Introduction
A rich literature on government breakdown has informed scholars that parties’ decisions to leave coalitions are strategic (Diermeier and Stevenson Reference Diermeier and Stevenson2000, Reference Diermeier and Stevenson1999). In their seminal article, Lupia and Strøm (Reference Lupia and Strøm1995) argue that when a government is plagued with internal policy differences, coalition parties consider how likely a government termination will result in them forming another government versus an early election, and their electoral prospects should such an election occur. Relatedly, So (Reference So2024) finds that a governing party does not lose votes when the government terminates for policy reasons. In fact, in a coalition setting, junior coalition partners are electorally rewarded for refusing to compromise on their policy positions. In contrast, voters punish government dissolution that stems from non-policy conflicts.
The implication is that government terminations stemming from policy conflicts are not perceived as events that damage parties’ electoral images; in contrast, terminations for non-policy reasons are. This type of government breakdown is the product of personal infighting or misconduct by members of government. Having to interrupt governance because of scandals should elicit disapproval from voters. This type of government termination not only necessitates a recalculation of how voters’ policy interests and preferences are represented in the legislature, but this need could have been prevented. Since the reason behind the breakdown is not tied to voters’ policy interests or preferences, voters may become dissatisfied with how democracy functions in their countries. This begs the question: How does this type of government termination affect voters’ evaluations of the performance of the government as a whole, and equally important, does it also affect voters’ satisfaction with the way democracy functions in their own countries?
In this paper, we examine voters’ reactions to a particular type of non-policy government terminations – those resulting from scandals. Although government terminations are a fact of life in parliamentary governance, scandal-based terminations are not a necessary byproduct of policy cooperation and compromise in government but rather the result of blunders. Thus, this type of termination may motivate citizens to evaluate the democratic functioning of their countries in a more negative manner. Surprisingly, existing work on coalition termination has largely been silent on the consequences of scandal-based coalition terminations. Although there is a plethora of works that examine the impacts of scandals and corruptions on citizens’ political attitudes and vote choices (e.g., De Vries and Solaz Reference De Vries and Solaz2017; Dewan and Dowding Reference Dewan and Dowding2005; Incerti Reference Incerti2020), as well as the strategic incentives behind coalition terminations (as the above literature attests), research on these two topics has developed only in tandem. While scandals can have deleterious effects on voters’ perceptions of the quality of democratic representation in their countries, it is not immediately clear if scandal-based terminations would exert the same effect, particularly as the government took accountability and dissolved. Thus, current research has yet to reveal how voters perceive coalitions that terminate because of scandals, which shatters the governing competence of the responsible party but not the competence of the innocent part(ies) (So Reference So2024).
We argue that scandal-based government termination alters individuals’ political attitudes in two distinct but complementary ways – their satisfaction with the way democracy functions in their countries and their propensity to vote for each of the coalition parties. Scandals and the ensuing necessity of having to form a new government and/or hold elections diminish voters’ perceptions that democracy is functioning well. Thus, despite the government taking accountability for the scandal by resigning, scandal-based terminations reduce citizens’ satisfaction with the state of democracy in their countries. Similarly, since scandals signal to individuals that the responsible party cannot be trusted to govern in a professional, efficient manner, scandal-based terminations reduce individuals’ perceptions of the party’s governing competence. Consequently, they should become less inclined to vote for the responsible party post-termination. In addition, we examine how this type of termination affects voters’ perceptions of the government’s general performance and offer two competing hypotheses for the relationship.
To test our argument, we turn to voters’ attitudinal reactions to the 2019 Austrian ‘Ibiza Scandal’, which caused the termination of the governing coalition between the Austrian People’s Party (ÖVP) and the Austrian Freedom Party (FPÖ). The nature of this affair was widely reported in the Austrian media. Thus, we can credibly assume that the average voter knows about the affair and could attribute fault to the responsible party – the FPÖ. Equally important, the 2019 Austrian National Elections Study (Aichholzer et al. Reference Aichholzer, Partheymüller, Wagner, Kritzinger, Plescia, Eberl, Meyer, Berk, Büttner, Boomgaarden and Müller2022) fielded relevant questions on vote intentions, satisfaction with democracy, and the government’s performance in the weeks before and after the termination. This creates the environment of a natural experiment and allows us to employ the Unexpected Event during Survey Design (UESD). We use the government termination as a treatment condition to test how it affects voters’ intention to vote for the governing parties, the government’s performance, and the way democracy functions in their countries. We find that the coalition termination reduces voters’ satisfaction with the way democracy functions in Austria and lowers their propensity to vote for the FPÖ – and only the FPÖ – but does not alter their perception of the performance of the ÖVP-FPÖ government.
We complement these results with an analysis of voters’ reactions to the corruption probe involving Sebastian Kurz, ÖVP leader and the prime minister at the time, using a similar UESD with data from the European Social Survey (ESS). The Kurz case differs from the ‘Ibiza’ scandal in that the coalition remained in place even as Kurz was coerced to resign from the prime minister’s position. This allows us to examine the differences between a scandal that breaks a coalition and one that does not. The results show that the scandal reduced voters’ satisfaction with democracy and their support for both coalition partners. Moreover, in contrast to the Ibiza scandal, the scandal also results in lower evaluations of the government’s performance.
Our findings suggest that voters are knowledgeable on salient political events and are able to discern why governments break down. They also support existing research findings that voters primarily attribute responsibility for the government performance output to the prime minister’s party, as well as being capable of isolating non-policy events that plague the government from the government’s overall policy performance. In addition, the finding that scandal-based coalition terminations reduce voters’ satisfaction with the way democracy functions in their countries suggests a negative externality of government breakdown. This highlights the need to revive a long line of research on government breakdown, particularly on its consequences on voters’ political and democratic attitudes.
Theoretical framework
We conceptualize scandals as non-policy-based, publicized conflicts that range from corruption, extra-marital affairs, or personal infighting between cabinet ministers to gaffes in media interviews (So Reference So2024). Even though there are no standardized definitions of scandals and that a scandal can consist of personal wrongdoing, corruption, or even gaffes, the core difference between scandals and other events that risk the breakup of a coalition is the former’s non-policy nature. That is, we are interested in how non-policy-based conflicts that threaten the integrity of a government can affect voters’ perceptions of the governing partners as well as their democratic attitudes.Footnote 1
We assume that voters in general possess enough political sophistication to understand that their perceptions of the party plagued by scandal, of the government’s performance, and their evaluations of the way democracy functions are distinct concepts – even if they are correlated. Current research provides some support to this assumption. Using responses from the European Social Surveys (ESS), Morgan-Jones and Loveless (Reference Morgan-Jones and Loveless2023) find that when prime ministers or presidents call early elections, citizens’ satisfaction with democracy drops significantly because the government is perceived to be evading democratic accountability. Although their theoretical focus is on early elections and not government terminations, their finding suggests the following. First, voters can discern governments’ strategic use of early elections. Second, their simultaneous rewarding of incumbents for early elections and becoming more dissatisfied about the way democracy functions in their country signals that voters prefer governments to be stable and carry themselves to full term, even as they reward governments for good economic performance. As such, voters may be politically sophisticated enough to distinguish between the performance output of a government and how the government measures up to the voters’ own democratic ideal. That is, they possess some degree of rationality in assessing the behaviors of governments. Even if they choose their parties based on social identities and partisan loyalty (Achen and Bartels Reference Achen and Bartels2017), voters do pay attention to the government’s performance to evaluate the incumbent’s competence and parties’ proposed policies (Besley Reference Besley2006; Ferejohn Reference Ferejohn1986). This opens the possibility that voters are able to distinguish between the policy performance of the government, the governing competence of individual governing parties, and their satisfaction with democracy in general.
