The winner-loser gap

Research consistently documents an attitudinal gap between electoral winners and losers and investigates varying contextual factors that affect the size of said gap. For instance, at the country level, institutions such as electoral systems moderate the extent to which a winner-loser gap manifests (Anderson and Guillory Reference Anderson and Guillory1997; Martini and Quaranta Reference Martini and Quaranta2019). To simplify a complex debate, under proportional systems, losers typically retain some representation, while in majoritarian systems, losing candidates have no such buffer and often result in a ‘the winner takes it all’ situation. In other words, in proportional systems, losers tend to lose less harshly, as their preferred party or an ideologically proximate party will still hold some seats and influence (Anderson Reference Anderson2005; Anderson and Guillory Reference Anderson and Guillory1997; Bernauer and Vatter Reference Bernauer and Vatter2012; Rich and Treece Reference Rich and Treece2018), while voters of a losing candidate in a majoritarian election do not benefit from such a buffer. Related to this, winning margins moderate the relationship: winners report greater satisfaction when victories are narrow, though margins of defeat do not affect losers (Howell and Justwan Reference Howell and Justwan2013).

Beyond electoral systems, the quality of democracy (Nadeau, Daoust, and Dassonneville Reference Nadeau, Daoust and Dassonneville2023), electoral fairness (Fortin-Rittberger, Harfst, and Dingler Reference Fortin-Rittberger, Harfst and Dingler2017; Magalhães Reference Magalhães2016; Moehler Reference Moehler2009), and the level of government on which the election happens, shape the size of the winner-loser gap. In higher-quality democracies, elections provide stronger signals of legitimacy, reducing losers’ discontent (Nadeau, Daoust, and Dassonneville Reference Nadeau, Daoust and Dassonneville2023), while fairly conducted elections lead to smaller attitude gaps as losers are more willing to accept their loss if elections are conducted fairly (Fortin-Rittberger, Harfst, and Dingler Reference Fortin-Rittberger, Harfst and Dingler2017; Magalhães Reference Magalhães2016; Moehler Reference Moehler2009). Notably, Daoust et al. (Reference Daoust, Nemčok, Broniecki and Loewen2024) find no significant winner-loser effects on democratic satisfaction immediately following elections in Norway, Iceland, and the Netherlands, all considered to have comparatively well-functioning institutions. However, perceptions of electoral fairness themselves are often affected by winning or losing elections, as losers have a higher likelihood of evaluating the conduct of elections as less fair compared to winners (Cantú and García-Ponce Reference Cantú and García-Ponce2015; Kernell and Mullinix Reference Kernell and Mullinix2019; Schnaudt Reference Schnaudt2023). Finally, evidence on spillovers from subnational elections is mixed: Singh, Karakoç and Blais (Reference Singh, Karakoc and Blais2012) find that winning or losing in subnational elections has an effect on national-level outcomes, while Daoust, Ridge, and Mongrain (Reference Daoust, Ridge and Mongrain2023) find no such spillover effect of subnational elections.

Focusing more directly on the mechanisms underlying the attitudinal differences between winners and losers of elections, two mechanisms are often cited to explain the gap, focusing on policy (Anderson and Tverdova Reference Anderson and Tverdova2001; Singh, Karakoç, and Blais Reference Singh, Karakoc and Blais2012) and psychology (Anderson Reference Anderson2005; Blais and Gélineau Reference Blais and Gélineau2007; Howell and Justwan Reference Howell and Justwan2013; Singh Reference Singh2014), respectively. The policy mechanism holds that losers expect fewer preferred policies to be enacted, as their preferred party has less influence over policy, lowering satisfaction, while winners expect the opposite (Anderson and Tverdova Reference Anderson and Tverdova2001; Singh, Karakoc, and Blais Reference Singh, Karakoc and Blais2012). As such, losers will be less satisfied with the policies and outputs of the system and will be less satisfied with democracy overall. Winners, in contrast, assume that more of their policy interests will be implemented and adjust their attitudes upwards.

The psychological mechanism assumes a less rational cost-benefit calculation of winners and losers of elections and rather builds on the idea of an emotional response. Like fans of a losing sports team, supporters of defeated parties experience negative emotions, leading to diminished political attitudes (Anderson Reference Anderson2005; Blais and Gélineau Reference Blais and Gélineau2007; Dalton Reference Dalton2018).

