2. Crisis and Health Policy in the EU

The crisis had a significant impact on EU health policy and public health. Unemployment, poverty, and income equality following economic recession had a negative impact on the latter, including rising suicide rates, illnesses, and deaths (Stuckler et al., Reference Stuckler, Basu, Suhrcke, Coutts and McKee2009, De Vogli, Reference De Vogli2014). This mainly resulted from a destabilized economy and job losses, accompanied by public frustration (Catalano, Reference Catalano2009). While austerity measures presented a frequently adopted but failed strategy in Europe to address this challenge and stabilize economic growth (McKee et al., Reference McKee, Karanikolos, Belcher and Stuckler2012), “cuts to public spending on health made in response to an economic shock typically come at a time when health systems may require more, not fewer, resources” (Mladovsky et al., Reference Mladovsky, Srivastava, Cylus, Karanikolos, Evetovits, Thomson and McKee2012b, 1). Consequently, government strategies in health expenditures are crucial for public health and social protection.

In times of crisis, public health expenditures increase less than in times of economic stability. Although this effect is a statistical average and not yet explicitly related to developments in single countries, the general tendency is that social insurance health care systems rather rely on cost-shifting towards private spending while tax-based systems employ a range of policy instruments to cut spending (Cylus et al., Reference Cylus, Mladovsky and McKee2012). Beyond overall analyses of post-crisis developments in the EU, single case studies explain the mechanisms of health policy responses (policy instruments) in the aftermath of crises. These also confirm that EU countries largely reacted to the crisis by reducing public spending the health sector (Quaglio et al., Reference Quaglio, Karapiperis, Van Woensel, Arnold and McDaid2013, McDaid et al., Reference McDaid, Quaglio, de Campos, Dario, Van Woensel, Karapiperis and Reeves2013, Merkur et al., Reference Merkur, Maresso and McDaid2012): for example, but not exclusively, Italy (Belvis et al., Reference Belvis, de, Ferrè, Specchia, Valerio, Fattore and Ricciardi2012), Greece (Kentikelenis et al., Reference Kentikelenis, Karanikolos, Reeves, McKee and Stuckler2014), and Ireland (Burke et al., Reference Burke, Thomas, Barry and Keegan2014).

One could therefore expect that the figures or at least the trends in EU countries regarding health expenditures are quite similar. For example, Greece, Italy, and Ireland are severely crisis-affected countries and in analyses of post-crisis health expenditure patterns put under the same label (Keegan et al., Reference Keegan and Thomas2013). However, if we look at the detailed numbers and actual changes in expenditures, the developments strike as quite different or even opposite. This presents an empirical puzzle for public policy scholars who are interested in pinpointing explanations for national and cross-national policies. Figure 1 visualizes this puzzle by presenting different measures of health expenditure developments (total, public, and private; in PPP (current international $) and current US$).

FIGURE 1. Health Expenditure Developments in EU Countries, 2008-2010

For example, if we compare Greece and Ireland, both intensively moderated by the international community and considered as victims of the crisis, we observe divergent adjustment in the share of public and private health expenditures. The countries also show divergent developments in health expenditures for both measures of health expenditures per capita in PPP or current US$. Looking at the measures in current US $, it is interesting to see that, across countries, there is a general decline in health expenditures, which most likely is rooted in the fallen exchange rate of the Euro compared to the US. Comparing the PPP measure of total health expenditures, according to which those of Ireland increased and those of Greece declined, the public health expenditures in Greece still increased and more so than in Ireland, where the public share of health expenditures diminished strongest within the EU. Different impacts of the crises on the Greek and Irish health sector have been studied so far, with reference to different developments of public and private shares but without reference to explanations of these divergent expenditure patterns (Hessel et al., Reference Hessel, Vandoros and Avendano2014, Burke et al., Reference Burke, Thomas, Barry and Keegan2014, Ifanti et al., Reference Ifanti, Argyriou, Kalofonou and Kalofonos2013). Trends in post-crisis health care expenditures measured in PPP therefore present an interesting and valid indicator for nation-specific developments in health policy, which are to be explained in this article. What explains these different developments of public expenses in at first sight similar countries? Or did every country just witness a country-specific process?

