Prior studies of SWD

Scholars argue that public support for democratic systems of government is critical, because broad satisfaction is associated with system stability, and its absence linked with problems for both the functioning and maintenance of democracy (Powell, Reference Powell1982; Anderson and Guillory, Reference Anderson and Singer1997). Simply put, ‘if citizens have positive attitudes towards the political system they live in, they are less likely to push for radical changes’ (Bernauer and Vatter, Reference Bernauer and Vatter2012: 435). Academic studies have identified several factors that influence SWD, including political culture (Almond and Verba, Reference Almond and Verba1963; Inglehart, Reference Inglehart1990; Norris, Reference Norris and Norris1999), and critically, institutional arrangements (e.g. Karp and Bowler, Reference Karp, Banducci and Bowler2001; Berggren et al., Reference Berggren, Fugate, Pruehs and Still2004; Matsubayashi, Reference Matsubayashi2007; Bernauer and Vatter, Reference Bernauer and Vatter2012). Lijphart (Reference Lijphart1984, Reference Lijphart1999) finds that citizens in consensual democracies tend to be more satisfied than those living in majoritarian systems, while Anderson and Guillory (Reference Anderson and Singer1997) investigate whether the nature of democratic institutions impacts attitudes towards the political system. Importantly, they find that the difference in satisfaction between winners and losers tends to vary across political systems: losers in consensual systems are more satisfied than losers in majoritarian systems because safeguards exist to ensure their voices will still be heard, whereas winners in majoritarian systems are more satisfied than their counterparts in consensual systems (see also Anderson et al., Reference Anderson, Blais, Bowler, Donovan and Listhaug2005). More recently, Berggren et al. have extended Anderson and Guillory’s findings by noting that different measures of institutional context (e.g. electoral district size, degree of unicameralism, electoral system disproportionality) affect the attitudes of winners and losers differentially (Berggren et al., Reference Berggren, Fugate, Pruehs and Still2004: 88; see also Bernauer and Vatter, Reference Bernauer and Vatter2012).

A related strand of literature analyzes the linkages between parties and citizens, emphasizing that enhancing the representational connection between the two should increase citizens’ SWD. Ezrow and Xezonakis (Reference Ezrow, De Vries, Steenbergen and Edwards2011) find that citizens are more supportive of democracy when the ideological dispersion of parties in a given party system more closely approximates the mean voter’s ideological position, while Curini et al. Reference Curini and Martelli(2012) demonstrate that, at the individual level, winners become more satisfied the more ideologically proximate to the government they are.

Valence-related studies

The focus on character valence and its potential ramifications for political behaviour ties into a growing body of research examining ‘valence’ dimensions of candidate/party evaluation. Developed by Stokes (Reference Stokes1963, Reference Stokes and .1992), who famously critiqued spatial models of party competition, there are two approaches to valence, derived from Stokes’ discussions. In the first, most prominently associated with scholarship emphasizing ‘issue ownership’ or ‘issue competence’, valence issues were a dimension of party/candidate evaluation along which all voters hold identical positions such as wanting economic growth (see Lewis-Beck, Reference Lewis-Beck1990; Anderson, Reference Anderson2000; Palmer and Whitten, Reference Palmer and Whitten2000). Accordingly, voters evaluate parties/candidates on the basis of which party/candidate is the ‘most capable’ at handling a given issue, rather than which has presented the policy position most proximate to their own preferred position (e.g. Petrocik, Reference Petrocik1996; Green, Reference Green2007; Belanger and Meguid, Reference Belanger and Meguid2008; Green and Hobolt, Reference Green and Hobolt2008; Clarke et al., Reference Clarke, Sanders, Stewart and Whiteley2009; Pardos-Prado, Reference Pardos-Prado2012; Green and Jennings, Reference Green and Jennings2012, Reference Green and Jennings2017).

