Abstract
Citizens are generally opposed to politicians receiving a high pay. We investigate the degree to which this aversion is moderated by citizens’ individual-level trust in politicians and whether institutional delegation can moderate the reactions to proposed changes in politicians’ pay. Using a survey experiment, we confirm that trust in politicians is a key predictor of attitudes regarding their pay. Distrust toward politicians seems to matter much more than general attitudes on income inequality when citizens form opinions on politicians’ pay. Furthermore, citizens’ aversion to high pay for politicians is affected by institutional delegation, but such delegation only lessen the opposition to pay raises modestly, leaving most citizens firmly against pay raises for politicians. Finally, while citizens’ trust in politicians matters greatly for their attitudes regarding politicians pay, proposed changes in politicians’ pay do not conversely affect citizen’s perceptions of the politicians.
Introduction
In 1816, protesters burned members of the US Congress in effigy after the congress members had enacted a pay raise for themselves (Theriault, 2004). Two centuries later, citizens’ dislike of high pay for politicians persists. Several studies across a diverse array of countries show that vast majorities of citizens in all of these countries believe that politicians should receive substantially lower salaries than they currently do (Kelley and Evans, 1993; Kiatpongsan and Norton, 2014; Mause, 2014). While this general aversion toward high pay for politicians is well-established, there are, however, several reasons to further investigate these attitudes.
First, the salary of politicians is an important policy issue in its own right. Public opinion on this issue is important because it may affect the actual salary levels of politicians, which may again have consequences for the recruitment and behavior of politicians (Braendle, 2015; Carnes and Hansen, 2016; Fisman et al., 2015; Hoffman and Lyons, 2013; Keane and Merlo, 2010).
Second, in contrast to nearly all other vocations, national politicians typically have the power to set their own wages (Brans and Peters, 2012; Mause, 2014). We argue that, because of this institutional feature, citizens’ attitudes on politicians’ pay are closely connected to perceptions regarding politicians’ self-interest and therefore also trust in the politicians, which is a key indicator of political legitimacy (Turper and Aarts, 2017). Building on the literature regarding political trust (e.g. Keele, 2007; Levi and Stoker, 2000), stealth democracy (Hibbing and Theiss-Morse, 2002), and deservingness (e.g. Feather, 1999; Petersen, 2012), we theorize that low levels of trust toward politicians is a key explanation for the publics’ aversion to high pay for politicians. In addition, we also suggest that changes in politicians’ pay could potentially be used as an instrument to increase trust in politicians.
Third, because of the relationship between citizens’ low trust in the politicians and their attitudes on these politicians’ pay, this issue provides a fitting test case on the advantages of policy delegation to an independent agency. Delegation of decision-making authority to independent agencies is often recommended as a way to increase credibility and trust by tying the hands of politicians, who could otherwise be tempted to act on their short-sighted self-interest (Gilardi, 2002; Majone, 1997). Indeed, politicians in a number of countries have chosen to delegate the authority to set their own wages to independent institutions (Brans and Peters, 2012; Mause, 2014). We propose that such delegation might abate citizens’ aversion to pay raises for politicians.
To investigate citizens’ reactions to proposed changes on politicians’ pay, we use a survey experiment (n~2000), where we randomly assign respondents to different scenarios regarding politicians’ pay. Our results show that attitudes toward politicians’ pay are indeed closely related to political trust, which is a more important predictor of pay attitudes than general attitudes on economic inequality. Furthermore, we show that delegating the issue of politicians’ pay to an independent agency can affect citizens’ reactions to such changes. However, institutional delegation does not change the fact that the average citizen remains staunchly opposed to pay increases for politicians. Finally, while changes in politicians pay has been proposed as a way of increasing the public’s trust in them (Hibbing and Theiss-Morse, 2002), our results do not indicate that such an initiative would be efficacious. A reduction of their own wages is not necessarily a silver bullet for politicians hoping to increase the electorate’s trust in them.
Attitudes on Pay for Politicians
Full-time politicians in modern democracies generally have substantially higher incomes than the average citizen (Brans and Peters, 2012; Hood and Peters, 1994; Mause, 2014). This high level of compensation seems to irk the public, and when politicians enact a pay raise for themselves, it is often followed by widespread condemnation and opposition from the news media and from citizens (Baimbridge and Darcy, 1999; Jones, 2007; Theriault, 2004). Furthermore, survey data from 40 countries show that, across all of these countries, majorities of voters think that politicians (specifically cabinet ministers) receive more than they ought to (Kelley and Evans, 1993; Kiatpongsan and Norton, 2014). 1 Thus, the current knowledge clearly suggests that it is reasonable to expect some aversion to pay raises for politicians, and our study takes departure in the following hypothesis:
Hypothesis 1. Citizens show higher support for a pay decrease than a pay increase for politicians.
While previous studies suggest that aversion to pay raises for politicians is to be expected, there is still little knowledge regarding the determinants and mechanisms behind this attitude. In the following section, we theorize how attitudes regarding proposed pay changes for politicians may depend on the sponsor of such proposals.