How do voters perceive government terminations and corruption scandals?
We begin our theoretical argument with the divergent effects of conflictual government dissolutions (So Reference So2024). First, voters do not punish the prime minister’s party for government breakdown that results from policy discord, and they even reward junior coalition partners for it. This suggests that policy-based terminations do not reduce voters’ satisfaction with the way democracy functions in their countries. Since a core function of parties is to represent the interests and preferences of their supporters, voters may nevertheless remain optimistic about parties’ ability to defend the interests of their supporters post-breakdown. Thus, the interruption in governance may well be perceived as a resetting of the government’s broader policy goals and a recalibration of whose policy interests the government should promote. Second, when a government terminates due to a non-policy conflict such as scandals, voters punish the part(ies) that are responsible for the conflict, regardless of whether the party is the prime minister’s party or a junior coalition partner. Unlike a recalibration of policy directions, since the interruption in governance stems from mistakes, blunders, or personal infighting, voters may perceive the party as lacking governing competence. If so, this loss of competence should motivate voters to evaluate the party and the quality of democracy in their countries less favorably (Easton Reference Easton1975; Powell Reference Powell2004b; Singh, Karakoç, and Blais Reference Singh, Karakoç and Blais2012). These findings also suggest that voters possess the necessary knowledge and rationality to distinguish between different types of government terminations and will react accordingly at the polls.
In order to examine how voters react to government terminations that result from scandals, we first need to understand how voters perceive and react to scandals. Clark (Reference Clark2009) finds that scandals that do not involve corruption, such as personal infighting or gaffes, hurt the offending parties’ valence, i.e., non-policy image, and therefore harm their electoral performances. When it comes to corruption scandals, the conclusions remain mixed. Some find that corruption harms the offending parties’ vote shares (Costas-Pérez, Solé-Ollé, and Sorribas-Navarro Reference Costas-Pérez, Solé-Ollé and Sorribas-Navarro2012; Rienks Reference Rienks2023) and that it reduces political participation, trust in government institutions, and partisans’ evaluations of their own candidates (Bowler and Karp Reference Bowler and Karp2004; Daniele, Galletta, and Geys Reference Daniele, Galletta and Geys2020; von Sikorski, Heiss, and Matthes Reference von Sikorski, Heiss and Matthes2020). These findings suggest that scandals motivate citizens to evaluate the functioning of democracy in their countries in less favorable ways. Yet, the effects of scandals do not always ‘stick.’ For instance, the electoral response to coalition scandals is muted (Jiménez and Cainzos Reference Jiménez, Cainzos, Garrard and Newell2006). This is in part due to partisan bias: supporters of the party at fault are more tolerant (Anduiza, Gallego, and Muñoz Reference Anduiza, Gallego and Muñoz2013). Also, voters’ reactions to the scandal are short-lived (Ares and Hernández Reference Ares and Hernández2017), suggesting that scandals do not always hurt the party responsible. Indeed, Pereira and Waterbury (Reference Pereira and Waterbury2019) show that in the US, voters discount corruptions that were publicized outside of election years; however, moral scandals (such as private misbehavior) and corruptions that occur in election years are met with voter punishment.
Equally important, parties can ‘recover’ the potential loss of votes by taking responsibility for the scandal. Doing so signals to voters that the party is taking the scandal seriously and that it acknowledges the need for redress, and also that the party is competent enough to understand that good governance should be free of scandals. Consequently, voters’ perceptions of the party’s governing competence should improve. Dewan and Dowding (Reference Dewan and Dowding2005) find that when a non-policy scandal hits the government, voters’ approval for the incumbent parties decreases. However, firing the minister responsible for the scandal exerts a corrective effect on the government’s approval rating. If one translates this logic to the coalition setting, when an individual minister becomes involved in a scandal, firing them may shield the coalition from negative evaluations from voters. When individual members are forced to resign and take the blame, the responsibility for the scandal becomes less diffuse. The remaining government can more credibly claim that the scandal was caused by ‘one bad apple’ rather than a systemic problem. That is, resignations by individual ministers shield the party from being blamed for the scandal; at the very least, it becomes difficult for voters to decide whether the parties can be blamed (De Vries and Solaz Reference De Vries and Solaz2017).
The above studies reveal that the effects of scandals on voters’ democratic attitudes and evaluations of parties are nuanced. Also, meta-studies and literature reviews suggest that the huge effects of corruption on vote shares are mostly found in survey experiments and rarely in the field (De Vries and Solaz Reference De Vries and Solaz2017). Survey experiments are especially prone to social desirability bias, and it is, in a sense, costless to change one’s vote in an abstract hypothetical scenario. Furthermore, survey experiments rarely capture the complexity of vote decisions in the real world (Incerti Reference Incerti2020). In contrast, using Unexpected Event during Survey Designs (UESD) to study voters’ reactions to scandals may offer a more accurate picture and provide much-needed external validity to this field. Two recent studies, both examining the Barcenas scandal from Spain in 2013, add important nuances to voters’ reactions to scandals. First, Solaz, De Vries, and De Geus (Reference Solaz, De Vries and De Geus2019) demonstrate that voters do not process information on corruption objectively. There are partisan and other in-group biases, which might lead voters to turn a blind eye (Anduiza, Gallego, and Muñoz Reference Anduiza, Gallego and Muñoz2013). Similarly, Ares and Hernández (Reference Ares and Hernández2017) used UESD to show that the effect of the corruption scandal on political trust was muted among supporters of the party at fault. They provide further evidence that the effect of the scandal decayed quickly, which likely explains why the electoral effects are relatively minor.
In sum, while the above works elucidate how voters perceive scandals as well as early elections (as one type of government termination), much less is known about how scandal-based government terminations affect other important ways in which voters perceive governing parties and their democratic attitudes. If a scandal prompts the coalition to terminate, does the act of termination achieve the same corrective effect as firing individual ministers? That is, do voters ‘forgive’ the governing party that committed the scandal, or do they punish the party at the polls? Do individuals also become dissatisfied with the way democracy functions in their countries? Our task here is to investigate the effect of scandal-based government terminations on citizens’ evaluations of the government’s performance and democratic attitudes.
