Varieties of Welfare Capitalism in Crisis: A Qualitative Comparative Analysis of Labour Market Reforms in 18 Advanced Welfare States

Varieties of Welfare Capitalism in Crisis: A Qualitative Comparative Analysis of Labour Market Reforms in 18 Advanced Welfare States Paper presented to the annual conference of the Sheffield Political Economy Research Institute, University of Sheffield, July 1-3. Faraz Vahid Shahidi faraz.vahidshahidi@spi.ox.ac.uk Department of Social Policy and Intervention University of Oxford

ABSTRACT Comparative welfare state scholars have long been concerned with explaining the nature and direction of cross-national welfare state developments. This paper contributes the literature on welfare state change by providing an empirical survey of and explanation for labour market reforms enacted across a sample of eighteen advanced welfare states during the period following the economic crisis that began in 2008. Drawing on a set-theoretic multi-method research design, the paper utilizes crisp-set qualitative comparative analysis to systematically test existing theories of welfare state change. The paper finds that the direction of cross-national labour market reforms implemented since the onset of the economic crisis can be explained principally in terms of the variable fiscal capacity of the state. In particular, the results indicate that (i) the presence of fiscal crisis has acted as a necessary (but insufficient) condition for the presence of recommodification; and (ii) the absence of fiscal crisis has acted as a sufficient (but unnecessary) condition for the absence of recommodification. In addition, the findings suggest that mainstream theories of welfare state change do not provide a compelling explanation for the cross-national diversity of contemporary crisis management.

I. INTRODUCTION The economic crisis that began in 2008 has provided students and scholars of the welfare state with the opportunity to study the political consequences of severe crisis precisely as they unfold. Advanced political economies have reacted to the recent economic crisis and its associated aftershocks by embarking upon one or another deliberate strategy of crisis management. By crisis management, I mean the variable policy responses that governments formulate in an effort to cope with crisis consequences. Modern welfare states have played a central role in the development and operation of architectures of crisis management. The specific contents of these architectures have varied considerably across space and over time (Farnsworth and Irving, 2011; Pontusson and Raess, 2012; Starke et al., 2013). It is therefore not surprising that contemporary crisis management has manifested itself in an uneven fashion across the landscape of advanced welfare states. Cross-national differences have persisted both in the degree to which countries have been exposed to the crisis, as well as in the responses they have formulated with which to confront it. Thus, despite the international scope of the recent economic crisis, its political consequences have taken a distinctly national form. This paper is concerned with explaining the cross-national variation in contemporary crisis management. It employs qualitative comparative analysis to investigate patterns of recommodification in recent labour market reforms across a sample of eighteen advanced welfare states. The central research questions are as follows. (i) How can we explain patterns of welfare state recommodification in the labour market policy trajectories of advanced welfare states during the aftermath of the recent economic crisis? (ii) What do contemporary patterns of recommodification suggest about the comparative political economy of the welfare state and, in particular, its presumed resilience in the face of present and future challenges. At its core, this paper constitutes an inquiry into the range of factors that have shaped and constrained the observed diversity of contemporary crisis management. In the process of providing an explanation for recent labour market reforms, however, it also offers a preliminary assessment of the extent to which the recent economic crisis has transformed some of the core architectures of advanced welfare states. A sample of eighteen advanced welfare states is included in the paper: Australia, Austria, Belgium, Canada, Denmark, Finland, France, Germany, Greece, Ireland, Italy, the Netherlands, New Zealand, Norway, Portugal, Spain, Sweden, and the United Kingdom. A number of advanced welfare states are excluded. Luxembourg and Iceland are excluded on the basis of restrictions on the availability of data. Switzerland and Japan have been

excluded on the basis that their unique systems of social concertation provide an inadequate foundation for meaningful comparison (Siaroff, 1999). Finally, the United States is excluded because of the presence of significant variation in the direction of welfare state change between individual states. II. EXPLAINING WELFARE STATE CHANGE DURING HARD TIMES Comparative welfare state scholars have long been concerned with explaining the nature and direction of cross-national welfare state developments. The existing literature offers a number of theoretical frameworks with which to attempt to explain welfare state change. According to a first theoretical framework, the welfare state is first and foremost a mechanism of social compensation that is presumed to expand (contract) whenever and wherever there is greater (lesser) need for it (Glatzer and Rueschemeyer, 2005). Advocates of this social compensation hypothesis suggest that socio-economic challenges be it globalization (Garrett, 1998), deindustrialization (Iversen and Cusack, 2000), or economic crisis (Vis et al., 2011) create a functional demand for welfare state expansion. Governments are expected to respond to an economic crisis by adjusting the supply of social welfare in order to reflect the objective need for it. Not all welfare states are created equal, however. Rather, the functional need for social compensation is predicted to be greatest in welfare states that are least equipped to cope with the socio-economic consequences of crisis. Accordingly, it is argued that the relationship between an economic crisis and welfare state expansion is moderated by the generosity of existing welfare state measures (Ramesh, 2009). We would therefore expect welfare state expansion to be most likely in small welfare states that lack strong automatic welfare state stabilizers. By contrast, we would expect the functional need for compensation to be the weakest where large automatic stabilizers provide a by and large adequate buffer to the crisis and its consequences. The social compensation hypothesis suffers from a number of theoretical and analytical problems. Chief among these is its functionalist tendency to rely somewhat singularly on impersonal socio-economic forces to explain cross-national patterns of welfare state change (Myles and Quadagno, 2002). The approach lacks a necessary attentiveness to the political and institutional factors that contextualize these forces. Partly in response to the inherent functionalism of the social compensation hypothesis, a second theoretical framework that is sensitive to political conflict has approached the question of welfare state change through the lens of partisan politics (Korpi and Palme, 2003). According to this second approach, the policy objectives of political parties differ systematically according to the perceived interests of their respective constituents. Consequently, even when faced with

