Kinder, gentler – and crisis-proof? Consensus democracy, inclusive institutions and COVID-19 pandemic performance

Abstract Which patterns of democracy perform best? Although a decades-long research tradition has fairly robustly suggested that consensual democracies outperform their Westminster-style majoritarian counterparts, the scope of previous studies has been limited to ‘normal times’. In this article, the endogenous context of the COVID-19 pandemic is leveraged to study whether the alleged superiority of consensualism also holds during crises. It is hypothesised that, in addition to consensus democracy, inclusive institutions – i.e. cabinet size and interest-group corporatism – enhance crisis-related performance. Drawing on new and original data, cross-sectional and hierarchical time-series regression analyses show that horizontal power-sharing and the number of ministers substantively reduced excess mortality, while the structure of the interest-group system had no effect. Although established consensus democracies can draw on their built-in institutional assets even during crises, our findings indicate that majoritarian systems may, as a compensatory performance-enhancing tool, flexibly gear up for crisis-induced necessities by adding more ministers to the cabinet.

that prove to be more capable of -and more suited to -'getting things done'? What kind of democracy has an edge over its less resilient peers? These are the crucial questions preoccupying political practitioners, the media and the public alike, as the world lurches from one crisis to the next.
The connection between political institutions and government performance lies at the centre of decades-old traditions of empirical research on different types of democracy (e.g. Bernauer et al. 2016;Bernauer and Vatter 2019;Doorenspleet and Pellikaan 2013;Finnegan 2022;Gerring and Thacker 2008;Roller 2005;Schmidt 2002). In the literature, the answer to the question 'Which types of democracy do better?' has been strongly influenced by Lijphart's (2012) seminal contribution -a contribution that, for all its weaknesses, 1 is still considered to be 'the single most influential typology of modern democracies' (Mainwaring 2001: 171). Lijphart (2012) measured a set of 10 political institutions using scales ranging from majoritarianism at one end to consensualism at the other. Based on a cluster analysis, two distinct dimensions were identified: the 'executive-parties' dimension and the 'federal-unitary' dimension, which map different 'patterns of democracy' . He not only shows that majoritarian and consensus democracies diverge with respect to the underlying rationality of whether 'the sheer majority' or 'as many people as possible' do the governing (Lijphart 2012: 4), but also addresses the notorious 'So what?' question, showing that institutional differences affect how well and how efficiently democracy works. His major finding can be boiled down to two appealing phrases: life in consensus democracies is 'kinder and gentler' , while government performance in consensus democracies is 'more effective' (Lijphart 2012: 274, 255). Subsequent generations of scholars have followed suit, connecting 'patterns of democracy' to a multitude of performance-related indicators. The main result has remained fairly robust: consensus democracies outperform majoritarian democracies in terms of, for example, the quality of democracy, government effectiveness, inflation containment, coherence of policy design and sustainable policy performance. Moreover, they are better able to 'act credibly and provide basic services for the most defenceless members of society' (Bernauer and Vatter 2019: 105;cf. Bernauer et al. 2016;Bernauer and Vatter 2019;Doorenspleet and Pellikaan 2013;Finnegan 2022;Jahn and Suda 2022;Roller 2005;Schmidt 2002). 2 However, the scope of these studies is limited to 'normal times' . 3 Contrary to what the obviously popular 'public enterprise' of linking institutional features to effective (or ineffective) crisis management suggests, there has, thus far, been no analysis of whether the alleged superiority of consensus democracies also holds outside periods of economic and/or political stability. On the one hand, decades of crisis and disaster research has consistently revealed that institutional settings do indeed shape crisis responses (e.g. Ansell et al. 2016;Boin and 't Hart 2007;Boin et al. 2016;Christensen et al. 2016;Rosenthal et al. 1989). On the other hand, given what crises are by definition, it is highly doubtful whether consensual institutions really support favourable (i.e. effective) crisis responses. While crises are characterised by, for example, time constraints, pressure, uncertainty and friction, consensus democracies slow down decision-making by involving 'the many' .
In this article, we leverage the endogenous context of a global crisis that is -given the severity of its multifaceted health-related, economic and social consequences -unprecedented in the post-war period to analyse whether patterns of democracy explain a given country's performance in fighting the spread of SARS-CoV-2. Hence, our aim is to assess whether consensus democracy also outperforms majoritarian democracy in times of crisis.