In the second school, Stokes defined valence as issues, ‘on which parties or leaders are differentiated not by what they advocate but by the degree to which they are linked in the public’s mind with conditions or goals or symbols of which almost everyone approves or disapproves’ (Stokes, Reference Stokes and .1992: 143). In other words, valence could refer to desirable ‘non-policy’ related characteristics such as honesty, trustworthiness, unity, competence etc. (Stokes, Reference Stokes and .1992). Thus, a second group of valence-related studies can be identified, which treat valence as a dimension of evaluation orthogonal to policy positions. Although this may include factors related to incumbent or campaigning advantage (Fiorina, Reference Fiorina1974; Burden, Reference Burden2004), of greater interest here are those studies which highlight the importance of party/candidate characteristics that would be intrinsically valued by the electorate. Several studies have robustly demonstrated the impact of valence-based attributes, including competence, integrity, and diligence, in U.S. Congressional elections (McCurley and Mondak, Reference McCurley and Mondak1995; Mondak, Reference Mondak1995; Stone and Simas, Reference Stone and Simas2010; Adams et al., Reference Adams, Merrill, Simas and Stone2011; Buttice and Stone, Reference Buttice and Stone2012; Butler and Powell, Reference Butler and Powell2014). Similarly, in a West European context, research has demonstrated the influence of parties’ character valence on political attitudes and behaviour (Clark, Reference Clark2009; Clark and Leiter, Reference Clark and Leiter2014; Zakharova and Warwick, Reference Zakharova and Warwick2014; Curini and Martelli, Reference Curini, Jou and Memoli2015; Leiter and Clark, Reference Leiter and Clark2015; Curini, Reference Curini2017). In sum, there is ample evidence that character valence can exert an important influence on individuals’ attitudes.

A conditional valence-related theory of SWD

The extant literature exploring SWD has strongly emphasized the role played by individual preferences. Perhaps, best known in this regard is Anderson and Guillory’s (Reference Anderson and Singer1997) study, which first suggested that SWD should vary between those individuals who were winners (those who voted for a party that won or maintained its place in government) and losers (those who voted for a party that did not win, or lost, office). Scholars have further investigated and refined the winner–loser dichotomy, while also paying closer attention to how different types of institutions influence individuals’ SWD (e.g. Bernauer and Vatter, Reference Bernauer and Vatter2012; Singh et al., Reference Singh and Thornton2012; Howell and Justwan, Reference Howell and Justwan2013; Ruiz-Rufino, Reference Ruiz-Rufino2013; Singh, Reference Singh2014; Dahlberg and Linde, Reference Dahlberg and Linde2016; Singh and Thornton, Reference Singh, Karakoc and Blais2016). In addition to this research, recent studies at both the individual and aggregate levels have drawn attention to how ideological positioning can play a role in influencing SWD (Henderson, Reference Henderson2008; Ezrow and Xezonakis, Reference Ezrow, De Vries, Steenbergen and Edwards2011) especially in regard to the winner–loser dichotomy, where ideological proximity reduces the negative attitudes induced by electoral loss (Curini et al., Reference Curini and Martelli2012, Reference Curini, Jou and Memoli2015).

In contrast, little attention has been paid to how valence-related dimensions of party/candidate evaluation – what we simply refer to as ‘character valence’ – impact political satisfaction. Given prior empirical research demonstrating the impact of character valence on various forms of political behaviour, we posit that parties’ character valence will impact SWD. More specifically, we argue that when political elites engage in behaviour that would detrimentally affect themselves and the parties to which they belong, then individuals’ assessments of their political system will be affected. From a normative perspective, one would expect that when political elites get into trouble and violate either formal or informal norms of conduct, this would impact how individuals assess their political system, plausibly contributing to feelings of animosity and cynicism. Our theoretical expectations in this respect are supported by a recent study in which the authors find that individuals’ political satisfaction is affected by the character valence of political parties (Leiter and Clark, Reference Leiter and Clark2015). In particular, the study finds that as the character valence of governing parties improves, respondents become more satisfied with democracy (and vice versa), but that this relationship is conditional on the character valence of opposition parties. In sum, empirical evidence suggests that the valence-related behaviour of parties provides individuals with cues as to how they should assess political parties and candidates, as well as the performance of their political system.