similar socio-economic circumstances, Left and Right parties are expected to display different preferences vis-à-vis the nature and direction of welfare state reform. Empirically speaking, this partisan politics hypothesis has been particularly successful at explaining the variegated consolidation of modern welfare capitalism during the immediate post-war period (Huber and Stephens, 2001). By contrast, considerable debates have emerged with regards to the salience of partisan politics during the ongoing era of permanent austerity (e.g. Hausermann et al., 2013; Kittel and Obinger, 2003). While very little literature has addressed this topic in the specific context of severe economic crisis, the literature that exists has given rise to inconsistent findings: while some observe the persistence of partisan effects during economic downturns (Cusack et al., 2008), others report its diminution (Lipsmeyer, 2011). There are important theoretical reasons to expect partisan politics to retain their significance during periods of economic crisis. While it is true that economic crises affect large portions of the population, they tend to affect the traditional constituencies of left and right parties differently. Thus, the strategies of crisis management they formulate can be expected to differ systematically. An alternative formulation, which we refer to as the constrained partisanship hypothesis, suggests that partisan differences are contingent upon the institutional context in which the crisis unfolds (Starke et al., 2013). More specifically, the salience of partisan politics during a period of economic crisis may depend on the strength of existing welfare state measures. According to this third hypothesis, partisan differences are most likely to appear under conditions where the weakness of automatic welfare state stabilizers demand that governments undertake active forms of crisis management. By contrast, in cases where strong automatic welfare state stabilizers serve as routine buffering mechanisms, passive crisis management may suffice. The result is the depoliticization of crisis management and a concomitant decline in the relevance of partisan politics and any policy differences that it might entail. In other words, the presence (absence) of a partisan effect on crisis management may reflect the absence (presence) of strong automatic welfare state stabilizers. The observation that partisan politics have experienced somewhat of a decline during the ongoing era of permanent austerity has led to the gradual, if partial, displacement of arguments about the old politics of the welfare state by those addressing its new politics (Pierson, 2001). According to this fourth theoretical framework, the welfare state has become an irreversible and immovable feature of contemporary capitalism. Advocates of this new politics hypothesis advance the argument that the welfare state has generated institutional and political deterrents to retrenchment that make it highly resilient to change. Institutionally

speaking, welfare states are said to create increasing returns whose positive feedbacks generate a stability bias in favour of continuity. In addition, the existence of broad welfare state constituencies is said to conspire against an unavoidably unpopular politics of retrenchment. On this basis, those who adhere to the new politics perspective argue that, even in the context of a severe economic crisis, welfare state retrenchment is difficult, if not impossible, to undertake (Kuhnle, 2000; Pierson, 1996). Accordingly, they expect that the radical curtailment of social welfare is unlikely to figure as a central feature of crisis management. Within the context of the emergence of the new politics of the welfare state, a fifth theoretical framework has emphasized the role of institutions in shaping the pathdependent contours of cross-national welfare state change. Institutionalist scholars have brought attention to the fact that, even in the face of a common challenge, welfare states will embark upon different and often divergent trajectories of reform (Esping-Andersen, 1996; Hall and Soskice, 2001; Iversen and Wren, 1998; Scharpf and Schmidt, 2000). Proponents of this path dependence hypothesis suggest that welfare states differ systematically in their logic of operation and, consequently, in their responses to a shared set of circumstances. In this sense, an economic crisis is presumed to generate distinct and pathdependent varieties or worlds of crisis management (Chung and Thewissen, 2011; Iversen, 2007; Lallement, 2011; Wood, 2001). Welfare states are expected to fall back on their established habits and exhibit regime-specific patterns of adjustment whose contents correspond neatly to the logic of pre-existing institutional configurations. Specifically, liberal market economies are expected to rely on market-oriented reforms and a strategy of deregulation. By contrast, non-liberal market economies are expected to reinforce the established principles of the welfare state in order to bolster the advantages that these are thought to confer. In summary, an economic crisis is not only predicted to generate a variegated set of responses, but the ex-post diversity of welfare state responses is explained primarily in terms of the ex-ante diversity of distinct varieties and worlds of welfare capitalism. Departing considerably from the expectations of the new politics and path dependence hypotheses, a final approach to welfare state change has argued that economic crises can cause the erosion of institutional and ideological constraints and give rise to abrupt and transformative social change. Advocates of this fiscal crisis hypothesis argue that economic crises have the capacity to interrupt and undermine the established principles of the welfare state, revive and reinforce earlier concerns about its long-term viability, and provide the necessary political and economic foundations for radical welfare state retrenchment (Clasen

