Political scandals – where political actors are perceived to have engaged in norm-breaking and/or unethical behaviour – can have a corrosive effect on political trust and support (Entman Reference Entman2012; Faris et al. Reference Faris, Roberts, Etling, Bourassa, Zuckerman and Benkler2017; Sikorski, et al. 2018). The most consistent and unsurprising finding across contexts is that political scandals damage the approval ratings of individual politicians who are the subject of them (Lanoue and Headrick Reference Lanoue and Headrick1994; Sikorski Reference Sikorski, Knoll and Matthes2018). More concerning, perhaps, is evidence of contagion: when the negative effects of scandals spill over from the original subject and citizens tar a wider pool of politicians and institutions with the same brush. A recent meta-analysis reveals plentiful evidence of such contagion (Sikorski Reference Sikorski, Knoll and Matthes2018). Scandals involving a single politician have affected citizens’ trust towards the government more broadly (Lee Reference Lee2018), the party of the scandal’s subject (Maier 2011; Leigh and Vulliamy Reference Leigh and Vulliamy1997; Sikorski, et al. 2020), and the wider political class of politicians (Bowler and Karp Reference Bowler and Karp2004; Pattie and Johnston 2012; Puente-Diaz 2015). Since trust – a bellwether for broader political support and legitimacy (Marien and Hooghe Reference Maier2011; Krupenkin Reference Krupenkin2021) – is generally fairly stable over time (Devine and Valgarðsson Reference Devine and Valgarðsson2023), the damage inflicted on it by a landmark scandal can be long-lasting. Understanding the nature and scope of contagion is therefore crucial to democratic health.
While the phenomenon of scandal contagion has been extensively studied horizontally – examining how scandals affect trust in political actors and institutions at the same level of government – the potential for vertical contagion across different levels in multilevel polities remains unexamined. This gap in the literature raises important questions about whether scandals can affect trust in government levels that differ in partisan composition and legislative competence from the scandal’s origin. This possibility is of particular concern for democratic health, with scandal contamination leaving citizens without a safe harbour of trusted governance at any level and potentially increasing susceptibility to anti-democratic sentiments.
In this article we provide a first empirical test of vertical scandal contagion using two survey experiments in Scotland, and make three substantial contributions to the literature. First, we provide a multilevel perspective on previous findings that typically focus on a single political echelon, allowing us to discern how the effects of scandals propagate across, upward, and downward through political hierarchies. Second, we provide a more nuanced understanding of contagion effects by considering the hierarchical nature of political systems. Substate governmental levels, often conceptualized as second-order arenas (Reif and Schmitt 1980; Schmitt and Teperoglou 2017), may be perceived as less consequential than state-level politics, potentially leading to differential propagation of scandal effects based on their point of origin. Third, we examine how individual-level factors, particularly alignment with salient identity in-groups and out-groups at various territorial scales (Turner et al. 1987; Simon and Klandermans 2001), influence reactions to scandals across governmental levels. Our findings contribute to broader discussions on political trust, democratic resilience, and the complexities of public opinion formation in federated or decentralized political contexts.
Existing Research
We have already noted the shortage of empirical evidence on the question posed in our study. However, past research – both theoretical and empirical – does provide guidance about when and how we might expect scandals to spill over to other levels of government. First, at a basic cognitive psychological level, Schwarz and Bless’s (1992) inclusion/exclusion model posits that spillover or contagion effects take place when incoming information seems applicable to both the subject and the wider category to which they belong. So in the context of political scandals, the critical question becomes: do voters view a scandal as indicative of broader issues within the category of politicians and, if they do, how far does that category stretch? More specifically for our study: does this perception of broader implications extend vertically across different levels of government?