The Effects of Institutional Delegation
There are substantial differences in the degree to which national politicians can set their own wages. In a majority of European national parliaments and in the US Congress, the politicians can set their own wages, albeit with some restrictions. However, in a few countries, politicians have delegated the right to determine their pay to independent entities. This is, for example, the case in Great Britain, where politicians’ pay are set by the Independent Parliamentary Standards Authority, in Sweden where the task has been delegated to Riksdagens arvodesnämnd (Brans and Peters, 2012; Mause, 2014), and partly in Denmark, where the task was temporarily delegated to an independent commission in 2014 (Vederlagskommissionen, 2016).
To what degree might such institutional delegation affect how citizens react to proposed changes in politicians’ pay? If citizens’ aversion to high pay for politicians was driven purely by attitudes regarding income and economic inequality, the sponsor of a proposed change in politicians’ wages should not matter. However, to the degree that citizens’ aversion to high pay for politicians is driven by their perceptions of—and attitudes toward—the politicians, the sponsor of such a proposed change in wages could affect attitudes regarding the proposal. To see how that might be the case, it is worth taking a closer look at the relationship between citizens’ perceptions of politicians’ self-interest and citizens’ trust toward the politicians.
Traditionally, studies of citizens’ trust in politicians have often focused on the degree to which such attitudes are affected by policy outputs and outcomes (Citrin, 1974; Keele, 2007; Whiteley et al., 2016). However, citizens do not just care about outputs, they also care about how decisions come about, also termed procedural fairness (Bøggild and Petersen, 2016). Hence, as argued by Hibbing and Theiss-Morse (2002), the low esteem to which politicians are held is not necessarily a result of dissatisfaction with the output of the political system. Rather, citizens are dissatisfied with the political processes preceding the output. A key issue regarding this political process is, according to Hibbing and Theiss-Morse (2002), the degree to which the politicians concern themselves with societal interests as opposed to special interests, including their own personal interests. People are “amazingly attuned, hypersensitive even, to the possibility that decision makers will attempt to improve themselves at the expense of everyone else” (Hibbing and Theiss-Morse, 2002: 85). Thus, the argument goes, current dissatisfaction with politicians is largely caused by the perception that politicians are overly self-interested.
Following this line of reasoning, we should expect citizens to react strongly against any action by politicians that may be construed as being driven by self-interest. The problem with a pay increase for politicians is therefore not simply the level of pay in itself, but rather that such a proposal sends a signal of self-interest if the proposal is sponsored by the politicians themselves. In contrast, a proposed pay increase for politicians coming from an independent entity is not in the same way a clear signal of politicians being self-interested, and citizens might therefore react less negatively to a proposal from such a sponsor. Conversely, if politicians themselves suggest that their own wages should be decreased, citizens might tend to see this as an indication of relative selflessness of these politician. Therefore, citizens may be particularly supportive of such a proposal coming straight from the politicians. Accordingly, we propose the following hypotheses:
Hypothesis 2. Citizens show lower support for a pay increase when the proposal is coming from an actor that appears to be closely connected to the politicians themselves (compared with a proposal coming from an actor that appears to not be closely connected to the politicians themselves).
Hypothesis 3. Citizens show higher support for a pay decrease when the proposal is coming from an actor that appears to be closely connected to the politicians themselves (compared with a proposal coming from an actor that appears to not be closely connected to the politicians themselves).
It is worth noting that our argument resembles the explanation for delegation found in agency theory: when policy-makers tie their hands and delegate powers to independent entities, this may increase the credibility and legitimacy of decisions because these decisions are thereby shielded from the short-term self-interests of these policy-makers (Gilardi, 2002; Majone, 1997: 139–140).
While hypotheses 2 and 3 focus on the effect that institutional factors may have for citizens’ reactions to proposals regarding pay for politicians, the next section addresses how individual-level characteristics of the citizens may also moderate their reactions to such proposals. Specifically, we theorize that citizens’ existing levels of trust toward politicians moderates their reactions to pay changes, and we consider another moderator, namely attitudes regarding economic inequality.
The Importance of Citizens’ Political Trust and Inequality Attitudes
In this section, we argue that citizens’ level of trust in politicians may be a key factor in explaining their reactions to pay changes for politicians. Before we do so, it is worth noting that recent decades have seen substantial drops in political trust across western democracies (Keele, 2007; Turper and Aarts, 2017; Whiteley et al., 2016). This development is potentially troublesome, because trust toward politicians is a key indicator of political legitimacy and often regarded as a prerequisite for well-functioning and stable political systems (Kelley and Zagorski, 2004; Turper and Aarts, 2017).