et al., 2012; Vis, 2009; Hay and Wincott, 2013; Streeck, 2011). In other words, economic crises are critical windows of opportunity during which political actors are awarded the necessary capacity to enact reforms that would, under different circumstances, be difficult if not impossible to undertake. Economic crises, so the argument goes, have the potential to fundamentally alter the structural opportunities and constraints that governments perceive themselves to be facing (Bieling, 2012). Significant problem pressures that result from rapidly rising levels of unemployment and declining levels of fiscal stability can generate perceptions real or imagined of a fiscal crisis of the welfare state (c.f. O Connor, 1973). The expectation is that governments will confront such an emergency by embarking upon an agenda of fiscal consolidation whose implications for the welfare state are likely to be dramatic. In summary, a range of theoretical frameworks exists with which to attempt to explain contemporary welfare state change. However, with some notable exceptions, individual frameworks and their corresponding hypotheses tend to highlight one or another determinant of welfare state change to the exclusion of all others. Nevertheless, the complexity of contemporary crisis management is likely to render each of them less than universal. Thus, while an individual framework may prove particularly useful in explaining one or another set of cases, it is unlikely that any of them supplies an adequately dynamic causal formulation with which to tackle broader patterns of cross-national variation. In other words, each theoretical framework is unlikely to provide anything more than a partial explanation for contemporary patterns of recommodification. A central goal of this paper is to use available methodological tools to confront the observed complexity of crisis management with the aim generating a more comprehensive account of its welfare state consequences. The next section of the paper presents qualitative comparative analysis as a method of comparative inquiry that is particularly suited for the analysis of causal complexity across an intermediate-size sample of cases. III. THE STRATEGY OF QUALITATIVE COMPARATIVE ANALYSIS Comparative scholars often face the task of meeting two seemingly contradictory goals: breadth of study and depth of study (Ragin, 1987). On the one hand, the comparativist is interested in undertaking extensive research that is capable of establishing generalizable conclusions across a wide sample of cases. On the other hand, the desire for breadth must be balanced against the demand for intensive research that is capable of attending to the complex and substantive relations that exist between the causal contexts underlying a given case. These seemingly contradictory tendencies have given rise to deep methodological

cleavages within the comparative social sciences; namely, between quantitative (i.e. extensive) and qualitative (i.e. intensive) strategies (Brady and Collier, 2004; King et al., 1996). Quantitative strategies in the comparative social sciences are variable-oriented and rely on the application of formal statistical techniques for the study of large samples of data. Despite having the capacity to provide researchers with highly generalizable results, variable-oriented strategies are not well-equipped for tackling the problem of causal complexity. In particular, they employ an additive approach to causation whose assumptions rely on the measurement of independent net effects. This approach the substantive relations that exist between causally relevant conditions (Ragin, 2008). Qualitative strategies, on the other hand, are case-oriented and facilitate the in-depth analysis of small-size samples of cases whose complexities are made apparent through the application of more intensive forms of inquiry. Despite offer a more holistic understanding of cases, case-oriented strategies rely on quasi-experimental designs that are limited in their ability to accommodate for causal complexity (Mahoney, 2000). Quasi-experimental research designs involve the systematic comparison of cases with the aim of attributing causation to a single condition (or combination of conditions). These designs are somewhat extreme to the extent that they overlook the possibility that multiple causal recipes can give rise to the same outcome (George and Bennett, 2005). In summary, both variable-oriented and case-oriented strategies are informed by assumptions that undermine their sensitivity to causal complexity. Accordingly, neither set of strategies appears to be particularly appropriate means through which to tackle the present research problem. Rather, an altogether different strategy is required whose epistemological and technical foundations are particularly suited for the study of causal complexity across an intermediate-size sample of cases. Among the methodological tools available to comparative social scientists, qualitative comparative analysis (QCA) represents a powerful instrument for the analysis of causal complexity (Rihoux and Ragin, 2009). As an approach, QCA draws on the extensiveness of variable-oriented strategies and the intensiveness of case-oriented strategies. By drawing on their respective strengths, the method facilitates a shift towards broader and more analytic questions while simultaneously maintaining a holistic understanding of cases. QCA therefore embodies a unique sensitivity to both theoretical generality and case complexity. As a formal technique of analysis, QCA is a configurational comparative method that

employs the logic of Boolean algebra and set theory to implement principles of comparison across small- and intermediate-size samples of cases. According to the procedure of crisp-set qualitative comparative analysis (csqca), the specific variant of QCA employed in this paper, each case is represented configurationally as a combination of causally relevant conditions whose variable presence or absence is said to result either in the occurrence or the non-occurrence of a given outcome of interest, such as recommodification. Cases and their corresponding configurations are systematically compared in order to uncover patterns of similarity and difference between them. Through the application of logical comparison, the method simplifies complex data structures with the overall aim of establishing the conditions of occurrence and non-occurrence of the outcome of interest (i.e. recommodification). Epistemologically speaking, configurational comparative methods rely on a conception of causality that is explicitly sensitive to causal complexity. csqca and related methods operate through the lens of multiple conjunctural causation (Berg-Schlosser et al., 2009). Multiple conjunctural causation implies a number of assumptions. First, it implies that there are multiple distinct causal pathways that can lead to the same outcome. Second, it implies that outcomes are often best explained by a combination of causally relevant conditions, rather than by any single condition alone. Because they are likely to offer meaningful explanation only when they are interpreted in conjunction with one another, the analysis of combinations of causally relevant conditions takes precedence over that of any individual condition. Finally, the concept of multiple conjunctural causation accounts for the possibility of causal asymmetry; that is, it does not assume a priori that the occurrence and the non-occurrence of a given outcome of interest is explained by the same causally relevant conditions. Partly as a result of its sensitivity to causal complexity, QCA and related configurational comparative methods are increasingly viewed as appropriate methodological instruments for comparative welfare state research (Emmenegger et al., 2013). IV. OPERATIONALIZATION AND CALIBRATION Recommodification Recommodification derives its meaning from the corollary concepts of commodification and decommodification. Commodification refers to the condition of labour power under capitalism; namely, as a commodity whose sale in the labour market becomes