In the research presented here, we seek to build on the findings of the aforementioned studies in two ways. First, we focus more directly on the ‘winner–loser’ dichotomy featured so prominently in the SWD literature by examining whether an individual’s status as either a winner or a loser moderates how they evaluate system performance. Second, and related, we evaluate how winners and losers respond to the character valence of both the parties they support and those they do not. Ultimately, we seek further to refine our understanding of the various political forces that condition how winners differ from losers in their attitudes towards the political system – as expressed in terms of SWD.

Extant studies of SWD tend to assume that ‘winners’ and ‘losers’ only care about their preferred parties, and in this respect the ‘winner–loser’ dichotomy is construed narrowly (e.g. Anderson and Guillory, Reference Anderson and Singer1997; Aarts and Thomassen, Reference Aarts and Thomassen2008; Bernauer and Vatter, Reference Bernauer and Vatter2012; Howell and Justwan, Reference Howell and Justwan2013). To an extent, we agree with this assumption – we expect winners to be primarily concerned with, and more satisfied when, the character valence of winning/governing parties improves, and less satisfied when their character valence worsens. The same goes for losers and losing/opposition parties. But beyond this, we consider what happens when we relax these basic assumptions and contend that, given previous findings regarding the influence of both governing and opposition parties on individuals’ political attitudes, it is somewhat short-sighted to assume that winners only focus on the character valence of the government, and perhaps more importantly, that losers only focus on their preferred parties. Similarly, in their examination of what kinds of electoral victories most engender political satisfaction Singh et al. note that it is essential to consider all political parties and how voters react to all election results (2012: 203).

By way of example, one could conceive of losers becoming increasingly satisfied (or at least, less dissatisfied) with the political system when the character valence of winning (governing) parties improves. Although their preferred parties may not be in power, losers could still take solace from the fact that the party or parties doing the actual governing are doing so with greater competence, integrity, unity, and so forth. They may not have backed a party now in government, but their focus shifts to hoping that the governing party or parties will do so in a credible fashion. High levels of character valence by these parties could plausibly send such a signal. A business-related analogy is perhaps helpful here – certain party supporters may not have gotten their preferred choice for CEO to run the company, but once he or she has taken the position they hope the new boss delivers the leadership the company requires and brings the returns those supporters ultimately hope for. Alternately, one could argue that the improved character valence of governing parties may only serve to alienate losers, because they have already suffered the ignominy of supporting parties that have failed to take office. Accordingly, good performance by governing parties in valence-related terms may only serve to lower the political satisfaction of losers (Buchler Reference Buchler2008). In line with existing studies, the focus on how losers respond to the performance of both governing and opposition parties is crucial, because, as noted ‘Only when losers overcome their negative experiences and consent to being governed by those they disagree with does democracy endure and flourish’ (Anderson et al., Reference Anderson, Blais, Bowler, Donovan and Listhaug2005: 13). As such, any evidence that losers’ attitudes towards the political system are positively affected by the character valence of governing parties will be an encouraging finding.

Thinking about the supporters of winning/governing parties specifically, our expectations are somewhat different. As noted earlier, the basic assumption is that winners will respond positively to the character valence of their preferred parties. However, it is also possible that because these individuals have already ‘won’ in that their preferred party or parties are in government then the character valence of these parties will not detract from this, and their political satisfaction will be unmoved accordingly. Further, there is no strong theoretical or normative expectation that winners’ political satisfaction will be affected by the character valence of losing/opposition parties. Again, because these individuals are the most satisfied, improved character valence by losing/opposition parties is unlikely to affect their political satisfaction in any great sense. After all, why should they care what parties outside of government are doing? Nonetheless, it is plausible that the political satisfaction of winners may be affected by the character valence of losing/opposition parties, particularly if it improves at a time when governing parties are struggling.

Accordingly, we test several hypotheses, which are detailed below. The intent behind these hypotheses is first to explore if winners and losers’ attitudes towards the political system are affected by parties’ character valence, and to what extent. Second, and more specifically, we seek to investigate if and how winners and losers’ political satisfaction responds to the character valence of their preferred parties, as well as in response to parties they did not support (i.e. winners responding to losing/opposition parties’ valence, and losers responding to winning/governing parties’ valence). Our first hypothesis is as follows:

The logic behind Hypothesis 1 is straightforward enough, and similar to myriad arguments made elsewhere (e.g. Anderson and Guillory, Reference Anderson and Singer1997; Curini et al., Reference Curini and Martelli2012). Winners, having seen their preferred party win office, will be more satisfied with current governing configurations than losers who supported a party that was removed from, or did not win, public office.