However, while the above-outlined mechanisms are argued to apply to both winners and losers of elections, several studies argue that the gap in attitudes is not driven by both winners and losers adjusting their attitudes. Instead, it is argued that winners adjust their attitudes upwards, while losers’ attitudes are assumed to remain stable (Esaiasson Reference Esaiasson2011; Nadeau, Daoust, and Dassonneville Reference Nadeau, Daoust and Dassonneville2023). Specifically, both groups may benefit cognitively from elections, which signal democratic functioning and provide voters with voice, offsetting negative emotions of defeat (Blais and Gélineau Reference Blais and Gélineau2007; Esaiasson Reference Esaiasson2011; Nadeau, Daoust, and Dassonneville Reference Nadeau, Daoust and Dassonneville2023; Van der Meer and Steenvoorden Reference Van der Meer and Steenvoorden2018).Footnote ¹ This positive boost should not only raise satisfaction with democracy but also translate into higher levels of trust toward systemwide institutions, as the conduct of an election signals to voters that political institutions continue to function properly and fairly. Accordingly, based on previous investigations, we expect that elections increase institutional trust overall (Esaiasson Reference Esaiasson2011; Van der Meer and Steenvoorden Reference Van der Meer and Steenvoorden2018):

However, since the resulting parliaments are filled with representatives of winning parties, their supporters should express higher institutional trust, as more trusted politicians now occupy the institution. Following Van der Meer (Reference Van der Meer2010), voters’ trust is based on four key dimensions: competence (belief in politicians’ ability to perform in line with voter expectations), commitment (intrinsic, based on shared values; or extrinsic, based on accountability), and predictability (consistency with past behavior). It is reasonable to argue that voters of a party evaluate their party higher on each of these traits compared to other parties, leading them to place higher trust in their party. Hence, if an individual’s preferred party wins, voters will place higher trust in an institution that they believe to be occupied by individuals in which they place higher trust. However, voters whose party loses an election may not experience the same upward adjustment of trust (Leiter, Clark, and Clark Reference Leiter, Clark and Clark2019).

At the individual level, research highlights several factors that shape attitudinal gaps after elections, such as valence perceptions (Leiter, Clark, and Clark Reference Leiter, Clark and Clark2019), whether a strategic or ideological vote was cast (Singh Reference Singh2014), and whether an individual is a first-time loser or has experienced electoral losses repeatedly (Delgado Reference Delgado2016). Reactions also depend on the ideological distance to the winning party: under proportional systems, defeats are softened when ideologically proximate parties win (Best and Seyis Reference Best and Seyis2021; Curini and Jou Reference Curini and Jou2016).

We argue that there is an additional factor beyond mere ideological distance that can affect voters’ reactions to winning or losing an election, namely, the ideological orientation of their preferred party.Footnote ² Particularly, voters of populist parties should react more strongly to losing or winning compared to voters of other parties. Recent work by Schäfer and Wenker (Reference Schäfer and Wenker2025) emphasizes that individuals with stronger populist attitudes react more negatively to electoral defeat due to the anti-pluralist and conflict-oriented nature of populist ideology, and so we advance a complementary argument by focusing on voters of populist parties and the mechanisms shaping their responses. Specifically, we argue that these voters are systematically exposed to anti-elite and anti-system narratives and therefore interpret electoral outcomes through this lens. Populist parties frequently employ narratives of a corrupt elite that is dominating political systems and that these elites are purposefully keeping anti-establishment, populist parties, from left or right, out of office (Holcombe Reference Holcombe2021), referring to negative polling results and even election results as ‘rigged’ and promoting conspiracy theories to explain results (Hameleers Reference Hameleers2021; Pirro and Taggart Reference Pirro and Taggart2023). Candidates of populist parties also express lower satisfaction with democracy and more negative attitudes towards elections in general (Gaio e Silva and Lisi Reference Gaio e Silva and Lisi2024), which finds expression in messages on alternative social media that denote an ‘us vs. them’ rhetoric during elections (Hameleers Reference Hameleers2020). Exposure to such messages, in turn, increases the populist attitudes of voters of populist parties (Hameleers Reference Hameleers2021) and likely further undermines perceptions of electoral integrity (Goldberg and Plescia Reference Goldberg and Plescia2024) and creates lower trust (Dahlberg and Linde Reference Dahlberg and Linde2016). Thus, the rhetoric to which voters of such parties are exposed should result in them interpreting electoral wins and losses through the lens of an ‘us vs. them’ and ‘corrupt elite’ narrative. Evidence from Central and Eastern Europe shows that populist supporters report higher trust when their party governs and significantly lower trust when it is in opposition (Kołczyńska Reference Kołczyńska2023).