The question of whether specific types of democracy are more crisis-resistant is highly relevant, both socially and scientifically. From a societal perspective, there is an indisputable interest in obtaining systematic answers, one that is fed by the insatiable curiosity of the public and the media to learn what could be done to gear up for even bigger crises on the horizon. The scientific relevance, for its part, lies in both the theoretical and empirical realm. In theoretical terms, we offer a novel theoretical account of how consensus democracy, as well as two additional institutions that are grounded in the 'spirit of inclusiveness' -namely cabinet size and interest-group corporatism -matter for crisis-related performance. On the empirical level, our study tackles the three shortcomings of previous analyses of the impact of institutional features on the management of the COVID-19 pandemic: First, although certainly insightful, existing studies are limited to only one aspect of the political system, such as federalism (e.g. Hegele and Schnabel 2021;Lecours et al. 2021), despite the fact that, following neo-institutional premises, political institutions can be understood to add up to a comprehensive set of 'rules' that jointly govern, incentivise and restrict human behaviour (Hall and Taylor 1996). Second, earlier studies are 'trapped' between small-N case-based research (e.g. Bandelow et al. 2021;Hegele and Schnabel 2021;Lecours et al. 2021) and large-N quantitative approaches (e.g. Toshkov et al. 2022;Wagschal 2022). The former struggle with generalisation, while the latter suffer from 'Y-centrism' , since although they factor in a large number of variables, they remain fairly silent about the actual theoretical mechanism at work. What is more, previous quantitative studies have usually been cross-sectional or limited to the 'first wave' (e.g. Engler et al. 2021;Toshkov et al. 2022). They miss out on undeniably strong 'pandemic dynamics' caused by, for example, the even deadlier Delta variant or the vaccination roll-out, both of which were 'game changer[s]' . 4 Third, existing political science research uses flawed crisis-performance indicators. Studies that equate policy measures with effective crisis management (e.g. Engler et al. 2021) have overlooked that 'more stringent policy' is not necessarily associated with better crisis handling (Herby et al. 2022). Authors who utilise (cumulative) numbers of confirmed COVID-19 cases or fatalities (e.g. Toshkov et al. 2022;Wagschal 2022), for their part, overlook crucial epidemiological warnings that case and/or death numbers are heavily skewed by differences in testing and reporting capacities (e.g. Aron et al. 2022;Beaney et al. 2020).
Hence, by offering a genuinely 'institutional account' (Finnegan 2022: 4), our novel theoretical approach ultimately answers the plea that 'formal political institutions' be considered as a 'key focus' (Greer et al. 2020(Greer et al. : 1413 for understanding the effectiveness of COVID-19 responses -or, more generally, crisis performance. Moreover, we tackle the three aforementioned shortfalls of previous research by: (i) studying the crisis performance of 'patterns of democracy' (rather than restricting our analysis to a single political institution); (ii) aiming for a homogenous medium-N sample of high-income democracies that combines a cross-sectional analysis of the extent to which inclusive institutions matter for level differences and hierarchical time-series cross-sectional analyses of pandemic dynamics across the entire first two years of the COVID-19 pandemic; and (iii) opting for the most robust epidemiological indicator available, namely excess mortality Beaney et al. 2020).
Our empirical findings establish that consensus democracies are more crisis-proof than their majoritarian, Westminster-style counterparts. They also show that larger cabinets are better able to manage crises, meaning that while established consensus democracies can continue to draw on their built-in institutional assets during crises, majoritarian systems may compensate by adding ministers to the cabinet, which provides them with flexibility in gearing up for the crisis.
The remainder of this article is structured as follows: the next section lays out our novel theoretical account of how consensus democracy, along with two inclusive institutions that have previously been neglected, matters for crisis-related performance; section three provides information on our data and methods; the fourth section presents the empirical results; and the final section concludes with a discussion of our findings in a broader context and sketches avenues for further research.

Theoretical argument: how consensus democracy and inclusive institutions matter for crisis-related performance
Throughout the prosperous post-war period, it has been argued that political institutions do 'make a difference for … the effectiveness of government' (Lijphart 2012: xviii;cf. Bernauer et al. 2016;Bernauer and Vatter 2019;Doorenspleet and Pellikaan 2013;Finnegan 2022;Gerring and Thacker 2008;Roller 2005;Schmidt 2002). More precisely, it has been assumed that consensus democracy outperforms its majoritarian, Westminster-style counterpart. These Lijphartian-inspired arguments emphasise the relative advantages of consensual political institutions over majoritarian ones and can be boiled down to three major claims. The first argument relates to 'input legitimacy' (Scharpf 1999): representative government must not only govern, but also represent. Not only the results matter, but also who has a say. However, in a Westminster-style majoritarian democracy, the electoral winner 'takes all' -over and over again. As a result, certain societal groups are systematically excluded from public policy-making and may eventually no longer be willing to bear the costs of implementation -or even start jeopardising governmental decrees by refusing to comply. Consensual institutions, such as multi-party cabinets, by contrast, ensure the greatest possible inclusion of contending groups. Inclusiveness fosters acceptance -and making widely acceptable rather than snap decisions is key for smooth governance.
The second argument touches upon the quality of decisions. Majoritarian democracies characterised by, for example, one-party majority cabinets endowed with a clear mandate to govern may be able to make decisions faster than in a consensual system. 'Retarding' consensual institutions, such as broad, encompassing cabinets, necessitate time-consuming 'amicable agreement' (Steiner 1974). That said, 'fast decisions are not necessarily wise decisions' (Lijphart 2012: 257). By consulting as many stakeholders as possible, consensus democracies make use of the 'wit of the many' for the 'common good' , that is, to make 'wiser' decisions. This notion can easily be extended by making explicit reference to the 'wisdom of crowds' . As psychology reminds us, '[t]he average quantitative estimate of a group of individuals is consistently more accurate than the typical [individual] estimate' (Herzog and Hertwig 2009: 231). Hence, '[i]ncreasing the crowd size improves the quality of the outcome' (Walter et al. 2022: 269). 5 Finally, the third argument for why consensus democracy outperforms brute majoritarianism concerns policy coherence. A coherent policy design manages to successfully 'integrate different understandings of a policy problem with different conceptions of the policy instruments to be utilized' (Peters and Fontaine 2022: 1). As the broad policy-design literature has thoroughly established, 'political cohesion' -i.e. 'leadership capacity for pursuing strategic directions and harnessing support' (Head 2022: 155) -is a crucial factor that contributes to fostering the emergence of coherent policy design. The latter is, in turn, essential for dealing with 'tricky' or 'messy' policy challenges in a chaotic world of crises and turbulence. Because a few extremely powerful people have a say, majoritarian governments may display short-term policy-design coherence. However, in majoritarian democracies, policy-design coherence is heavily undermined, or even 'negated' , by frequent, sweeping alternations in government. If the cabinet (government) and the 'shadow cabinet' (opposition) often swap places, politics becomes a matter of 'stop and go' . 6 Successful government, however, requires a 'steady hand' (Finer 1975). In the long run, consensus democracies are thus not only 'better able to provide coherent, centrist policy-making' . Rather, coherent policy is also more likely to be implemented successfully and to remain 'on course' (Lijphart 2012: 257;cf. Finnegan 2022).
