On 6 January 2021, supporters of President Donald Trump stormed the Capitol in Washington, DC. Inside Congress, they mocked democratic symbols, threatened staff and attacked security forces. Five people died during the attack, and another in its aftermath. The protest was fuelled by Trump’s repeated claims that the election had been ‘stolen’ and that only large-scale fraud could have brought Joe Biden to office. Beyond the protesters, many Republican voters also believed the electoral process had been rigged and questioned the legitimacy of the incoming president.
Almost exactly two years later, supporters of Brazil’s defeated outgoing president, Jair Bolsonaro, rioted in the capital, attempting to storm government buildings and ultimately overthrow newly elected President Lula da Silva. These events are among the most visible examples of a growing trend: election losers who refuse to accept the results. For them, defeat itself is proof that the election was fraudulent.
Coming to terms with defeat is difficult, as losers must accept a process that leaves them in a vastly different position from the winners. While winners have every reason to welcome the election outcome, losers see their preferred candidates fail to take office and cannot expect their favoured policies to be implemented in the near future (Anderson et al. Reference Anderson2005: 5; Nadeau and Blais Reference Nadeau and Blais1993: 553). However, even a positive evaluation of an election outcome simply because one’s side has won can be harmful to democracy. Before the 2016 election, Trump famously stated that he would ‘totally accept’ the results – ‘if I win’ (Lewis etal. Reference Lewis, Jacobs and Siddiqui2016). Yet, in a democracy, losers should not be less satisfied with the way democracy works just because they have lost – and winners should not be satisfied just because they have won. One’s assessment of democracy should not be based purely on instrumental considerations but on a fundamental acceptance of democratic procedures.
However, not all groups react to winning and losing in the same way. There are strong theoretical reasons to expect populist voters, in particular, to be sore losers. Populism is a worldview that: (1) pits ordinary citizens against a corrupt, self-serving elite that is out of touch with the people (anti-elite); (2) frames politics as a struggle between friends and enemies (a Manichean outlook); and (3) denies the legitimacy of disagreement (anti-pluralist) (Mudde Reference Mudde2004). If these three elements together define populism (Hawkins and Rovira Kaltwasser Reference Hawkins, Rovira Kaltwasser and Hawkins2019; Wuttke etal. Reference Wuttke, Schimpf and Schoen2020), then losing an election is not only difficult to accept but hard even to conceive. Given these ideological dispositions, the very fact of losing signals that democracy is not functioning as it should.
To examine how losing affects citizens’ satisfaction with democracy, we use three sets of data. First, we analyse data from the Comparative Studies of Electoral Systems (CSES) to test whether the impact of winning and losing on satisfaction with democracy differs for populist and non-populist voters in 39 countries. In addition, we draw on the V-Party data to analyse if voting for more populist parties affects the satisfaction of democracy of winners and losers. This allows us to expand the analyses to 200 elections in 57 countries from 1996 to 2019. Turning to a more rigorous research design, we analyse panel data from the German Longitudinal Election Studies (GLES) and the Dutch Longitudinal Internet Studies for the Social Sciences (LISS) to examine how electoral defeat shape the attitudes of more and less populist voters. The use of panel data and a longitudinal research design allows us to address potential endogeneity issues, such as reverse causality. For instance, citizens who are less satisfied with democracy may be more likely to adopt populist attitudes or support populist parties. By measuring respondents’ satisfaction with democracy before and after the election and regressing the latter on the former, we can test whether populism amplifies citizens’ reactions to electoral victory or defeat.
We contribute to the existing literature in three ways. First, we integrate the theoretical literature on populism into the study of losers’ consent, arguing that the more people adhere to a populist worldview, the more strongly they reject electoral defeat. Second, we move beyond a dichotomous measure of populism by employing a continuous scale – both at the level of individual attitudes and for competing parties. Third, we draw on both cross-sectional and panel data, allowing us to better identify the causal effect of losing for different types of voters.
Our analyses confirm these theoretical expectations. Across countries, electoral losers are generally less satisfied than winners are with the way democracy works. Moreover, citizens with stronger populist attitudes express lower satisfaction than those with weaker populist attitudes. Most importantly, the effect of losing is amplified among citizens with more populist attitudes and those who vote for more populist parties.
The article is structured as follows. In the next section, we review existing research on losers’ consent and summarise its main findings. We then outline why we expect populism to magnify the effects of electoral victory and defeat. In the third section, we present the data and methods for our empirical analyses. Then we report the results from the cross-sectional comparison using CSES data, as well as from the two panel datasets. The fifth section concludes.