Given this, an electoral loss should be perceived as a confirmation of negative attitudes towards the system and a decrease in expressed trust in the institution, as expectations of the ‘corrupt elite’ hindering ‘the political will of the masses’ are reinforced, further undermining the trust that individuals have in institutions. Indeed, research on, e.g., electoral fairness shows that voters of populist parties in general believe elections to be less fair (Goldberg and Plescia Reference Goldberg and Plescia2024) and that voters of such parties are more likely to believe in and consume so-called ‘fake-news’ or ‘alternative news’ in which narratives of rigged elections are more often presented (Schulz, Wirth, and Müller Reference Schulz, Wirth and Müller2020; Van Kessel, Sajuria, and Van Hauwaert Reference Van Kessel, Sajuria and Van Hauwaert2020).Footnote ³

In contrast, a win could be interpreted as finally having overcome the prejudice and hurdles that the corrupt elites have put into place and a perception that the opinions of the ‘true people’ will finally find expression in parliaments. This positive effect could be funded both in emotions (i.e., the emotion of winning after a long fight) and policy-related thoughts – most voters of populist parties might hold fundamentally different beliefs than the ones expressed by governing parties and might believe that only their preferred party can bring about policy change in the desired direction. To exemplify, populist radical-right parties often hold issue ownership on migration issues, and voters of populist parties believe their party to be the only one able to address these issues appropriately (Down and Han Reference Down and Han2020). An electoral win should thus lead to an even stronger effect on the attitudes of supporters of populist parties compared to voters of other parties.

Moreover, populist actors emphasize majoritarian understandings of democracy over consensus oriented or pluralist approaches. Their claim to represent ‘the will of the people’ (Mudde and Kaltwasser Reference Mudde and Rovira Kaltwasser2017) is often linked to calls for more direct democratic instruments and a weakening of intermediary institutions such as representative legislatures and administrative bodies (Mansbridge and Macedo Reference Mansbridge and Macedo2019). As a result, election outcomes potentially carry particularly high symbolic and political significance for populist voters in comparison to voters of parties supporting democratic pluralist parties and can be interpreted as validations or rejections of popular sovereignty by them.

Although lower-trust individuals are more likely to support populist parties (Doyle Reference Doyle2011; Hooghe, Marien, and Pauwels Reference Hooghe, Marien and Pauwels2011), our design mitigates concerns about reverse causality. The quasi-random timing of survey fieldwork before and after elections allows us to isolate electoral effects on trust. While the arguments we develop here pertain to populist parties at large, independent of their ideology, in our design, right-leaning parties are dominant (eight right-leaning parties, three left-leaning parties).

While we, so far, mainly discussed the theoretical underpinnings of the winner-loser gap after elections as well as the potential moderating effect of populist ideology, our study also addresses two empirical shortcomings in the literature. First, most previous research leverages observational data, either cross-sectional data with outcomes measured far after an election, which could introduce confounding effects, or panel surveys (e.g., several contributions from the ‘Making Electoral Democracy Work’ project by Blais (Reference Blais2010)) conducted before and after an election to investigate potential winner/loser gaps. While valuable, these cross-sectional and panel designs limit causal inference. By contrast, we leverage the quasi-random implementation of the European Quality of Government Index survey, which enables causal estimation of short-term electoral effects, an approach that we detail below. This focus on causality is crucial, as individuals with lower prior political support may be more likely to vote for parties that are systematically excluded from government, thereby making it unclear whether losing reduces support or low support leads individuals to become electoral losers in the first place.