Importantly, these arguments in favour of consensus democracy were all formulated during, and certainly influenced by, exceptional, prolonged periods of post-war economic and political stability. It remains unclear whether consensualism's superiority in terms of input legitimacy, quality of decision-making and policy coherence endure when a crisis hits. A crisis is usually defined as 'a perceived threat against the core values or life-sustaining functions of a social system that requires urgent remedial action in uncertain circumstances' (Rosenthal et al. 1989: 10;cf. Boin et al. 2016;Christensen et al. 2016). At first sight, crises thus seem to be the gravediggers of consensus democracy. Urgency, acute time pressure and the need for immediate action seem to fundamentally undermine the advantages of consensus democracy in times of crisis. If 'retarding' consensual institutions may be an asset under ordinary conditions, 'not moving' is a fatal error in crises. 'The many' may eventually arrive at wiser decisions, but, in an emergency, time for deliberation is lacking. Speed trumps perfection, since otherwise settlements will flood, enemies will flow across the borders and lethal viruses will spread uncontrollably. In short, crises are exceptional situations that require an exceptional 'crisis approach' (Boin and 't Hart 2007), in the sense of a governmental crackdown rather than steady guidance.
Against these objections, however, we argue that the performance-enhancing attributes of consensus democracy work even during a crisis, offering a competitive advantage over its Westminster-style majoritarian counterparts. However, crisis-related specificities must be accounted for in two respects: First, we must clarify which set of political institutions in consensus democracy are actually responsible for better crisis management. Second, crisis brings additional 'inclusive institutions' to the fore -institutions that provide a means to foster inclusion in practice and to promote it as a normative ideal (e.g. Bogaards 2020). We theorise that in addition to consensus democracy, cabinet size and interest-group corporatism also influence crisis-related performance.
To start with, we expect the relative advantage of consensus democracy to be driven by (i) the so-called 'executive-parties' dimension, i.e. horizontal power-sharing. Already in 'normal times' , the impact of the 'federalist-consensual' dimension on effectiveness has been found, respectively, to be 'uniformly minimal' or 'extremely weak and statistically insignificant' (Lijphart 2012: 256, 272;cf. Bernauer and Vatter 2019). The reason for this is that the performance-enhancing and performance-inhibiting effects of vertical political institutions cancel each other out. We presume that crises make their 'Janus face' even more visible. Federalism is a case in point: on the one hand, it allows for 'experimental' region-specific public policy well-tailored to particularly affected regions, e.g. regions that are marked by a high incidence of environmental disasters; on the other hand, federalism encourages 'patchwork responses' , non-coordination and blame-shifting (e.g. Hegele and Schnabel 2021;Lecours et al. 2021). There is obviously a trade-off between desirable and undesirable effects. Accordingly, we theorise that only the executive-parties dimension of consensus democracy is associated with better crisis performance, while no systematic relationship can be found in terms of vertical power-sharing.
(ii) Cabinet size is the first of the two inclusive institutions that we believe make a difference. We posit that not only the type of cabinet, but also its size, matters for effective crisis management. It is well understood that crises are the 'hour of the executive' . It is the government that determines special emergency measures (e.g. first aid, provision of recovery funding). Such government decisions are often based on emergency decrees that work at the expense of the legislature (see Bolleyer and Salát 2021;Cormacain and Bar-Siman-Tov 2020). That is why executives deserve special attention. However, previous research on patterns of democracy has only considered who: 'executive inclusiveness' -a core indicator of horizontal power-sharing -has been limited to the coalition in place and/or the governing parties (e.g. Bernauer and Vatter 2019;Ganghof 2010;Lijphart 2012). In doing so, how many has been completely left aside, although large cabinets are also a 'proxy of the inclusiveness of the government' (Indridason and Bowler 2014: 383). Such an omission is a severe shortcoming, as studies in political economy and/or executive politics have robustly and consistently established that the number of ministers is 'politically and economically consequential' (Indridason and Bowler 2014: 383;cf. Baskara 2013;Shpaizman 2022;Vall-Prat and Rodon 2017;Wehner 2010;Woo 2003). Accordingly, we hypothesise that larger cabinets enhance crisis-related performance in terms of both preparedness and government capacity. Governmental decision-making that is distributed across many ministers increases expenditures, since in order to survive frequent 'cabinet reshuffles' (Indridason and Kam 2008), ministers have every incentive to use their departments to serve their own ambitions by engaging in 'signalling' towards, for example, their political party and/or constituency. The most straightforward way to do so is by spending large amounts of money. In larger cabinets, these 'individual-level incentives to boost spending' (Wehner 2010: 631) multiply. More ministers equals more (welfare) spending, especially in health care and public safety (Baskara 2013: 367;cf. Perotti and Kontopoulos 2002;Vall-Prat and Rodon 2017;Wehner 2010;Woo 2003). This expenditure, which is what promoters of fiscal discipline are afraid of, gives larger cabinets the edge over smaller and thus less free-spending ones. In 'normal times' , larger cabinets provide for better quality health-care services (e.g. more hospital or ICU beds) and better trained medical staff -preparedness that pays off as soon as crises hit (e.g. Abrams et al. 2021). Yet larger cabinets are not only better prepared, but also more capable of handling crises effectively. Specialisation is key here: the doctrine of 'individual ministerial responsibility' presupposes that a cabinet minister bears ultimate responsibility for the actions of the corresponding ministry. Every minister is an accountable expert in their own domain -and is expected to take the blame by resigning. The existence of many departments implies that there are many different, highly specialised portfolios. If a multifaceted crisis hits, ministers are thus strongly incentivised to throw their technical skills into the ring to help advance the 'common good' for the sake of their own survival. Once again, in larger cabinets, these individual-level incentives multiply.