Studies on perceptions of deservingness have found that key factors in people’s judgment of the deservingness of others are their like or dislike of the other and perceptions about the others’ moral character (Feather, 1999, 2015). As noted by Feather (1999), these perceptions consist, among other thing, of assessments regarding the others’ integrity and trustworthiness. Recent studies have confirmed that assessments regarding individuals’ deservingness have a strong effect on people’s willingness to support the poor and unemployed (Petersen, 2012; Petersen et al., 2012), and it also seems to play an important role for people’s judgments of the deservingness of high-income groups (Hansen, 2018; Ragusa, 2015). Thus, such perceptions regarding deservingness seems to have strong effect across different strata of society. In so far as our moral assessment of other individuals are associated with our trust in these individuals, as suggested by Feather (1999), it is therefore also reasonably to expect citizens’ trust in politicians to play a role in their attitudes regarding pay for these politicians. In the context of this study, we would therefore expect that citizens with low levels of trust toward politicians find these politicians to be less deserving of a high pay. Conversely, citizens with a high level of trust should be relatively more accepting of a high pay for politicians. Consequently, we propose the following hypothesis:
Hypothesis 4. Citizens with a high level of political trust show lower relative support for a pay decrease than citizens low on political trust.
In order to assess the relative importance of trust, we now turn our focus toward what we view to be the primary alternative explanation for citizens’ aversion to high pay for politicians, namely general aversion to economic inequality. National politicians typically receive a pay that puts them well above the mean of the income distribution (Brans and Peters, 2012; Hood and Peters, 1994). Aversion to such a high pay for politicians may therefore also to some degree be driven by a general aversion toward economic inequality. Numerous studies across countries and time periods show that large parts of the populations find current levels of inequality and the wages of high-income groups such as CEOs to be unreasonably high (Bartels, 2005; Kelley and Zagorski, 2004; McCall, 2013; Osberg and Smeeding, 2006). To the extent that a preference for lower pay for politicians is explained by such a general aversion to high levels of inequality, there is nothing particularly unique about the preferences regarding pay for politicians; politicians are merely part of a larger high-income group, and citizens with relatively egalitarian preferences would prefer lower incomes for this group. Following this line of reasoning, we propose the following hypothesis:
Hypothesis 5. Citizens with a high level of aversion to economic inequality show higher relative support for a pay decrease than citizens low on inequality aversion.
Reversing the Causal Arrow: Can Changes in Pay Affect Trust?
If distrust toward politicians is driven by the perception that they are excessively self-interested, could politicians potentially increase the publics’ trust in them simply by decreasing their own pay? Notably, Hibbing and Theiss-Morse (2002: 217) highlighted reforms of remuneration to elected officials as the main place to look when considering potential remedies to the low political trust. Similarly, that the pay received by politicians could affect trust in them had also been suggested by Hood and Peters (1994). However, Hood and Peters did not empirically test this proposed relationship, nor did Hibbing and Theiss-Morse (2002), perhaps because they had, in the end, little faith in the efficacy of such an initiative. While they found pay decreases for politicians to be worthy of consideration, they ended up stating, rather pessimistically, that “We do not believe the people can be convinced, even after serious reform, that politicians are behaving in a non-self-interested manner” (Hibbing and Theiss-Morse, 2002: 220). Nevertheless, the possibility of pay changes affecting citizens’ trust in politicians has had such a prominent role in the literature that an empirical test seems sensible. Therefore, we also investigate the potential downstream effects of pay changes on trust.
When looking at these potential effects, it is also worthwhile elaborating on the exact nature of trust. In the numerous existing investigations of political trust, this concept is usually treated as a unidimensional construct. However, when we judge the trustworthiness of other individuals or groups, we generally judge them on two dimensions. The first dimension involves the commitment to act in the interests of the truster, and when someone is called trustworthy, it is often a reference to this commitment. However, there is also a second dimension to trustworthiness, namely the trustee’s competence in a given domain (Levi and Stoker, 2000). 2
This two-dimensional perspective on the concept of trust aligns very well with the stereotype content model (SCM), according to which people assess other individuals or groups on two fundamental dimensions, competence and warmth (Judd et al., 2005). The dimension of warmth reflects traits related to an individual’s (or group’s) intentions, such as friendliness and helpfulness, whereas the dimension of competence captures traits related to the individual’s abilities, such as intelligence and skills in general (Koch and Obermaier, 2016).
There are good reasons to investigate both dimensions of trust in this study. According to Hibbing and Theiss-Morse (2002), peoples’ distrust toward politicians is primarily distrust toward their intentions, that is, their warmth. Following their perspective, we expect that proposals regarding higher pay for politicians will primarily affect trust in politicians’ warmth. In the interest of completeness and in order to compare potential effects on the two dimensions of trust, however, we also test for any effects of trust in politicians’ competence. We therefore propose the following hypotheses:
Hypothesis 6. Citizens exposed to a proposed pay increase have lower levels of trust in politicians’ warmth than citizens exposed to a proposed pay decrease.
Hypothesis 7. Citizens exposed to a proposed pay increase have lower levels of trust in politicians’ competence than citizens exposed to a proposed pay decrease.