Theoretical expectations
Expectations at the party level
Our theoretical starting points are 1) nonpolicy-based conflictual government terminations are met with voter punishment against the part(ies) responsible for the conflict (So Reference So2024), and 2) government scandals reduce voters’ political trust (Anderson and Tverdova Reference Anderson and Tverdova2003; Bowler and Karp Reference Bowler and Karp2004; von Sikorski, Heiss, and Matthes Reference von Sikorski, Heiss and Matthes2020). We first argue that the need to terminate the government due to a scandal reduces individuals’ propensity to vote for the coalition party responsible for the scandal. A scandal-based coalition termination can send one of two possible messages from the responsible party. Either the party affected by the scandal supports the minister(s) responsible for the scandal and is willing to lose office benefits by doing so, which would send the signal that the party lacks basic understanding of how to govern in a competent manner, or the scandal is so severe that firing the individual minister(s) alone is insufficient in redressing the problem, which would send the signal that the party failed to choose a competent minister and that the party was not able to prevent the minister from committing the scandal.
Both would imply the following. Compared to cases in which the party remains in a coalition post-scandal by firing the minister responsible, which would help the party recover its electoral image (Dewan and Dowding Reference Dewan and Dowding2005), scandal-based terminations lower individuals’ perceptions of the governing competence of the responsible coalition party. As Green and Jennings (Reference Green and Jennings2012) argue, a party’s governing competence constitutes a large part of voters’ calculus. In the case of the economy, bad economic performance results in a loss of perceived governing competence in general (Green and Jennings Reference Green and Jennings2017). If a scandal has impaired governance to such a degree that the government needed to terminate, the party responsible for the scandal should be perceived as unfit to govern. This loss of competence, then, should diminish individuals’ inclinations to vote for the party. Therefore:
Voter Support Hypothesis for the Responsible Party (H1a): A scandal-based coalition breakdown lowers voters’ propensity to vote for the coalition party responsible for the scandal.
At the same time, coalition termination prevents voters from perceiving other coalition parties as losing governing competence – particularly if blame for the scandal can be attributed to one coalition partner (So Reference So2024). Termination signals that the coalition partner is distancing itself from the scandal-clad party. This is important to protect the coalition partner’s image of governing competence: by refusing to continue governing together, the partner relays to the electorate that it disapproves of the scandal and is willing to risk office benefits to ‘do the right thing.’ Thus, voters’ propensity to vote for the partner part(ies) should remain unchanged:
Voter Support Hypothesis for Coalition Partner (H1b): A scandal-based coalition breakdown does not affect their propensity to vote for the non-responsible coalition partner(s).
This is distinct from the transfer effect of votes from the scandal-laden party to its governing partner(s). Although withholding their voter support for the scandal-clad party means that voters decide to choose another party, it is not necessary that these voters then transfer their votes to a governing partner. It may be possible, for instance, that voters decide not to vote altogether or vote for an opposition party. That is, while voters should be less willing to vote for the scandal-laden party, their propensity to vote for its coalition partners should be unchanged. For instance, these voters may decide to transfer their votes from the scandal-hit party to an opposition party. Thus, the key here is that the scandal does not make the voters less likely to turn away from the coalition partner(s).
Expectations at the government level
We now examine the effect of scandal-based coalition termination on the way voters perceive the government’s performance and do so by considering the role of the perceived governing competence of the governing parties. We conceptualize governing competence as a form of valence, i.e., what Stokes (Reference Stokes and Dennis1992) defines 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 Dennis1992). As discussed, scandals affect parties’ perceived governing competence (Clark Reference Clark2009). Seen in this light, a government’s performance is inevitably tied to their governing competence. Now, since scandals destroy voters’ perception of the responsible party’s governing competence, the fact that the party co-governs with its partners may motivate voters to also perceive its governing partners as ‘guilty by association.’ Indeed, research on coalition policymaking has consistently shown that voters blur the policy positions of coalition parties (Fortunato and Stevenson Reference Fortunato and Stevenson2013; Klüver and Spoon Reference Klüver and Spoon2017; Spoon and Klüver Reference Spoon and Klüver2017). This implies that in a coalition, poor policy performance by one governing party may reflect badly on their governing partners. Via the same mechanism, then, a scandal may reduce the perceived general governing competence of all coalition parties. If so, this would motivate voters to evaluate the government’s performance less favorably. If so, then scandal-based coalition termination would lower voters’ evaluations of the government’s performance:
Tainted Government Performance Hypothesis (H2a): A scandal-based coalition termination lowers voters’ perception of the exiting government’s overall performance.
On the other hand, perhaps less intuitively, scandal-based coalition terminations, particularly those that are not the result of policy failure, may not necessarily affect voters’ perceptions of the government’s performance. If the coalition remains in place, voters may perceive these other coalition parties as willing to accept the scandal and value office spoils over ‘doing the right thing.’ This should harm these parties’ perceived governing competence. The perceived lack of governing competence, in turn, should taint voters’ evaluations of the government’s performance – despite the other coalition parties bearing no responsibility for the scandal. In contrast, if the scandal-free governing partners decide to exit government, they would convey to voters that they do not wish to continue governing with a party laden with scandal. This sends the signal of governing competence. Thus, even though a governing party and/or an individual minister is responsible for the scandal, the coalition partners’ decision to discontinue governing with the scandal-laden party tells voters that these parties retain a moral standard of good governance. As such, the act of termination may draw a clear line between the government’s performance and the scandal.
Moreover, if our H1a and H1b are valid, that is, if voters are able to distinguish between the governing competence of the scandal-laden party and its coalition partners, then they may also distinguish between the factors affecting a party’s governing competence and those that affect the government’s overall performance. That is, voters may be politically sophisticated enough to discern that the government’s policy performance is not necessarily tied to the scandal. And, since termination signals to voters that the governing partner(s) are unwilling to let the ‘guilty’ party taint the government’s image of integrity and effective governance – and willing to sacrifice office benefits to protect this image – voters may perceive the scandal as an act committed by a ‘bad apple’ and thus not alter their evaluations of the government’s pre-termination performance. In other words, dissolving the government preserves voters’ impressions of the government’s overall pre-scandal performance and prevents the scandal from tainting their evaluations of the performance.
This leads to the competing hypothesis:
Untainted Government Performance Hypothesis (H2b): A scandal-based coalition termination does not lower voters’ perception of the exiting government’s overall performance.