In addition, in line with our arguments explicated earlier, and to expand further on the research of Leiter and Clark (Reference Leiter and Clark2015) and Singh et al. Reference Singh and Thornton(2012), we examine the following four hypotheses that analyze how the satisfaction of winners and losers is moderated by the character valence of winning/governing and losing/opposition parties:

In Hypothesis 2a, we assume that winners do not care about either the character valence of either governing or opposition parties, and so we expect parties’ character valence to have negligible effects on winners’ levels of satisfaction. As noted earlier, the reasoning here is that, having supported a winning party, these individuals are satisfied regardless of the character valence of either governing or opposition parties. Alternatively, Hypothesis 2b presents a differing expectation in that winners’ levels of satisfaction are assumed to be motivated by the character valence of the parties they backed – parties that won office and entered into government in the most recent election. So, if governing parties supported by winners perform well in valence-related terms then winners’ levels of satisfaction will increase (and vice versa). Not caring for the opposition, winners’ levels of satisfaction will not be affected by the character valence of those opposition parties. Hypotheses 2c and 2d outline similar expectations, but with regard to losers, that is, those who supported a party that did not enter into government. Hypothesis 2c assumes that losers’ satisfaction will be positively affected by improvements in the character valence of both governing and opposition parties. The logic behind Hypothesis 2c is that losers’ dissatisfaction can potentially be lessened by the high character valence of either governing or opposition parties. As discussed in the prior section, higher character valence of governing parties at least signals to losers that they will be getting a government that is relatively more honest, competent, and unified (or some combination thereof), whereas an improvement in the character valence of opposition parties signals to losers that the parties they care about most have at least made gains along an important dimension of evaluation. At last, Hypothesis 2d takes the alternate perspective – given that losers must accept a party (or parties) taking office that they do not support, the expectation is that governing parties’ character valence will lower losers’ levels of satisfaction, but opposition parties’ character valence (including their preferred parties) will lead to increases in levels of satisfaction. The reasoning here is that losers will not take any comfort from improvements in performance – valence or otherwise – of parties in government, and parties they did not willingly advocate for with their votes. On the other hand, as in Hypothesis 2c, higher character valence of opposition parties signals to losers that the parties they care about most have at least made gains in one important respect, and this should improve losers’ outlooks regarding political satisfaction.

Taken together, these hypotheses provide a stern test of the importance of valence dimensions of evaluation. In particular, if losers’ satisfaction responds to the character valence of governing parties, this would provide further empirical evidence of the significance of valence dimensions of evaluation on political behaviour. In line with the arguments put forth by Anderson et al. Reference Anderson, Blais, Bowler, Donovan and Listhaug(2005) and others regarding the significant role played by electoral losers in the democratic process, this would be an important finding. In addition, the expectations outlined previously dovetail with theoretical and empirical predictions drawn from the psychological motivation literature discussed earlier, which argues that individuals show bias when processing information related to their preferred parties/candidates, responding positively to evidence and information that fits with, or reinforces, their existing views, and rejecting or ignoring that which does not (Taber and Lodge, Reference Taber and Lodge2006; Claassen and Ensley, Reference Claassen and Ensley2016).

Applied here, motivated reasoning would predict that supporters of winning/governing parties would process positive valence-related information about these parties by becoming increasingly satisfied with democracy, while dismissing negative valence-related information, and thus not similarly updating their views on the political system. In a similar vein to how party supporters respond to scandal, motivated reasoning would also suggest that supporters of winning/governing parties would discount positive valence-related information in relation to losing/opposition parties, while reacting positively (reporting greater SWD) to negative valence-related information linked to these parties, because it would essentially confirm pre-existing notions regarding these parties. Government supporters may not, for example, respond to poor valence of the opposition because they cannot lower their opinion any further (Vonnahme, Reference Wells and Krieckhaus2014). A similar logic holds for supporters of losing/opposition parties. In the next section, we present the results of statistical analyses designed to test our hypotheses and to consider predictions associated with motivated reasoning.