Second, the definition of electoral winners and losers remains contested. Many studies define winning as government entry, yet voters may still feel like winners when their party gains votes or seats, even without joining government (Kern and Kölln Reference Kern and Kölln2022; Stiers, Daoust, and Blais Reference Stiers, Daoust and Blais2018; Van der Meer and Steenvoorden Reference Van der Meer and Steenvoorden2018). Our design allows us to explore the idea that winning and losing seats matters and extends these approaches by also considering whether parties outperform or underperform relative to pre-election polls, i.e., whether a party outperforms what electoral polls predict or underperforms in comparison to polling. Voters rely on polls to form their expectations about a party’s performance (Barnfield Reference Barnfield2023; Irwin Reference Irwin2002; Morwitz and Pluzinski Reference Morwitz and Pluzinski1996; Blais and Bodet Reference Blais and Bodet2006; Dahlgaard et al. Reference Dahlgaard, Hedegaard Hansen, Hansen and Larsen2016), and therefore, any change in the expected performance should also be reflected in the update of attitudes towards institutions. Particularly, as polls will shape voters’ expectations before election day on how many seats their party will gain and how likely they are to enter government afterward, a change in performance compared to expectations should result in a change in attitudes. Hence, we argue that voters are more myopic than previous research argues and update attitudes not only based on government entry or seat share but also on short-term performance relative to polls. By defining winning and losing based on expectations, we also build directly on Plescia (Reference Plescia2019), who operationalized winning and losing based on whether a small party was expected to clear an electoral threshold and whether voters’ expectations that a party would become the largest party after an election were confirmed. Our measure of winning and losing based on poll measures follows the same logic of disappointed or surpassed expectations but is applicable to the entire set of parties in a party system. Leveraging different operationalizations is also relevant in the context of causality: voters with low levels of trust might tend to vote to a larger extent for parties that tend to be excluded from government. It is, hence, imaginable that low support determines people’s vote choice and, by doing so, turns them into losers – our measure of under- and over-performing in combination with our causal design allows us to address this concern. Using three operationalizations of winning and losing provides robustness, though we believe our theoretical elaboration of different measures of winning and losing is valuable, we do not have strong priors regarding which effect will be strongest and instead remain agnostic regarding effect magnitudes.

Research design

Our research design relies on survey data from the latest two rounds of the European Quality of Government Index (‘EQI’; see Charron et al. Reference Charron, Lapuente and Bauhr2024, Reference Charron, Lapuente, Bauhr and Annoni2022).Footnote ⁴ The survey, fielded roughly every three years, inquires about citizen attitudes, perceptions, and experiences, with a focus on corruption, impartiality, and quality of local/regional public services. It also includes a set of questions on trust in national institutions, political attitudes and partisanship, and demographic information, inter alia. Crucially for this study, both rounds contain a respondent-level indicator of the exact day on which the survey was administered and completed.

Sample, data, and variables

Due to the design, the sample employed in this study relies on respondents from a subset of EQI countries and regions where an election took place during the fieldwork. In round 5 (2023), the fieldwork took place during elections in three countries − Poland, Slovakia, and Luxembourg, along with the regional elections in two German Länder − Hessen and Bavaria. Previously, in round 4 (2020), the fieldwork was done during parliamentary elections in Romania. Thus, in total, there are 19,059 individual respondents in these four countries and two regions used in the analyses. Leveraging cases from diverse electoral contexts, our sample enables investigation of the winner-loser gap under different conditions: for instance, Luxembourg implements compulsory voting, which may moderate the gap (Singh Reference Singh2023), while the two German regional elections allow further exploration of subnational dynamics (Daoust, Plescia, and Blais Reference Daoust, Plescia and Blais2023; Singh, Karakoc, and Blais Reference Singh, Karakoc and Blais2012).Footnote ⁵ More details on the EQI survey and sample, along with summary statistics of all variables, are found in the online Appendix section 1.

Table 1 summarizes the sample cases, along with election dates and number of observations by country/region, and hence provides information relevant to the ‘unexpected event’ assumption. In all countries, polls suggested highly uncertain election outcomes between the top two projected contenders (parties or coalitions) in the latest national polls prior to the election, whereby the differences between the two largest contenders were essentially within the margins of error. For example, in Poland, experts stressed that the high-stakes election was highly uncertain,Footnote ⁶ while in Luxembourg, the incumbent government was predicted to hold a slim majority but ultimately did not. In Slovakia, the polls were extremely narrow, leading to high uncertainty of any electoral winners,Footnote ⁷ and also in Romania (2020), polls predicted just a 1.5% difference between the two largest contenders, PSD (Partidul Social Democrat, Social Democratic Party) and the incumbent PNL (Partidul Național Liberal, National Liberal Party), the latter of which was predicted by polls to be the largest party yet finished second. Thus, while the elections themselves were expected, the results were not. By contrast, results in the two German Länder largely matched pre-election polls, meaning the assumption of ‘unexpectedness’ is less plausible there; we return to this issue later.