a necessary pre-requisite for securing the means of subsistence (Room, 2000). Accordingly, decommodification is said to occur when a person can maintain an adequate standard of living without having to rely on the sale of their labour power. Decommodification is therefore used as a term to describe the process whereby the means of subsistence is rendered as a matter of right, rather than as a matter of one s performance in the labour market (Esping-Andersen, 1990). By contrast, recommodification describes the process whereby state-market relations are reconfigured in such a way as to imply a greater reliance on or heightened exposure to the market as a prerequisite for securing the means of subsistence (Holden, 2003). In other words, one s ability to secure these means depends to a greater extent upon one s participation in the labour market. Thus, while decommodification aims to protect citizens from the market, recommodification aims precisely at orienting them towards it. Recommodification has figured as a central feature of recent welfare state transformations and, in particular, the emergence of contemporary activation regimes (Peck, 2001). Recommodification is operationalized along three dimensions of labour market policy change. The first dimension of recommodification accounts for labour market policy change that reinforces the obligation to work as a condition of unemployment protection. The reinforcement of work-related obligations implies a process of recommodification insofar as it strengthens the emphasis on paid employment as a pre-requisite for securing the means of subsistence (Clasen and Clegg, 2007). In the analysis, work-related obligations are said to have been reinforced when (i) mandatory participation in job search activities as a condition for the receipt of unemployment benefits is introduced or intensified; (ii) mandatory participation in active labour market policy schemes as a condition for the receipt of unemployment benefits is introduced or intensified; (iii) the definition of what constitutes a suitable job offer is tightened; or (iv) punitive sanctions resulting from the failure to comply with benefit conditionalities are introduced or intensified. The second dimension of recommodification accounts for labour market policy change that reduces the generosity of unemployment protection. By reducing the generosity of unemployment protection, one restricts alternatives to paid employment as a pre-requisite for securing the means of subsistence (Pierson, 2001; Van Vliet, 2010). In the analysis, the generosity of unemployment protection is said to have been reduced when (i) the level of unemployment benefits is reduced; (ii) the maximum duration of unemployment benefits is reduced; (iii) access to unemployment benefits is reduced through the tightening of qualifying conditions or the lengthening of the waiting period before benefits can be claimed.

The first two dimensions of recommodification account for the variable intensity of recommodification experienced by those who are not engaged in paid employment. A third dimension of recommodification has been included in order to account for the variable intensity of recommodification experienced by those who are engaged in paid employment. This final dimension of recommodification accounts for labour market policy change that reduces the strictness of employment protection. The analysis focuses on changes to the regulatory architectures that govern the parameters of lawful and unlawful dismissals. Accordingly, employment protection is said to have been reduced when (i) the parameters of lawful dismissal is relaxed; or (ii) the punitive consequences associated with the use of unlawful dismissal are relaxed. Notably, employment protection and additional labour market regulations are by and large neglected in existing indices of decommodification (e.g. Esping-Andersen, 1990; Scruggs and Allan, 2006). Nevertheless, they are centrally linked to the experience of commodification and are therefore crucial elements in its reconfiguration vis-à-vis the welfare state (Papadopoulos, 2005). A comprehensive review of labour market reforms enacted across the eighteen advanced welfare states between January 2008 and March 2013 is conducted in order to collect the necessary data on labour market policy change. Temporary labour market reforms are excluded from the analysis on the basis that the aim of the study is to generate insight into the substantive impact of the economic crisis on the long-term trajectories of advanced welfare states. The data were derived from two databases of the European Commission (2013a/2013b) and a database of the International Labour Organization/World Bank (2013). These sources were supplemented with relevant case-based knowledge where such information was available in the existing literature. Table 1 summarizes the results of the comprehensive review of recent labour market reforms across the eighteen advanced welfare states. A more detailed outline of the reforms is provided in Appendix A. The results of the comprehensive review indicate that fourteen of the eighteen cases included in the study enacted at least one measure of recommodification between January 2008 and May 2013. These findings suggest that recommodification may not be as difficult as is often suggested by advocates of the new politics hypothesis (c.f. Pierson, 2001; Kuhnle, 2000). Among the fourteen cases that enacted measures of recommodification in their respective labour market policy architectures, considerable empirical diversity is observed. Cross-national variation is observable both in the relative scale of labour market reforms, as well as in the specific policy instruments they have targeted. A detailed comparison of the