The research literature attributes declining public health expenditures to symbiotic negative financial developments, such as borrowing from the Monetary Fund, rising debt, or a tax-financed health care system (Reeves et al., Reference Reeves, McKee, Basu and Stuckler2014). Studies thereby often focus on one measurement of changing expenditures like shifting percentages of total government expenditures or total health expenditures. In welfare state research on public policy, the development of shares of public spending is a central interest (Esping-Andersen Reference Esping-Andersen1990, Häusermann et al., Reference Häusermann, Kurer and Traber2018). What determines public spending in the welfare state is however a contested question. There is both supportive and disclaiming evidence for the role of partisanship in extending the welfare state (Falkenbach et al., Reference Falkenbach, Bekker and Greer2019, Green-Pedersen and Jensen, Reference Green-Pedersen and Jensen2019), convergence trends across Europe towards competition and user-empowerment (increase of private spending) (Pavolini and Ranci, Reference Pavolini and Ranci2008), and the role of institutions and path dependency (Del Pino and Ramos, Reference Del Pino and Ramos2018, Walter, Reference Walter2018). To proceed in this welfare state research, it is therefore indispensable to take into account or even combine several of these explanatory factors (Starke et al., Reference Starke, Kaasch and Van Hooren2014).

What this article adds to this research is the required more nuanced view on the dynamics of public health expenditures. Whereas a first intuitive look in the attempt to explain health expenditures would be devoted to the different policy instruments used by state government to react to the crisis, the research literature and an empirical review of policy instruments adopted in the EU countries allow only for an inconclusive judgment of how policy instruments affect levels of health expenditures. Instead, it is necessary to draw on theoretical perspectives that shed light on how changing expenditures patterns resulted from different political and institutional predispositions.

3. Theoretical Perspectives in Comparative Public Policy Research

Explanations for public policy outcomes rest on six main pillars that can be traced back to a range of public policy literature (Myles and Quadagno, Reference Myles and Quadagno2002, Wenzelburger et al., Reference Wenzelburger, König and Wolf2019). We start with socioeconomic parameters, which are the first explanatory factor for developments in social policy. Given the assumption that policies are responses to current social and economic developments, several studies use socioeconomic explanations for health policy, such as analyses of regional health expenditures in Canada (Di Matteo and Di Matteo, Reference Di Matteo and Di1998) or Switzerland (Crivelli et al., Reference Crivelli, Filippini and Mosca2006), stating that GDP per capita and the share of elderly people to the total population are important determinants for rising health expenditures. These considerations lead to four hypotheses on the influence of socioeconomic developments on health expenditures. When governments face financial constraints, they are expected not to increase public spending on health care. Financial constraints are operationalized with reference to the parameters of the Maastricht criteria (Soukiazis and Castro, Reference Soukiazis and Castro2005), including a decline in GDP and an increase in public debt and deficit. In addition, as unemployment rises, per capita public health expenditures are expected to decline due to lacking financial resources, both in the form of taxes and/or wage-based contributions to health care. However, if the proportion of elderly people increases, public health expenditures will rise because the proportion of the working population decreases and governments have to allocate more resources to their health care.

Turning to aspects of interest representation and influence on public policy, power resource theory gives explanations for why certain interest groups manage to put their preferences into policy. Organized interest groups with a high ability to deal with conflicts can influence policy decisions in the direction of their own interests (Rothstein et al., Reference Rothstein, Samanni and Teorell2012). In some countries, the systematic involvement of interest groups is much more distinct than in others; that is, the degree of corporatism in the country is much higher (Molina and Rhodes, Reference Molina and Rhodes2002). In such cases, interest groups are expected to influence policy-making in ways that benefit them. What benefits a sectoral interest group are increased financial resources, so that the corporatism hypothesis states:

Related to the power resource theory, the institutional design can be an important precondition for actors’ behavior and policy change. In health policy, various associations muster relevant actors of the health sector, among them doctors, hospitals and pharmacists. However, this is only one example of an institutional obstacle that governments might need to overcome. Institutional conditions are not necessarily of structural nature in the sense that legal rules constrain the scope of action, which a system perspective would propose; they can also be actor-related, thereby strongly oriented towards the concept of veto players (Tsebelis, Reference Tsebelis2002, Zohlnhöfer, Reference Zohlnhöfer2009). Institutional possibilities and constraints may furthermore be formal or informal. Comparative studies ascribe an important role to the explanatory power of institutional settings (De la Maisonneuve et al., Reference De la Maisonneuve, Moreno-Serra, Murtin and Martins2017), their interaction with other theoretical strands (Pierson, Reference Pierson1995), and of the distinction between majoritarian and consensus democracies (Schmidt, Reference Schmidt2002). In addition to that, a dominant government in a parliamentary system might face less opposition than a president in a presidential system.