Table 1.

Sample and observations

Table 1. Long description

A table comparing election dates, poll differences, and results across countries and regions. The table has 7 rows and 5 columns. The columns are labeled Country, Election date, Avg. Poll Diff., Diff. vs., and Result. The rows are as follows: Row 1: Country, Election date, Avg. Poll Diff., Diff. vs., Result; Row 2: Poland, 15/10/2023, 3.9, PiS vs. KO, R to cen. L; Row 3: Slovakia, 30/09/2023, 0.7, Smer vs. PS, R to cen. L; Row 4: AQ20 ourg, 08/10/2023, 2*, Gov vs. Opp., cen. L to cen. R; Row 5: Romania (2020), 06/12/2020, 1.5, PSD vs PNL, cen. L to cen. R; Row 6: Regions Bayern, 08/10/2023, 21.6, CSU vs. Green, CSU hold; Row 7: Regions Hessen, 08/10/2023, 16.5, CDU vs. Green, CDU hold. The table provides information on the average poll differences and the results of elections in various countries and regions, highlighting the uncertainty and variability in election outcomes.

Note: ‘Avg. poll diff.’ refers to the mean difference in the two largest groups, highlighted under ‘diff vs.’, according to the three opinion polls in each country/region closest to the election date. In the case of Luxembourg, the difference refers to the projected seat difference between the previous government and the opposition, where the government was predicted to hold its majority in Parliament (31 to 29). ‘L’ and ‘R’ stand for ‘left’ and ‘right’ broadly speaking, according to EU parliament affiliation. The ‘result’ column shows whether there was a shift in government and in what direction or an incumbent hold.

Regarding how country-sample selection might affect the results, previous research shows that party volatility is generally higher in Eastern and Central Europe (Powell and Tucker Reference Powell and Tucker2014), implying that partisan loyalty – and partisan gains or losses – may be less salient, especially in the three former communist countries. However, political polarization has been rising in countries such as Poland, Slovakia, and Romania (Coppedge et al. Reference Coppedge, Gerring, Knutsen, Lindberg, Teorell, Marquardt, Medzihorsky, Pemstein, Fox, Gastaldi, Pernes, Rydén, Römer, Tzelgov, Wang and Wilson2024; Haughton et al. Reference Haughton, Pomorska, Malová and Deegan-Krause2024; Horonziak Reference Horonziak2022), which could amplify the trust gap between winners and losers. It is therefore an open empirical question whether winner-loser effects will be stronger or weaker in these former communist states compared to elections in Luxembourg and the German Länder.

Our main outcome of interest is trust (or confidence) in the national parliament. Respondents were asked to indicate their level of confidence on a 1–10 scale, with higher values reflecting greater confidence (see online Appendix for exact question formulation).Footnote ⁸ This outcome differs somewhat from many studies in the literature, which mainly rely on questions concerning democratic satisfaction, an outcome not measured in the EQI survey. On the one hand, we view both confidence in parliament and democratic support as indicators of greater latent ‘systems support’, as opposed to ‘specific support’ (Easton Reference Easton1965), and studies have shown that the two indicators are strongly correlated at the individual level (Zmerli and Newton Reference Zmerli and Newton2007). On the other hand, as Norris (Reference Norris1999, Reference Norris2017) notes, trust and confidence in parliament are evaluations of regime institutions in practice, rather than abstract democratic principles, through which citizens express broader ideals about the democratic system. We therefore treat our outcome as complementary to existing measures of system support, while distinct enough to offer a novel angle in the literature.