labour market reforms reveals that measures belonging to the first dimension of recommodification have been the most common variety of policy change. Work-related obligations in national unemployment protection systems were reinforced in twelve of the eighteen advanced welfare states. They were reinforced through the intensification of existing work-related obligations (e.g. Australia, Belgium, Canada, Denmark, Finland, and Portugal); through their extension to additional categories of benefit claimants (e.g. France, Ireland, and New Zealand); through the implementation of new work-related obligations (e.g. Belgium, France, Italy, and Spain); or through the implementation of new punitive sanction measures (e.g. Finland, Ireland, Italy, and the United Kingdom). In addition to work-related obligations, nine of the eighteen cases enacted reforms that reduced the overall generosity of national unemployment protection systems. Generosity was reduced through the reduction of benefit levels (e.g. Belgium, Greece; Ireland, Portugal, Spain, and the United Kingdom); through the reduction of the duration of benefits (e.g. Denmark, France, Ireland, and Portugal); through the abolition of benefit schemes (e.g. Germany); through the tightening of qualifying conditions (e.g. Denmark and Ireland); or through the extension of the waiting period before which benefits can be claimed (e.g. Belgium). Table 1 Overview of Labour Market Reforms, 2008-2013 Change that reinforced the obligation to work Change that reduced the generosity of unemployment protection Change that reduced the strictness of employment protection Australia X Austria Belgium X X Canada X Denmark X X Finland X France X X X Germany X Greece X X Ireland X X Italy X X Netherlands New Zealand X Norway Portugal X X X Spain X X X Sweden United Kingdom X X Sources: European Commission (2013a/2013b); International Labour Organization/World Bank (2013).

Finally, six of the eighteen cases reduced the strictness of employment protection by loosening the legal parameters of lawful individual dismissal (e.g. France, Greece, and Spain); by loosening the legal parameters of lawful collective dismissal (e.g. Greece and Spain); or by reducing the extent of sanctions applied to employers who are found to have transgressed the legal parameters of dismissal (e.g. Italy, Portugal, and Spain). In order to derive crisp membership scores, the outcome of interest is calibrated such that each case is described as belonging either to the subset of cases displaying the outcome [1] or to the subset of cases displaying the negation of the outcome [0]. For the purpose of the present analysis, a distinction is made between cases that display strong evidence of recommodification and those that display limited or no evidence of recommodification. Strong evidence of recommodification is said to be observable when a case tests positively for more than one of the dimensions of recommodification. These cases are assigned a score of [1] for the outcome and are formally operated in the csqca as the subset of cases that display the presence of recommodification. By contrast, cases that test either positively for only one of the dimensions of recommodification or negatively for all three of them are said to offer limited or no evidence of recommodification. These cases are assigned a score of [0] for the outcome and are formally operated in the csqca as the subset of cases that display the absence of recommodification. The results of the calibration of recommodification along with the cases that correspond either to its presence or its absence are summarized in Table 2. Table 2 Calibration Results for the Outcome of Interest Cases Recommodification Belgium, Denmark, France, Greece, Ireland, Italy, Portugal, Spain, United Kingdom Present [1] Australia, Austria, Canada, Finland, Germany, Netherlands, New Zealand, Norway, Sweden Absent [0] Causally Relevant Conditions A perspectives approach is applied to the selection of causally relevant conditions (Amenta and Poulsen, 1994). This approach aims to combine the core propositions of different theoretical hypotheses, such as those outlined above. It has the advantage of facilitating a combinatorial analysis of multiple hypotheses with the aim of deriving a more complex and, by extension, more comprehensive account of the phenomenon under

investigation. On the basis of a perspectives approach, a total of five causally relevant conditions are included in the formal csqca. The raw data corresponding to these five causally relevant conditions is summarized in Table 3. In order to derive crisp membership scores, thresholds are derived for each of the causally relevant conditions according to which the raw data can be recoded into binary form. Whether or not a case meets a defined threshold determines whether it is assigned membership in the subset of cases displaying the condition [1] or in the subset of cases displaying the negation of the condition [0]. The process of calibration is informed by theoretical and substantive criteria (Rihoux and De Meur, 2009). Existing knowledge suggests that labour market policy change during an economic downturn may depend on the partisan composition of government (Cusack et al., 2008; Starke et al., 2013). Accordingly, the first causally relevant condition [RIGHT] accounts for the partisan composition of government. Data on the partisan composition of government were derived from the Comparative Political Data Set (Armingeon et al., 2012). A score of [1] is assigned to countries in which right-leaning parties dominated during key periods of crisis management. A score of [0] is assigned to countries in which left-leaning parties, centrist-parties, or government coalitions involving parties of opposing partisan tendencies dominated during key periods of crisis management. The direction of labour market policy change during an economic downturn may also depend on the fiscal capacity of the state (Bieling, 2012). The current account balance of a national political economy provides a strong indication of the fiscal maneuverability of a country (Claessens et al., 2012). Accordingly, the second causally relevant condition [DEF] accounts for the strength of national current account balances as a proportion of GDP. Data on current account balances were derived from the OECD (2013). A score of [1] is assigned to countries that reported current account deficits in both 2007 and 2008. A score of [0] is assigned to countries that reported a current account surplus in both 2007 and 2008. A third factor that may shape the direction of labour market reforms during an economic downturn is the level of unemployment (Streeck, 2011). Accordingly, the third causally relevant condition [UNEMP] accounts for the change in national levels of unemployment over the course of the economic crisis. Data on unemployment rates were derived from the OECD (2013). Figure 1 shows the percentage change in levels of unemployment between 2007 and 2011 for the eighteen cases. Case-based knowledge suggests that the United Kingdom and Portugal have experienced qualitatively different jobs crises. The United Kingdom has experienced an unexpectedly mild, though far from