On the one hand, there is plentiful general evidence of interdependencies in multilevel systems (Schmitt and van der Eijk 2008; Muñoz 2017). Looking specifically at trust, citizens’ ratings across different levels of government tend to be clearly positively correlated (Muñoz 2017), suggesting a general assessment of politics and politicians as a category that applies at multiple levels. If a scandal reshapes that judgement, the effect is likely to filter up or down the system. Even if a scandal is tied to a specific party rather than politicians in general, there is still scope for that filtering insofar as that party operates across levels. Some circumstantial evidence suggests this possibility: in the decade following the Watergate scandal, trust in all levels of US government declined, but the federal government suffered a good deal more than the state and local levels (Zimmer 1979; Jennings Reference Jennings1998). These arguments indicate scope for contagion. On the other hand, different levels of government are often prominent and distinct, and citizens are well able to maintain separate evaluations of them, including trust (Anderson Reference Anderson2006; Fitzgerald and Wolak Reference Fitzgerald and Wolak2016; Elsas et al. Reference Van Elsas, Brosius, Marquart and De Vreese2020; Denemark and Sharman Reference Denemark and Sharman1994; Henderson Reference Henderson2010). This points more to containment than to contagion, where effects on evaluations are limited to the specific level in question.
Given these conflicting predictions, it is more useful to classify the circumstances under which contagion is more or less likely. One inhibitor of contagion is the distinctness or remoteness of a certain level, such as the European Union compared to national institutions (Desmet, et al. Reference Desmet, van Spanje and de Vreese2012; Harteveld, Meer and Vries Reference Harteveld, van der Meer and De Vries2013; Elsas et al. Reference Van Elsas, Brosius, Marquart and De Vreese2020). Another situation in which contagion seems less likely is between two levels which each have considerable power and importance, such as federal and substate governments in systems like Canada, Germany, or our own case of Scotland within the UK. In sum, the more distinct and salient the other arena(s) to which a scandal might spill over, the more likely that its politicians will feel as if they belong in a separate category – such as ‘EU bureaucrats’ or ‘those politicians down in London’ – and the more likely is containment rather than contagion.
Those examples hint at another limit on vertical contagion. Trust judgements are likely to be coloured by identity or affective considerations: that is, strong feelings of belonging to a given level, in some cases accompanied by hostility towards a ‘rival’ level (Muñoz 2017; Berg and Hjerm Reference Berg and Hjerm2010). It is hard to imagine, for example, that the Gürtel scandal, which ravaged Spain’s governing People’s Party (PP) in Madrid, did anything to erode trust in politics at the Catalan level. Those on either side of the constitutional divide are likely to have Spanish-level and Catalan-level institutions in entirely separate categories. This is a particularly febrile example but there are many instances in which citizens have at least mild preferences for or against the subnational, national, or supranational level and will either reject or embrace contagion accordingly. This argument runs parallel to the familiar point that reactions to scandals are also driven by partisan identities and preferences (Fischle Reference Fischle2000; Sikorski, et al. 2020). Indeed, the two converge in those cases – like the Gürtel scandal and the cases studied here – where the partisan and territorial or ‘cross-level’ divides overlap. Nationalist sympathizers in the Catalan electorate had still more reason to dismiss any contagion from a scandal in Madrid given that the party engulfed was the quintessentially Spanish nationalist PP.
In cases where relations between levels are rivalrous and overlaid with partisan animus – as is the case in Scotland – not only is containment far more likely than contagion but there is even a third possibility: a contrast effect. Far from belonging to the same generic category, politicians across the two rival levels are liable to be seen as belonging to opposing categories, and a scandal casting one level in a bad light could make the other appear more favourably by comparison. Certainly, those hostile to the scandal-affected level have a strong incentive both to link the scandal to that level as a whole and to suggest, implicitly or more likely explicitly, that the same thing would not happen at their preferred level (Taber and Lodge 2006; Redlawsk, et al. 2010). But such a contrast effect may also occur among those without strong prior commitments to criticise one level and praise another. Provided that they see the two levels as alternatives or rivals, they may interpret a scandal in one as reason to place more trust in politicians at the other.