Finally, we argue that (iii) the interest-group system is also conducive to enhanced crisis performance. In line with recent advancements in empirical research on patterns of democracy, we consider corporatism as an inclusive institution that captures 'the inclusion of business and labour' (Armingeon 2002: 84), rather than belonging to the executiveparties dimension (see Armingeon et al. 2021;Giuliani 2016). It measures how cooperative and consensus-oriented the nexus between state and non-state actors is. Hence, corporatism fits perfectly within the 'consensual logic' of inclusiveness (Finnegan 2022). Faring well in crises not only requires intra-cabinet bargaining, but also that the political elite involve non-governmental collective organisations, such as first-aid responders or business associations to obtain the relevant information 'on the ground' . When facing 'super wicked problems' with devastating, yet multifaceted health, social and/or economic consequences, decision-makers are 'naturally' more 'inclined to seek input from organisations that are best equipped to provide information that is relevant to meet … new needs' (Junk et al. 2022(Junk et al. : 1379). Yet, more information from sectoral lobbies may also result in a fragmented and incoherent policy response, as lobbies, by definition, prioritise vested interests over the 'common good' (e.g. public health). Hence, the presumed superiority of corporatism over pluralism: as a more 'closed' interest-group system, it offers fewer 'access points' for fewer sectoral lobbies. Corporatism is less vulnerable to organised interests that torpedo public authorities' efforts to effectively manage crises. Instead, corporatist networks are based on 'frequent face-to-face interactions' and mutual 'trust' (Finnegan 2022: 11, 12). In a similar way to consensus democracy, albeit with the help of different (i.e. non-governmental) players, corporatism therefore helps the 'wisdom of crowds' to crystallise.
Bringing all these strands together, we expect that consensus democracy, if limited to the executive-parties dimension outperforms majoritarian democracy even in times of crisis. Yet we also theorise that, in times of crisis, two additional inclusive institutions matter as well, namely cabinet size and corporatism.

Data and methods
We tested our theoretical assumptions for the full sample of 22 long-established democracies, i.e. democracies with an uninterrupted democratic history since 1976 and with at least one million inhabitants. 7 This reflects the state-of-the-art medium-N approach in the 'So what?' tradition of empirical research on democracy (see Finnegan 2022 for a recent example). Moreover, such a homogeneous most similar sample of high-income democracies is not skewed by different publication standards for real-world data. 8 In contrast to previous studies (e.g. Engler et al. 2021;Toshkov et al. 2022;Wagschal 2022), our data is not limited to the first wave(s) of the COVID-19 pandemic, but covers the first two years in full (1 March 2020-31 December 2021. Hence, our data consists of 22 different points in time and a total of N = 484 observations. We drew on a new and original dataset that triangulates data from various sources, further supplemented by our own data collection. In terms of our dependent variable (i.e. effective management of the COVID-19 pandemic), the most obvious indicators all have specific downsides. Across countries, case numbers and mortality estimates are seriously affected by differences in testing and reporting capacities (Omori et al. 2020). There are still many ambiguities concerning what a 'COVID-19 death' actually is, with the coronavirus often being 'a contributing factor, but not the cause leading directly to death' (Beaney et al. 2020: 329). Vaccination shares, for their part, are only available for the period after roll-out began, starting from early 2021. Epidemiologists thus univocally recommend the usage of 'excess mortality' as the most reliable and comprehensive metric, i.e. 'the gold standard' (Beaney et al. 2020: 329;cf. Aron et al. 2022). Excess mortality reflects how the number of deaths during the period of the COVID-19 pandemic compares to the number of deaths one would have expected had the pandemic not occurred, a crucial quantity that cannot be known but that can be estimated in several different ways. Given that the raw number of excess deaths is 'less comparable across countries due to large differences in population' , we use the P-score ). This is a highly trustworthy measure of excess mortality proposed by an interdisciplinary group of researchers at the University of Oxford, which has been used by, and published in, numerous well-received, peer-reviewed articles representing a diverse range of disciplines (e.g. De Geyter et al. 2022):

P score
Reported Deaths Projected Deaths Projected Deaths *100 We specify our core independent (i.e. institutional) variables by relying on novel data provided by the Comparative Political Data Set (CPDS; Armingeon et al. 2021), which offers a time-variant proxy for the executive-parties dimension (lfirstp). The data is composed of the moving averages of 10 years of four indices (2010-2019): the proportionality of electoral systems (dis_gall multiplied by [−1]), the number of effective parties in parliament (effpar_leg), the absence of minimal winning and single-party majority cabinets (dummy variable; gov_type), and a measure for cabinet dominance, established on the basis of the number of changes in government per year (gov_change). The moving averages of the four indices were z-standardised and added up. The sum was then z-standardised again to yield the country score for the updated executive-parties dimension. Importantly, the CPDS time-variant proxy does not include the structure of the interest-group system, which is in line with our theoretical argument (previous section). For the measurement of interest-group pluralism, we reverted to a time-variant corporatism index provided by Jahn (2019). 9 We utilised CorpCORE, the most reliable index for 29 countries, which is based on z-scores and is available on an annual basis and smoothed time series. Following the 'Who governs?' dataset, we operationalised cabinet size as the number of core ministers who are formally members of the cabinet, excluding unoccupied positions, positions held by the same person and posts that are not considered core positions (Nyrup and Bramwell 2020;cf. Indridason and Bowler 2014 for a similar approach). In order to account for frequent 'cabinet reshuffles' (Indridason and Kam 2008), the data has been averaged over time. This measurement produces a time-invariant variable, which corresponds well to our conceptualisation of cabinet size as an inclusive institutional feature.