Table 3 Raw Representation of the Causally Relevant Conditions Partisanship Current Account Balance (% GDP) 2007 2008 Percentage Change in Unemployment, 2007-2011 (%) Variety of Capitalism Automatic Stabilizers Australia Left -6.1-4.5 16 LME 5.6 Austria Mixed 3.5 4.8-6 CME 5.4 Belgium Mixed 1.9-1.3-4 CME 10.1 Canada Right 0.8 0.3 24 LME 6.7 Denmark Right 1.3 2.6 99 CME 10.5 Finland Mixed 4.0 2.6 13 CME 8.5 France Right -0.9-1.7 16 SME 6.9 Germany Right 7.4 6.1-31 LME 7.5 Greece Left -14.3-14.6 113 SME N/A Ireland Right -5.3-5.6 209 LME 10.5 Italy Right -2.4-2.8 38 SME 4.5 Netherlands Right 6.7 4.2 24 CME 10.6 New Zealand Right -8.1-8.7 78 LME 4.6 Norway Left 12.4 15.9 30 CME 12.9 Portugal Left -10.0-12.6 60 SME N/A Spain Left -9.9-9.6 162 SME N/A Sweden Right 9.1 9.0 24 CME 10.6 United Kingdom Right -2.2-1.0 48 LME 6.6 Sources: Partisanship (Armingeon et al., 2012); Current account balances (OECD, 2013); Unemployment rates (OECD, 2013); Variety of capitalism (Hall and Gingerich, 2009; Schmidt, 2009); Automatic stabilizers (Scruggs, 2004).

insignificant, increase in unemployment (Clasen et al., 2012; Leschke and Watt, 2010). By contrast, in Portugal, recent labour market reforms have in fact exacerbated an already steep increase in unemployment (Rodrigues and Reis, 2012). The threshold is therefore placed between the United Kingdom (48%) and Portugal (60%). A score of [1] is assigned to countries where the increase in unemployment between 2007 and 2011 amounted to 50% or more. A score of [0] was assigned to countries where the increase in unemployment between 2007 and 2011 was less than 50%. Figure 1 Change in Unemployment, 2007-2011 (%) 250 200 150 100 50 0-50 Germany Austria Belgium Source: OECD (2013). Finland Australia France Canada Netherlands Sweden Norway Italy United Kingdom Portugal New Zealand Denmark Greece Spain Ireland A fourth causally relevant condition [LME] is included in order to account for the possibility that labour market policy responses to the economic crisis may vary systematically between differently organized models of welfare capitalism (Lallement, 2011; Iversen, 2007). According to the varieties of capitalism and welfare regimes approaches, when faced with an economic shock, liberal market economies are more likely than nonliberal market economies to rely on a market-oriented strategy of crisis management. A score of [1] is assigned to liberal market economies. A score of [0] is assigned to non-liberal market economies. The distinction between liberal and non-liberal market economies is informed by the empirical analysis of Hall and Gingerich (2009) and Schmidt (2009). Finally, existing knowledge suggests that the presence or absence of automatic welfare state stabilizers may shape the direction of welfare state change during an economic downturn (Ramesh, 2009; Starke et al., 2013). Accordingly, a fifth causally relevant condition [STAB] is included in order to account for the strength of automatic welfare stabilizers. In the analysis that follows, the generosity of unemployment benefits is taken as a proxy for the strength of automatic welfare state stabilizers. Data on the generosity of unemployment benefits were derived from the Comparative Welfare Entitlement Dataset (Scruggs, 2004). Figure 2 lists the generosity scores for the eighteen countries. Case-based knowledge

suggests that a significant qualitative distinction can be made between the national unemployment benefit systems of Finland and Germany. While the former is characterized by a comparatively generous and inclusive system of unemployment benefits, the latter is characterized by an exclusive system of unemployment insurance whose generosity has experienced considerable decline over the past decade (Dingeldey, 2011; Picot, 2012). The qualitative threshold is therefore placed between Germany (7.5) and Finland (8.5). A score of [1] is assigned to any country whose unemployment benefit generosity score is 8 or higher. A score of [0] is assigned to countries whose generosity score falls below 8. Figure 2 Automatic Welfare State Stabilizers 15 10 5 0 Italy New Zealand Australia Source: Scruggs (2004). Austria United Kingdom Canada France Germany Finland Belgium Denmark Ireland Netherlands Sweden Norway In the analysis that follows, partisan effects on crisis management are said to be constrained when the case is assigned a score of [1] for the condition [STAB]. In addition, fiscal crisis is said to be present when a case is assigned a score of [1] for either the condition [DEF] or the condition [UNEMP]. By contrast, fiscal crisis is said to be absent when a case is assigned a score of [0] for both the condition [DEF] and the condition [UNEMP]. V. TRUTH TABLE ANALYSIS Based on the above calibration of the outcome and the causally relevant conditions, it is possible to generate a configurational representation of the cases and their corresponding membership scores. This is depicted in Table 4. The derived matrix of scores is reformulated into a truth table. The truth table is the fundamental unit of analysis in configurational comparative analysis (Rihoux and De Meur, 2009). It lists all of the logically possible combinations of causally relevant conditions, the cases that correspond to them, and the empirically observed outcomes with which they are associated. Due to limited diversity in the world of actually observed cases of the phenomenon, not all of the logically possible combinations listed in the truth table exist in reality. Table 5 depicts a partial truth table. It is