Finally, the degree to which proposed changes to politicians’ pay affect trust in these politicians could depend not just on the proposed change, but also of the sponsor of such a proposal. Specifically, voters should be expected to update their perceptions regarding politicians more when exposed to a proposal coming from these politicians rather than some other, independent, actor. Thus, the final hypotheses:
Hypothesis 8. The effect of a proposed pay change will be higher when the proposal is coming from an actor that appears to be closely connected to the politicians themselves.
Experiment
To test our hypotheses, we conducted a survey experiment in Denmark. The income of politicians in the Danish parliament is approximately 1.7 times larger than the salary of the average Danish citizen. Comparatively speaking, this is a relatively small pay gap, and all but four countries among 27 European countries (EU) countries have larger pay gaps between national politicians and the average citizen (Mause, 2014). Furthermore, despite recent declines, Danish voters still have comparatively high levels of political trust (Torcal, 2017; Van Erkel and Van Der Meer, 2016). As such, the Danish case can be seen as a least likely case when it comes to opposition toward high pay for politicians. While there are, just like in other countries, sometimes stories in the Danish news media about the pays and perks of politicians and other high income groups, there were no salient news stories on this issue at the time of our experiment.
Prior to any data collection, the study was preregistered on Open Science Framework. 3 The experiment was conducted in a commercial web panel (Voxmeter), where panel members were invited by email to participate. 4 A total of 2196 respondents started the survey and 2001 respondents completed the questionnaire, corresponding to a completion rate of 91.1%. More important for our ability to use the experimental data for casual inferences, most of the drop off occurred prior to exposure of the experimental stimuli. Among the 2033 respondents exposed to the experimental stimuli, 98.4% completed the questionnaire. The miniscule drop off after exposure to the experimental stimuli did not differ significantly between the experimental conditions (χ2(4, N = 2033) = 1.74, p = 0.78). The final sample was approximately representative of the adult Danish population on gender (50.2% female) and age (M = 49.8 years, standard deviation (SD) = 17.8), while the educational level in the sample was somewhat higher than in the general population. For sample and population characteristics, see Appendix A (available online).
Design
At the start of the survey, respondents were asked standard questions on demographics and vote choice, and they were asked to place themselves on an 11-point political left-right scale, ranging from 0 to 10. (In our analyses, all continuous variables have subsequently been standardized to range from 0 to 1.) Respondents were then asked two questions regarding their specific attitude on economic inequality. Together, respondents’ left-right self-placement and these two specific inequality questions formed a reliable index on inequality aversion (Cronbach’s α = 0.71). Next, respondents were asked about their general level of trust toward politicians on an 11-point left-right scale, ranging from 0 to 10. (The distributions of respondents’ values across these two measures are shown in Appendix A, available online.)
Respondents were then exposed to the experimental stimuli. They were asked to read a brief description of a commission, which suggested changes in the wages of the politicians in the Danish national parliament. Following our hypotheses, we manipulated two factors in the experiment: (1) whether respondents were exposed to a scenario were the commission proposed a decrease or an increase in wages and (2) whether this commission was presented as being closely connected to the politicians themselves or independent of these politicians. In the conditions where the commission was described as being closely aligned with the politicians, respondents were informed that the commission had been set up by the politicians in the Danish parliament, that the current politicians had appointed the commissions members, and that a majority of these commission members were themselves former politicians. In the remainder of the article, we use the shorthand “political commission” for this description of the commission. In the conditions where the commission was presented as independent of the politicians, respondents were told that the commission was completely independent, that members of the commission were experts with knowledge of the conditions on the Danish labor market, and that the members were not members of the parliament themselves. 5 We use the term “independent commission” when referring to this description of the commission. (The exact wording of all conditions is included in the survey questionnaire in Appendix B, available online.)
These different descriptions of the commission were all in line with an actual commission, Vederlagskommissionen, which was set up in Denmark in 2014 and presented their proposals regarding politicians’ pay approximately 2 years before our experiment was conducted. Most members of this commission were indeed former politicians but they were also selected for their expertise. Thus, our varying descriptions of the commission were not deceptive. Rather, they were simply different ways of framing the commission. (We did however change the exact content of the commission’s proposal in order to maximize simplicity and stringency.) 6
We chose to expose respondents to a scenario closely resembling real world processes regarding politicians pay in order to increase the external validity and relevance of the experiment. A potential disadvantage to this choice is that it may have limited the strength of the stimuli, for example, by presenting the members of the political commission as former, rather than current, politicians. We assume, however, that respondents will tend to see both former and current politicians as part of the same group, in particular because our stimuli highlighted that members of the political commission were selected by current parliamentarians. In other words, they acted on behalf of the current politicians, their former colleagues. Another potential disadvantage of using experimental stimuli based on real life events is that respondents might react to current discussions about this event rather than just the experimental stimuli. However, while there had been some debate surrounding the commission’s proposal almost two years prior to our experiment, the proposal of the commissions had all but completely vanished from the public agenda at the time of our experiment. 7
Our experiment was full factorial, so this manipulation of the two factors resulted in four experimental conditions. In addition, a control group was not exposed to any proposal or description of a commission, meaning that the experiment had a total of five conditions.