We factored in several controls, all of which were derived in light of core insights from a multitude of disciplines studying the determinants of SARS-CoV-2 transmission (e.g. epidemiology, human geography, public health and economics). As many countries experienced unusual levels of 'executive dominance' (e.g. Bolleyer and Salát 2021; Cormacain and Bar-Siman-Tov 2020), we controlled for the specific pandemic-related institutional configuration by factoring in the PanDem index (Edgell et al. 2021). This index comprehensively captures the degree to which government responses to the COVID-19 pandemic departed from a given country's 'ordinary' levels of horizontal and vertical power-sharing. The PanDem index assesses, among others, the 'emergency role' of the legislature (e.g. limitations to its law-making powers), legislative approval of government measures and subnational variation in pandemic-related measures. To account for public policy, we have included the Oxford COVID-19 Government Response Tracker (OxCGRT), or 'COVID-19 Stringency index' for short (Hale et al. 2022). The number of hospital beds per 1000 inhabitants controls for the preparedness of the health-care system. 10 Drawing on epidemiological findings, we also checked whether pandemic performance is influenced by geographical scope conditions. Island nations can exploit a natural advantage when it comes to stopping travellers importing infectious disease (dummy variable; located on an island 1/0). 11 Finally, we accounted for the share of people who reported having confidence in the national government (OECD 2021; cf. Erhardt et al. 2021).
We pursued a two-step empirical strategy. First, we performed various cross-sectional OLS regression models to explain differences in levels. To this end, time-sensitive variables were averaged over the full investigation period (i.e. excess mortality, number of ministers, stringency index, PanDem index). Regression diagnostics (e.g. multicollinearity, heteroscedasticity, outliers), as well as numerous robustness tests, are reported in the extensive online documentation. Additional analyses show that excluding outlier cases (i.e. Australia and the UK) and influential control variables (e.g. island dummy, trust in government) does not change any of the main results of the cross-sectional models. In order to validate and qualitatively corroborate our cross-sectional findings, we drew on 'nested analysis' (Lieberman 2005), a multi-method strategy for gaining maximum leverage when combining statistical analyses with case-based research.
The hierarchical time-series cross-sectional analyses performed in the second step account for pandemic dynamics. We estimated hierarchical linear models for monthly data (Gelman and Hill 2007), where j indicates a country and t a specific month: The parameter a j is the random intercept, while the coefficients of the explanatory variables remain constant over time ( β ) . We used excess mortality, the PanDem index and the COVID-19 Stringency index (squared) as level-1 predictors, since they vary over time, whereas the institutional variables, as well as trust in government, island location and the number of hospital beds per 1000 inhabitants were considered time-invariant 'contextual' variables (level-2). The implication of this modelling is that these time-invariant variables will hardly explain any short-term variation in excess mortality, although they may be related to systematically varying levels of our dependent variable. This fits nicely with our theoretical expectations. We used time-fixed effects ( t ) to account for (monthly) time dynamics in excess mortality. This approach is particularly suitable for capturing the different pandemic waves (see our extensive online documentation for more information on this). 12

Cross-sectional analysis: how consensus democracy and inclusive institutions matter for level differences
Does consensus democracy, along with cabinet size and interest-group pluralism, matter for crisis performance? Table 1 reports the empirical results of the cross-sectional analysis.