partial because it excludes logical remainders; that is, combinations of causally relevant conditions that are not associated with actually observed cases. Table 4 Configurational Representation of the Cases RIGHT DEF UNEMP LME STAB OUTCOME Australia 0 1 0 1 0 0 Austria 0 0 0 0 0 0 Belgium 0 1 0 0 1 1 Canada 1 0 0 1 0 0 Denmark 1 0 1 0 1 1 Finland 0 0 0 0 1 0 France 1 1 0 0 0 1 Germany 1 0 0 0 0 0 Greece 0 1 1 0 0 1 Ireland 1 1 1 1 1 1 Italy 1 1 0 0 0 1 Netherlands 1 0 0 0 1 0 New Zealand 1 1 1 1 0 0 0 0 0 1 Norway 0 0 0 0 1 0 Netherlands 1 0 0 0 1 0 Portugal 0 1 1 0 0 1 New Zealand 1 1 1 1 0 0 Spain 0 1 1 0 0 1 Norway 0 0 0 0 1 0 Sweden 1 0 0 0 1 0 Portugal 0 1 1 0 0 1 United Kingdom Spain 1 0 1 1 0 1 1 0 0 0 1 1 Note: [1] denotes presence; [0] denotes absence. Sweden 1 0 0 0 1 0 United Kingdom 1 1 0 1 0 1 Table 5 Partial Truth Table for the Presence and Absence of Recommodification RIGHT DEF UNEMP LME STAB OUTCOME Cases 0 1 0 1 0 0 Australia 0 0 0 0 0 0 Austria 0 1 0 0 1 1 Belgium 1 0 0 1 0 0 Canada 1 0 1 0 1 1 Denmark 0 0 0 0 1 0 Finland, Norway 1 1 0 0 0 1 France, Italy 1 0 0 0 0 0 Germany 1 0 0 0 1 0 Netherlands, Sweden 0 1 1 0 0 1 Greece, Portugal, Spain 1 1 1 1 1 1 Ireland 1 1 1 1 0 0 New Zealand 1 1 0 1 0 1 United Kingdom

Truth table analysis proceeds through the process of logical minimization. Broadly speaking, logical minimization refers to the systematic application of Boolean logic with the aim of reducing the complexity of the truth table. The end result of logical minimization is a Boolean solution whose constituent elements reveal minimally relevant causal relations that explain the data within the truth table. The notation and operators involved in the application of Boolean algebra are listed in Box 1. Logical minimization involves the paired comparison of different truth table rows that display the same outcome. The paired comparison of different rows is informed by a simple rule that reflects the system of logic first proposed by Mill (1843). Ragin (1987) summarizes the rule as follows: if two Boolean expressions differ in only one causal condition, yet produce the same outcome, then the causal condition that distinguishes the two expressions can be considered irrelevant and can be removed to create a simpler, combined expression (p. 93). Box 1 Boolean Notation and Operators Upper case letters (e.g. [RIGHT]) represent the [1] value for a given condition. Lower case letters (e.g. [right]) represent the [0] value for a given condition. Logical AND is represented by the [*] symbol. Logical OR is represented by the [+] symbol. Adapted from Rihoux and De Meur (2009). Depending on the strategy of minimization employed, a single truth table can produce several Boolean solutions of varying complexity (Rihoux and De Meur, 2009; Schneider and Wagemann, 2010). Solutions vary in their complexity based on whether or not they exclude or include logical remainders in the process of paired comparison. If they are excluded, a purely descriptive solution is derived. This descriptive solution does not go much beyond the world of actually observed cases. In order to further reduce the complexity of the truth table, logical remainders can alternatively be included in the process of logical minimization to generate a parsimonious solution. The latter strategy relies on the use of simplifying assumptions. These are simplifying because they allow for a greater number of paired comparisons and, by extension, facilitate the derivation of more parsimonious solutions. However, they are also assumptions because they rely on counterfactual cases that are not observed in the empirical world. Parameters for the use of simplifying assumptions are established on the basis of technical and theoretical criteria (Ragin and Sonnett, 2005). Truth table analysis was conducted for the presence and the absence of recommodification using Tosmana (version 1.3.2). For both the presence and the absence of the outcome, the analysis proceeds in a two-fold manner, beginning with the analysis of