Expectations at the system level
Next, we establish the link between scandal-based government terminations and individuals’ satisfaction with how well democracy functions in their own countries (SWD for short). While our point of interest is on citizens’ SWD and not their diffuse or specific support for democracy, we take our theoretical inspiration from the argument by Easton (Reference Easton1975) on citizens’ diffuse support for democracy. Easton categorizes support for democracy as either diffuse or specific (Easton Reference Easton1975). Diffuse support refers to citizens’ evaluation of the polity itself, particularly how democratic processes represent their preferences and interests (Easton Reference Easton1975; Golder and Stramski Reference Golder and Stramski2010). This is different from specific support for democracy, which refers to citizens’ evaluations of democratic institutions as instruments for ensuring their well-being. A government’s performance output is tied to citizens’ specific support for democracy (Anderson and Guillory Reference Anderson and Guillory1997; Easton Reference Easton1975; Gilley Reference Gilley2006). The worse a government performs, the less satisfied citizens are with democracy as an instrument for delivering policy goods (Clarke, Dutt, and Kornberg Reference Clarke, Dutt and Kornberg1993). In other words, effective representation of citizens’ interests entails governments being able to safeguard citizens’ general welfare. In contrast, diffuse support for democracy is not dependent on the government’s performance output. That is, citizens who are dissatisfied with the government’s performance can nevertheless be satisfied with democracy if they believe that the democratic process can still represent their interests effectively. Likewise, citizens who are satisfied with the government’s performance can still be dissatisfied with the way democracy functions in their countries if they believe that the democratic processes are tainted, as demonstrated by Morgan-Jones and Loveless’s (Reference Morgan-Jones and Loveless2023) study of early elections and citizens’ satisfaction with democracy.
We argue that scandal-based government termination lowers citizens’ satisfaction with how well democracy is functioning in their countries. Scandals signal to voters that a certain political representative or a party does not adhere to the norms of democratic governance. Personal misconduct or an act of corruption means that a member of the government abused their position of power for their own enrichment, for their party to receive an unfair advantage, or as a means to satisfy their own needs, among others. These acts shatter voters’ trust in the government (Bowler and Karp Reference Bowler and Karp2004; Solé-Ollé and Sorribas-Navarro Reference Solé-Ollé and Sorribas-Navarro2018), which is important for their support for democratic institutions (Van der Meer and Hakhverdian Reference Van der Meer and Hakhverdian2017). Equally important, they signal to voters that the democratic institutions in their countries cannot prevent these acts. That is, their faith in the country’s democratic institutions as bulwarks against non-democratic behavior is weakened. In addition, since governments are expected to respond to citizens’ policy demands (Powell Reference Powell2004a,b; Soroka and Wlezien Reference Soroka and Wlezien2010), the fact that the scandal-based government termination interrupted governance means that the government is not responding to citizens’ demands. Thus, they should become less satisfied with the way democracy functions in their countries.Footnote 2
Dissolving the government post-scandal demonstrates to voters that the government as a whole is holding itself accountable. Although this reassures voters that politicians are being held accountable for their actions, we argue that scandal-based government terminations reduce voters’ satisfaction with the way democracy functions in their countries. For one, voters would still perceive democratic institutions as unable to prevent scandals despite the government taking action by dissolving. That is, even if the responsible party and/or its governing partner(s) redress the scandal by exiting the government, voters would still perceive that democratic norms have been violated by a member of government and/or a governing party. For another, the fact that governance needs to be interrupted because of the scandal signals to voters that democracy is not functioning optimally in their countries. Not only were the responsible party/member(s) of government able to circumvent democratic institutions and violate norms of good governance, but it also became necessary to recalibrate the composition of the government, either through negotiation with parties or by holding early elections.
Both have important consequences on how effectively voters’ interests and policy preferences are being represented. At a minimum, the need for a caretaker government, which cannot propose new policies, the time needed to organize new elections, or the time a new government needs to adjust to governing together would reduce both the efficiency and effectiveness in legislating the policies that voters would want. And, since voters do not have a say in the termination or what follows after termination (i.e., formation of a new government or holding early elections), the government’s decision to terminate effectively halts the process of legislating policies in a meaningful manner. This should be the case even if voters would likely prefer termination over having corrupted parties remain in government, and even if (in the case of early elections) they are able to vote for their preferred parties. As such, it is the fact that the termination is not the result of policy disagreement but rather the result of misbehavior by members/party in government that robs the legislature of the time devoted to the meaningful legislative policymaking process. This should reduce voters’ satisfaction with the way democracy functions in their countries, at least in the short termFootnote 3 :
Satisfaction with Democracy Hypothesis (H3): A scandal-based coalition termination lowers voters’ degree of satisfaction with the way democracy functions in their countries.
Empirical testing: the Ibiza scandal
We test our hypotheses by examining voters’ reactions after the ‘Ibiza Scandal,’ using panel data from the Austrian National Election Study (Aichholzer et al. Reference Aichholzer, Partheymüller, Wagner, Kritzinger, Plescia, Eberl, Meyer, Berk, Büttner, Boomgaarden and Müller2022). The scandal began with the publication of a video recorded in July 2017 by two German newspapers, the Süddeutsche and Der Spiegel. The video was published on 17 May 2019, almost two years after the recording. Two important politicians could be seen in this video: the vice chancellor and leader of the FPÖ, Heinz-Christian Strache, as well as Johann Gudenus (also FPÖ). They were talking to a woman who pretended to be the niece of a Russian oligarch, wanting to invest in Austria. Specifically, they were talking about her buying the widely circulated Austrian tabloid newspaper, the Kronen Zeitung. Additionally, there were talks about favorable reporting in the newspaper in exchange for public contracts if the FPÖ became a member of the government after the 2017 election. As such, the Ibiza scandal can be considered a political scandal (Perlot et al. Reference Perlot, Oberluggauer, Praprotnik, Kritzinger, Zeglovits, Glinitzer, Aichholzer and Wagner2021). This threatens a core value of liberal democracy – a free, independent press. If even part of the press is controlled by a national party, the objectivity of news reporting would come into question. Law-breaking activities from the party would have been covered up, and scandals would have been hidden.
When the video was released, conflict arose within the ÖVP-FPÖ government. The day after the scandal was published, Strache resigned, and the chancellor, Sebastian Kurz, announced early elections to be held in September of the same yearFootnote 4 . Over the next few days, calls for the resignation of the interior minister, Herbert Kickl, from the FPÖ finally led to the resignation of all FPÖ ministers by May 20th. This scandal did not only result in an early termination of the government but also had consequences for the performance in the upcoming federal election in September. While the ÖVP was able to gain around six percent in comparison to the 2017 election, the FPÖ suffered major losses: they ended up with only around 16 percent of the vote, a drop of almost ten percent compared to 2017 (Eberl, Huber, and Plescia Reference Eberl, Huber and Plescia2020).
The case we focus on can be considered a hard test of the hypotheses. The FPÖ’s status as a radical right party means that voters’ opinions of the party are polarized. Given their populist and anti-immigration attitude, there is a sizable number of voters who would refuse to vote for them under any circumstances, thus creating a substantial floor effect. Those who supported the party would have done so because of their anti-establishment stances and promotion of anti-immigration policies, which raises the threshold for these voters to choose another party. As such, we expect voters to be less likely to alter their support for or against the party. That is, compared to other parties, the FPÖ should be more immune to loss of support from scandals and resulting coalition terminations. Also, since a radical right party was in government, voters may perceive politics in a less positive light, particularly given their greater hostility toward radical right parties than others (Gidron, Adams, and Horne Reference Gidron, Adams and Horne2023; Reiljan Reference Reiljan2020). Thus, one can reasonably assume that for voters who do not support the FPÖ, the party’s entrance into government may prompt them to be less satisfied with the way democracy functions in their countries regardless of any scandals – more so than if another party that they do not support had entered government. Therefore, positive findings here would suggest that other scandal-based coalition terminations would result in the same attitudinal effects.