Losers’ consent and populism
The extent to which citizens uphold the demanding democratic standard of consent in the face of electoral defeat has been the subject of extensive research. In general, losers are less satisfied with the way democracy functions than winners and tend to have lower levels of trust in politicians (Anderson etal. Reference Anderson2005: 43; Anderson and Guillory Reference Anderson and Guillory1997; Anderson and LoTempio Reference Anderson and LoTempio2002; Blais and Gélineau Reference Blais and Gélineau2007; Mauk Reference Mauk2020). Moreover, the effects of losing persist over time (Hansen etal. Reference Hansen, Klemmensen and Serritzlew2019) and become more pronounced after repeated defeats (Anderson etal. Reference Anderson2005: 61–67; Kern etal. Reference Kern, Marien and Muradova2024). However, the magnitude of these negative effects depends on both individual characteristics and broader contextual factors.
At the macro level, several findings stand out. In younger democracies, electoral outcomes have a stronger impact on satisfaction with democracy than in more established ones (Anderson etal. Reference Anderson2005: 105; Rich Reference Rich2015; see, however, Farrer and Zingher Reference Farrer and Zingher2019 for contrasting findings). The higher the quality of democracy and the integrity of the electoral process, the smaller the gap in democratic satisfaction between winners and losers (Fortin-Rittberger etal. Reference Fortin-Rittberger, Harfst and Dingler2017; Lago and Martinez i Coma Reference Lago and Martinez i Coma2017; Nadeau etal. Reference Nadeau, Daoust and Dassonneville2021b, Reference Nadeau, Bélanger and Atikcan2021a). Proportional representation and consensus-based democracy mitigate the negative effects of losing (Anderson and Guillory Reference Anderson and Guillory1997; Bernauer and Vatter Reference Bernauer and Vatter2012), while in countries with high social inequality electoral defeat produces a stronger impact on satisfaction than in those with medium or low levels of inequality (Han Reference Han2016). Additionally, when the stakes of an election appear particularly high – such as in the referendum in the United Kingdom on leaving or remaining in the European Union – losing has an especially strong impact on democratic satisfaction (Nadeau etal. Reference Nadeau, Bélanger and Atikcan2021a; van der Eijk and Rose Reference van der Eijk and Rose2021).
At the individual level, resources, behaviour and attitudes play a crucial role. Citizens with greater resources tend to be more satisfied with democratic processes, while higher levels of political interest and participation are associated with greater satisfaction (Nadeau and Blais Reference Nadeau and Blais1993). Previous electoral victories can also buffer the negative effects of losing; having won in past elections reduces the impact of defeat in subsequent ones (Curini etal. Reference Curini, Jou and Memoli2012). Conversely, perceptions of unfair electoral procedures intensify losers’ dissatisfaction (Blais etal. Reference Blais, Morin-Chassé and Singh2017; Esaiasson Reference Esaiasson2011). Finally, the negative impact of losing is more pronounced among individuals with strong party identification or more extreme political views (Anderson etal. Reference Anderson2005: 89; Janssen Reference Janssen2023)– a dynamic that already hints at the potentially amplifying effect of populism.
Building on these well-established findings, we formulate our first, basic hypothesis:
Hypothesis 1: Losers are less satisfied with the way democracy works than winners.
While the impact of election outcomes on voter attitudes has rightly received significant scholarly attention, relatively few studies have examined whether populist voters react differently from their less populist counterparts – despite strong reasons to expect that populists may struggle more with electoral defeat. Following the widely accepted framework established by Cas Mudde (e.g. Mudde Reference Mudde2004; Mudde and Rovira Kaltwasser Reference Mudde, Rovira Kaltwasser, Mudde and Kaltwasser2012, Reference Mudde and Rovira Kaltwasser2017), we define populism based on three core elements (see also Wuttke etal. Reference Wuttke, Schimpf and Schoen2020):
a) a Manichean and moral cosmology; b) the proclamation of ‘the people’ as a homogenous and virtuous community; and c) the depiction of ‘the elite’ as a corrupt and self-serving entity. In other words, something is defined as ‘populist’ if it has all three of these conditions. (Hawkins and Rovira Kaltwasser Reference Hawkins, Rovira Kaltwasser and Hawkins2019: 3)
Partly due to these characteristics, populists are often seen as being uneasy with the democratic systems they live in (Mudde Reference Mudde2004). Compared to their non-populist counterparts, they tend to have a more negative view of politics and place less value on the performance of existing democratic institutions (Dahlberg and Linde Reference Dahlberg and Linde2017). In particular, they perceive democratic procedures as biased against ordinary citizens, believing that a distant elite prevents the ‘silent majority’ from having its way. A defining feature of populist ideology is the sharp divide between those in power and the people. Populists assume that democratic procedures fail to hold decision-makers accountable to the will of the people, who are presumed to share a common political will.
Given these ideological dispositions, we argue that citizens with stronger populist attitudes are more likely to be dissatisfied with the functioning of representative democracy. This leads to our second hypothesis:
Hypothesis 2: The stronger their populist attitudes, the less satisfied citizens are with democracy.
