https://orcid.org/0000-0003-1481-7268
Abstract
Citizens value competence in politicians. However, while offering high pay is a key strategy when recruiting competent candidates in the job market, most citizens are highly averse to paying politicians higher salaries. We argue that this aversion arises from one fundamental concern among citizens: self-interested politicians. Using data from a large-scale preregistered survey experiment, we show that citizens are affected by the argument that higher salaries may attract self-interested politicians, whereas they are not affected by the argument that higher salaries will attract competent politicians. Surprisingly, the more positively citizens view politicians, the more they are affected by the argument about self-interested politicians. These results suggest that citizens may view modest salaries as a guard against self-interested politicians.
Politicians have important jobs. Holding the reins of state power, politicians make decisions with profound consequences for millions of citizens. Thus, it matters to citizens that politicians are competent (Funk, 1996; Huddy & Terkildsen, 1993; Kinder et al., 1980; McGraw, 2011). One might therefore expect that citizens would support higher salaries for politicians. After all, paying a high salary is an important way to attract and retain the most competent candidates in the job market (Murphy & Zabojnik, 2004; Weiss, 1980). However, despite citizens’ desire for competent politicians, the idea of paying politicians high salaries is not popular. Large majorities of citizens think that the salaries of politicians are already too high, and suggestions of salary increases for politicians are exceedingly unpopular (Kelley & Evans, 1993; Pedersen & Pedersen, 2020; Theriault, 2004).
These attitudes regarding politicians’ salaries are important in their own right, as they may affect the actual salary levels of politicians and thereby affect the subsequent recruitment and behavior of politicians (Braendle, 2015; Carnes & Hansen, 2016; Fisman et al., 2015; Hoffman & Lyons, 2013). Furthermore, citizens’ attitudes regarding the salaries of politicians are also relevant because they can provide us with insights into more fundamental questions about what citizens perceive to be the right type of politician. Consequently, in this article we argue that politicians’ salaries present a dilemma for the public: If salaries to politicians are raised, this may attract politicians who are more competent, but higher salaries might also attract politicians who are mainly driven by their own self-interest.
We test several preregistered hypotheses pertaining to this dilemma with a survey experiment conducted on a large and demographically diverse sample in Denmark (n = 2,520). Our results show that citizens are affected by the argument that higher salaries carry the risk of attracting self-interested politicians, while they are not affected by the argument that higher pay will attract competent politicians. Moreover, citizens with the most positive perceptions of politicians’ warmth and competence are most affected by the argument that higher salaries risk attracting self-interested politicians. Furthermore, consistent with these experimental results, exploratory analyses show that perceptions and attitudes towards politicians’ warmth are key predictors of citizens’ attitudes on politicians’ salaries. By contrast, perceptions of politicians’ competence and the view that competence is important are not significantly associated with attitudes regarding politicians’ salaries.
We conclude by discussing the implications of these results. In many countries, politicians are authorized to set their own salaries (Brans & Peters, 2012; Mause, 2014), but as citizens are ultimately the principals of elected officials, public approval is important. Our findings suggest that when people with the most positive impressions of politicians are also the most responsive to arguments highlighting the risks of self-seeking politicians, it will be very difficult to initiate an open public debate to gather support for paying politicians more.
Public aversion to high salaries for politicians
Citizens want their politicians to be competent. When describing the ideal presidential candidate, people describe a competent candidate (Kinder et al., 1980), and people evaluate politicians more positively if they are knowledgeable, intelligent, and skilled (Funk, 1997; Laustsen & Bor, 2017; McGraw, 2011). Given that politicians are routinely up for election, the demand for competent politicians could be addressed by increasing politicians’ pay. This proposition builds on the classic tenet of labor market theory that higher salaries will attract the best candidates (Caselli & Morelli, 2004; Murphy & Zabojnik, 2004; Weiss, 1980). Whether this also holds for politician is contested. While several studies find that high salaries can increase the competence levels and performance of politicians (Braendle, 2015; Diermeier et al., 2005; Gagliarducci & Nannicini, 2013), other studies question these results (Carnes & Hansen, 2016; Fisman et al., 2015; Keane & Merlo, 2010).