With regard to the explanatory variables, a key question is how to operationalize electoral ‘winners’ and ‘losers’, as previously this has been shown to be subject to the perceptions of voters rather than purely objective (Plescia Reference Plescia2019; Stiers, Daoust, and Blais Reference Stiers, Daoust and Blais2018). In most observational studies, these categories are typically clear because enough time has elapsed to form a government. In our case, however, respondents were surveyed immediately after the elections, when the status of winners and losers may have been less obvious. Due to this uncertainty, we categorize parties as winners, losers, or ‘other’ via three different approaches: First, using the power of hindsight to code what eventually became the government (‘winner’) and opposition (‘loser’) parties in each case post-election, we categorize winners and losers as those who enter government or serve in the opposition, respectively, with ‘winners’ having at least one cabinet position, which previous research has shown to be a highly salient predictor of democratic satisfaction (Singh, Karakoç, and Blais Reference Singh, Karakoc and Blais2012). Second, given that this might not have been clear to voters in all cases just days/weeks after the election, we use a simple categorical measure of seats gained (or lost) in parliament. In other words, under this operationalization, all parties that gained any seats are ‘winners’, those that lost any are ‘losers’ (regardless of eventual government status), and we also distinguish those with no difference and non-voters. In addition, we also show the continuous effects of seat gains/losses in the Appendix. Finally, it might be difficult for some voters to recall their party’s relative gains/losses compared to the previous election; thus, they rely on more recent heuristics to assess their party’s success in an election. Hence, we take data from the most recent ‘poll of polls’ published by Politico ^{Footnote 9} before each election and create a three-category variable, whereby ‘1’ equals underperforming, ‘2’ equals over-performing, and ‘3’ equals election results that match the polls. Hence, ‘winners’ in this case are those parties that outperformed the polls, while ‘losers’ are those that underperformed in terms of the percentage of the vote.Footnote ¹⁰

Although the winner-loser measures correlate in the expected direction, they do not perfectly overlap. The seat and government winner variables demonstrate the strongest correlation, while the polling and government party measures show the lowest degree of correspondence.Footnote ¹¹ For example, in the Polish election, supporters of Lewica (Left) are ‘winners’ in government formation yet lost seats and underperformed relative to the polling, while supporters of Konfederacia are winners in the seat measure yet ‘losers’ in government formation and also underperformed in polling. Therefore, we argue that using several different measures of this variable will provide a more robust empirical understanding of the effects of elections on political trust. For each of these three measures, we code individuals based on their self-reported vote choice.Footnote ¹² Specific codings of each measure by party for each election are found in Appendix Table A3, and polling predictions and sources by party are provided in Appendix Table A4.

Table 2 presents a pooled balance table to assess the ‘as if’ random criterion of the design, highlighting differences in means across a number of commonly used covariates along with the dependent variable (trust) and several other perceptions of institutions.Footnote ¹³ First, in terms of the country-level sample, we observe that there is overall empirical support for the notion that elections enhance political support (i.e., Esaiasson Reference Esaiasson2011; Van der Meer and Steenvoorden Reference Van der Meer and Steenvoorden2018), whereby we observe (slightly) higher national trust and we see less economic dissatisfaction. In terms of relevant covariates, there is a balance among some covariates (gender and unemployment), yet the treatment group is on average two years younger, more educated, more urban, and less rural compared with the pre-election control group. In addition, there is a higher proportion of ‘winners’ and non-voters in the treatment group than in the control group, though the ‘loser’ and ‘other’ groups are statistically balanced. Following recommended practice (Muñoz, Falcó-Gimeno, and Hernández Reference Muñoz, Falcó-Gimeno and Hernández2020), we adjust for these imbalances by including the relevant covariates in our models. Finally, we note that the distribution of the number of observations across groups is skewed, with roughly 82% and 18% of the respondents falling into these categories, respectively. This imbalance varies by country – Slovakia and Luxembourg, in particular, contribute relatively few post-election observations (see Appendix section 2 for individual country tables). While not necessarily a problem for the validity of the results per se, this imbalance will lead to less statistical power in these cases, resulting in higher standard errors and more conservative inferences.

Table 2.

Pooled balance table − country and regional level samples

Table 2. Long description

The table presents a comparison of country-level and regional-level samples across various covariates. It includes pre-election and post-election data, highlighting differences in means. The table has 21 rows and 10 columns. The columns are labeled as Pre, Post, and Diff. for both Country-level sample and Regional level sample. The rows are labeled with different covariates such as Nat. trust, Female, Age, University, Urban, Rural, Unemployed, Populist, Non-voter, and several others related to government formation and election outcomes. Each row provides values for Pre, Post, and the difference (Diff.) between these values for both country-level and regional-level samples. Notable trends include changes in national trust, age, education level, urban and rural populations, and political affiliations. The table also includes the number of observations for each category.

Note: Pooled means of respondents pre (‘control’) and post (‘treatment’) national and regional elections, respectively. See Appendix for balance tables by each country and region. ***p < 0.01, **p < 0.05, *p < 0.1.