Originating in macroeconomic theory (Downs Reference Downs1957, Hibbs Reference Hibbs1977), the partisan theory largely assumes that political parties act corresponding to their interest of office-seeking, which they achieve through votes, and that the vote-seeking behavior that they adopt because of their interests is guided by their political program in exchange with the preferences of their electorate (Zohlnhöfer, Reference Zohlnhöfer2003). Health policy is an important part of welfare state expansion or retrenchment, and left parties as well as their voters are traditionally not in favor of retrenchment (Allan and Scruggs, Reference Allan and Scruggs2004). However, research results indicate a contested influence of partisan effects on retrenchment or whether voters correspondingly adjust their voting behavior (Giger and Nelson, Reference Giger and Nelson2011). Although partisan effects have been found to be rather low and even vanish in the case of health policy, they might be relevant in combination with – as a moderator for or moderated by – other factors, such as Europeanization (Potrafke, Reference Potrafke2009). Controlling for institutional factors and provided that economic growth is sufficient, Sirén (Reference Sirén2020) finds that public expenditures still increase with left party governments. Thus, the following hypothesis is derived from the literature:

Sometimes, policies have no other reason than having been there all along. Governments and interest groups would like to alter them, socioeconomic developments seem to challenge them, institutional conditions would allow them but transaction costs are simply too high. In these cases, path dependency is the explanation for long-term policy stability. Obviously, it is suitable to explain maintenance instead of change. In health policy, path dependency has often been used to explain why health system reforms happen merely incrementally or, in other words, how it impairs big change (Wilsford Reference Wilsford1994). Besides the scope of reforms, case studies indicate the lack of efficiency and control of health care costs that result from path dependency (Bevan and Robinson, Reference Bevan and Robinson2005) but outline the structural stability it brings about (Vrangbæk and Christiansen, Reference Vrangbæk and Christiansen2005). Combined with other explanatory factors like decentralization and social power (Rico and Costa-Font, Reference Rico and Costa-Font2005) path dependency might even lose explanatory power. In order to evaluate whether a policy follows its previous path, which could also result in an enforcement of the previous strategy, researchers often identify the dominant path by referring to welfare state typology (Esping-Andersen, Reference Esping-Andersen1990, Marmor and Wendt, Reference Marmor and Wendt2012) presenting ideal types of welfare states that exist consistently over a long period of time. However, Marmor and Wendt (Reference Marmor and Wendt2012, 17) argue that it is rather the national health system being a crucial explanation for health spending and advocate the use of health system typologies for comparative studies instead of welfare typologies. In addition to that, many scholars have criticized the applicability of welfare state typology to Southern, Central or Eastern European countries (e.g. Castles, Reference Castles1995). This paper therefore includes the traditional Bismarck-Beveridge classification (Bonoli Reference Bonoli1997). Bismarckian systems are characterized by an insurance system based on earning-related contributions. Beveridge systems collect necessary contributions by taxes and distribute services via a national health service. The financial crisis leading to unemployment may create the need for further public expenditures in Bismarckian systems due to the cuts in contributions that are dependent on earnings. Beveridge systems may cut public health expenditures more easily (Reeves et al., Reference Reeves, McKee, Basu and Stuckler2014).

Equally belonging to institutional explanations of public policy is the typology labelled as Varieties of Capitalism (Hall Reference Hall2014, Hall and Soskice, Reference Hall, Soskice and Hancké2010) that distinguishes coordinated (CME) and liberal market economies (LME). While institutional systems may impede policy change, this is dependent on the interaction between institutions and politics (Hammond and Butler, Reference Hammond and Butler2003), calling for a combination of the institution hypothesis with other political explanations.

Since the European Union is sharing a common market, open market conditions are the same in every member state and therefore not perfectly suitable to capture differences in social policy outcomes. However, evidence shows that Europeanization induces fiscal restraints that force governments to cost-containment (Leibfried, Reference Leibfried, Wallace, Pollack and Young2015). Considering international fiscal constraints imposed on national governments, the financial crisis indeed provides a tying knot. Since this contribution highlights the short-term reactions of governments to the financial crisis and there was a high uncertainty with regard to its financial impact on national systems, it can be assumed that this uncertainty was constant across countries. As a result, it is concluded that this aspect can be kept constant in the analysis.