Data and measurement

To examine the relationship between winners and losers, and parties’ character valence, we rely on the ‘Mannheim Eurobarometer Trend File 1970–2003’, which captures consistent survey items across the Eurobarometer surveys. For parties’ character valence, we use an original data set created using content analysis of Keesing’s Records of World Events, which provides us with measures of parties’ character valence for nine European countries (Clark, Reference Clark2009; Abney et al., Reference Abney, Adams, Clark, Easton, Ezrow, Kosmidis and Neundorf2011; Clark and Leiter, Reference Clark and Leiter2014).

Key independent variables: parties’ character valence

To capture parties’ character valence, we rely on an original data set created using content analysis of the media coverage of parties’ valence events.Footnote ⁷ We do so because very few surveys offer consistent cross-national measures of parties’ character valence. In addition, it can be challenging to separate out other important political evaluations, such as party identification or issue proximity, from valence evaluations using survey items. As such, we use an exogenous measure of parties’ character-related valence attributes constructed from content analysis of media-based news reports (Clark, Reference Clark2009). The source of these data is Keesing’s Record of World Events, which records global events by drawing on international news and internet sources.

There are nine countries covered in this data set, including: Britain, France, Denmark, Germany, Greece, Italy, the Netherlands, Portugal, and Spain. Each party-related news report in Keesing’s was content-analyzed for events that reflected on parties’ competence, integrity, or unity. These events included political scandals, abuse of official powers, various forms of misconduct, reports discussing poor performance in office (e.g. the handling of some kind of crisis or emergency), and intra-party divisions. In theory, the scores assigned to valence-related events appearing in Keesing’s news reports could range from +2 (for reported events that reflected very positively on parties’ competence, integrity, or unity) to −2 (for reported events that reflected very negatively on the party). However, it was extremely rare for the reported events appearing in Keesing’s to cast political elites in a positive light, so the scores that were awarded were virtually always negative.

Parties’ valence scores were created by adding together the scores awarded to each party for each valence event covered in Keesing’s for the entire inter-electoral period, defined as the period following the previous election up until the day before the current election. We take the total valence score from the inter-electoral period for each party. From this measure, we created character-valence scores for both government and opposition parties. For each party i in government (or opposition), we weight their valence score for a given inter-electoral period by their proportion of the popular vote received in the previous election (t − 1) as follows:

$${\rm{Weighted}}\,{\rm{government}}\,{\rm{valence}} = \sum\limits_{i = 1}^{{n_g}} {\left( {{{{\rm{Valenc}}{{\rm{e}}_i} \times {\rm{Voteshar}}{{\rm{e}}_i}} \over {{n_g}}}} \right).}$$

$${\rm{Weighted}}\,{\rm{opposition}}\,{\rm{valence}} = \sum\nolimits_{j = 1}^{{n_0}} {\left( {{{{\rm{Valenc}}{{\rm{e}}_j} \times {\rm{Voteshar}}{{\rm{e}}_j}} \over {{n_o}}}} \right).}$$

Our justification for relying on weighted valence scores by the size of the parties in government or opposition turns on the fact that larger parties are more likely to attract public/media attention and blame than smaller parties are. A government where one small niche party engages in poor behaviour is quite different from a government where the party of the prime minister is a lightning rod for scandal, and we expect the same is true for parties in opposition. By weighting the parties’ valence scores by size, we ensure they reflect this emphasis. However, the results of our analyses are substantively and significantly robust to the use of non-weighted government valence.

Evaluating the winners and losers hypotheses

We present two logit analyses, with country-fixed effects and standard errors clustered by country, on more than 37,000 respondents across nine countries, including Britain, Denmark, France, Germany, Greece, Italy, the Netherlands, Portugal, and Spain, with an average of 4.6 election cycles, and 4184 respondents per country in Table 1. We rely on logit models as we have collapsed ‘SWD’ into a dichotomous measure. In model 1, we examine the effect of simple winner status (supporters of parties in the current government), and government and opposition valence, as measured by media coverage of the party, on SWD. In model 2, we interact the basic ‘winner–loser’ dummy variable with both government and opposition parties’ valence scores.Footnote ⁸

Table 1. Logit Analysis of Valence and Winners–Losers on Satisfaction with Democracy (SWD)

Robust standard errors in parentheses, clustered by country. *P < 0.05, **P < 0.01, ***P < 0.001.