Irrespective of the scholarly debate about the fruitfulness of paying politicians high salaries, it is clear that citizens have a strong aversion towards high salaries for politicians. Studies covering multiple countries show that most citizens think that politicians’ salaries are too generous, and most citizens are correspondingly also against salary increases for politicians (Kelley & Evans, 1993; Pedersen & Pedersen, 2020; Theriault, 2004). Citizens’ aversion to high salaries for politicians could reflect a broader aversion to large-scale societal economic inequality. A desire for lower levels of inequality is prevalent among citizens (McCall, 2013), and full-time politicians have higher incomes than the average citizen (Brans & Peters, 2012; Hood & Peters, 1994; Mause, 2014). However, while inequality aversion is correlated with opposition to pay raises for politicians, citizens with very low levels of inequality aversion are still firmly against increased salaries for politicians (Pedersen & Pedersen, 2020). This suggests that citizens’ opinion formation process on the issue of politicians’ salaries differs from how they form opinions about high-income groups such as CEOs. For such high-income occupations, political predispositions and support for free market ideology have been shown to have a substantial impact on attitudes regarding their salaries (Burak, 2018; Burak & Cumberworth, 2020).
Competence and warmth as a trade-off
We propose that in order to understand attitudes towards politicians’ salaries, we need to understand citizens’ concerns about politicians’ personalities and character traits. Here, traits pertaining to competence are important, but traits that reflect intentions (i.e.,“warmth”) are even more important (Cuddy et al., 2008; Fiske et al., 2007; Levi & Stoker, 2000). Social psychology research has shown that people across cultures are highly sensitive to cues reflecting whether others are well-intentioned, because these cues help determine whether another person intends to act in a way that will benefit others or serve herself at the expense of others (Cuddy et al., 2008; Fiske et al., 2007). Traits that relate to the warmth dimension include friendliness, helpfulness, sincereness, and morality, whereas being selfish, self-interested, deceitful, and dishonest belong to the opposite pole of the warmth dimension (Goodwin et al., 2014; Zhang & Wang, 2018). The competence dimension instead reflects traits related to abilities, for example, intelligence, industriousness, and determination (Cuddy et al., 2008; Koch & Obermaier, 2016). Extant research argues that people spontaneously form impressions of other individuals—and social groups such as the rich, welfare recipients ethnic groups, and politicians—along the warmth and competence dimensions (Cuddy et al., 2008; Fiske et al., 2007; Funk, 1997; Laustsen & Bor, 2017). Traits on the warmth dimension are strongly associated with trustworthiness because they pertain to a person’s intentions. Competence tends to be secondary, as competence traits reflect whether others are capable of carrying out their—good or bad—intentions (Cuddy et al., 2008; Levi & Stoker, 2000).
It should be noted that research sometimes divide warmth into several subdimensions such as moral character and social warmth (Goodwin et al., 2014). Similarly, much research within political science has also used more dimension when looking at personalities (Hardy, 2017), but research also finds there are two overall dimensions on which these traits cluster together: competence and warmth, although the latter is sometimes termed character (Bittner, 2011, 2018). It should also be noted that warmth and competence do not encapsulate all the traits that voters asses politicians on, but several studies confirm that voters do assess politicians along these dimensions, and perceptions on the two dimensions have substantial effects (Laustsen, 2017; Pedersen, 2017). Consequently, we use these two overarching dimensions in this study.
Given that warmth and competence are two distinct dimensions, they do not necessarily go hand in hand. A highly competent person may also be a cold person who behaves in a selfish manner. Vice versa, a warm and prosocial person might be incompetent. Consequently, people often face a basic interpersonal dilemma when they evaluate others and when they manage the impression they display to others (Cuddy et al., 2011; Swencionis et al., 2017). This dilemma is particularly pertinent when it concerns the use of monetary incentives. On the one hand, exposure to money entices people to display competence; on the other hand, people also become less prosocial and more unethical when exposed to money (Kouchaki et al., 2013; Vohs, 2015). Thus, citizens’ preferences for warm and competent politicians may come into conflict when citizens develop opinions regarding salary increases for politicians. While higher salaries can attract competent people (Gagliarducci & Nannicini, 2013), there is also a risk that monetary incentives for politicians could crowd out politicians’ genuine desires to serve public interests (Besley, 2004), and people with cold personality traits are generally more likely to be attracted to organizations that offer higher pay (Boddy, 2010).
These considerations regarding salaries are also in play in discussions about salaries for public office: In a qualitative discourse analysis across 17 countries, Dekker (2013) identifies both the “motivation selection argument,” which operates on the assumption that officials should not be in it for the money, and the conflicting “optimal performance argument,” which holds that achieving the best possible performance from high public officeholders requires having the most capable individuals sit in such positions. The question is how do citizens solve this dilemma in the context of politicians pay?