If we look at mean excess mortality, consensus democracies do indeed seem to have been better at handling the pandemic's most severe health-related consequences. The significant coefficient implies a substantial effect: an increase in a country's 'executive-parties' dimension score by 1 index unit reduces excess mortality by 1.9-2.2 percentage points. This is a noteworthy effect size, given that -on average and across the 22 high-income countries in our sample -excess mortality increased by 6.2 percentage points during the pandemic. This significant, substantial cross-sectional result holds even when controlling for 'natural' geographical advantages (i.e. located on an island), (mis)trust in government and health-care system preparedness in terms of number of hospital beds. Nor does it disappear once we control for crisis-specific institutional configuration, as captured by the PanDem index (model 4). It is also robust if the non-significant policy measures adopted to fight the spread of the coronavirus are factored in (model 5). Accordingly, even if consensual democracies experienced almost full empowerment of the executive branch (especially in the first waves, when vaccinations were not yet available), they are still able to benefit from their ordinary levels of horizontal power-sharing. The assets of consensus democracy -namely the greatest possible inclusivity (input legitimacy), quality of decisions  note: *p < 0.1; **p < 0.05; ***p < 0.01. and policy coherence -seem to continue to be effective even during crises. This finding can be validated and given qualitative support by contrasting the crisis performance of consensus democracies with the record of Westminster-style majoritarian democracies. The Nordic countries stand out for both their high consensual scores and comparatively low excess mortality. 13 Denmark, Finland, Norway and Sweden are well-known for their predominantly tax-financed health-care systems that provide universal coverage to all residents. It is certainly fair to believe that public responsibility, high quality standards and their 'social-democratic welfare state' (Esping-Andersen 1990) gave them an edge in terms of preparedness. It is important to note, however, that the Nordic countries also embodied the principle of broad expert involvement and evidence-based policy-making from the onset of the pandemic. Convinced that 'wise' policy response requires not only the 'wit of the many' , but also 'the wit of the right people' , Denmark, Finland, Norway and Sweden made '[e]xtended use of ad hoc experts groups' that supplemented national 'advisory agencies' (Saunes et al. 2022: 425;cf. Laage-Thomsen and Frandsen 2022). Whereas (growing) scepticism towards scientific expertise or even science denialism severely exacerbated the health-related consequences of the coronavirus elsewhere, the Nordic countries exploited the tendency towards inclusive, encompassing public policy-making that is 'naturally' built into their model of democracy. The Nordic 'consensual democracies' are indeed famous for their 'distinctively Scandinavian culture of consensus and … structures for conciliation and arbitration' (Elder et al. 1988: 221;cf. Lijphart 2012). Westminster-style majoritarian democracies in the English-speaking world, above all the UK and the US, were among the most severely affected countries. Although on paper, the UK and US were the best prepared to deal with a pandemic (Global Health Security Index 2021), 14 this assessment 'did not account for the political context': 15 The strong institutional position of their leaders led to 'appalling political decisions … [that] consigned many of their citizens to the grave' . 16 Whereas in the UK, the single-party majority cabinet did not allow for 'course correction' through institutionalised inter-party bargaining, the former US administration 'rejected the value of science' (Cairney and Wellstead 2021: 7) and hence refrained from consulting 'the many' . 17 There is therefore strong case-based evidence in support of our first empirical finding that consensus democracies seem to outperform their majoritarian counterparts. Furthermore, in line with our theoretical hypotheses, the federalist dimension consistently and robustly failed to attain statistical significance (Table 1). This harmonises with our expectation that the performance-enhancing and performance-inhibiting effects balance each other out, resulting in an ultimately insignificant relationship. That neither the federal-unitary dimension nor federalism 18 per se are related to government effectiveness is an important finding that provides further evidence that crises often lead to blame-shifting: federalism becomes a scapegoat (e.g. Lester and Krejci 2007).
When it comes to the two inclusive institutions that were argued play a role in addition to horizontal power-sharing, our empirical findings are nuanced. To reiterate, we hypothesised that cabinet size and interest-group corporatism both enhance crisis performance as well. Our results show that cabinet size -an indicator that has been completely neglected in empirical research on different types of democracy -clearly matters. One additional minister in the cabinet significantly drags down excess mortality, by almost one percentage point. The structure of the interest-group system does not, however, reach conventional levels of statistical significance.

Hierarchical time-series cross-sectional analysis: how consensus democracy and inclusive institutions matter for pandemic dynamics
Crises unfold over time: accelerating in a self-reinforcing manner at one moment and slowing down at another (e.g. Boin and 't Hart 2007;Boin et al. 2016). In order to account for such crisis-related dynamics, we used monthly data to test whether our findings from the cross-sectional analysis also hold when taking into account temporal dynamics, particularly the different pandemic waves. We present the results of the hierarchical time-series cross-sectional analysis in Table 2.
The findings are highly consistent across different model specifications and confirm the picture already revealed by the cross-sectional analysis: even if we account for pandemic dynamics and the highly volatile stringency of policy measures aimed at halting the spread of the coronavirus, consensual democracies outperform majoritarian democracies in terms of pandemic management. Any increase in a given country's score on the executive-parties dimension corresponds to a 2.2 to 2.4 percentage-point decrease in excess mortality, which is a substantial effect size.
The mixed picture concerning inclusive institutions is corroborated as well. On the one hand, cabinet size is significantly associated with a more limited number of deaths during the COVID-19 pandemic, as compared to the deaths one would have expected had the pandemic not occurred. On the other hand, interest-group corporatism is not related to crisis-related government performance. From our hierarchical time-series cross-sectional analyses, it is thus safe to conclude that consensus democracy and cabinet size do matter -even in times of crisis.   note: *p < 0.1; **p < 0.05; ***p < 0.01.