Unexpected events during survey design
We rely on Unexpected Events during Survey Design, henceforth UESD, and argue that whether our respondents were interviewed before or after the scandal-based coalition termination was pseudo-random. The ninth wave of the AUTNES (Aichholzer et al. Reference Aichholzer, Partheymüller, Wagner, Kritzinger, Plescia, Eberl, Meyer, Berk, Büttner, Boomgaarden and Müller2022) was in the field between May 6th and May 25th 2019Footnote 5 . Neither the survey researchers nor the participants had any way of anticipating the scandal that hit on May 17th and resulted in a scandal-based coalition termination on May 20th. Thus, we argue that the scandal represented a genuine external shock, and whether a particular respondent answered the survey before or after was as if it were random. We code all observations before May 17th as untreated and all observations after May 20th as treated. The 702 responses collected between the 17th and 20th are excluded from the sample since exposure to the scandal on these days is ambiguous.
There are several aspects that make UESD an attractive design. In contrast to traditional observational studies, the exogenous assignment to treatment increases the internal validity, but at the same time the fact that the stimulus is naturally occurring increases the external validity (Muñoz, Falcó-Gimeno, and Hernández Reference Muñoz, Falcó-Gimeno and Hernández2020). There are two key assumptions of the model: excludability and temporal ignorability, which we will discuss in turn.
First, excludability requires that any difference between respondents should be attributable only to the scandal-based coalition termination, or in other words, that the timing of the interview should affect the outcome only through this event (Muñoz, Falcó-Gimeno, and Hernández Reference Muñoz, Falcó-Gimeno and Hernández2020). The Ibiza scandal triggered a number of events of interest, most notably the government crisis and the breakdown of the coalition. This could be considered a collateral events, but this is a feature instead of a bug in our research design. We are interested in the joint effect of the scandal and the subsequent reactions, and not the scandal itself. Rather than fearing collateral events, we are more concerned that voters respond to only the scandal, while the termination had no effect. To seriously consider this possibility, we run a second supplementary study examining the effects of a similar scandal, which did not result in coalition termination.
Furthermore, the precise time that a corruption scandal breaks in the media is unexpected by almost all citizens, but of course not by the journalists involved (Muñoz, Falcó-Gimeno, and Hernández Reference Muñoz, Falcó-Gimeno and Hernández2020, 190–191). After all, they had been in possession of the video material for almost two years but decided to break the news at this particular point in time. This provides a challenge to the excludability assumption, especially if the journalist could be considered a politically motivated actor who strategically manipulated the time to maximize the damage to the incumbent. We doubt that the journalist had these motivations. In any case, to alleviate concerns related to excludability, we include graphs of how the dependent variable has developed over time. We estimate linear trends independently for the days before and after the Ibiza scandal. If there are pre-existing time trends, this would constitute a violation of the exclusion restriction (Schraff Reference Schraff2021), but this is evidently not the case. Following best practices, we also inspect pre-existing time trends by testing placebo treatments. We divide the control group by its empirical median and test for the absence of an effect at that point (Muñoz, Falcó-Gimeno, and Hernández Reference Muñoz, Falcó-Gimeno and Hernández2020, 197).
Secondly, the assumption of temporal ignorability requires that selection of the timing of the interview should be as good as random. Treatment assignment is determined by decisions related to data collection (Muñoz, Falcó-Gimeno, and Hernández Reference Muñoz, Falcó-Gimeno and Hernández2020). In this case, the fieldwork was organized based on quota sampling to create a representative sample. This can create potential imbalances, as the respondents interviewed in the later stages of the field work are mostly those who match an incomplete quota. We check that by running extensive balance checks. Usually, in UESD, all variables are measured post-treatment, but because AUTNES is a long-running panel study, we can test the balance of our study on a wide range of background variables.
Table 1 shows the mean or proportion of 30 variables before and after the scandal broke, the difference, and the p-value for a two-tailed t-test of differences. We find no significant differences between the control and the treatment group on key socio-demographic background variables, including gender, age, education, region, and employment status. Next, we look towards political variables measured in the previous wave. Here, we find no significant difference in political interest and left-right self-placement. Because we have access to these pre-treatment variables, we can also explore heterogeneous effects of the scandal-based termination. These are included online in Appendix C (Tables C1 and C2). Overall, we find very little evidence of heterogeneous effects: only the interaction between left-right self-placement and PTV for FPÖ reaches significance at the 0.05 level. We discuss this finding in Section Propensity to Vote.
Balance check on background variables and dependent variables measured a month prior (wave 8). Absolute numbers in parentheses. Data: AUTNES
***p < 0.01; **p < 0.05; *p < 0.1. Two-tailed test.
In addition, Table 1 shows no difference in the proportion of voters for most parties. The only exception we find is that there were significantly more Neos voters in the control group than in the treatment group. Finally, we look at the lagged version of our four dependent variables and find that the control group was significantly more satisfied with democracy than the treatment group even before the scandal. That two out of thirty variables display significant differences could easily be a product of the stochastic process – we are operating with a significance level of 0.05. Nevertheless, we follow best practices (Muñoz, Falcó-Gimeno, and Hernández Reference Muñoz, Falcó-Gimeno and Hernández2020) and include these two covariates as controls. In this way, we can rely on a more plausible assumption of conditional ignorability.
Another key assumption, under the heading of ignorability, is compliance. We assume that all respondents interviewed after the event were exposed to the treatment, i.e., aware of the event. To test this, we include two different pseudo-manipulation checks in the results section. First, we test the effect of the scandal on the question: ‘When you think about the coalition between the ÖVP and the FPÖ: How united or divided do you think it is internally?’ Answers were given on a scale from ‘completely united’ (0) to ‘completely divided’ (10). Since the scandal caused intense strife within the government, we ought to see a clear effect here. Secondly, respondents were asked to rank the three most important issues in the country, and corruption was one of the answer options. We code this into a 4-point variable: (0) if corruption was not mentioned and (3) if it was given the highest priority.
Variables and measurement
We rely on our main independent variable, exposure to the scandal-based coalition termination, to predict a number of dependent variables. First, we predicted the propensity to vote (PTV) for the two members of the terminated coalition. The variables were measured with the survey question: ‘How likely is it that you would ever vote for each of the following parties?’ and answers were given on a scale from ‘highly unlikely’ (0) to ‘very likely’ (10). Secondly, we look at the evaluations of the incumbent government. This was measured with the question: ‘Concerning the performance to date of the current federal government (ÖVP-FPÖ coalition), are you…?’. Finally, we consider individuals’ satisfaction with democracy (SWD), which is measured with the survey question: ‘On the whole, how satisfied are you with the way democracy works in Austria?’. For both questions, answers were provided on a scale from ‘very dissatisfied’ (1) to ‘very satisfied’ (4). Although Easton (Reference Easton1975) distinguishes between diffuse and specific support for democracy, Powell (Reference Powell2004b) argues that it is difficult to disentangle these two aspects of satisfaction with democracy; in addition, democracy may mean different things to different individuals even within a country. As such, it should be acceptable to conceptualize satisfaction with democracy as an amalgamation of both diffuse and specific support for democracy. In Appendix F, we test whether our results for the Kurz corruption probe are robust to alternative ways of measuring satisfaction with democracy.