4. Fuzzy-Set Qualitative Comparative Analysis (QCA)

Fuzzy-set QCA provides the possibility to assess the configurational power of the individual hypotheses due to its specific strength to determine several explanations for the same outcome, called equifinality, and – as opposed to multi-value and crisp-set QCA – to provide a nuanced operationalization of degrees of membership in subsets and reveal even complex patterns of country responses immediately following crises (Saltkjel et al., Reference Saltkjel, Ingelsrud, Dahl and Halvorsen2017, Shahidi, Reference Shahidi2015). The main reason for applying QCA lies in the fact that existing research points towards complex interdependencies that effect post-crisis health policy and thus indicates that configurations of different conditions in their combination resulted in nation-specific strategies and outcomes. Identifying these configurations is the core research interest of this article. Thus, we are not interested in the average effect that e.g. left shares of government seats had on health expenditures across Europe but we are interested in the conditions under which (and in combination with which) left-wing participation in government enables health (spending) policies. Thereby, the analysis aims at understanding the immediate country-specific reactions and outcomes, because knowledge about each case is high, and at the same time tries to reveal systematic patterns that allow for prediction of government responses and outcomes in future moments of crisis. Lastly, QCA is most appropriate if samples are of medium size, such as in this case 28 EU countries.

In fuzzy-set QCA, outcomes and explanatory conditions for each case are coded as fuzzy values, i.e. the degrees of membership in subsets are distinguished stepwise between 0 (non-membership) and 1 (full membership). For analyses of post-crisis developments, an accurate time frame is indispensable. Scholars suggest the use of data from 2010 and 2011 for identifying short-term trends (Mladovsky et al., Reference Mladovsky, Srivastava, Cylus, Karanikolos, Evetovits, Thomson and McKee2012b, 11), but emphasize the need for a long-term observation of developments – up to the year 2014 (Del Pino and Ramos, Reference Del Pino and Ramos2018, Ronchi, Reference Ronchi2018). Given that this article’s focus lies in detecting the short-term health policy responses to crisis and uncertainty, the empirical analysis concentrates on the expenditures trends between 2008 and 2010. To provide a holistic picture and embed the results in a long-term perspective, however, we also perform a separate analysis for long-term reactions to account for the time needed by governments to react to developments initiated by the crisis.

The operationalization of the outcome and conditions and the assignment of fuzzy values are based on theoretical grounds. As concerns outcome, data on public health expenditure trends is derived from the WorldBank (2020). After an intense discussion of the dynamics of expenditures in section 2, the developments of public expenditures in PPP (current international $) are taken as outcome, to account for the fact that power purchasing parity has strongly changed in the course of the crisis and following imbalances in the Eurozone (Talani, Reference Talani and Talani2016). Such an operationalization profits from its independence of developments in dynamic public/private shares of total health expenditures, of GDP, and exchange rates in countries outside the Eurozone. The analysis of the changes in public health expenditures from 2008-2010 and from 2008-2014 is based on the percentage changes to adequately reflect proportionality, whereby the extent to which health expenditures increased is depicted by the fuzzy values 0.4 (less than 2.5 %), 0.6 (more than 2.5 % increase), 0.8 (more than 3 % increase), and 1 (more than 10 % increase), while the value 0 indicates a decrease in public health expenditures. Based on the striking upwards trend of developments of public health expenditures during the preceding decade, it can be theoretically argued that an increase of more than 2.5 % or 3 % is considered considerable and should be coded with a set membership in increasing health expenditures of more than 0.5. An increase of more than 10 % then depicts an enormous augmentation, thus coded as full set membership.