Post-Treatment Measures
After reading the description of the commission and their proposal, respondents were asked to indicate their support or opposition to the proposal on a 5-point scale, ranging from “strongly opposed” to “strongly in favor.” Next, trust in politicians’ warmth and competence were measured with two questions batteries, containing items adapted from previous trait, and trust studies (Bøggild, 2016; Fridkin and Kenney, 2011; Funk, 1999; Harring, 2015; Koch and Obermaier, 2016; Pedersen, 2017). An index for trust in warmth was based on four of these items (6d, 6e, 6f, and 7c), forming a reliable scale (α = 0.84). 8 The index for trust in competence was based on five items (6a, 6b, 6c, 7a, and 7b), also forming a reliable scale (α = 0.89). The distributions of respondents’ values across these two measures are shown in Appendix A (available online). While our post-treatment measures of trust specifically measure the two dimensions underlying trust (warmth and competence), our pre-treatment measure of trust only addresses general political trust (c.f. hypothesis 4). This feature of the survey was based on considerations regarding the risk of respondent fatigue and survey cost (all three measures of trust correlate strongly, see Appendix C, available online).
Respondents were then asked two questions that served as checks of our experimental manipulation. First, respondents were asked to indicate the degree to which they thought that the politicians in parliament had influenced the proposal of the commission. Second, respondents were probed on their ability to remember the content of the proposal. Finally, respondents were debriefed (see Appendix B, available online, for the entire survey questionnaire).
Results
As specified in our preregistration analysis plan, the tests of the hypotheses are based on regression models (ordinary least squares (OLS)), which include as predictor variables: dummies for experimental conditions, the pre-treatment measures of political trust and inequality attitude, as well as interaction terms of the experimental conditions and trust, and the experimental conditions and inequality attitude. Detailed results for all the planned models are shown in Appendix D (available online).
The Main Effects of Proposal Content and Delegation
We start the analysis by looking at the effect of proposal contents on support for proposal. Figure 1 shows that—just as expected—there are strong preferences for pay decreases for politicians relative to pay increases (Figure 1 is based on Model 1 in Appendix D, available online). Respondents exposed to the suggested pay increases, from either the political commission or the independent commission, exhibited very low support for the proposal. In contrast, respondents exposed to a proposed decrease in politicians’ pay were much more strongly in support for this proposal. 9 As evident from Figure 1, these differences between the first two condition and the last two conditions are highly significant (p < 0.001). These results are clearly in line with hypothesis 1.

Support for Proposal, Conditional on Treatment.
The results illustrated in Figure 1 also allow us to test hypotheses 2 and 3. In line with the prediction in hypothesis 2, pay increases are even more unpopular when suggested by the political committee rather than an independent committee. A proposed increase from the political commission had a mean value on the measure of support of just 0.25 (95% confidence interval (CI) = (0.22–0.27)), while support was 0.30 (95% CI = (0.27–0.32)) when the independent commission suggested the increase (the two means differ significantly from each other, p = 0.007). This result suggests that aversion to high pay for politicians to some degree may be driven by an aversion to politicians being overly self-interested, not just the high pay in itself. Accordingly, delegation of such decisions to independent entities can in fact lessen opposition. However, it should be noted that while the effect of the sponsor is statistically significant, it is relatively small, moving support on the scale ranging from zero to one with just 0.05 points (95% CI = (0.01–0.08)). Furthermore, we do not see the same pattern when looking at the two groups being exposed to proposed decreases in pay. Here, support for the proposal seems to be completely unaffected by the sponsor of the proposal. Support for a pay decrease suggested by the political commission is 0.68 (95% CI = (0.65–0.70)) just as it is 0.68 (95% CI = (0.65–0.70)) when proposed by the independent commission. (The differences between means is just 0.0005 points and insignificant, p = 0.978.). Thus, while we find support for hypothesis 2, the data do not support hypothesis 3.
Political Trust as a Moderator
Next, we turn to the moderating role of political trust. An estimation based on Model 1 shows that the expected support for an increase in pay is significantly higher for higher levels of trust, regardless of whether this proposal originates from the political commission or the independent commission. Conversely, expected support for a pay decrease is significantly lower when trust level are high, again regardless of whether the proposal came from the political or independent commission. (The marginal effect of trust is significant, p < 0.001, in all conditions, see Appendix D, Table D2, available online.) These results are clearly in line with hypothesis 4. Furthermore, the differences in responses across trust levels are not just statistically significant but they are also substantive in size. To illustrate this, Figure 2 displays the expected support for the proposals for an individual with a low level of trust (1 SD below the mean) and a high level of trust (1 SD above the mean). 10

Support for Proposal, Conditional on Treatment and Trust.