Some of our dependent variables are prone to floor and ceiling effects. For instance, some respondents might be very opposed to the FPÖ from the outset and would answer that they would never vote for the party even before the scandal. In that case, we would not expect to find any additional negative effect of the treatment. Following Singh, Karakoç, and Blais (Reference Singh, Karakoç and Blais2012), we can explore the floor and ceiling effects by controlling for the pre-scandal (lagged) variable. For this purpose, we use the previous wave 8 fielded less than a month before, between April 12th and April 28th, 2019.
Results
Manipulation check
Figure 1 plots answers to our pseudo-manipulation check at different points in time. We add linear regression lines, estimated separately based on the control group, the treatment group, and the intermediate group, to illustrate the time trends. The intermediate group is highlighted in red and should be disregarded in the visual analysis of the plot. We clearly see that there is no trend before the scandal broke. This is followed by a much higher level of perceived division in the treatment group, but it appears to decay rather quickly – something that has also been observed in similar scandals (Ares and Hernández Reference Ares and Hernández2017).
Pseudo-manipulation check: Perceived division within government plotted against time of interview. Trend lines estimated for control, treatment, and intermediate period separately. Data: AUTNES.
Next, we include (1) the perceived division between the coalition members and (2) the ranking of corruption as the most important problem, as the two dependent variables in an OLS regression. We controlled for having voted Neos and lagged satisfaction with democracy (wave 8) because these two variables did not pass our balance check. Furthermore, we include lagged perceived division (wave 7) and lagged corruption ranking (wave 8) and their interaction with the scandal treatment. The results are available in Table 2. As expected, there are strong significant treatment effects.
Pseudo-manipulation check
***p < 0.01; **p < 0.05; *p < 0.1. OLS regression including political and demographic weights.
Respondents in the treatment group on average perceived the coalition members as 4 units more divided on an 11-point scale. This is a very strong effect, which increases even more if we consider the strong interaction. The more unified a voter perceived the coalition to be a month earlier, the stronger the effect of the scandal. Similarly, we find a strong effect on how highly corruption was ranked among the most important problems. However, there is no significant interaction effect in model 4.
Propensity to vote
Moving on to the dependent variables of more theoretical interest, we start by graphically examining the time trends. Figures 2 and 3 clearly show that there was no trend before May 17th. Both regression lines are virtually flat. For FPÖ in Figure 2, the propensity to vote has decreased after the crisis, but this appears to recover quite quickly in the treatment group. For ÖVP in Figure 3, there is no clear development in propensity to vote in either of the periods.
Propensity to vote for FPÖ plotted against time of interview. Trend lines estimated for control, treatment, and intermediate period separately. Data: AUTNES.
Propensity to vote for ÖVP plotted against time of interview. Trend lines estimated for control, treatment, and intermediate period separately. Data: AUTNES.
In line with hypothesis 1, we find a strong negative effect of the scandal-based coalition termination on the propensity to vote (PTV) for FPÖ in model 5, Table 3. Respondents who answered the survey after the scandal broke indicated that they were less likely to vote for the party. It is worthwhile to recall that all respondents were surveyed within a time frame of nineteen days. The short time frame suggests that the decline in support for FPÖ that can be attributed to either reversion to the mean or the cost of governance should be very similar. Thus, it is not unreasonable to conclude that the average difference between those surveyed before the government breakdown and those surveyed after is due to dissolution and not merely due to the cost of governance.
Propensity to vote for responsible party (FPÖ) and coalition partner (ÖVP). Data: AUTNES
***p < 0.01; **p < 0.05; *p < 0.1. OLS regression including political and demographic weights.
We include an interaction with the lagged propensity to vote FPÖ (wave 8) and, as expected, find a significant negative interaction effect in model 6, Table 3. For voters who already strongly disliked FPÖ, the scandal had no significant effect. In other words, there is a significant floor effect. We find further evidence of this floor effect when we interact exposure to the scandal with left-right self-placement. The results are available in Appendix C, Table C1, model A11. For FPÖ, there is a strong significant negative interaction effect, while the coefficient for scandal exposure is positive and insignificant. This means that left-wing voters had a very low propensity to vote FPÖ regardless of the scandal – and thus the treatment has no effect. However, for right-wing voters there is a strong decline in their PTV.Footnote 6
This pattern does not extend to the coalition partner ÖVP. For this party, we see no significant effect on the change in PTV. This clearly indicates that voters were able to attribute blame to the responsible party, while the senior member of the coalition was not held accountable. This is also mirrored in the actual election results in September of 2019, in which the FPÖ lost around 10 per cent of their votes compared to the 2017 election (Eberl, Huber, and Plescia Reference Eberl, Huber and Plescia2020).
The marginal effects are illustrated in Figures 4 and 5. Once again, they reveal that the higher the PTV was for FPÖ in the previous wave, the more negative the treatment effect is. In contrast, the treatment effect is statistically significant for individuals’ PTV for ÖVP, regardless of their PTV for the party in the previous wave.
Marginal effect of scandal-based coalition termination by lagged PTV for ÖVP.
Marginal effect of scandal-based coalition termination by lagged PTV for FPÖ.
Government performance and SWD
Table 4 shows similar regression models with the perceived performance of the government and satisfaction with democracy as the dependent variables. In line with hypothesis 2, we find no significant effect of scandal-based coalition termination of respondents’ perception of the government’s overall performance. There is neither any significant direct effect nor any interaction effect. In line with hypothesis 3, we find that individuals who experienced a scandal-based coalition termination had a significantly bigger decrease in SWD compared to individuals who did not have this experience. Being exposed to the treatment is associated with a 0.1 unit decrease in SWD. In model 8, we also find a very strong significant interaction effect between the lagged dependent variable and the treatment indicator. Following Singh, Karakoç, and Blais (Reference Singh, Karakoç and Blais2012), we interpret this as a sign of a ceiling effect. Respondents who are already very satisfied (4) can only lower or maintain their satisfaction between the two periods. Marginal effect plots illustrating the moderating effects of the lagged dependent variable can be found in Appendix B, Figures B1 and B2.
Perceived government performance and satisfaction with democracy. Data: AUTNES
***p < 0.01; **p < 0.05; *p < 0.1. OLS regression including political and demographic weights.
Tables A1 and A2 from Appendix A display the results from placebo tests, where scandal exposure is replaced with the empirical median of the control group. We find no significant effect. Similarly, both Figures 6 and 7 show a horizontal trendline in the pre-treatment period, indicating that satisfaction with the government performance and with democracy in general was stable until the scandal broke. Thus, we are not worried that we are picking up pre-existing time trends.