Hypotheses
To investigate how citizens’ concerns about politicians’ competence and warmth shape public opinion about politicians’ salaries, we test several hypotheses. Given that people both value competence and warmth in politicians, we hypothesize that people become less opposed to higher salaries for politicians if exposed to the argument that higher salaries would attract competent politicians (Hypothesis 1) and people become more opposed to higher salaries for politicians when exposed to the argument that higher salaries would attract self-interested politicians (Hypothesis 2). Given that trust in politicians is a fundamental concern in the relationship between citizens and politicians (Levi & Stoker, 2000), we also hypothesize that people are more affected by the argument that higher salaries attract self-interested politicians than the argument that higher salaries attract competent politicians (Hypothesis 3). This hypothesis suggest that people mainly solve the dilemma of politicians pay by prioritizing the concern about their self-interest, which builds on the key finding that trustworthiness stems from perceptions of character traits pertaining to warmth (Cuddy et al., 2008; Goodwin et al., 2014) and that people tend to evaluate politicians and the elite more on their warmth than on their competence (Hansen, 2018; Laustsen & Bor, 2017).
To form predictions about moderating effects, we build on the classic notion that people are most persuaded by an argument if it matches their underlying values and preexisting beliefs (Chong & Druckman, 2007; Taber & Lodge, 2006; Zaller, 1992). Thus, our overall hypothesis is that the effects of arguments regarding competence and self-interest are moderated by peoples’ attitudes and perceptions regarding the competence and warmth of politicians (Hypothesis 4). It is important to note that these respondent characteristics are not experimentally manipulated, and our analyses do not estimate the direct causal effects of such characteristics. Rather, we can only analyze how these personal characteristics modify the effects of the experimental treatments—see Keele et al. (2020) for the distinction between causal interaction and effect modification. We test Hypothesis 4 in several ways:
First, whether the competence argument persuades participants may depend on the degree to which participants view the occupation of a politician one that only few people are able to perform. We term this perceived “skill uniqueness,” and we introduce a novel measure of this perception in our survey. This enables us to test whether citizens’ reasoning resembles the economic notion that the problem of a short supply of unique skills in the labor market can be addressed by raising salaries (Weiss, 1980). Thus, we hypothesize that people rating the job of politicians to have a high level of skill uniqueness are more affected by the competence argument, compared to people rating the job of politicians to have a low level of skill uniqueness (Hypothesis 4a).
Second, whether people are persuaded by the competence and warmth arguments may depend on the degree to which they rate warmth and competence to be important traits for politicians to have. Previous work has shown that voters may place different emphases on such traits (Laustsen, 2017). We expect that people attaching high importance to politicians’ competence are more moved by the competence argument, compared to people attaching low importance on politicians’ competence (Hypothesis 4b), whereas we expect people attaching high importance to politicians’ warmth are more moved by the self-interest argument, compared to people attaching low importance on politicians’ warmth (Hypothesis 4c).
People’s reactions may also depend on the degree to which they find the current politicians to be lacking in competence or warmth. We therefore test whether people perceiving current politicians to have a low level of competence are more moved by the competence argument, compared to people perceiving current politicians to have a high level of competence (Hypothesis 4d). This hypothesis is based on the notion that people already perceiving politicians to be competent would generally find little need to increase competence through even higher pay. In contrast, among people perceiving the current politicians to have low levels of competence, the argument that higher pay attracts competence should resonate with their preexisting belief about the need for more competent politicians. Finally, we test whether people perceiving current politicians to have a low level of warmth are more moved by the self-interest argument, compared to people perceiving current politicians to have a high level of warmth (Hypothesis 4e). The reasoning underlying this hypothesis is similar to the reasoning underlying Hypothesis 4d. The argument that higher pay could be detrimental to warmth might resonate stronger with people concerned about low levels of warmth among politicians.1
Method
We test our hypotheses with an experiment embedded in a survey. Prior to data collection, all hypotheses were preregistered along with a preanalysis plan at the Open Science Framework (see Supplementary Appendix C).2 We conduct the experiment within the context of Danish politics. The Danish parliament, Folketinget, has 179 members. These members receive a salary approximately 1.7 times greater than the salary of the average Danish citizen. Compared to other European countries, this reflects a relatively modest difference in salaries (Mause, 2014). Furthermore, the salaries of politicians have increased at a lover pace than salaries among senior civil for some decades (Vederlagskommissionen, 2016). Nevertheless, there is widespread aversion to salary increases for Danish politicians (Pedersen et al., 2019). The experiment was conducted with a commercial web panel (Voxmeter) in January and February 2020. While 2,586 respondents started the survey, 2,520 completed the entire survey (completion rate: 97.4%). Virtually all of the attrition occurred prior to exposure to the experimental stimuli, as just two respondents (0.08%) left the study after exposure to the stimuli. Our final sample is a demographically diverse sample: 49.7% of the sample is female, and age ranges from 18 to 90 years old (M = 49.1 SD = 16.0). The sample also shows substantial variation in political attitudes, and the sample is approximately representative of the Danish population on party choice (for sample and population characteristics, see Supplementary Appendix A). Furthermore, we note that treatment effects in survey experiments are generally highly consistent across probability and nonprobability samples (Coppock et al., 2018).