As shown in Figure 2, an individual with a low trust level has a very high level of expected support for a pay decrease, regardless of whether the proposal is coming from the political commission or the independent commission. Averaging across these two conditions, a low trusting individual has an expected support of 0.76 on the 0–1 scale. In contrast, when presented with a proposed pay increase, an individual with a low level of trust has an expected support for this proposal of only 0.17 (again, averaging across the two conditions presenting pay increases). Thus, for such individuals with low trust, the relative preference for a pay decrease (i.e. the difference in support for a decrease and an increase) is estimated to be 0.59 points. In contrast, individuals with a high level of trust have a relative preference for a pay decrease of just 0.21 points (for an individual with a high level of trust, the expected level of support is 0.59 for a pay decrease and 0.38 for a pay increase). This marked difference of 0.39 points in relative preferences is significant, p < 0.001.
It is important to note that while these results align with our theoretical reasoning on the moderating effects of trust, a causal interpretation of the results is of course based on the assumption that the apparent effects are not confounded by unobserved variables that simultaneously affect trust and support for this proposal. Such confounders cannot be ruled out, but it is also worth noting that the analysis is based on a regression which controls for the effect of inequality attitudes. (Political trust and inequality aversion correlate negatively, although not very strongly, r = –0.19, p < 0.001; cf. Appendix C, available online.) In addition, an exploratory analysis shows that if we add to the model additional covariates (gender, age, education, and party choice) as well as interactions between these covariates and treatment, the moderating role of trust remains substantially unchanged (see Appendix D, Table D2, available online).
Inequality Attitudes as a Moderator
As the next step, we look at the moderating role of inequality aversion. Notably, the expected support for a pay increase for politicians does not increase significantly with higher levels of acceptance of economic inequality. The marginal effect is insignificant when respondents are exposed to an increase proposed by the political commission (p = 0.811), and it is also insignificant for respondents exposed to an increase proposed by the independent political commission (p = 0.060). However, for the respondents exposed to a proposed decrease, the marginal effect of inequality attitudes is significantly negative, regardless of whether the proposal is coming from the political or the independent commission (p = 0.002 in both cases). Thus, the relative support for a pay decrease is higher among individuals with a high level of inequality aversion, as proposed by hypotheses 5. Once again, these results are robust to the inclusion of additional covariates (see Appendix D, Table D2, available online).
Again, to get a better grasp of the substantive differences, we can compare the expected values for an individual low on inequality aversion (1 SD below the mean) versus an individual high on inequality aversion (1 SD above the mean). Figure 3 illustrates support for the proposal, conditional on experimental treatment and aversion to economic inequality. The measure of inequality aversion is scaled 0–1, where a high value indicates a high level of inequality aversion.

Support for Proposal, Conditional on Treatment and Inequality Aversion.
Here, a respondent with a low level of aversion to economic inequality would have an expected relative preference for a pay decrease of 0.35 points, whereas a respondent placed 1 SD above the mean would have an expected relative preference for a pay decrease of 0.46 points. (These effects do differ significantly from each other, p < 0.001.) Compared with the moderating role of trust, the moderating role of inequality aversion seems modest. A formal test confirms this: partial omega squared for the interaction between experimental condition and trust is 0.125 (95% CI = (0.095–0.155)), while it is just 0.010 (95% CI = (0.001–0.021)) for the interaction between experimental condition and inequality attitude.
So far, we have seen that, overall, respondents have strong preferences for politicians receiving a pay decrease relative to a pay increase (in line with hypothesis 1). This aversion is even more pronounced when an increase is being suggested by the political commission (in line with hypothesis 2), whereas the sponsor did not have a significant effect on the popularity of a pay decrease (disconfirming hypothesis 3). Furthermore, we have seen that this relative preference for a pay decrease is moderated by political trust (in line with hypothesis 4) and inequality attitudes (in line with hypothesis 5) and that the moderating role of trust is larger than the moderating role of inequality attitudes. Thus, taken together it seems that trust really does matter a great deal for attitudes on pay. Low trust toward politicians is clearly associated with preferences for low pay. In the last part of the analysis, we see whether this relationship between trust and pay could perhaps work in the other direction.
From Pay Changes to Trust
Are proposed changes in politicians’ pay an efficacious way of affecting citizens’ trust in these politicians? The results of our study suggest that the answer to this question is no. As illustrated in Figure 4, none of the experimental treatments or the control group differed from each other when comparing the mean levels of trust in politicians’ warmth (all differences between conditions are statistically insignificant, p > 0.05). Similarly, none of the conditions, differed from each other when comparing the mean levels of trust in politicians’ competence. These results are not in line with hypotheses 6 and 7, which posited that citizens exposed to a proposed pay increase should end up with lower levels of trust in politicians’ warmth and competence than citizens exposed to a proposed pay decrease.

Trust in Politicians’ Warmth and Competence, Conditional on Treatment.