Conclusion
Which types of democracy perform best in terms of government effectiveness? Ever since Lijphart's (2012) seminal analyses of how 'patterns of democracy' matter for numerous performance-related indicators, the answer has generally been that consensus democracy, which disperses power into the hands of the many, outperforms its majoritarian, Westminster-style counterparts in terms of, for example, government effectiveness, economic performance and societal inclusiveness. However, empirical research has so far been limited to 'normal times ' , i.e. (post-war) periods of economic and/or political stability. In this paper, we have therefore leveraged the endogenous context of an unprecedented global crisis to scrutinise whether consensus democracy also offers a competitive advantage in times of crisis. More precisely, we have studied whether consensus democracies differed from majoritarian democracies with respect to excess mortality -the 'gold standard' (Beaney et al. 2020: 329) when systematically comparing public health consequences -over the two most severe years of the COVID-19 pandemic (January 2020-December 2021).
On the theoretical level, we have argued that the relative superiority of consensus democracy -namely the greatest possible inclusivity (i.e. input legitimacy), the high quality of decisions and policy coherencemanifest themselves even during a crisis. Yet we have theorised that crisis-related specificities must be accounted for in two respects. First, we expected the relative advantage of consensus democracy to be driven by the executive-parties dimension (i.e. horizontal power-sharing). Second, we argued that crises bring to the fore two additional inclusive institutions that help 'the spirit of inclusion' to materialise, namely cabinet size and the interest-group system, hypothesising that these would impact crisis-related performance on top of consensus democracy.
Drawing on a new and original dataset that triangulates data from various sources, further supported by our own data collection, we subjected our assumptions to a rigorous empirical test for a most similar sample of 22 high-income democracies (see Finnegan 2022 for a similar approach). Our findings, obtained from both cross-sectional and hierarchical time-series multivariate regression analyses, are consistent and highly robust across several different specifications: consensus democracy is significantly and substantively associated with lower excess mortality. Hence, consensus democracy not only outperforms its majoritarian, Westminster-style counterparts in 'normal times' , but it does so even in times of crisis. The federal-unitary dimension, in contrast, is not systematically related to crisis performance. This is an important finding, because it starkly contrasts with the public and media discourse, which often assumes that 'federalism is a bad prescription for handling … [crises]' . 19 As regards the two other inclusive institutions we theorised about, the results are mixed, yet consistent and robust. Cabinet size is significantly associated with a better crisis-related performance: the more ministers, the lower the excess mortality. In times of crisis, the number of ministers is not only economically consequential, in the sense that it incentivises more (welfare) spending at the individual level -thus enhancing preparedness, especially in health care and public safety (e.g. Baskara 2013; Indridason and Bowler 2014;Shpaizman 2022;Vall-Prat and Rodon 2017;Wehner 2010;Woo 2003)but it is also politically consequential, as there are strong individual-level incentives at play. Every single minister who is responsible for a highly specialised portfolio knows what is at stake: if they fail to get a grip on developing crises, it will be necessary to accept responsibility and resign. This is why, for the sake of their own survival, ministers instead join forces, helping to advance the 'common good' . Since, in larger cabinets, these individual-level incentives multiply, an increase in the number of ministers leads to better crisis performance at the aggregate level too.
The interest-group system, by contrast, fails to reach conventional levels of significance. When it comes to explaining the crisis-related performances of different types of democracy from both a cross-sectional and hierarchical time-series perspective, it is the executive-parties dimension (which is constitutive for consensus democracy) and cabinet size that matter. Corporatism does not.
Our study has important implications for at least four different research strands that have, thus far, rarely been brought together. First, it contributes to answering the 'So what?' question in the ever-appealing empirical study of 'patterns of democracies' (Lijphart 2012) by 'broadening the empirical scope' (Borman 2010: 5) from 'normal times' to times of crises. After all, limited applicability has been one of the major criticisms levelled at the Lijphartian typology of different types of democracies. The finding that consensus democracies are not only 'kinder and gentler [and] more effective' (Lijphart 2012: 274, 255) during phases of stability, but also handle crises more effectively, significantly buttresses the external validity of the superiority of consensus democracy over majoritarian democracy. As regards conceptualisation and measurement, our study stresses the importance of examining the structure of the interest-group system as a separate inclusive institution, rather than as an integral part of horizontal power-sharing (see Armingeon 2002;Armingeon et al. 2021;Finnegan 2022;Giuliani 2016). Second, our findings make a contribution to political economy, where the number of ministers has proved to have robust and substantial explanatory power in relation to government spending (e.g. Baskara 2013;Wehner 2010;Woo 2003). Beyond these economic consequences, however, the genuinely 'political side' of political economy has received hardly any attention. Our results stress the need to closely investigate political effects as well. Not only coalition forms and/or the ideological composition of the government (Ganghof 2010), but cabinet size, too, should be conceived of, and studied as, a defining feature of the executive -one that clearly matters for crisis-related performance. Third, our results contribute to crisis and disaster research (e.g. Boin et al. 2016;Boin and 't Hart 2007;Christensen et al. 2016;Rosenthal et al. 1989), we have not only shown that institutional settings are relevant, but also theorised about the actual mechanisms at work -a theoretical innovation the aforementioned literature can benefit from. Finally, on the epidemiological level, it has been shown that rather than the stringency of non-pharmaceutical interventions (e.g. physical distancing, stay-at-home orders), it is the polity that helps to prevent and control SARS-CoV-2 transmission -a finding that aligns well with existing meta-studies which indicate that 'lockdowns have had little to no public health effects' (Herby et al. 2022: 2).