Pretreatment covariates
Respondents were first asked about basic demographics and vote choice. Participants were then surveyed about self-placement on the political left–right scale and three specific questions on their attitudes towards economic inequality. Together, these questions formed a reliable scale on inequality attitude (Cronbach’s α = 0.79). To measure participant perceptions of the skill uniqueness of politicians, the respondents were asked the following: “Think about the job as a politician in parliament. In your opinion, how many people in society would be qualified to do this job, if they were given the chance?” Subsequently, trust in politicians was measured with three survey items, forming a reliable index (α = 0.82).
To measure the importance ascribed to politicians’ competence and warmth, we asked the participants to rate how important it was for politicians to possess different traits associated with these two dimensions. The scale for importance of competence was based on the traits intelligent, skilled, knowledgeable, and hardworking (α = 0.80), while the scale for importance of warmth was based on the traits likeable, friendly, and caring (α = 0.80). A fourth item, conscientious, was dropped from this scale in accordance with our preregistration plan, as it was found to decrease the internal reliability of the scale. All of these trait items have previously been used to measure the dimensions of competence and warmth (Bøggild, 2016; Fridkin & Kenney, 2011; Koch & Obermaier, 2016; Pedersen, 2017).
Finally, to measure perceptions of competence and warmth of current politicians, the participants were asked to rate politicians in parliament on the same trait items. These items also formed highly reliable scales for politicians’ competence (α = 0.85) and warmth (α = 0.83).3 Prior to analyses, all continuous measures were rescaled to range from 0 to 1.
Experiment
Following the measurements of our pretreatment variables, our survey participants were randomly sorted between four experimental conditions. All participants were presented with the following statement: “The salary of the Danish parliamentarians consists of a remuneration and a supplement for expenses. In total, their monthly pay before taxes is approximately DKK 62,000. It was recently proposed that the salary of the parliamentary politicians be increased in the future.” Depending on the experimental condition assigned to them, participants were then presented with the arguments (Table 1).
Experimental conditions.
| Condition A: Control | Condition B: Competence argument | Condition C: Self-interest argument | Condition D: Competence and self-interest arguments |
|---|---|---|---|
| [empty, no argument] | One of the arguments in favor of a salary increase is that if the salary were increased, we would in the future get more politicians that are competent and able to make the decisions that are best for the Danish society. | One of the arguments against a salary increase is that if the salary were increased, we would in the future get more politicians that think more about themselves than about what is best for the Danish society. | [Participants in this condition were exposed to the arguments from Conditions B and C. The ordering of the two arguments was randomized.] |
| Condition A: Control | Condition B: Competence argument | Condition C: Self-interest argument | Condition D: Competence and self-interest arguments |
|---|---|---|---|
| [empty, no argument] | One of the arguments in favor of a salary increase is that if the salary were increased, we would in the future get more politicians that are competent and able to make the decisions that are best for the Danish society. | One of the arguments against a salary increase is that if the salary were increased, we would in the future get more politicians that think more about themselves than about what is best for the Danish society. | [Participants in this condition were exposed to the arguments from Conditions B and C. The ordering of the two arguments was randomized.] |
Experimental conditions.