The lack of significant effects could potentially be due to our measures of trust in politicians’ warmth and competence. However, we note that both of these measures correlates fairly strongly with the more traditional pre-treatment measure of trust in politicians (cf. Appendix C, available online). Furthermore, several of the items used in our scales of trust in warmth and trust in competence have previously been used as standard measures of political trust. 11 Even if we look separately at these typical measures of political trust, the null-finding stands.
Finally, without any significant effects on either dimensions of trust, there is no support for hypothesis 8, which proposed that effect sizes would be larger when a proposal came from the political commission: differences in effects size are insignificant for trust in warmth (p = 0.95) as well as for trust in competence (p = 0.84).
Why are the respondents’ trust in politicians so unaffected by the proposed pay changes? One explanation could be that our experimental treatment simply wasn’t strong enough. For example, respondents were exposed to a potential, not an actual, change in pay. Another possible explanation for this lack of a significant effects on the two types of trust could be that respondents had not paid attention or remembered the content of the proposals presented to them. However, at the end of the survey, when asked to recall the proposed percentage change in politicians pay among four alternatives, 81.6% of respondents correctly remembered the percentage (9.0% of the respondents replied “don’t know,” while 9.4% of the respondents gave an incorrect answer). 12 We suggest that at least some part of the explanation for these null-effects could also be that respondents’ had a tendency to interpret the proposals on pay changes in ways which confirmed their generally skeptical view of the politicians. Related to this, Figure 5 shows the respondents’ mean assessment of the extent to which the parliamentary politicians influenced the proposed pay change. As one can see, respondents did respond to the description of the commissions. Conditional on the other experimental factor (i.e. pay decrease versus increase), respondents did perceive the independent commission as being less influenced by the politicians than the political commission. As such, our manipulation of this factor worked as intended.

Perceptions of Influence, Conditional on Treatment.
However, the content of the proposal also had substantial effects on respondents’ perceptions regarding the independence of the commission. For example, when the independent commission suggested a pay increase, the mean assessment of influence from the parliamentary politicians (i.e. the politicians benefiting from such an increase) was 0.56 (95% CI = (0.53–0.59)), while it was just 0.37 (95% CI = (0.34–0.40)), when the same commission suggested a pay decrease. (We see the same pattern when comparing the increase and decrease from the political commission.) Thus, whenever a pay increase is suggested, respondents generally assume that politicians had some influence on this proposal, and such proposals may therefore simply confirm peoples’ prior notions of self-interested politicians. In contrast, when a pay decrease is suggested, respondents tend to believe that politicians had a limited influence on the proposal, even when the proposal came from the political commission. Respondents with prior notions of politicians being self-interested, may thereby retain this notion by assuming that the politicians had relatively little to do with this proposal.
This response pattern seems to us to be a typical example of directional motivated reasoning, where previously held attitudes structure how new information is interpreted (Leeper and Slothuus, 2014). This explanation of the null-effects on trust is further corroborated by an additional exploratory analysis: the perceptions of political influence on the pay proposal are particularly affected by the content of the proposal among the respondents with low pre-treatment levels of trust. In contrast, respondents with a high pre-treatment level of trust generally respond more strongly to the description of the commission (see Appendix E for this additional analysis, available online).
Conclusion and Discussion
The relatively strong preference for lower pay for politicians found among the participants in our study corresponds well with earlier studies, all showing majorities of voters to be averse to high pay for politicians (Kelley and Evans, 1993; Kiatpongsan and Norton, 2014; Theriault, 2004). Thus, our study adds further weight to the assertion that aversion to high pay for politicians is a general phenomenon across different contexts. In fact, our finding on the importance of political trust for these pay attitudes suggests that Denmark, where trust in politicians is still comparatively high, should be one of the countries where we should have been least likely to find a strong aversion to higher pay for politicians.
As shown by our experiment, this aversion toward pay raises for politicians can be lessened through institutional delegation. Reactions to proposed pay changes do depend on the sponsor of such proposals, as opposition to a pay raise for politicians was weaker when it was proposed by an independent commission instead of a commission closely aligned with the politicians. However, while the effect of such delegation was statistically significant, it is important to note that it was substantially modest. The limited size of the effect may partly be due to the fact that people’s perceptions regarding politicians’ influence on pay proposals are heavily affected by the content of such proposals. Whenever people were presented with a proposal containing a pay raise for politicians, they generally assumed some influence over this proposal from the politicians themselves, even when the commission was described as being independent. In contrast, politicians generally get far less credit for proposals containing a pay decrease.
Furthermore, this tendency to view the independence of the commission in light of their proposals was most pronounced among respondents with low levels of political trust. Thus, these people that one might expect to be most strongly in favor of institutional delegation—because it would take decisions out of the hands of ostensibly untrustworthy politicians—are also the people that tend to distrust that this delegation is genuine. This pattern relates to a key finding of our study: the role of political trust. Citizens with low levels of political trust exhibit much stronger preferences for pay decreases for politicians than citizens with high levels of trust. Unsurprisingly, general economic attitudes also matter for citizens’ attitudes on pay, but trust seems to be a more important predictor.