Country-fixed effects suppressed for space.

In column 1, we examine the effect of being an electoral winner (a supporter of a party currently in government) on SWD. As predicted in Hypothesis 1, respondents who state they currently support a party in government are significantly more likely to be satisfied with democracy than those respondents who support parties in opposition (P < 0.01). In addition, and in line with findings reported in Leiter and Clark (Reference Leiter and Clark2015), as the parties in government improve in terms of their character valence, respondents’ SWD increases, while an increase in opposition parties’ character valence decreases satisfaction. Taken together, these results strongly suggest that both political affiliation (being a winner or a loser), and the character valence of political parties impact individuals’ overall satisfaction with the political system.

Moving on to consider Hypotheses 2a–2d, a respondent’s party affiliation may also have an indirect effect on SWD by moderating the impact of parties’ valence as reflected in media coverage. In column 2, we interact the winner–loser dummy variable with both governing and opposition parties’ character-valence scores, and these interactions are visually presented in Figure 1. Two patterns emerge from our analysis. First, and perhaps unsurprisingly, regardless of the character valence of government or opposition parties, winners are much more likely to be satisfied with democracy than losers. The predicted probability that winners will be satisfied never falls below 70% in the range of observed valence values, whereas the highest predicted probability of satisfaction for losers is just over 60%. Although this gap persists across government and opposition performance, this does not mean that parties’ character valence do not moderate political satisfaction.

Note: Figure presents predicted probability of SWD over values of Government Valence (left-hand figure) and opposition valence (right-hand figure), with all other variables set to their means or medians.

Figure 1. Predicted probability of satisfaction with democracy (SWD) by level of valence and party support.

The second pattern that emerges from these results is that, in general, individuals are not responding to media coverage of the character valence of the parties they support; instead, they are responding to the character valence of the parties they do not. The substantive meaning of this pattern is clearer when we examine the differences between winners and losers. Somewhat in line with our prediction in Hypothesis 2a, winners’ SWD is not statistically significantly affected by the character valence of governing parties. However, counter to the expectations of Hypothesis 2a, our results do show that winners respond to the character valence of opposition parties. Thus, the empirical support for Hypothesis 2a is somewhat mixed. Of particular note here is that our findings show that winners are significantly less likely to be satisfied with democracy as the character valence of opposition parties improves, although the substantive effect is relatively small (about 13% lower probability of being satisfied between the minimum and maximum values of opposition parties’ character valence).

These findings also bear on Hypothesis 2b, which states that winners should be positively affected by improvements in the character valence of governing parties, but not opposition parties. We do not find empirical support for the hypothesis, as winners’ reported SWD does not change substantively as a result of improvements in the character valence of governing parties, and indeed the effect is not statistically significant. Counter to Hypothesis 2b, and as noted previously, winners do respond to the character valence of opposition parties, becoming increasingly less satisfied with democracy as opposition character valence improves. Essentially, winners are largely unmoved by the character valence of either governing or opposition parties, but when they do respond the effects are far more notable with respect to opposition parties.

Although the evidence in support of our hypotheses is mixed so far, taken together these empirical findings fit closely with predictions drawn from theories of motivated reasoning (Kunda, Reference Kunda1990; Nyhan and Reifler, Reference Nyhan and Reifler2010; Claassen and Ensley, Reference Claassen and Ensley2016). First, motivated reasoning would suggest we should not see winners responding to any negative information regarding the character valence of governing parties, because the cognitive dissonance caused by this information would lead to its dismissal (Taber and Lodge, Reference Taber and Lodge2006). Winners are those who support governing parties, and information that runs counter to their pre-existing views of their preferred parties will be rejected. That winners do not become increasingly satisfied in response to information suggesting positive character valence by governing parties is a little surprising. However, this is likely indicative of a ‘ceiling effect’ – winners have essentially ‘maxed out’ in terms of satisfaction.Footnote ⁹ Second, motivational reasoning would suggest that winners should become increasingly satisfied as the character valence of opposition parties worsens, and this is indeed what we see. Winners do not support these parties, and affirmation of their pre-existing biases against these parties increases satisfaction. In other words, winners appear to be taking pleasure (or ‘schadenfreude’ as an anonymous reviewer aptly stated) from the valence-related failings of these parties.