We follow Jan-Werner Müller (Reference Müller2016) in arguing that anti-elite orientations are a necessary but not sufficient component of populism. Populist parties and movements are also fundamentally anti-pluralist, he contends, because they claim to be the sole legitimate representatives of ‘the people’. For populists, political conflict is particularly intense because, in their view, it should not exist at all – that is, if the established parties did not violate common sense and disregarded the volonté générale. When legitimate disagreement is seen as unacceptable, electoral defeat becomes not just a setback but a fundamental threat. As Carl Schmitt (Reference Schmitt2015: 26) famously argued, politics is ultimately defined by the distinction between ‘friend’ and ‘enemy’. In this struggle, violence and even (civil) war can be considered ultimate means of defending one’s ‘essential way of life’ (Schmitt Reference Schmitt2015: 27). Of course, most political conflicts do not escalate to this extreme, but when politics is framed in such existential terms, accepting the outcome of an election – unless one’s side wins – becomes profoundly difficult.
Given these three defining elements of populism, electoral defeat is likely to be especially significant for populist voters. When one adheres to a worldview that divides society into ‘us’ and ‘them’, viewing ‘them’ as an out-of-touch elite, traitors or even enemies, and ‘us’ as a unified people with a singular political will, losing is not simply a matter of policy preference – it signifies rule by those whom populists deeply resent. Moreover, if political disagreement is considered illegitimate and the ‘true people’ are presumed to stand in unanimous opposition to the elite, the very idea of electoral defeat becomes inconceivable. For populists, losing an election signals that the system is rigged, reinforcing suspicions of electoral fraud. While populists may claim to support democratic values, they see those values as inadequately realized in the existing democratic system. Thus, for populist voters, electoral defeat serves as proof of democracy’s dysfunction, whereas for non-populists, satisfaction with democracy is less directly tied to electoral outcomes.
A growing body of research supports this perspective. One set of studies finds that political dissatisfaction is a stronger predictor of vote choice when populist parties are in opposition, but less important when they are in government (Cohen Reference Cohen2020; Kriesi and Schulte-Cloos Reference Kriesi and Schulte-Cloos2020; Muis etal. Reference Muis, Brils and Gaidytė2022; Rooduijn and van Slageren Reference Rooduijn and van Slageren J2022). However, because these analyses mainly rely on cross-sectional data, they cannot distinguish between two possible mechanisms: whether dissatisfied voters shift their support to other parties or abstain from voting (sorting), or whether they actually become more satisfied when a populist party takes office (attitude change) (for a discussion, see Muis etal. Reference Muis, Brils and Gaidytė2022: 755).
Empirical evidence from several countries suggests that populist electoral victories can reshape patterns of democratic satisfaction. In Norway, Atle Haugsgjerd (Reference Haugsgjerd2019) finds that after the populist Progress Party joined a coalition government in 2013, the gap in satisfaction with democracy became smaller because satisfaction decreased more slowly for its supporters than for other voters. Similarly, in Brazil, Jair Bolsonaro’s victory reduced the difference between winners and losers in terms of support for the political system, while simultaneously widening the gap in tolerance for anti-democratic behaviour (Cohen etal. Reference Cohen2023). Christina-Marie Juen (Reference Juen2024) demonstrates that political trust fluctuates in Austria more strongly among populist winners and losers than among other voters. Finally, Eelco Harteveld etal. (Reference Harteveld2021) find that after a right-wing populist party enters government, satisfaction with democracy increases among those who hold a nativist ideology. Building on and extending these studies, we examine how populist attitudes and voting for populist parties shape satisfaction with democracy, depending on election outcomes. Based on the above discussion, the third hypothesis follows:
Hypothesis 3: Losing has a stronger negative effect on satisfaction with democracy for voters with more populist attitudes.
Data and methods
In the first part of the empirical section, we use cross-sectional data from the Comparative Studies of Electoral Systems (CSES).Footnote 1 In all waves, the CSES has asked respondents about their perceptions of the performance of the democratic regime they live in. For the following analyses, we use satisfaction with democracy as the dependent variable, which is the standard in the literature on losers’ consent. In the CSES, it is asked with the following wording: ‘On the whole, are you very satisfied, fairly satisfied, not very satisfied, or not at all satisfied with the way democracy works in [COUNTRY]?’Footnote 2 Respondents can answer on a scale from 0 to 10, where 0 means ‘not at all satisfied’ and 10 means ‘very satisfied’. We use this variable as the dependent variable in the following analyses.Footnote 3
The fifth wave of the CSES measures populism based on seven items measuring individuals’ attitudes towards political elites, attitudes towards representative democracy and majority rule and attitudes towards out-groups (Electoral Systems TCS of 2023). The exact wording of the items can be found in the Appendix in the Supplementary Material. We calculate the mean of all seven items to determine the respondent’s level of populist attitude.Footnote 4 The CSES also contains information on how respondents voted in the last national election in their country. Based on this information, we construct a variable that indicates whether someone is a ‘winner’ or a ‘loser’. If the party voted for ended up in opposition, the respondent is coded as a loser.Footnote 5 Information about cabinet composition comes from the ParlGov dataset (Döring and Manow Reference Döring and Manow2021) and the V-Party dataset (Düpont etal. Reference Düpont2022). Thus, the level of populist attitudes and whether the respondent’s preferred party ended up in government or opposition are the main explanatory variables in the following analyses. To account for previous findings, we control for age, gender, education, place of residence (rural versus urban), political interest and party identification. Table A1 in the Appendix provides descriptive statistics for all the variables that are included in the subsequent analyses. Using this dataset we are able to cover elections in 39 countries.