In the regional subsample from Hessen and Bavaria, balance is generally stronger, with the only statistical differences being among gender and education, whereby the treatment groups are less female and have less tertiary education. We thus control for these variables in the regional analysis. All other covariates are balanced, along with the control and treatment groups being closer in overall sample size than the country-level sample. Interestingly, we observe a loss in overall national-level trust as well as an increase in economic dissatisfaction immediately following the election, which suggests that perhaps the positive electoral effects on political support apply to national elections only.

Concluding remarks

What is the effect of an election on trust in political institutions among citizens? In this study, we test several hypotheses regarding the electoral effects of democratic elections on citizens’ evaluations of political institutions using recent data from the European Quality of Government Index survey (Charron et al. Reference Charron, Lapuente and Bauhr2024, Reference Charron, Lapuente, Bauhr and Annoni2022). Leveraging cases in which survey fieldwork coincided with national or regional elections, we employ a quasi-experimental design to test three primary hypotheses. First, we find evidence of an overall immediate boost in trust post-election, supporting our first hypothesis. Second, there are indeed winner-loser effects, which are mostly consistent across different operationalizations of ‘winner/loser’ (H ₂ ). In this case, we find strong ‘winner-effects’ across the majority of cases that are robust to the various operationalizations. Additionally, we find some evidence for negative ‘loser-effects’, yet these are more case- and measure-specific, and the effects are smaller in magnitude than the increase in trust among winners. Finally, we test whether the effects of winning or losing elections are stronger among supporters of anti-establishment populist parties. We find consistent support that populist ‘winners’ increase political trust at significantly higher rates than non-populist partisans. Moreover, losing has a substantially greater negative effect on trust among populists compared with non-populists, supporting our third hypothesis and in line with recent work by Schäfer and Wenker (Reference Schäfer and Wenker2025).

This study contributes to research on the winner-loser gap in several ways: First, by leveraging the quasi-random implementation of the EQI survey, we provide causal estimates of the immediate electoral effect of winning and losing on political trust; we contribute to previous research, which mainly relied on observational data with a longer time lapse from the elections included so far. By demonstrating that winning and losing elections results in a causal updating of attitudes towards institutions and that this update is not unequivocally positive, we also draw into question the extent to which measures of trust in institutions truly capture perceptions of the institution itself. They may instead also reflect voters’ evaluations of the actors who populate the institution.

In addition, we extend previous conceptualizations of winning and losing beyond participation in government or a change in seat share by arguing that voters also evaluate their party’s performance relative to recent polls. Indeed, our results show interesting variation depending on the operationalization of winning and losing, such that eventual government participation (or anticipation thereof) has stronger effects on trust than a change in seat share. Performance relative to polls also shows larger effects, particularly among voters of populist radical-right parties. Our findings also demonstrate some heterogeneity across cases, highlighting the importance of the context and level (country or region) in which elections take place. Thus, while the overall pooled effects tend to be robust, the effect magnitude varies across countries and elections. Finally, by theorizing why populist radical-right voters are particularly sensitive to winning or losing, we contribute to theoretical development in the field.

These findings also have important implications for democratic politics. That elections can produce immediate increases in institutional trust suggests that democratic contests continue to function as moments of renewal. At the same time, the asymmetric effects of winning and losing, particularly among populist voters, raise concerns that elections could deepen affective polarization.

While our study makes important contributions to the field of winning and losing elections, it is not without shortcomings. First, it cannot assess the long-term effects of winning or losing − i.e., we cannot conclude whether voters continue to update their trust once a new government takes office.Footnote ²⁰ Similarly, our study cannot identify the mechanisms underlying changes in institutional trust. We cannot investigate whether voters update their attitudes based on experiencing the emotion of winning or losing or whether they base their updated attitudes on policies in line with their preferences being implemented by individuals in whom they have higher trust. Similarly, we cannot establish whether the exposure to populist messages drives populist voters’ attitude change or whether the effect is driven by other mechanisms. Future research should examine the drivers of these attitude updates, building on the work of Daoust, Plescia, and Blais (Reference Daoust, Plescia and Blais2023) and investigating why populist, anti-establishment voters respond more strongly.

With respect to future research, understanding whether these immediate shifts in trust persist, and through which mechanisms, is critical both for refining theories of political behavior and for assessing the durability of democratic legitimacy.