necessary conditions and ending with the parsimonious analysis of sufficient conditions. Truth Table Analysis for the Presence of Recommodification The analysis of necessary conditions is concerned with determining whether any condition (or combination of conditions) is necessary for an outcome to occur. In other words, it aims to establish whether the outcome is a subset of a condition (or combination of conditions). The analysis of necessary conditions in configurational comparative methods relies on two measures: consistency and coverage (Bol and Luppi, 2013; Ragin, 2008). Consistency refers to the degree to which a given subset relation is approximated by the data. It is measured as the proportion of cases displaying a given condition (or combination of conditions) among the subset of cases displaying the outcome. Coverage refers to the empirical relevance of a subset relation that has already been determined to be consistent. It is measured as the proportion of cases displaying the outcome among the subset of cases displaying a given condition (or combination of conditions) (Goertz, 2006). Table 6 summarizes the consistency and coverage scores of the causally relevant conditions for the presence of recommodification. The analysis of necessary conditions is conducted using a consistency threshold of 0.900 (Schneider and Wagemann, 2007). Conditions whose consistency scores fall above this threshold are said to be necessary. A test for necessity reveals that none of the individual conditions are individually necessary for recommodification to occur. A test for necessity can also be performed for combinations of conditions. However, such a test is plausible only where there are theoretical arguments to justify it (Schneider and Wagemann, 2007). The fiscal crisis outlined above states that recommodification should occur in the presence of either current account deficits or a significant increase in unemployment. In other words, while the hypothesis holds that fiscal crisis is necessary for the presence of recommodification, there are two causally relevant conditions [DEF] or [UNEMP] that individually constitute sufficient but unnecessary conditions for the presence of fiscal crisis. Accordingly, the combination of conditions [DEF + UNEMP] is included in the analysis of necessary conditions. The consistency score for the configuration [DEF + UNEMP] is 1.000, indicating perfect consistency. Furthermore, the combination of conditions reports a somewhat high coverage score of 0.818, revealing that the necessity relation is far from trivial. In other words, fiscal crisis, represented by the combination [DEF + UNEMP], is a necessary condition for the presence of recommodification. A case must display a current account deficit in 2007 and 2008, an increase in unemployment of 50% or more between 2007 and 2011, or both in order for recommodification to occur.

Table 6 Analysis of Necessary Conditions for the Presence of Recommodification Condition Tested Consistency Coverage RIGHT 0.556 0.500 right 0.444 0.500 DEF 0.889 0.800 def 0.111 0.125 UNEMP 0.556 0.833 unemp 0.444 0.333 LME 0.222 0.400 lme 0.778 0.538 STAB 0.333 0.500 stab 0.667 0.500 DEF + UNEMP 1.000 0.818 Table 7 Parsimonious Analysis of Sufficient Conditions for the Presence of Recommodification Solution DEF*lme + UNEMP*STAB + RIGHT*DEF*unemp Case Coverage Belgium, France, Greece, Italy, Portugal, Spain Denmark, Ireland France, Italy, United Kingdom Raw Coverage 0.667 0.222 0.333 Unique Coverage 0.444 0.222 0.111 Consistency 1.000 1.000 1.000 Solution Consistency 1.000 Solution Coverage 1.000 The analysis of sufficient conditions is concerned with determining the configurations of conditions that are sufficient for an outcome to occur. The analysis of sufficient conditions is conducted through the logical reduction of the truth table using the process of paired comparison described above. Table 7 summarizes the results of the parsimonious analysis of sufficient conditions. The parsimonious solution suggests that there are three causal paths that lead to the presence of recommodification. Each causal path is composed of a unique combination of conditions. However, each combination of conditions is individually sufficient to ensure the presence of recommodification. The empirical relevance of a solution term is assessed by way of its coverage score (Rihoux and De Meur, 2009). Both raw and unique coverage scores are listed in Table 7. The first term [DEF*lme] in the parsimonious solution combines a current account deficit and non-liberal market arrangements. The second term [UNEMP*STAB] combines a significant increase in unemployment and strong automatic welfare state stabilizers. The third and final term [RIGHT*DEF*unemp] combines a right-leaning government, a current account deficit, and the absence of a significant increase

in unemployment. Notably, either [DEF] or [UNEMP] appears in all three of the parsimonious solution terms, confirming the earlier finding that fiscal crisis is a necessary condition for the presence of recommodification. The truth table of the analysis for the presence of recommodification reveals that fiscal crisis a condition that is sufficiently though not necessarily met by the presence of either [DEF] or [UNEMP] is a necessary but insufficient condition for the occurrence of recommodification. Fiscal crisis is insufficient insofar as it must operate in combination with the other causally relevant conditions to generate the causal recipes that are sufficient for the presence of recommodification. More specifically, fiscal crisis combines with either (i) the presence of non-liberal market arrangements; (ii) the presence of strong automatic welfare state stabilizers; or (iii) the presence of a right-leaning government and the absence of a significant increase in unemployment in order to create the sufficient conditions for the presence of recommodification. On this basis, [DEF] and [UNEMP] are said to form a pair of SUIN conditions (Mahoney et al., 2009). In other words, they are individual conditions that are sufficient but unnecessary parts of a broader configuration (i.e. fiscal crisis) whose presence is itself insufficient but necessary for the occurrence of recommodification. Truth Table Analysis for the Absence of Recommodification The analysis of necessary conditions for the absence of recommodification proceeds in a similar fashion to that for the presence of recommodification. Table 8 lists the consistency and coverage scores of the causally relevant conditions for the absence of recommodification. All of the causally relevant conditions report consistency scores below the threshold of 0.900. Accordingly, no single condition appears to be individually necessary for the absence of recommodification to occur. Table 9 lists the results of the parsimonious analysis of sufficient conditions for the absence of recommodification. The parsimonious solution indicates that there are three minimally relevant paths that lead to the absence of recommodification. Each of the paths consists of a unique combination of conditions. The first term [def*unemp] in the parsimonious solution combines a current account surplus and the absence of a significant increase in unemployment. The second term [right*unemp*lme] combines a right-leaning government, the absence of a significant increase in unemployment, and liberal market arrangements. The third and final term [UNEMP*LME*stab] combines a significant increase in unemployment, liberal market arrangements, and low levels of automatic welfare state