Because the first four waves of the CSES did not include a populism scale, we use the degree of populism of the party a respondent voted for as a proxy for populist attitudes– even though not everyone with populist attitudes votes for a populist party, and not everyone who votes for a populist party has populist leanings. However, this is the only way to use the entire CSES data in the analyses. For every election from 1970 to 2019, the V-Party data provide a continuous populism index for each (major) party.Footnote 6 This index is computed as the harmonic mean of two indicators. The first indicator measures the extent to which the party uses anti-elite rhetoric. The second measures the degree of people-centredness – that is, whether party leaders glorify ordinary people and identify themselves as part of them. The full wording of the indicators can be found in the Appendix.Footnote 7 Using this measure offers two advantages over a binary indicator distinguishing only between populist and non-populist parties. First, while there are some clearly populist parties, the literature also acknowledges borderline cases (Mudde Reference Mudde2007: 57–58). These cases are difficult to handle using a binary variable which separates parties in distinct categories. Second, the V-Party index, and thus our populism index, is available for each major party for each national election. This allows us to look at changes in the populist stance of parties over time. In turn, this approach allows us to extend the time period under investigation by analysing 200 elections in 57 countries from 1996 to 2019 (seeFigure 1 for an overview of countries and years). Descriptive statistics can be found in Table A2 in the Appendix.
Countries Included in the Cross-Sectional Analyses
Beyond the cross-sectional analyses, we also use panel data from the GLESFootnote 8 for the 2017 and 2021 Bundestag elections. To the best of our knowledge, the GLES panel is the most suitable data for the analysis because the data allow us to compare pre-election and post-election levels of satisfaction with democracy while also measuring individuals’ populist attitudes. The use of panel data allows us to address potential endogeneity problems regarding Hypothesis 3. One potential issue is reverse causality; namely that citizens who are less satisfied with democracy and have higher levels of populist attitude vote disproportionately for losing parties. As we are able to measure the respondents’ level of satisfaction with democracy before and after the election, we can directly test the effect of the election on the respondents’ attitudes. Again, we use the item for satisfaction with democracy as the dependent variable, which is covered using the following wording by the GLES panel: ‘On the whole, how satisfied or dissatisfied are you with the way democracy works in Germany?’ (1=‘very satisfied’ to 5=‘very dissatisfied’).
Respondents’ populist attitudes are measured by eight items that capture core elements of populist ideology, emphasizing anti-elitism, people-centredness and preferences for direct democracy. Again, the full wording can be found in the Appendix. We control for age, gender, education, place of residence (rural versus urban), political interest and party identification, while also including a control variable that distinguishes whether the respondent lives in West or East Germany. Table A3 in the Appendix provides descriptive statistics. Pre-election values for the satisfaction variable are taken from the fifth wave of the GLES panel fielded in August 2017 and from the eighteenth wave in August 2021. The elections took place in September 2017 and September 2021, respectively, so there is only a very small time interval between the initial measurement of respondents’ satisfaction with democracy and the election. After the 2017 elections, the formation of a new government took an unusually long time. Therefore, we use citizens’ satisfaction with democracy from the ninth wave of the GLES panel, whose fielding period started on 15 March 2018, as post-election values. For the 2021 election, we use the twenty-first wave, fielded in December 2021, immediately after the formation of the new government, as post-election values of satisfaction with democracy.
Finally, we use the LISS panelFootnote 9 administered by CentERdata (Tilburg University, the Netherlands). The LISS panel is a representative sample of Dutch individuals who participate in monthly surveys. The panel is based on a true probability sample of households drawn from the population register.Footnote 10 This datasetallows us to cover three national elections (2010, 2012, 2017).Footnote 11 From 2010 to 2012, Geert Wilders’s right-wing populist Party for Freedom (PVV) tolerated a centre-right minority government. This government resigned in 2012 because the PVV did not support its actions in the European financial crisis. After the next election, a government was formed without the PVV’s support. Thus, we have data for PVV supporters as an opposition party, as a tolerating party, and again as an opposition party. Since we cannot measure populist attitudes directly, we again use the V-Party populism index as a proxy. Satisfaction with democracy is measured as follows: ‘And how satisfied are you with the way in which the following institutions operate in the Netherlands? – Democracy’ (0=‘extremely dissatisfied’ to 10=‘extremely satisfied’).