When we examine the relationship between ‘losers’ and the valence-related performances of governing and opposition parties drawn from media reports, we basically see the reverse pattern. Hypothesis 2c states that losers will be positively affected (in terms of their SWD) by improvements in the character valence of both governing and opposition parties, and as before, we find mixed support for the hypothesis. Figure 1 shows that improvements in opposition parties’ character valence does not lead to any discernible increase in losers’ probabilities of being satisfied with democracy (the ‘predicted’ line is virtually flat). On the other hand, Figure 1 also shows that losers clearly respond to improvements in the character valence of governing parties, and this effect is character valence both significant and substantively strong. Our model predicts that as governing parties’ character valence improve from the minimum value to the maximum value, the probability that an individual who supported an opposition party will be satisfied increases by almost 20% – coming close to the predicted levels of satisfaction of winners. Finally, Hypothesis 2d stated that losers’ satisfaction would be negatively affected by improvements in the character valence of governing parties, but positively affected by opposition parties. Losers are basically unresponsive to the character valence of their preferred parties in any direction, whereas intriguingly they respond strongly to the character valence of governing parties. In light of these findings, we find no support for Hypothesis 2d.

With respect to theories of motivated reasoning, these findings also broadly fit with expectations. Losers, like winners, appear unresponsive to either positive or negative media coverage of their preferred parties. That losers do not appear to update their views towards the political system in response to positive valence-related information is, again, suggestive of a ceiling effect, with levels of satisfaction appearing to hit an upper boundary. Similarly, as with winners, the notion of cognitive dissonance likely accounts for why losers do not respond to negative character valence, because this information is predicted to be dismissed as inconsistent with losers’ pre-conceived notions regarding their preferred parties. At last, motivated reasoning would predict that losers should become increasingly satisfied as the character valence of governing parties worsens, while being unresponsive to positive valence-related information, which would be dismissed. But our empirical findings do not fit with these predictions. As noted, losers become increasingly satisfied when the character valence of governing parties improves, and less satisfied when it worsens. Moreover, the effect is the strongest for all the relationships we examine. This result highlights something important – although partisanship may filter information and attitudes, once in government, these parties may be viewed differently, even by those voters who did not initially support them.