Besides these important contributions to four different and rarely integrated fields of study (i.e. empirical research on democracy, political economy, crisis and disaster research and epidemiology), our article also has important practical implications. Consensus democracy (or the degree of consensualism) remains remarkably stable across time. The electoral system, the party system, executive-legislative relations and the coalition type that form part of the executive-parties dimension are path-dependent and inherently coupled. The party system, for example, evolves as a short-and long-term consequence of the electoral system. That is why consensus democracy is not something that can just be adopted whenever a crisis hits (and abandoned when things start to normalise again). Cabinet size, for its part, is an inclusive institution that is 'usually not fixed in the constitution' (Perotti and Kontopoulos 2002: 217). 20 Cabinet reshuffles happen frequently (Indridason and Kam 2008), meaning that, in an emergency, the president or the prime minister can easily, flexibly and rapidly increase the number of ministers in the cabinet, including additional, likely crisis-specific, portfolios in the government. That is good news for both consensus democracies and majoritarian democracies -and a relevant policy implication worth considering: on the one hand, consensus democracies, such as those of the Nordic countries, can benefit from their 'ordinary' levels of horizontal power-sharing. The relative advantage of a more consensual executive-parties dimension manifests itself even when controlling for crisis-specific changes to the institutional framework (e.g. centralisation of decision-making powers -see Bolleyer and Salát 2021;Cormacain and Bar-Siman-Tov 2020). On the other hand, majoritarian democracies such as the UK or the US, which are -even amidst an unprecedented global emergency -plagued by adversarial politics, partisanship and/or mutual distrust, have a compensatory performance-enhancing tool at their disposal, namely increasing the size of the cabinet. 21 Further research is needed to flesh out the actual 'causal mechanism' (e.g. Hedström and Swedberg 1998), since although our article has established that consensus democracy and larger, more inclusive cabinets are more crisis-proof, our design cannot explain why this is the case. Nonetheless, our study hints at potentially relevant 'causal mechanisms' at work, e.g. higher 'input legitimacy' (Scharpf 1999) through trust in government, higher quality decision-making and/or more coherent policy design. A promising way forward would be to adopt 'theory-testing process-tracing' in a comparative case study design (Beach and Pedersen 2013), a method that enables us to study the causal connections between consensus democracy and inclusive institutions, on the one hand, and better (crisis-related) government performance, on the other.

Notes
Jahn 2019; Lijphart 2012 for a similar approach). As an alternative specification, we have tested the explanatory relevance of 'sectoral corporatism' , measured through the health-care system (Reibling et al. 2019). As reported in the online appendix, neither the full CorpCORE index nor 'sectoral corporatism' matter for crisis-related government performance. 10. Note that we have also tested alternative health-care variables (e.g. health-care expenditure, expenditure on outpatient care). As reported in the online appendix, they are not significantly related to our dependent variable. At the same time, our key findings from the hierarchical time-series cross-sectional models remain unchanged. This confirms that utilising the number of hospital beds (per 1000 inhabitants) is not only conceptually, but also empirically the better choice (see Engler et al. 2021;Toshkov et al. 2022 for a similar approach). 11. Geography has established that different 'island definitions' can be discerned. Measured against the standards and state of the art in the field of geography, we use the criterion of evolutionary-tectonic divergence to measure an 'island nation' as a tract of land surrounded by water, split off from land masses (i.e. Australia, Ireland, Japan, New Zealand, UK). To account for additional factors that may influence excess mortality, we have tested for, among others, population density (people per km 2 ), as SARS-CoV-2 transmission is potentially more likely to occur in cities with higher population densities, GDP per capita, the share of people aged 65 or older and whether a national election was held. None of these indicators reached conventional levels of statistical significance, nor did they affect our main empirical results (see online appendix). 12. We also estimated models with a metric variable with and without a random (i.e. country-specific) slope, but the fixed-effect approach proved to better fit the data. Estimations using different time specifications (e.g. time trend, with and without random slope, lagged dependent variable) as well as the respective model fit (ANOVA) are reported in the online appendix. 13. A note on Sweden is in order: for the first months of the global pandemic, Sweden's unconventional approach seemed out of step with much of the world, given its voluntary, rather than mandatory, measures. While these lax policies resulted in one of the worst per capita COVID-19 mortality rates in Europe in the first wave(s), Sweden's strategy in the later stages of the pandemic 'no longer [stood] out' (Financial Times, 12 November 2021). 14. Retrieved from https://www.ghsindex.org/report-model/. 15. Time, 1 July 2020. 16. Ibid. 17. That said, some of the majoritarian democracies in the English-speaking world fared much better than the UK and the US, notably Canada, Australia and New Zealand (see Bromfield and McConnell 2021). 18. Even if the 'federalism' variable of Lijphart (2012) and Bernauer and Vatter (2019) is factored in individually (instead of the aggregate 'federal-unitary' score), it remains insignificant. The RAI score by Hooghe et al. (2016) is also insignificant. 19. The New York Magazine, 1 April 2020. 20. Countries where linguistic cabinet parity is prescribed (e.g. Belgium) and/ or the number of ministers is constitutionally enshrined (e.g. Switzerland) represent an exception.
21. Note that we have argued that the number of ministers may reach a certain tipping point, as cabinets that are too large may lead to performance-inhibiting coordination problems. Further analyses have shown that the squared number of ministers is not significant.