| Condition A: Control | Condition B: Competence argument | Condition C: Self-interest argument | Condition D: Competence and self-interest arguments |
|---|---|---|---|
| [empty, no argument] | One of the arguments in favor of a salary increase is that if the salary were increased, we would in the future get more politicians that are competent and able to make the decisions that are best for the Danish society. | One of the arguments against a salary increase is that if the salary were increased, we would in the future get more politicians that think more about themselves than about what is best for the Danish society. | [Participants in this condition were exposed to the arguments from Conditions B and C. The ordering of the two arguments was randomized.] |
| Condition A: Control | Condition B: Competence argument | Condition C: Self-interest argument | Condition D: Competence and self-interest arguments |
|---|---|---|---|
| [empty, no argument] | One of the arguments in favor of a salary increase is that if the salary were increased, we would in the future get more politicians that are competent and able to make the decisions that are best for the Danish society. | One of the arguments against a salary increase is that if the salary were increased, we would in the future get more politicians that think more about themselves than about what is best for the Danish society. | [Participants in this condition were exposed to the arguments from Conditions B and C. The ordering of the two arguments was randomized.] |
Following these arguments, all participants were asked, “Are you opposed to or in favor of this proposal to increase the salaries of the parliamentary politicians in the future?” (the entire survey questionnaire, including the experiment, can be found in the Supplementary Appendix B, while descriptive statistics for our measures are shown in Supplementary Appendix D, Tables D1 and D2). It is important to note that the experiment was not designed with the expectation that any argument could turn a majority of participants into supporters of a salary increase for politicians. On average, citizens are so strongly against salary increases for politicians that we would also expect this to be the case regardless of the arguments presented to them in our experiment. Rather, the point of the experiment was simply to investigate the degree to which the respondents responded to the arguments.
Results
We report our results in three sections. First, we investigate main effects of the arguments regarding higher pay for politicians (Hypotheses 1–3). Second, we investigate whether these effects are moderated by individual-level factors among the respondents (Hypothesis 4). Third, we report exploratory results based on the observational measures.
What argument has the strongest impact on attitudes regarding politicians’ salaries?
Following our preregistration plan, the analyses regarding the effects of our experiment are based on ordinary least squares regression models (OLS). Our primary model for testing Hypotheses 1 to 3 is a simple model that includes indicator variables for the conditions of the experiment (Model 1, Supplementary Appendix D). Based on this model, Figure 1 shows mean support for the proposal to increase politicians’ wages, conditional on experimental conditions.
Support for proposal, conditional on treatment.
Following Hypothesis 1, respondents should become relatively more supportive of higher wages when exposed to the argument that higher wages could attract more competent politicians (Condition B). However, the results do not support this hypothesis. In fact, respondents exposed to this argument are slightly less supportive of the wage increase than respondents in the control condition (Condition A), although the difference between these two conditions is not statistically significant (p = .407). In contrast, the argument that a wage increase attracts more self-interested politicians resonates more with the respondents. In line with Hypothesis 2, respondents exposed to this argument (in Condition C) are significantly less supportive of the wage increase than respondents in the control condition (p = .005).
As stated in Hypothesis 3, we expected the self-interest argument against higher wages to have a stronger effect than the competence argument in favor of higher wages. This is clearly the case. As noted above, the competence argument was wholly ineffective, resulting in a nonsignificant decrease in support, while the self-interest argument had a significant effect in the expected direction (mean support for the proposal also differs significantly between Conditions B and C, p = .044). Further supporting Hypothesis 3, respondents also become significantly more opposed to the wage raise when exposed to both the competence argument and the self-interest argument under Condition D (support for the proposal is significantly lower in Condition D than in Condition A, p = .045).
While the self-interest argument clearly had stronger effects than the competence argument, the argument did not fundamentally alter overall sentiments on the issue. Compared to the control condition, where mean support is .34, the self-interest argument decreases mean support to .29 in Condition C (an illustration of marginal effects are shown in Supplementary Appendix D, Figure D1). This is arguably a relatively modest effect on a scale ranging from 0 to 1, corresponding to a Hedge’s g effect size of 0.18. Nevertheless, it is worth keeping in mind that it is difficult to sway people when they have strong and crystallized opinions on a topic. The issue of salaries for politicians is likely to be such an issue in line with attitudes towards other elites, for example, the wealthy, which are generally stable and difficult to change (Burak, 2013).
In line with our preregistration analysis plan, we also conducted a supplementary analysis in which several prespecified covariates were added to our model (Model 2, Supplementary Appendix D). While covariates are not necessary to achieve unbiased effect estimates from the experiment, they were expected to have substantial explanatory power (Pedersen & Pedersen, 2020). Including them could therefore potentially more precise effect estimates (Mutz, 2011). The models with and without covariates yielded very similar estimates (Table D3), and the differences between the models are inconsequential for our conclusions regarding Hypotheses 1–3. Results for both models are illustrated in Supplementary Appendix D, Figure D2.