Regarding the explanatory conditions, the socioeconomic factors are also taken from the WorldBank (2020). Economic strength and problem pressure through financial constraints are measured by the developments of GDP, public debt and public deficit. Again, the percentage changes of these parameters are coded according to the theoretically derive hypotheses, which assume decreases in GDP (< 0 %) and increases in public debt and public deficit (> 0 %) to affect public health expenditures. The fuzzy values are then assigned in quarter steps (for the detailed coding of fuzzy values, see Table A2 in the appendix). As a last socioeconomic condition, it is coded whether the share of the elderly population has decreased (0) or increased (1). Corporatism values of 1 (highly corporatist) and 0 (highly pluralist) are taken from Lijphart (Reference Lijphart2012), Siaroff (Reference Siaroff1999), Ost (Reference Ost2000), Woolfson, Kallaste, and Bernzins (Reference Woolfson, Kallaste and Bernzins2011), Tanasoiu (Reference Tanasoiu2012), and Ioannou and Kentas (Reference Ioannou and Kentas2009). Operationalization of party strength is usually proxied by referring to the share of government seats in the years of analysis (Vatter and Rüefli, Reference Vatter and Rüefli2003). Due to the requirements of QCA, dominance of left-wing parties is operationalized by the threshold of seats that left parties held in cabinet in 2009 (for short-term responses) and on average between 2008 and 2014 (for long-term effects) (data taken from the Comparative Political Data Set (CPDS); Armingeon et al., Reference Armingeon, Isler, Knöpfel, Weisstanner and Engler2016). Institutional constraints are operationalized with reference to majoritarian/consensus and federalist/unitary democracies (Lijphart, Reference Lijphart2012, Roberts, Reference Roberts2006) and varieties of capitalism (Knell and Srholec, Reference Knell, Srholec, Lane and Myant2007, Pegasiou, Reference Pegasiou2013). Path dependency is – as already discussed – operationalized by the dichotomous classification of Bismarck (coded as 0) and Beveridge (coded as 1) types) (Bonoli, Reference Bonoli1997, Schubert et al., Reference Schubert, de Villota and Kuhlmann2016), but also by the trend of public health expenditures in the previous decade (from 2000-2008). At this point, it is striking that such an operationalization of path dependency reveals that the general trend of public health expenditures across the EU is positive, and that public health expenditures decreased in none of the analyzed countries. Table A3 in the annex depicts the final truth table (including theoretically derived thresholds for fuzzy values), which is used for the QCA.

5. Configurations of Post-Crisis Responses Towards Public Health Expenditure Patterns

Following the standards of good practice for QCA (Schneider and Wagemann, Reference Schneider and Wagemann2010), each theoretically derived condition is firstly tested as a necessary condition before a truth table algorithm attempts to identify configurational conditions that are sufficient for a certain outcome (i.e. public health expenditure development). This procedure repeats itself for the two time periods under study (2008-2010 and 2008-2014), keeping in mind that the original research interest of this article lies in detecting short-term reactions to the crisis and that the second analysis serves as a re-evaluation of the empirical results and theoretical argument derived. All analyses are performed by means of the software fs/QCA developed by Ragin (Reference Ragin2016). The tables with detailed values of the analysis of necessary conditions are annexed (table A4 and A5), including also the test for negated conditions and including both periods under study (2008-2010 and 2008-2014). Only one condition, the development of the share of elderly people, fulfills the consistency criterion of being considered as a necessary condition. Consequently, this condition will be considered together with other conditions near the threshold (particularly unemployment and public debt and left parties in government) for the later truth table analysis by means of the Quine-McClusky algorithm.

Nevertheless, a trial-and-error procedure of including theoretically relevant conditions and excluding empirically repeatedly irrelevant conditions will also test the conditions derived from the various public policy perspectives to make sure that potential configurations of conditions are uncovered. For this purpose, the conditions are tested for potential configurational sufficiency. The results show that the influence of political and socioeconomic conditions on post-crisis health responses does indeed depend on institutional conditions, although these could not be identified as necessary conditions. As a result, the following solution formula emerges as both the most theoretically logical and the empirically parsimonious solution.

While Table 1 depicts the final solution, formula based on the truth table algorithm, which was selected due to its parsimony, consistency and coverage values, and theoretical soundness, Figure 2 assigns all countries according to their outcome and value in the respective solution conditions. In the sense of QCA, the solution formula(s) provide(s) a systematization of configurational explanations of single cases but ask(s) for a qualitative discussion of each case with the goal of revealing the mechanisms per country. The discussion of single cases regarding their empirical configuration of conditions presents a regular exchange of qualitative and quantitative evaluation and eventually a bigger picture of post-crisis health policy. Summarizing these findings results in several interesting statements.