Based on these considerations, we propose three competing hypotheses when examining the vertical effects of political scandals:Footnote 1
-
Contagion: A scandal will negatively affect attitudes towards multiple levels of governmentFootnote 2
-
Containment: A scandal will negatively affect attitudes towards the level implicated, but not towards other levels
-
Contrast: A scandal will negatively affect attitudes towards the level implicated, but positively affect attitudes towards another level
Cases
We test these hypotheses in two survey experiments, each fielded in Scotland: a historic nation within the UK with its own parliament and government with substantial fiscal powers and responsibility over major policy areas. These ‘devolved’ institutions are familiar and prominent features on the political landscape, meaning that most citizens make a clear distinction between this Scottish level and the UK or Westminster level (Johns Reference Johns2011; León and Orriols Reference León and Orriols2019). Indeed, for many, the substate Scottish level of government is the primary political arena (Henderson Reference Henderson2022). The two levels are divided politically, too. This is partly because the independence referendum in 2014 hugely intensified debate about how power should be divided between the levels. It is also because, for more than a decade, the two levels were governed by the two parties – the devolution-sceptic pro-union Conservatives at Westminster and the secessionist Scottish National Party (SNP) at Holyrood (the seat of the devolved Parliament of Scotland) – representing the poles of that debate. This combination has been shown to generate the most partisan bias in evaluations of the different levels of government (Rico and Liñeira 2018). All in all, then, the Scottish context appears to represent a ‘hard’ test for vertical contagion. A scandal at either level seems quite likely to be attributed to that level rather than to politics more broadly, either because the two levels are clearly distinct in the public minds (like McDonald’s and Domino’s) or because citizens see the levels as competitors (like McDonald’s and Burger King) and want to assign blame in line with their preferences among the levels.
Our experimental treatments prime respondents to think about two real-world political scandals that occurred in the UK and Scotland between 2021 and 2023. Study 1 focuses on the Westminster-based ’Partygate scandal’. This concerned allegations that the UK Prime Minister Boris Johnson and various other senior officials had broken COVID-19 lockdown restrictions by hosting parties at Downing Street throughout 2020 and 2021. The information became public in January 2022, with civil service and police investigations taking place between January and May of 2022, and very likely contributed to Johnson’s resignation later that year. Study 2 focuses on a substate political scandal: the 2023 SNP finance scandal. On 15 February 2023, Nicola Sturgeon resigned as First Minister of Scotland. While there had been an ongoing investigation into the SNP’s finances, Sturgeon cited personal reasons for her resignation. However, in April 2023 the investigation led to the arrest,Footnote 3 questioning, and release of senior party officials, including the party CEO, who is the First Minister’s husband, and the party’s treasurer, as well as a high-profile searches of the First Minister’s residence and party headquarters. It also led to the seizure of a campervan at the home of the First Minister’s mother-in-law, earning the ordeal the label ‘Campervangate’ from various sources. Both scandals generated intensive coverage across the media for several months, a point illustrated in Figure 1 for the four largest UK-wide broadsheet newspapers but one no less true of the Scottish-based media – indeed, the latter gave even more attention to Campervangate.Footnote 4
Number of newspaper articles per month that included ‘Partygate’ or ‘SNP finance probe’ in the four largest UK broadsheets (The Times, The Daily Telegraph, The Guardian, and the Financial Times).
Research Design
Both studies were carried out as part of the Scottish Opinion (SCOOP) Monitor and were administered online by YouGov. YouGov uses a sampling frame to approximate the demographic composition of the Scottish population, and provides post-stratification weights so that model estimates can be interpreted as nationally representative.Footnote 5
Experimental Design
The studies are identical in design, using a simple split sample vignette. Our samples were randomly divided into two groups: control and treatment. Both groups were asked whether they trusted various political actors to behave ethically. Prior to being asked this question, the treatment groups were shown a piece of text about a recent political scandal that had occurred at either the statewide or substate level. The treatments are outlined in Table 1.
Description of experiment treatments
Outcome Measures
We estimate the effect of our treatments on a battery of survey items, each measured using a seven-point Likert scale: ‘To what extent do you trust the following groups to behave ethically in their job? (‘Not at all’ to ‘To a very large extent’). Given that scandals are generally understood to involve a behaviour that is unethical (Dziuda and Howell Reference Dziuda and Howell2021), this outcome gives a very direct test of the effects of scandal contagion across levels – more direct than would be obtained by asking about, say, trust in general. The distribution of control group responses (see Supplementary Materials) demonstrates considerable variation in responses to different categories. The base levels of trust are considerably different, too, with all UK-level actors substantially less trusted to behave ethically than their Scottish counterparts.