For our longitudinal analysis, we use the pre- and post-election waves for the three national elections from the LISS ‘politics and values’ core studies, so we use six waves in total.Footnote 12 From the LISS background variables, we draw basic socio-demographic and socio-economic information. We control for age, gender, education, political interest and place of residence. The LISS dataset does not provide a variable for party identification; therefore we use party membership as a substitute (for descriptive statistics see Table A4 in the Appendix).
For all analyses, we have rescaled the variables for satisfaction with democracy and the populism indicators so that they range from 0 to 1. Higher values indicate greater satisfaction with democracy and more populist attitudes (or more populist parties). For the cross-country analyses using the CSES dataset, we use a multilevel model with citizens nested within countries. For the longitudinal analyses using the GLES and LISS panels, we use a multilevel residualized change model. As part of the residualized change model, we use the level of satisfaction with democracy after the election, while controlling for its level before the election. We pool the data from different elections with different winning and losing parties in one dataset. Because there is a possibility that we are treating the same individual’s behaviour for different elections separately, we also use a multilevel model that nests individual election cases within individuals, while also using fixed effects for the respective election. For replication purposes, we also use cross-sectional analyses of the panel data. In this case, we use multilevel regression analyses, again nesting citizen elections within citizens and applying election fixed effects. All multilevel models are run using restricted maximum likelihood (REML) estimators with robust standard errors. We also use alternative statistical methods for robustness checks. We find that using these alternative models does not change our main results.
We believe that the three datasets we use in the following and the combination of cross-sectional and panel analyses allow us to systematically test the effect of winning and losing for voters with varying levels of populist attitudes or who vote for parties with different degrees of populism.Footnote 13
Analyses
In this section, we begin with cross-sectional analyses using data from the CSES before turning to the GLES and LISS panel studies. Overall, our findings confirm that while populist voters tend to be more satisfied with democracy after electoral victories, they become significantly more dissatisfied after defeats.
The effect of winning and losing for populist voters (CSES)
As the fifth module of the CSES – an international cross-sectional election study covering 39 countries – includes a battery measuring populist attitudes, it is well suited to study the effect of winning/losing on different kinds of voters’ evaluation of democracy. Figure 2 provides an initial overview of the data. The horizontal axis represents voters’ populist attitudes, while the vertical axis indicates satisfaction with democracy. Colours distinguish between winners and losers in the most recent election. Three key patterns emerge. First, respondents with stronger populist attitudes tend to be less satisfied with democracy, regardless of whether their preferred party recently won or lost. Second, as expected, winners are generally more satisfied than losers. This is visually evident: blue dots (winners) are concentrated in the upper half of the graph, while red dots (losers) are more prevalent in the lower half. Finally, and most notably, the gap in democratic satisfaction between winners and losers widens as populist attitudes increase. Among the most populist respondents, dissatisfaction with democracy is particularly pronounced after electoral defeat.
Scatter Plot of Populist Attitudes and Satisfaction with Democracy
In the next step, we run multilevel linear regression models with respondents nested in countries to test our three hypotheses. Our main variables of interest are whether a respondent was a loser in the last general election (Hypothesis 1), the respondent’s level of populist attitude (Hypothesis 2) and the interaction effect of these two variables (Hypothesis 3). The left panel ofFigure 3 plots the coefficients of the key variables of interest for the model without interaction effects (see Table A5 in the Appendix for full results). Consistent with Hypothesis 1 and in line with previous studies, we observe that losers are less satisfied with democracy than winners. We also find support for Hypothesis 2, as more populist citizens are less satisfied with democracy than their less populist counterparts. Finally, to test the core argument of this article, we examine the top right panel, which presents the results of the second regression model, including the interaction term. The findings reveal a statistically significant effect: populist attitudes indeed amplify the negative impact of losing.
Coefficient Plot and the Marginal Effect of Losing: Data from CSES V
The impact of populist attitudes on the experience of losing is substantial. In the lower half ofFigure 3, we illustrate the marginal effects of this interaction. While losers with very low levels of populist attitudes are less satisfied with democracy than non-populist winners, the gap widens significantly as populist beliefs become stronger. Comparing highly populist citizens with those who have minimal populist attitudes, we find that the negative effect of losing is nearly three times larger. Thus, the analyses so far confirm all three hypotheses.
To confirm these results, we replicate these analyses using all five modules of the CSES. This allows us to extend the time period under investigation by analysing 200 elections in 57 countries from 1996 to 2019. However, since the first four waves of the CSES do not include a populism scale for individual respondents, we use the party-populism index provided by the V-Party data as a proxy. Looking at the data from the fifth CSES module, we find a significant but low correlation between the individual populism score and our proxy of r=0.33.