Are the effects moderated by attitudes and perceptions regarding politicians?
For the second part of the preregistered analysis, we investigate whether the experimental effects are moderated by individual-level attitudes and perceptions of politicians. We do so by adding to our simple OLS model the relevant perceptions and attitudes for each hypothesis and interactions between this perception or attitude and the experimental conditions. These models are shown in Supplementary Appendix D.
The relevant test for our moderation hypotheses is whether the interaction terms within each model differ significantly from each other. By this criterion, the data do not support Hypothesis 4a on the moderating impact of the perceived skill uniqueness of politicians (Model 4 in Supplementary Appendix D). Believing that the vocation of a politician involves a high level of skill uniqueness does have a positive impact on support for the proposed salary increase, but people with this belief are not significantly more responsive to the competence argument than people who do not hold this belief. We also illustrate the results of this model in Supplementary Appendix D, Figure D3, by showing expected support for the proposal by individuals perceiving politicians to have a low or high level of skill uniqueness. (Supplementary Appendix D, Figure D4 illustrates the corresponding marginal effects.)4 Similarly, the results are not in line with Hypothesis 4b on competence importance or Hypothesis 4c on warmth importance (Models 5 and 6 in Supplementary Appendix D, illustrated in Supplementary Appendix D, Figures D5–D8). However, perceptions regarding the current competence and warmth of politicians do interact with the arguments. Figure 2 illustrates support for the proposal to increase politicians’ wages conditional on experimental conditions and the perceived level of competence among politicians.5
Support for proposal, conditional on treatment and perceived competence.
As illustrated in Figure 2, respondents perceiving current politicians to have a low level of competence are largely unaffected by the competence argument. Support among this group is virtually identical in Conditions A (No argument) and B (Competence argument) as well as in Conditions C and D. In contrast, respondents perceiving current politicians to have a high level of competence are affected by the arguments. Compared to those of Condition A, their relative levels of support for the proposal are significantly lower in Conditions C and D, where they are exposed to the self-interest argument. (In other words, the marginal effect of competence is significantly higher in Condition A than in Condition C [p = .039] and Condition D [p = .007], Model 7 in Supplementary Appendix D.) This is not the pattern predicted by Hypothesis 4d, according to which we would only have expected respondents perceiving current politicians to have a low level of competence to be more affected by the competence argument than respondents perceiving current politicians to have a high level of competence. (The marginal effect of competence perceptions is insignificantly smaller in Condition B than in Condition A, p = .165.)
Perceptions of warmth also interact with the arguments regarding salary increases for politicians. This is illustrated in Figure 3. Respondents perceiving current politicians to have a low level of warmth are largely unaffected by the arguments, while respondents perceiving current politicians to have a high level of warmth become significantly less supportive of the proposal when exposed to the self-interest argument under Condition C. (The marginal effect of warmth perceptions is significantly higher in condition A than in condition C, p = .005, cf. Model 8 in Supplementary Appendix D.)6
Support for proposal, conditional on treatment and perceived warmth.
This interaction, however, is in the opposite direction from what is predicted by Hypothesis 4e. According to this hypothesis, one would expect respondents perceiving current politicians to have a low level of warmth to be more affected by the self-interest argument than respondents perceiving current politicians to have a high level of warmth. Thus, the results are not consistent with the classic notion that people are most persuaded by an argument if it matches their underlying values and beliefs (Taber & Lodge, 2006; Zaller, 1992). Instead, citizens with the most positive impressions of politicians (i.e., citizens who perceive politicians as being warm and competent) are most affected by the argument that higher salaries risk attracting self-interested politicians.
These interactions may reflect floor effects. Among all the participants of the study, we find that 30.5% are strongly opposed to the proposed salary increase, which is the lowest possible level of support measurable by our question, and this share is even higher for participants perceiving politicians to have a low level of warmth or competence. If we regard the potential floor effect as an issue of measurement, it can be addressed by using Tobit models (McBee, 2010). These Tobit models, which were not planned in our preregistration, are shown as Models 9 and 10 in Supplementary Appendix D. The results from these models are substantively similar to the results from our planned OLS models (the interaction between the self-interest argument and competence, which is significant in the OLS models [p = .039], becomes marginally significant [p = .055], while all other estimates remain significant).