Empirical Strategy
For main effects, we present difference in group means with the average treatment effect and visualize the entire range of the outcome variables in our randomly formed treatment groups (Coppock Reference Coppock, Druckman and Green2021). We also test for heterogeneous effects across the treatments. Previous research by Sikorski, et al. (2020) has highlighted how partisan-motivated reasoning may alter how scandals are perceived among partisans. In Scotland, the constitutional divide has acted as the major political cleavage, with substantial numbers of the population taking on constitutional labels as social identities (Henderson Reference Henderson2022). Constitutional preferences may therefore moderate a respondent’s response to each scandal. Supporters of independence should be more likely to engage in contrasting in the Partygate scandal, using it to affirm their hostility to Westminster and their preference for Holyrood, with the same logic applied to supporters of the Union in the Campervangate scandal. To account for this, we also interact our treatment variable with a binary variable indicating whether a respondent is in favour of Scottish independence or not. In Study 1, 42.2 per cent of respondents (n = 519) supported independence, with 46 per cent opposing (n = 569). In Study 2, 40.9 per cent (n = 422) supported independence, while 44 per cent of respondents opposed (n = 459). Respondents who answered ’Don’t know’ were excluded from this analysis. In the supplementary appendix, we perform the same test but interact treatment with incumbent support.
Main Results
The results of both studies are presented in Figure 2. Focusing first on Study 1 (upper panel), we observe almost a mirror effect across levels of governments. Exposure to the Partygate treatment caused a reduction in trust in MPs (at the ninety per cent level), UK Ministers, and UK civil servants – all statewide actors. Conversely, the treatment resulted in an increase in levels of trust towards MSPs and Scottish ministers. The findings provide support for the contrast hypothesis. It is noteworthy that the positive effects on trust in actors at the Scottish level are equivalent to the negative effects on trust in the UK actors. While this may owe in part to a floor effect – there being only so much lower that trust in Westminster could go – it is nonetheless a strikingly powerful instance of a contrast effect.
Effect of treatment on trust in X to behave ethically in their job with 95 per cent confidence intervals.
These results did not replicate in Study 2 (lower panel). Priming respondents to think about a scandal at the substate Scottish level (Campervangate) had very little effect on trust evaluations, and certainly there was no contrast effect whereby UK institutions benefited from the Scottish-level scandal. Given the limited effects on show, it is hard to classify the results neatly per the hypotheses. The fact that the only two significant reductions in trust were at the scandal-hit level – Scottish civil servants and, at the ninety per cent level, Scottish government ministers – is more consistent with the containment hypothesis. But there is a hint of contagion given that the non-significant results for UK-level actors are at least all in a negative direction.
Heterogeneous Effects: Do Constitutional Preferences Shape Reactions?
How far does the divide between supporters and opponents of independence explain the pattern of response to the two treatments? Figures 3 and 4 visualize the interaction between constitutional preferences and the experimental treatment. In Study 1, there is little clear evidence of constitutionally motivated reasoning, with only one significant negative effect observed among Independence supporters’ attitudes towards UK civil servants. However, we observe a containment effect among unionists, with negative effects observed for every UK level actor and no effect among Scottish actors. These findings reveal a nuanced pattern of response across the electorate, wherein the most pronounced negative evaluations of UK political actors paradoxically emanate from those traditionally expected to be their staunchest defenders.
Study 2 also produces little evidence of motivated reasoning in respondents’ reaction to a scandal. Among unionists, the only significant effect we observe is one of containment towards Scottish Civil Servants. Among independence supporters, we observe no significant changes in evaluations for UK or Scottish actors.
Study 1: Effect of treatment on trust in [X] to behave ethically in their job interacted with support for independence.