Government performance plotted against time of interview. Trend lines estimated for control, treatment, and intermediate period separately. Data: AUTNES.
Satisfaction with democracy plotted against time of interview. Trend lines estimated for control, treatment, and intermediate period separately. Data: AUTNES.
Supplementary analysis: Kurz’ corruption probe
As discussed above, our treatment in the analysis above consists of collateral events: first the Ibiza scandal breaks, next the coalition government is thrown into a crisis, and finally there is a scandal-based coalition termination. This is a necessary feature of both our theory and research design, but it does imply that we cannot fully disentangle which part of the effect can be attributed to the corruption scandal alone and which part is due to the scandal-based termination. Because these terminations are immediately preceded by scandals, voters may focus on the scandal itself and punish the responsible party, while largely ignoring the fact that the coalition terminated as a result. If so, the consequences of a scandal-based termination on voters’ perceptions of politics, the government, and coalition parties would be the same as the consequences of a scandal that did not result in termination. Since in these cases the scandal is committed by government members, voters may lose confidence in the state of democratic governance in their countries and thus become less satisfied with their country’s democratic functioning. Or, voters may incorporate both the scandal and the resulting termination in their evaluations of the government’s overall performance and the way democracy functions in their countries.
To improve the plausibility of our hypotheses, we conduct additional analyses, using a recent corruption scandal and government crisis in Austria. On 6 October 2021, government agents raided the Federal Chancellery and the headquarters of the ÖVP as part of a corruption probe targeting Chancellor Sebastian Kurz and several of the ÖVP party members in his inner circle. The investigations are still ongoing, but prosecutors suspect that individuals close to Kurz had used public funds from the Finance Ministry to bribe several Austrian news outlets, including Krone, Österreich, and Heute, to bring fabricated public opinion polls along with favorable reporting. The polls essentially served the purpose of propaganda, helping Kurz in his bid to replace Reinhold Mitterlehner as the party leader of ÖVP in 2017. Kurz allegedly later rewarded his allies with prestigious political posts. The two scandals are connected only because confiscation of hard drives after the Ibiza scandal allowed investigators to access incriminating text exchanges from an old cell phone backup.
While opposition parties were trying to gather support for a motion of no confidence, the junior coalition member during this period, the Greens, wanted to continue the coalition if Kurz personally and voluntarily resigned his office. Following the raid, opposition parties unanimously demanded Kurz’ resignation and called a special session of the Parliament to vote on a motion of no confidence. The Greens pondered supporting the motion if Kurz were unwilling to voluntarily step down but were also supportive of continuing the coalition cabinet if Kurz were replaced. On 9 October 2021, Kurz resigned the chancellorship, and on 11 October 2021, he was officially removed from office by the president, who simultaneously appointed his successor, also from ÖVP.
This case offers the opportunity to examine how voters react to scandals in cases where the coalition partner effectively signals support for the scandal-clad party by continuing formal cooperation. If we find that the scandal reduced vote intention for the Greens, this would suggest that the perceived reduction in governing competence, caused by the scandal, has spilled over to the junior partner. If the Kurz corruption scandal is associated with less satisfaction with democracy, the implication is that scandals lower voters’ satisfaction with democracy even if they do not cause the government to dissolve. Last but not least, if the scandal resulted in lower ratings of the government’s performance, then this would suggest that since the coalition partner did not break from the scandal-clad party, voters would perceive the government as ‘business as usual’ and cast the corruption shadow over it, motivating them to evaluate the whole government’s performance less favorably.
The AUTNES was not in the field during this period, but the European Social Survey (ESS) was. Unlike the AUTNES, the ESS is not a panel study, which means that all variables are measured post-treatment for the treatment group. Thus, it is especially important to choose covariates that are not subject to post-treatment bias (Muñoz, Falcó-Gimeno, and Hernández Reference Muñoz, Falcó-Gimeno and Hernández2020). In Table 5, we include the same background demographics as above as well as vote recall. The latter is somewhat problematic as people might have a hard time recalling correctly and it might be subject to social desirability bias. However, we find that the two groups are balanced in terms of party supporters with the exception of Green voters. This is perfectly in line with the overrepresentation of younger, urban, and highly educated voters in the control group. Overall, the two groups in the ESS sample are not as well-balanced as those in the AUTNES. This means that we include many more control variables in our analysis below. Furthermore, there was no PTV question in the ESS, so we use a dichotomous measure of whether a respondent claimed to identify with the party. This is a more conservative estimate. We use it as the dependent variable in a linear probability model.
Balance check on background variables. Absolute numbers in parentheses. Data: EES
***p < 0.01; **p < 0.05; *p < 0.1. Two-tailed test.
The results are detailed in Table 6 . We show a graph of the dependent variable plotted against the time of the interview, with separate trend lines for the control, intermediate, and treatment groups, in the online appendix. The results show significant effects on the performance of the government and SWD. In addition, their results reveal that voters polled after the scandal are less satisfied with democracy and that there are significantly fewer respondents who identify with the ÖVP after the scandal, but there is no effect on the Greens. In particular, the findings suggest that voters are politically sophisticated enough to pinpoint the party responsible for the scandal and direct their reactions accordingly.
Effects of scandal-based coalition termination on perceived government performance, satisfaction with democracy, and identification with the responsible party (ÖVP) and coalition partner (Greens). Data: ESS
***p < 0.01; **p < 0.05; *p < 0.1. OLS regression including poststratification weights.
Unlike the AUTNES, the ESS included a number of other outcome variables that are relevant to our project. This allows us to conduct a series of robustness tests, available in Appendix F, and show that our results cannot be explained by specific question wording but do in fact convey important and substantial changes in public opinion. We find that being exposed to the Kurz corruption probe had a significant negative impact on trust in parliament, trust in parties, and trust in politicians. While these questions do not distinguish between government and opposition, they do capture the evaluations of the decision-makers in a similar way. In addition, we find a negative effect on answers to the question ‘To what extent you think governing parties in Austria are punished in elections when they have done a bad job’ (specific support) but no effect on answers to the question, ‘How important do you think it is for democracy in general that governing parties are punished in elections when they have done a bad job?’ (diffuse support). Finally, there is also a small negative effect on the question, ‘How important is it for you to live in a country that is governed democratically’ which could perhaps reveal that respondents think the country that they do in fact live in is now slightly less democratically governed.
In sum, in terms of voter reactions, the main difference between the scandal-based coalition termination in 2019 and the corruption scandal in 2021 is how it affected the evaluation of government performance. Although the junior partner is not held responsible, the overall evaluation of the government suffers from the scandal. Replacing the individual at fault, in this case the chancellor, is seemingly not considered a sufficient response. Since formal cooperation between the two parties continued, the implication is that the perception of a corrupt government continued to remain, which damaged voters’ perceptions of the government’s overall performance. This finding calls for further research to investigate the mechanisms of how scandals and scandal-based government terminations affect voters’ perceptions of the government’s performance.