TABLE 1. Solution Formula: Substantial Increase in Public Shares of Health Expenditures 2008-2010 (in PPP, current international $), (1= more than 10 percentage, 0.8= more than 5 percentage, 0.6= more than 4 percentage)

Source: Formatted QCA output; solution consistency: 0.84; solution coverage 0.89

FIGURE 2. Visualization of Outcomes According to Solution Formula (Table 3)

Source: Own Depiction Based on Truth Table

Firstly, the combination of left party membership in government and coordinated market economy is a sufficient condition for an increase in public health expenditures. Furthermore, in almost all cases where public health expenditures increased substantially (more than 10 percent, countries in bold and underlined), left parties hold seats in the governmental cabinet. This makes left party membership in government almost a necessary condition for these major increases in public health expenditures. 23 countries in total show a greater membership in the rise of public health expenditures over 2 percent. 19 of them had left partisan influence in government. The other four, Denmark, France, Greece, and Malta, had a right-wing government without participation of left parties but coordinate market economy emerged as an explanatory condition here. In Denmark, the influence of long-standing trade unions favors rising public spending on social policy, which in a coordinated market economy leads to crisis reactions that increase public health expenditures. In France, the right-wing government is in fact statist and in the course of a programmatic group acting in this period of analysis (Genieys and Hassenteufel, Reference Genieys and Hassenteufel2015), the strengthening of public involvement in the health sector can be explained by a condition that has not been included theoretically in this paper, as it must be identified in in-depth qualitative field work including expert interviews. Malta is often excluded from analyses of welfare state policies due to its special status and size. Therefore, it presents a special case here, which should be analyzed again in a qualitative case study. Greece has been affected most by the crisis and therefore needed to increase public health expenditures to cope with the heavy fall in private expenditures. This was a reaction made necessary by the severe impact of the crisis.

Secondly, despite not being necessary conditions for the outcome, both left party participation in government and coordinated market economies generally lead to the outcome. The exceptions, in which either one is present and the other is absent but the outcome is absent, too, are Estonia, Slovenia, and Sweden. Like Denmark, Sweden has a history of strong trade unions but it had just experienced the electoral success of a right-wing government after a long tradition of left-wing governments. Thus, the slow increase in public health expenditures (1.6 percent; +) compared to the other EU countries seems to be explained by a clear partisan turn in government, thus confirming the “parties-do-matter” hypothesis (Zohlnhöfer, Reference Zohlnhöfer2009) within a coordinated market economy that guards the social security system to a certain extent from austerity measures in times of crisis. Estonia aimed at upholding primary care for as many citizens as possible, by extending entitlement and relying on insurance funds reserves and newly introduced taxes for financing, which corresponds to left-wing preferences and policy-making. To generally decrease levels of spending, price reduction for pharmaceutical products and incentives for the cheaper outpatient care as opposed to inpatient care were promoted, by e.g. reducing user charges except for inpatient and specialist care. The multi-party minority government aiming at achieving consensus in crisis reactions enabled the government to perform changes across all areas without significantly augmenting public expenses for health. In Slovenia, the crisis led to a policy gridlock, during which structural health reforms failed to be adopted and general problems of mutual trust hindered the political activity towards increasing public spending on health (Filipovič Hrast and Kopač Mrak, Reference Filipovič Hrast, Kopač Mrak, Schubert, de Villota and Kuhlmann2016).

Lastly, all countries which neither have left party participation in government nor an institutional frame of coordinated market economy did not experience increases in public health expenditures over 2 percent. This underlines the finding that a combination of partisanship and institutional constraints enables public spending in times of crisis. The cases of Sweden, Slovenia, and Estonia emphasize, however, that it is indispensable to qualitatively look at single cases if these are not meeting the identified conditions configurations and the relating outcome.

If the same analytical steps are carried out for the period between 2008 and 2014, different results come to light. Institutional conditions like the type of health systems, market economies, and the dichotomy of majoritarian and consensus democracies slightly increase in their consistency values as necessary conditions. Conducting an in-depth case analysis confirms some postulated mechanisms, e.g. that public health expenditures remain stable during crises and thus are not subject to austerity measures of Bismarckian health care systems (Hassenteufel and Palier, Reference Hassenteufel and Palier2007). Indeed, one important limitation of the empirical results presented here is that they cannot be generalized for long-term policy responses and developments over several years after a crisis outbreak, as other studies emphasize (Petmesidou et al., Reference Petmesidou, Pavolini and Guillén2014).