Study 2: Effect of treatment on trust in [X] to behave ethically in their job interacted with support for independence.
Discussion
It hardly helps to rebuild general trust if a specific scandal in one part of the political forest rebounds on political actors and institutions more broadly. There are signs of such contagion in previous research, suggesting that scandals are attributed not just to those involved but also to politicians as a class or category. But does that category extend vertically as well as horizontally, across as well as within levels of government? The experimental studies presented here suggest that it does not. Scottish citizens primed to consider a scandal at the heart of Westminster did not think significantly less of political actors at Holyrood, and the same is true vice versa. This is not surprising given that these are two distinct, important, often contrasted and sometimes hostile levels of government. This explains why, when a scandal broke at Westminster, the Scottish level actually stood to benefit from the contrast rather than to suffer from any contagion.
Three points should be noted about this contrast effect. One is its novelty: ours is the first study to report such positive contagion. The second is its size. Priming treatment effects are hard to assess in absolute terms, especially given the dampening impact of pre-treatment: that is, the fact that these prominent scandals will already have left their mark on the attitudes of both control and treatment groups. However, we can assess them in relative terms and it is striking that the positive effect of Partygate on attitudes towards Scottish actors looks as strong as its negative effect on evaluations of the implicated UK level. The third point, however, is that this contrast effect did not manifest itself in the case of Campervangate. There are no signs of Westminster politicians profiting from the Scottish government’s troubles. Why the divergence across the two cases? And does it lie in the differences between the specific scandals or the differences between the two levels in terms of public readiness to read from one to the other?
We contend that the latter is more likely. The scale, prominence, and egregiousness of the scandal did not vary much across the two cases. Both were landmark scandals, extensively covered in the media, which left large dents in the affected party’s poll ratings and were at least implicated in the departure of the head of government. Where the differences lie, we think, is in two related variables: the entrenchment and the positivity of perceptions of the two levels. Westminster institutions and, in particular, Conservative governments are wearily familiar features of the political landscape to a large majority of Scots. Less than twenty-five years since devolution, perceptions of the Scottish institutions are probably less crystallized – as well as being appreciably more positive. This evaluations gap is partly driven by the large anti-Conservative majority in Scotland, but it is about national as well as party identities: there is a pronounced skew towards Scottish over British identity and, reflecting the strong link between national identity and constitutional preferences, this skew also translates to an advantage for the Scottish level in terms of trust in political institutions. In the light of this, the lack of a contrast effect after Campervangate is unsurprising. Views on Westminster politics were too well entrenched to be shifted by events at Holyrood, and in any case, there was likely to be motivated resistance to any positive contagion. The reaction of at least some citizens to Partygate seemed to be that ‘things could be done better elsewhere’. The reaction to Campervangate looks more like ‘maybe they’re all as bad as each other’.
As these arguments imply, there are scope conditions on our findings. Partygate could be seen as a most likely case for a contrast effect, partly because of the politicized rivalry between the two levels and, related, partly because of the alignment of the scandal-hit parties – the Conservatives and the SNP – with the level or arena in which they offended. Negative contagion seems the likelier outcome in a case where the same party is in government at both levels. Contagion also seems likelier from a first-order arena scandal to a more obviously subordinate and non-rival arena, such as local government, although it is also quite possible that local politicians are seen as a distinct category in a way that would minimize spillover. Overall, it seems that citizens do distinguish vertically between institutions and levels, which makes scandal contagion more unusual than we might expect from the common portrayal of citizens as arbitrary in their punishments and ignorant of constitutional distinctions.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/S0007123425000043.
Data availability statement
Replication data for this article can be found in Harvard Dataverse at: https://doi.org/10.7910/DVN/9OOKLP.
Acknowledgements
The authors thank the editor and anonymous reviewers for their feedback and suggestions on the initial draft. Previous versions of this paper were presented at the EPOP Annual Conference in 2022 and we thank panel participants for their comments.
Financial support
This research was supported by grants from the Economic and Social Research Council (Award number: ES/V01000X/1).
Competing interests
None for all authors.
Open access