While the results from these Tobit models suggest that the interactions are not just statistical artifacts caused by censoring on the survey response scale, it is in a sense not necessarily wrong to interpret these interactions as results of substantive floor effects. The arguments against salary increases are largely inefficacious among respondents with low perceptions of politicians’ competence and warmth because these respondents are already firmly against any salary increase for politicians. In contrast, among respondents perceiving politicians to have a high level of competence and warmth, the argument against higher salaries can affect them because they were initially—on average—only moderately skeptical of providing higher salaries for politicians.
Exploratory results based on observational measures
While the primary purpose of our study was to test causal hypotheses, we also provide an exploratory analysis of observational data from the survey. Specifically, we investigate the associations between support for pay increases to politicians and several pretreatment covariates. Our exploratory analysis is based on Model 3 in Supplementary Appendix D. The results from this exploratory model are shown in Figure 4.
Support for proposal, associations with pretreatment covariates.
As shown at the top of Figure 4, perceptions of politicians’ warmth are predictive of salary attitudes. This is consistent with the results of our experiment. People who perceive current politicians as warm tend to be relatively less opposed to raising their salaries, whereas people who perceive current politicians as cold are strongly opposed to raising their salaries. Similarly, attitudes on the importance of politicians’ warmth are also predictive of salary attitudes, as people who express that it is important that politicians have warm traits are highly opposed to raising politicians’ salaries. This corresponds well with the notion that people view modest salaries for politicians as protection against self-serving politicians.
In contrast to pretreatment perceptions of warmth, perceptions of politicians’ competence do not have a significant association with support for the proposed salary increase. Similarly, attitudes on the importance of politicians’ competence are not at all predictive of support for higher salaries. However, our measure of skill uniqueness has some predictive power. Respondents believing that few people are qualified to be politicians are relatively more in favor of salary increases for them. This measure seems to capture perceptions of politicians not captured by the traditional measures of warmth and competence, and it is worth noting that the measure is weakly and insignificantly correlated with the measures of warmth (r = −.02, p = .393) and competence (r = .02, p = .279). Finally, our data replicate previous findings regarding the importance of attitudes for economic inequality and trust in politicians (Pedersen & Pedersen, 2020). Inequality aversion is predictive of opposition to raising politicians’ salaries, but trust in politicians is a more powerful predictor.
In sum, the exploratory analyses are consistent with the experimental results. When people form opinions about politicians’ salaries, they are concerned with whether politicians have good intentions and not with whether they are competent.
Conclusion and discussion
While citizens value competence in political candidates, they also show strong aversion towards high salaries for politicians (Kinder et al., 1980; Theriault, 2004). We have argued that politicians’ salaries confront citizens with a fundamental dilemma: They want politicians who are competent, but they also want politicians who are not too self-interested and motivated by pecuniary rewards. How do citizens solve this dilemma? In their seminal work, Hibbing and Theiss-Morse (2002) argue that citizens are “amazingly attuned, hypersensitive even, to the possibility that decision makers will attempt to improve themselves at the expense of everyone else” (p. 85). Our findings suggest that one manifestation of this concern is a strong aversion against higher salaries for politicians, even when pay raises are presented as a way of attracting more competent people to the profession. This is the key finding of our experiment and the additional exploratory analyses.
Interestingly, citizens’ responsiveness to arguments made in the debate concerning politicians’ salaries is not conditional on the extent to which the arguments match citizens’ underlying values and beliefs, as proposed by classic persuasion theory (Taber & Lodge, 2006; Zaller, 1992). From this perspective, we would expect that people should be swayed by the competence argument if they strongly value competence, believe that only a few people are qualified for the job as politicians (skill uniqueness) or perceive current politicians as incompetent. Instead, the results show that regardless of how much citizens view politicians as competent and value competence, they remain unaffected by the competence argument. Instead, citizens with the most positive impressions of politicians, that is, citizens who perceive politicians as warm and competent, are those who are affected by the argument that higher salaries risk attracting self-interested politicians.
The fact that those with the most positive impressions of politicians are persuaded by arguments about the risks of self-seeking politicians has important implications. In political communication research, a central notion is that politicians can persuade voters by appealing to their values and preexisting beliefs (Aarøe & Jensen, 2015; Sniderman & Theriault, 2018; Zaller, 1992). In the case of politicians’ salaries, however, our findings suggest that a communication strategy of appealing to the value of political competence would fail. Moreover, if the competence argument is countered by the argument of self-seeking politicians in a political debate, it may even backfire.