The relatively importance of trust and inequality attitudes could potentially have been different, had we informed our respondents about the existing levels of politicians’ pay. Having knowledge about the exact pay levels of politicians could potentially make it easier for respondents to rely on their attitudes regarding economic inequality when evaluating pay changes. Future studies may therefore want to investigate how respondents react when provided with such information. However, studies often find that peoples’ attitudes on policy questions can be remarkably unaffected by relevant numbers (Hopkins et al., 2018; Kuklinski et al., 2000; Lawrence and Sides, 2014; although see Mérola and Hitt, 2015; Schueler and West, 2015). We suspect the same might be the case, when citizens are asked to make up their mind on politicians’ pay. Rather than carefully considering whether pay levels are in line with their own principles and attitudes on income inequality, people may respond more viscerally and base their response on their general affect toward the politicians.
What then are the potential consequences of such a strong opposition to high pay for politicians among the voters? One potential consequence could be that politicians, fearing the reactions of their voters, abstain from pay increases, and thereby gradually let their pay levels drop behind. This could potentially affect the skill-levels, effort, and representativeness of politicians, although recent studies have found somewhat differing results on the consequences of pay levels for politicians (Braendle, 2015; Carnes and Hansen, 2016; Fisman et al., 2015; Hoffman and Lyons, 2013; Keane and Merlo, 2010). A second, perhaps more insidious consequence of the public aversion to high pay for politicians, might be that politicians opt for less transparent reward structures for themselves. The fate of the actual proposal from the Danish commission on politicians’ pay is an illustrative example of this.
The Danish commission on politicians’ pay, Vederlagskommissionen (2016), presented their proposal after almost 2 years of work. On the face of it, their proposal contained a large raise in politicians’ pay. However, the proposal also contained the removal of several tax exempt salary supplements and adjustments of the lucrative pension schemes enjoyed by the politicians. The actual gross change in pay and pensions for the politicians were therefore a modest raise for the members of parliament and a decrease in pay for cabinet ministers. Nevertheless, despite a prior agreement to accept the proposal of the commission, the main parties in the Danish parliament balked at the appearance of a large pay increase, and they ended up largely rejecting the proposal, thereby continuing with an opaque reward structure (Rohde and Jørgensen, 2016).
Finally, our experiment suggests that while voters may have very strong opinions regarding proposed changes in politicians pay, such proposals do not seem to affect subsequent levels of trust in politicians. The absence of such an effect in our experiment could perhaps be explained by the notion that talk is cheap, since respondents in the experiment were exposed to proposals regarding changes in politicians’ pay rather than actual changes in politicians’ pay. This is a limitation of our study, and citizens might react differently to actual changes in pay initiated by current politicians with a more obvious self-interest in pay levels. However, we tend to suspect that even such actual changes in politicians pay might be interpreted with the same level of skepticism and reliance on prior perceptions of politicians as our respondents seemed to do. While pay raises confirm perceptions of self-interested politicians, people seem loathe to interpret decreases in politicians pay as genuine selflessness on behalf of the politicians. In so far as that is the case, this suggests that changes in politicians’ pay will not have an effect on people’s trust in politicians. Thus, the relationship between trust and politicians pay may very well be a one-way street: People with low trust toward politicians clearly want politicians to make less than they currently do, but lowering the salaries of politicians—or delegating such decision to independent entities—does not seem to abate the distrust toward them.
Supplemental Material
PSX850073_Supplemental_material – Supplemental material for Citizen Attitudes on Politicians’ Pay: Trust Issues Are Not Solved by Delegation
Supplemental material, PSX850073_Supplemental_material for Citizen Attitudes on Politicians’ Pay: Trust Issues Are Not Solved by Delegation by Rasmus T Pedersen and Lene Holm Pedersen in Political Studies
Footnotes
Acknowledgements
A previous version of this paper was presented at APSA 2018, Boston, and DPSA 2018, Vejlefjord. For very thoughtful and constructive comments, we thank Mia Costa, Jens Olav Dahlgaard, Malte Rokkjær Dahl, and Kristina Jessen Hansen.
Funding
The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work is funded by the Danish Council for Independent Research, Social Sciences [Grant no: DFF-6109-00052].
Supplemental Material
Supplemental material for this article is available online.
Figure A1: Distribution of respondents’ values across key variables.
Figure E1: Perceptions of influence, conditional on treatment and trust.
Figure E2: Perceptions of influence, conditional on treatment and inequality aversion.
Table A1: Comparison of sample respondents and Danish population.
Table C1: Correlations.
Table D1: Models planned in pre-registration analysis plan.
Table D2: Marginal effects of trust and inequality aversion on support for proposals.
Notes
Author Biographies
References
Supplementary Material
Please find the following supplemental material available below.
For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.
For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.