The results of the replicated analyses using the party-populism score and all modules of CSES data are reported in Table A7 in the Appendix. As above, the upper part ofFigure 4 presents the regression coefficients for our main variables of interest. Again, our three hypotheses are confirmed. Losers are less satisfied with democracy than winners, while citizens who vote for more populist parties are also less satisfied. Most importantly, we again see a significant interaction effect, indicating that the gap between winners and losers is more pronounced for voters of more populist parties.
Coefficient Plot and the Marginal Effects of Losing: Data from CSES I–V
To ensure the robustness of our findings, we conduct several checks. First, while the CSES datasetallows us to analyse a large number of elections, our results could be driven by individual countries. To account for this possibility, we perform a jackknife procedure, systematically excluding one country at a time and re-running the regression analysis. As shown in Figure A2 in the Appendix, the interaction between individual populist attitudes and losing remains negative and statistically significant, regardless of which country is omitted. Additionally, we replace the multilevel model with a specification that includes country-fixed effects and standard errors clustered by country. As Table A6 in the Appendix demonstrates, this alternative model yields results consistent with our main findings.
To further validate our results, we incorporate party-level populism scores. First, we replicate our original analysis using only the fifth module of the CSES. The results closely match our original conceptualization, reinforcing confidence in our approach (see Models 3 and 4 in Table A7). Since our argument applies to both left- and right-wing populism, we also test whether our results hold when analysing these types separately. As shown in Table A9 in the Appendix, both main and interaction effects remain statistically significant and point in the expected direction.
Finally, we extend our robustness checks by utilizing all five waves of the CSES. We again apply jackknife procedures, sequentially removing individual countries (FigureA3), continental regions (Figure A4) and entire waves (Figure A5) from the analyses. While these adjustments lead to moderate variations in the regression coefficient of the interaction term, it consistently remains negative and statistically significant, confirming the robustness of our results.
Furthermore, while populist citizens are more likely to support populist parties, there remains a significant group of populist-leaning voters who back non-populist parties – and vice versa. To disentangle these effects, we create a categorical variable distinguishing respondents based on whether they (a) hold populist attitudes, (b) vote for a populist party, (c) do both or (d) do neither. Our findings show that both citizens who hold populist attitudes but vote for a non-populist party and those who vote for a populist party despite not holding populist attitudes are less satisfied losers than those who neither hold populist views nor vote for a populist party. Most notably, the negative effect of losing is strongest among citizens who both hold populist attitudes and vote for a populist party. The full results are presented in Table A10 in the Appendix.
To further validate our findings, we replicate the analysis using party identification instead of vote choice, recognizing that voters may not always cast their ballots for the party they most strongly support – particularly in the case of minor parties (Nemčok Reference Nemčok2020; Ridge Reference Ridge2024). This alternative conceptualization does not change our main results, as shown in Table A8 in the Appendix.
Finally, we extend our analysis using an additional cross-sectional dataset. We rerun our models on the first 10 waves of the European Social Survey (ESS), again using the party-populism score as a proxy. As confirmed in Table A11 in the Appendix, the results remain robust across this dataset as well. Taken together, this set of additional analyses reinforces the findings of our initial cross-sectional models.
German Longitudinal Election Study data: the causal effect of losing for populist citizens
In this second step, we analyse data from the GLES covering the 2017 and 2021 national elections. The most salient issue in the 2017 election was the handling of the so-called ‘migration crisis’ (Arzheimer and Berning Reference Arzheimer and Berning2019; Dostal Reference Dostal2017), while the response to the coronavirus pandemic dominated the 2021 election (Rensmann and de Zee Reference Rensmann and de Zee2022). In both cases, the populist radical-right Alternative for Germany (AfD) positioned itself as a vocal critic of government policies, opposing both the perceived openness to refugees and the restrictive measures implemented during the pandemic.
The 2017 election marked the AfD’s first entry into the Bundestag after it had failed to pass the electoral threshold in 2013. In 2021, the party secured representation in the Bundestag for the second consecutive time. According to the V-Party dataset, the AfD was by far the most populist party in Germany, scoring 0.952 on a scale from 0 to 1. In contrast, all other parties exhibit much lower levels of populism (ranging from 0.054 to 0.169), with the exception of ‘The Left’, which scores a moderate 0.323. The AfD also attracted the voters with the most populist attitudes (0.729), although the difference in populist attitudes among voters of different parties is smaller than the difference between the parties themselves (ranging from 0.549 to 0.636 for all other parties).
The GLES panel data are particularly well suited for our analysis, as they include a battery of questions measuring populist attitudes and track citizens’ satisfaction with democracy both before and after the elections. This structure allows us to rigorously test the effect of losing by analysing post-election satisfaction while controlling for pre-election levels.
We begin by replicating the cross-sectional analyses to check that the GLES data produce results consistent with those found earlier. Regression models closely mirroring our previous analyses confirm that populism amplifies the negative effect of losing (see Table A12 in the Appendix). Next, we extend our analysis by incorporating pre-election levels of satisfaction with democracy as a control variable (see Table A13 in the Appendix). This panel model provides a more robust test of Hypothesis 1 and Hypothesis 3. The upper part ofFigure 5 presents the regression coefficients for our key variables of interest, first without and then with the interaction term.
