materials available upon request.
Open access