The overall narrative emerging from these results is that winners, regardless of the character valence of the political parties in the system, are more satisfied than losers – a finding in line with existing research on this topic. Importantly, however, this does not imply that parties’ character valence are irrelevant to individuals’ political satisfaction. When the media reports the highest levels of governing parties’ character valence, the difference in the probability of being satisfied with democracy reduces to only a 12% difference between winners and losers, compared to a more than 35% difference in probability at the lowest levels of governing parties’ character valence. However, it is not just governing parties’ character valence that is important, because neither winners nor losers appear to react to their preferred parties’ character valence. Our predictions were confounded with regard to how winners would respond to the character valence of opposition parties, and in terms of how losers responded to the character valence of governing parties. In these respects, however, we could look to expectations drawn from theories of motivated reasoning, and in this light our finding that winners’ probabilities of being satisfied with democracy decrease as a result of a decline in opposition parties’ character valence makes a lot of sense. Indeed, our results in this respect allow us to make an empirical contribution to this body of literature because winners display the ‘cognitive dissonance’ theorized by scholars working in the psychological motivation literature (e.g. Kunda, Reference Kunda1990; Taber and Lodge, Reference Taber and Lodge2006). Finally, we come back to the findings with regard to losers and governing parties – there is a clear relationship between losers’ reported levels of SWD and improvements in the character valence of governing parties. This finding is not only interesting in of itself (because it ran counter to both our hypotheses and the predictions of motivated reasoning), it is also important in the context of concerns over losers’ relationships to the political system and its legitimacy. Referenced earlier, Anderson et al. Reference Anderson, Blais, Bowler, Donovan and Listhaug(2005) make a strong case that losers are a kind of ‘linchpin’, and ‘only when losers consent to being governed by those they disagree with does democracy endure and flourish’ (2005: 13). Our empirical findings with regard to losers and governing parties’ character valence suggest that we should not simply focus on ideological factors (Anderson et al., Reference Anderson, Blais, Bowler, Donovan and Listhaug2005; Henderson, Reference Henderson2008; Curini et al., Reference Curini and Martelli2012), institutions (Anderson and Guillory, Reference Anderson and Singer1997; Aarts and Thomassen, Reference Aarts and Thomassen2008; Bernauer and Vatter, Reference Bernauer and Vatter2012), or even the closeness of elections (Howell and Justwan, Reference Howell and Justwan2013). Indeed, although satisfaction is often assumed to stem from concerns over the likelihood of having more of one’s preferred policies implemented by government, our findings suggest that valence-related dimensions of evaluation may provide one plausible pathway for losers to move beyond their electoral and policy-related disappointments and express their consent for governing parties. After all, it is perhaps easier to warm to a government consisting of a party or parties that are acting more competently, more honestly, and in a more unified fashion. Parties’ character valence may also serve to counter any policy-related consequences losers experience as a result of seeing their preferred parties left out of government. In this regard, our findings with respect to losers and governing parties’ character valence are supported by recent research by Dahlberg and Linde (Reference Dahlberg and Linde2016) who find that ‘high quality of government’ narrows the gap between electoral winners and losers in European democracies. Their results suggest that losers will be far less dissatisfied if they regard government as performing well, and broadly speaking, in their interests.

Concluding remarks

We embarked on this study with the goal of contributing to the literature examining factors which affect SWD. Extant studies have already shown a bright spotlight on how being a ‘winner’ or a ‘loser’ affects attitudes towards satisfaction (Anderson and Guillory, Reference Anderson and Singer1997), and scholars have also begun to examine how parties’ character valence can affect individuals’ political attitudes (Leiter and Clark, Reference Leiter and Clark2015), but no study to date examines how these two factors interact. Generally speaking, we find empirical support for a relationship but not always in the manner expected.

As with prior studies, the results of our statistical analyses confirm that winners and losers have differing attitudes regarding SWD, with winners being far more satisfied. When we add parties’ character valence into the equation we find that individuals’ satisfaction does not respond to the character valence of their preferred parties but to the character valence of parties that they do not. Winners are relatively unresponsive to character valence in general, but do report less SWD when opposition parties’ valence improves. Taken together, these observations would perhaps suggest that winners are much less concerned with parties’ character valence, which makes intuitive sense – their party has won office after all. As for losers, we find their satisfaction is affected by the character valence of governing parties quite dramatically. This may seem odd at first blush, and unexpected in light of motivated reasoning considerations. But having seen their preferred party lose it is reasonable to assume that losers would become more satisfied when evaluating parties they had previously dismissed, and have no choice but to live with, improve their character valence. As such, governing parties can make important steps to improve overall citizen satisfaction, and not just among their supporters, by improving their own valence-related behaviour in office. The results of our analyses suggest that the ‘consent of losers’ can be at least partially won by avoiding the kinds of actions and behaviours that so often places political elites in the headlines. Of course, this is easier said than done.

One final point to note relates to the work of Singh (Reference Singh2014) who draw attention to the fact that winners and losers should not be viewed in such absolute terms, differentiating instead between ‘optimal winners’ and ‘non-optimal winners’ as well as ‘optimal losers’ and ‘non-optimal losers’. Arguably, our empirical findings with respect to winners and losers could also be motivated, in part, by this notion of optimality. For example, with regard to losers and their responsiveness to governing parties’ valence, our findings could be as a result of the fact that many of those losers may have been ‘non-optimal losers’ and therefore far less ‘offended’ by the parties who ultimately formed the government. Because of both data and space limitations, it remains a task for future studies to be able to more directly explore and test these assumptions.