Coefficient Plot and the Marginal Effects of Losing: Data from GLES Panel
Again, these analyses support our hypotheses, as we find that losers are less satisfied with democracy than winners (Hypothesis 1), and more populist attitudes dampen satisfaction with democracy (Hypothesis 2). We also find evidence supporting Hypothesis3. The results clearly show that the effect of losing is more pronounced for more populist citizens – even when we control for their pre-election level of satisfaction with democracy.
Interestingly, the right-hand panel ofFigure 5 does not show a statistically significant effect of losing on satisfaction with democracy when we include the interaction term in the model. This suggests that losers with the least populist attitudes are not less satisfied with democracy than non-populist winners – their evaluation of democracy does not seem to be driven by election results at all. To better understand this phenomenon, we plot the marginal effects of the interaction effect in the lower part ofFigure 5. For citizens with low levels of populist attitudes, the gap between winners and losers disappears. For the most populist citizens, however, we find a substantial effect of about 12 percentage points on the satisfaction with democracy scale.
Populist losers remain less satisfied with the way democracy works even when we account for baseline effects. To further validate our findings, we conduct several robustness checks. First, we test whether our results hold when excluding voters of individual parties. As shown in Figure A6 in the Appendix, the results remain statistically significant regardless of which party is excluded – except for the right-wing populist AfD. This exception is expected, as the AfD is both the most populist party and the primary choice of the most populist citizens.
We also examine whether our results are driven by one particular election by rerunning the analyses separately for 2017 and 2021 using ordinary least squares (OLS) regressions. As Figure A7 in the Appendix confirms, our conclusions remain consistent across both elections. Additionally, replacing vote choice with party identification as our key independent variable produces nearly identical results, as demonstrated in Table A14 in the Appendix. Finally, we replicate our analyses using the party-populism index instead of individual populist attitudes, with models 3 and 4 reported in TableA13 in the Appendix further confirming our findings.
Longitudinal Internet Studies for the Social Sciences data: populist toleration
In this final step, we analyse LISS data from the Netherlands. While these data do not include a battery for populist attitudes, they provide panel data over three national elections (2010, 2012 and 2017). Both the 2010 and 2012 election were early elections. In early 2010 the previous coalition was unable to find a compromise on the Dutch military mission in Afghanistan. The electoral campaigns were mostly about the financial crisis and its implications for the federal budget and possible cuts for government spending. In addition, issues of migration and integration were put forward in the party manifestos – especially for the populist radical-right PVV (van Holsteyn Reference van Holsteyn JJM2011). This government ultimately failed as it was unable to find an agreement over the cuts in the federal budget, which later also became a central aspect of the following election campaign (van Holsteyn Reference van Holsteyn2014). Just as in the German case, the issue of immigration or ‘Dutch identity’ was very important for the 2017 election (van Holsteyn Reference van Holsteyn2018). For all three elections, the PVV was by far the most populist party according to the V-Party dataset, with a value of 0.957. All other parties are much less populist (ranging from 0.075 to 0.109) with the only exception being the left-wing ‘Socialist Party’, which features a medium level of populism (0.586).
As before, we use a measure of populism at the level of parties. What makes the Dutch case particularly interesting is that the right-wing populist PVV supported the centre-right minority government between 2010 and 2012. Although the PVV was not formally part of the government, it was able to influence policy decisions for a short period of time. After 2012, it became an opposition party again. Thus, we can observe the effects of winning and losing for voters of a highly populist party.Footnote 14
We first replicate the cross-sectional analyses from above, to check that the LISS data lead to comparable results. Running regression models that mirror those earlier as closely as possible, we find a clear interplay between winning/losing and the level of populism of the respective party. In accordance with this finding and our previous results, Table A15 shows that Dutch losers are less content than winners overall. Looking at the influence of populism, voters of more populist parties are less satisfied with democracy in the Netherlands. Finally, voting for a more populist party exacerbates the effect of losing.
As these results align well with the previous analyses, we move on to the next step and apply a longitudinal model regressing on citizens’ level of satisfaction with democracy after the election while controlling for its level before the election, as we did with the GLES panel. As the upper part ofFigure 6 demonstrates, these results again confirm our hypotheses. Losers are less satisfied with democracy than winners, and this effect increases the more populist a party is, even when controlling for pre-election satisfaction with democracy. Once more, the marginal effects plot at the bottom of Figure6 shows a substantial effect of losing for those respondents who voted for a more populist party. Analogous to the results based on the GLES panel, we find that for voters of non-populist parties it does not make a difference whether they win or lose in terms of their satisfaction with democracy.