Discussion
The main goal of this paper was to test how scandal-based coalition termination influences individuals’ propensity to vote for the parties involved and their evaluation of government performance, as well as their degrees of satisfaction with democracy. There is a plethora of works on governing parties’ electoral considerations and government terminations. One focus has been on early elections as a strategic tool that incumbent parties employ to maximize their electoral performances (Schleiter and Tavits Reference Schleiter and Tavits2016; Smith Reference Smith2003). These findings both highlight the ability of incumbent parties to take advantage of early government terminations to reap electoral benefits (or avoid electoral punishment). Yet, we know less about how voters react to terminations that are essentially the result of scandals that threaten the integrity of the government. Unlike strategic government terminations, the governing parties essentially have their hands tied, as the primary motivation for termination is not to boost their electoral fortunes but rather to contain the damage incurred by the scandal. How voters perceive the scandal-based coalition terminations, which may be different from the effects of scandals in and of itself, remains underexplored. To investigate this, we relied on a natural experimental setting where a scandal regarding one of the government parties in Austria was published while the AUTNES (Aichholzer et al. Reference Aichholzer, Partheymüller, Wagner, Kritzinger, Plescia, Eberl, Meyer, Berk, Büttner, Boomgaarden and Müller2022) was in the field. This creates the ideal conditions for an Unexpected Event during Survey Design. Our placebo tests show that there were no pre-existing time trends, while our extensive balance checks confirm that exposure to the scandal-based coalition termination was as if random, and our pseudo-manipulation check reveals that knowledge about the scandal and the ensuing intra-coalition conflict was widespread.
We find support for our hypotheses: the propensity to vote for the FPÖ decreased after being treated with the news of the scandal, and the overall satisfaction with democracy also dropped after the scandal was published. However, there was no significant effect on the propensity to vote for the senior member of the coalition, the ÖVP, or on the overall evaluations of the government’s performance. We contrast these results with a similar supplementary analysis of a high-level corruption scandal, which did not lead to a coalition termination. In this case, evaluations of the government’s performance deteriorated after the scandal, while individuals became less satisfied with the way democracy functions in Austria and are less inclined to vote for the offending party, the ÖVP. To our knowledge, there are no other corruption cases in which the offending party was the PM party and resulted in coalition termination while also offering the opportunity to employ UESD on existing survey data. Thus, we cannot rule out the possibility that the less favorable evaluations of the government’s performance are due to the offending party’s PM status. In any case, the results do not rule out the possibility that voters may react differently to a government scandal followed by continued governance versus a scandal-based government termination. They highlight the need for deeper investigations into how the PM status of the offending party may moderate or mediate voters’ perceptions and the attitudinal consequences of scandal-based terminations.
This paper complements the work by So (Reference So2024), amongst others, and further suggests that different types of terminations have different consequences for the parties involved. It adds quasi-experimental evidence to the field of research. In the future, it might be worthwhile to run survey experiments to tease out the underlying mechanisms even better. Our results reveal that even though the scandal itself is an exogenous shock, the hands of the government are rarely truly tied. What the consequences of the scandal should be, whether a single minister, a coalition member, or the entire government should fall, is ultimately at the discretion of the prime minister. In the case of the Ibiza scandal, it appears that the ÖVP used the opportunity to throw their coalition partner under the bus. It lowered the PTV for the FPÖ but had no impact on the PTV or the perceived government competence of the ÖVP itself. Since coalition members are often each other’s most proximate competitors, the downfall of one might even benefit the other; this is undoubtedly a contributing factor to the electoral success of the ÖVP in the election in October that same year (Eberl, Huber, and Plescia Reference Eberl, Huber and Plescia2020; Perlot et al. Reference Perlot, Oberluggauer, Praprotnik, Kritzinger, Zeglovits, Glinitzer, Aichholzer and Wagner2021).
This has a number of important implications. Our findings suggest that scandal-based coalition terminations do affect voters’ attitudes and that this has important consequences for how voters perceive the functioning of the democratic system as such. The negative effect of the scandal-based termination on respondents’ satisfaction with the way democracy functions in Austria is particularly relevant. While we expect and find that scandals without coalition terminations also exert similar effects, our finding raises the question of whether and how much dissolving the government post-scandal attenuates the negative impact of scandals on citizens’ trust in democratic institutions. It also inspires questions on whether and how non-scandal-based government terminations, such as those arising from policy conflicts and parliamentary defeats, affect voters’ perceptions of the way democracy functions in their countries. Thus, our results highlight the need for a revival of a long line of research on government breakdowns. In particular, more research on the consequences of government terminations is needed in order to arrive at more generalizable conclusions about the quality of democratic representation in the coalition government setting.
Additionally, the findings regarding government performance and propensity to vote suggest that voters pay attention to politics and possess enough political knowledge to correctly attribute responsibility to individual parties. In other words, voters’ difficulty in differentiating the policy positions of coalition parties does not imply that they are unaware of salient political events. Rather, the deficit in knowledge of coalition parties’ policy positions may stem from the nature of coalition policymaking. Both implications hint at the need to study the relationship between coalitions’ policymaking process, voters’ understanding of these processes, and their evaluations of the democratic functioning of coalition governance.
Last but not least, our scope condition for scandal-based coalition terminations includes only scandals that have been publicized, in which voters are aware whether the scandal is contained within a single governing party, multiple governing parties, or all governing parties. A natural question to ask is, how are scandals portrayed in public in the first place? Although we are agnostic about this question, how the media and politicians frame each scandal no doubt affects voters’ perceptions of the scandal’s severity as well as how each scandal affects the integrity of the government. While it is outside the scope of our paper, our results highlight the importance of examining the effect of media framing of scandals on the probability of coalition termination, as well as on how the scandal affects voters’ perceptions of governing parties’ competence and their satisfaction with the way democracy functions in their countries. It would be beneficial for future research on coalition politics to examine how the media portrays critical events within the coalition, i.e., scandals and policy disagreements within and between governing parties, and how these portrayals in turn affect the durability of coalitions as well as voters’ attribution of blame for these events.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/S1475676526101194.
Data availability statement
The data can be found under the Austrian Social Science Data Archive (AUSSDA).
Acknowledgments
The authors are indebted to Svenja Krauss, who provided invaluable knowledge on the ‘Ibiza scandal’ and insightful feedback on the theory and empirical components of the paper. The authors are also grateful for the constructive comments and feedback from the organizers and participants at the Workshop on Party Politics at the Department of Political Science, Aarhus University (Oct. 2024), as well as the Workshop on the Political Psychology of Party Credibility, organized by Radboud University Nijmegen and Montreal University (Jul. 2024).
Funding statement
Ida B. Hjermitslev’s working time on the manuscript was funded by the Carlsberg Foundation: CF23-0407. Florence So’s working time on the manuscript was funded by the Marie Curie Individual Fellowship, grant number 892779.
Competing interests
The authors declare no competing interests.
Open access