One caveat warrants mentioning, the experimental design does not enable us to disentangle specifically why people are not swayed by the competence argument. It may be because people do not believe the premise for the argument, that the argument not common enough in the public discourse to be familiar or that it does not resonate sufficiently with people’s preexisting concerns (Chong & Druckman, 2007). This caveat notwithstanding, our results align with an emerging body of research suggesting that people tend to evaluate society’s elites more on their warmth than on their competence (Hansen, 2018; Laustsen & Bor, 2017).
Our results add to this literature by demonstrating that in the case of politicians’ pay, only warmth concerns matter, while competence concerns have no significant explanatory power. The self-seeking risk argument is likely influential despite the crystallized nature of opinions about rewarding the elite because it resonates with a strong psychological mechanism that prompts people to be vigilant towards exploitation from leaders (Bøggild, 2016; Bøggild et al., 2021; Hibbing & Alford, 2004). Thus, the results imply that it will be extremely difficult to garner public support for higher salaries for politicians because initiating such a debate would entail that those who already have negative impressions of politicians remain firmly against higher salaries for politicians, while those who view politicians positively will turn against higher salaries regardless of their values. If politicians openly present a proposal to raise their own salaries, public backlash seems inevitable. This might motivate politicians to refrain from discussing the issue publicly and choose a less transparent legislative process to raise their own salaries. Another possibility is that politicians tacitly accept that their pay levels over time tend to decrease relative to other occupations because they worry too much about the reactions of the electorate. Such a drop in politicians’ salaries could influence the composition and performance of future politicians (Braendle, 2015; Carnes & Hansen, 2016; Fisman et al., 2015).
At first glance, being authorized to set one’s own salary can be viewed as an enormous privilege, yet our research suggests that this can also be a liability. It seems that politicians choosing to raise their own salaries inevitably face backlash from the public because such pay raises confirm one of the most potent concerns of representative governments: the self-serving motives of politicians.
Supplementary Data
Supplementary Data are available at IJPOR online.
Footnotes
As pointed out by a reviewer, one could argue that people perceiving current politicians to have high warmth levels would be more concerned about changes, whereas people perceiving politicians to have low warmth levels would find that there was little to lose.
The preregistration is available at: https://osf.io/5tyxw. We deviate from the pre-analysis plan in one way. While planned sample was 2,000 respondents, our final sample include 2,520 respondents. The reason is an error made by the polling company. Respondents should have been randomly sorted with equal probability into four conditions: A, B, C, and D, with respondents in D then randomly sorted between D1 and D2. The polling company randomized with equal probability between A, B, C, D1, and D2, resulting in fewer observations than planned in A, B, and C. Upon discovery, the survey was fielded to 500 additional respondents, assigned to the conditions with the same probabilities as in the first round (Had we only randomized respondents between conditions A, B, and C in the supplementary round, assignment of treatment would be confounded by time-period of survey).
In line with previous studies, we find that perceptions of warmth and competence correlate fairly strongly (r = .66). However, a factor analysis confirms that the warmth and competence traits load on two distinct dimensions.
In all figures, a low level of the variable is defined as one standard deviation below the mean, while a high level is defined as one standard deviation above the mean. This definition of high and low values has no bearing on the conclusions regarding Hypotheses 4a–4e, as the conclusions are strictly based on the interaction terms included in Models 4–8.
A marginal effect plot for perceived competence is also shown in Figure D9, Supplementary Material, Appendix D.
A marginal effect plot for perceived warmth is also shown in Figure D10, Supplementary Material, Appendix D.
Funding
This work is supported by the Danish Council for Independent Research, Social Sciences (Grant No: DFF-6109-00052).
Conflicts of interest
None declared.
References
Rasmus T. Pedersen is a senior researcher at The Danish Center for Social Science Research (VIVE). He holds a Phd in political science from University of Copenhagen. His main research areas are political behavior and psychology, public opinion, heuristics, biases, and framing.
Kristina J. Hansen is an assistant professor at the Department of Politics and Society, Aalborg University. She holds a PhD in political science from Aarhus University. Her research field is political psychology, and her research focuses on emotions, psychological biases and political attitude formation, for example, in relation to public support for economic redistribution.
Lene H. Pedersen is professor in public management at the University of Copenhagen, Department of Political Science, and a director of research at The Danish Center for Social Science Research (VIVE). She holds a PhD in political science from University of Copenhagen.