Coefficient Plot and the Marginal Effects of Losing: Data from LISS Panel
The LISS data allow for additional robustness checks by adding variables to the models. First, we include a variable for whether a respondent voted for a party that was in opposition or in government before the election and also interact this variable with being a current winner or loser. This captures the possible effect that voters of some parties may have experienced a change from government to opposition or vice versa more often than other voters, which could influence the results. Table A16 in the Appendix shows that this is not the case and our results remain unchanged. In addition, since panel data provides us with information about voting behaviour over multiple elections, we control for respondents’ experience of being a winner or loser in the penultimate election (see Table A16 in the Appendix). However, controlling for this additional variable does not change our main results either.
Conclusion
Elections are inherently challenging because they inevitably produce winners and losers. Some parties will govern, while others will not. No matter how much democratic systems seek to dignify the role of the opposition – for example, in the UK, where it is officially recognized as His Majesty’s Loyal Opposition – losing remains frustrating. It means that a government one did not vote for will set the political agenda, potentially enacting policies that are not only disagreeable but also directly contrary to one’s interests.
In many countries, populist parties, politicians and movements have gained ground– often refusing to accept democratic outcomes that do not favour them. For populists, election results are not just a reflection of democratic competition but a test of legitimacy itself. As exemplified by Donald Trump and his supporters, the perception of democracy’s fairness is contingent on the outcome: winning affirms the system’s integrity, while losing is taken as evidence that the election was rigged.
Building on the losers’ consent literature, we have shown that the impact of losing increases with citizens’ populist attitudes. For populists, politics is a battle between an out-of-touch elite and ordinary citizens, between them and us, or between friend and foe, and when the wrong side wins, populist leaders often cast doubt on the legitimacy of democratic procedures. Using cross-national data from five waves of the CSES and panel data from Germany and the Netherlands, we demonstrate that losers and populist citizens are generally less satisfied with the way democracy works in their country. While the cross-sectional analyses allow us to cover a wide range of 200 elections in 57 countries from 1996 to 2019, the use of panel data allows us to test our hypotheses more rigorously by addressing potential endogeneity. In all cases, the negative effect of losing is significantly more pronounced for more populist citizens. Losing more severely affects the attitudes of populist voters due to core features of populist ideology.
In this article, we did not test whether losing affects support for democratic principles. However, our findings contribute to an emerging body of research examining the strength of citizens’ commitment to democracy. Our results suggest that populist losers tend to value electoral outcomes only when they win. This aligns with other studies indicating that supporters of populist parties are often willing to sacrifice democratic principles for partisan gain (Janssen Reference Janssen2023; Lewandowsky and Jankowski Reference Lewandowsky and Jankowski2023).
More broadly, research suggests that when political stakes are high, voters may prioritize partisan interests over democratic norms (Fossati etal. Reference Fossati, Muhtadi and Warburton2022; Graham and Svolik Reference Graham and Svolik2020; Simonovits etal. Reference Simonovits, McCoy and Littvay2022; Svolik Reference Svolik2020). When citizens strongly support a particular policy, they may not only tolerate violations of democratic principles but also justify politicians’ anti-democratic actions (Krishnarajan Reference Krishnarajan2023). While surveys indicate widespread support for democracy in principle, at least some segments of society appear willing to abandon these commitments in practice.
One limitation of this study is that we could not analyse how losing power affects the attitudes of populist voters when their party has led government. In countries such as Poland and Hungary, supporters of populist ruling parties tend to be more satisfied with democracy than opposition voters. However, it remains unclear how their satisfaction changes if their party loses power. To the best of our knowledge, no suitable panel data are currently available to examine this question.
Additionally, while our findings are broadly applicable, the geographical scope of this study is limited. Many Latin American countries, for instance, have seen both left- and right-wing populists compete for power and enter government at various times. Future research could test how electoral victories and defeats shape voters’ attitudes towards democracy in these contexts.
We also focused on the most commonly used dependent variable in the literature on losers’ consent – satisfaction with democracy – but future studies could extend the analysis to other indicators, such as trust in democratic institutions or perceptions of electoral integrity. If our theoretical arguments hold, we would expect a stronger effect of losing among more populist voters across these measures as well.
Finally, longer-term panel data could shed light on whether the negative effects of losing persist over time or fade after an initial disappointment. Repeated electoral defeats, in particular, could deepen populist voters’ alienation from democracy and fuel support for more radical alternatives.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/gov.2025.10.
Data availability
Replication data for this article is available at Harvard Dataverse under 10.7910/DVN/V2AQUJ.
Acknowledgements
We would like to thank three anonymous reviewers for their detailed and helpful comments. We would also like to thank Lea Elsässer, Julian Garritzmann, Sven Hillen, Claudia Landwehr, Nils Steiner and participants in workshops in Frankfurt, Lüneburg, Mainz, Mannheim, Saarbrücken and at the Max Planck Institute for the Study of Societies and at the Villa Vigoni for generous feedback on earlier versions of this paper.
Open access
