Introduction
Active labor market policies (ALMPs), which spend for public employment services and administration, labor market training, youth measures, measures for the disabled and subsidized employed, are a central part of established democracies’ socio-economic policymaking. Specifically, the countries of the organisation of economic cooperation and development (OECD) spend between 0.12% (the United Kingdom in 1985) and 2.85% (Sweden in 1994) of their gross domestic product on ALMPs. In the mid 1990s, both the OECD and the European Union (EU) embraced these policies as a means to improve access to the labor market and to enhance workers’ possibilities to get a job and therefore recommended their increased use (OECD, 1994; see Armingeon, Reference Armingeon2007). The idea that benefits should be ‘active’ is now widely endorsed among developed democracies. On the individual level, most people prefer active programs to passive ones (OECD, 2006). Moreover, political parties from different sides of the political spectrum agree on the value of ‘activation’. For example, the Dutch social democrats, Christian democrats as well as the conservative liberals stressed the importance of labor market integration and activation in their party manifestos for the 2010 national parliamentary elections. The parties vary, however, in their emphasis on the different sorts of ALMPs, such as training or subsidized work, and on the importance of these policies vis-à-vis other – at least partly competing – policies such as poverty reduction or housing.
Against the backdrop of the electorate’s and the mainstream parties’ appreciation of ALMPs and the push by the OECD and the EU to increase the emphasis on such policies, why do we nonetheless find substantial variation across governments in increases in spending on these policies? The focus of different political parties on different types of ALMPs could be one of the reasons for this. Yet, empirical reality does not support this hypothesis univocally. Despite their different political colors, all Dutch governments since the mid-1980s have increased the share of active spending. However, in countries such as Norway and Switzerland governments have variously increased and decreased the share of active spending over this time period. This empirical pattern amounts to an interesting puzzle. Given the typical support for ALMPs, how does one account for the variation in spending on these policies across countries and – perhaps even more interestingly – across governments?
This study presents new data for 53 governments in 18 countries between 1985 and 2003 to answer this question. These data reveal a substantial variation in changes in spending on ALMPs across governments. This study proposes that the socio-economic situation is critical for explaining this variation. Specifically, I hypothesize that because of the high costs of ALMPs and their low electoral rewards, governments increase ALMP spending only in improving socio-economic situations. To test this hypothesis, this study employs a settheoretical approach – fuzzy-set qualitative comparative analysis (fsQCA) – that allows for identifying necessary and/or sufficient combinations of conditions (Ragin, Reference Ragin1987, Reference Ragin2000, Reference Ragin2008). This feature makes it particularly suitable for testing this study’s hypothesis. FsQCA can also reveal the various possible routes toward the outcome (Bennett and Elman, Reference Bennett and Elman2006), something statistical techniques are not well suited for Mahoney and Goertz (Reference Mahoney and Goertz2006), Shalev (Reference Shalev2007), but see Clark et al. (Reference Clark, Gilligan and Golder2006). Given that the combinations of conditions under which governments increase their spending on ALMPs likely vary – and that there are thus different routes toward activation, this feature of fsQCA also makes the approach apt for this study.
The fsQCA analysis demonstrates that an improving socio-economic situation, in the form of decreasing unemployment, is indeed almost always necessary for governments to increase spending on ALMPs. It is only sufficient for triggering this outcome, though, when combined with other conditions (the presence of trade openness or the absence of corporatism in the case of leftist partisanship or the presence of corporatism in the case of rightist partisanship).
The article’s contribution is twofold. First, it presents new data on the public policymaking activities of 53 cabinets in 18 countries, spanning almost 20 years. Since many claims in comparative welfare state research pertain to what governments do, such data are very useful. These data are still rare, as most studies use either the country (in cross-section analysis) or the country year (in pooled time series analysis) as unit of analysis (exceptions include Armingeon and Giger, Reference Armingeon and Giger2008; Vis, Reference Vis2010). If we understand reform as an intentional political action to alter policy output, the government is the appropriate unit of analysis since it is the only political actor with the formal authority to make binding decisions. Even if social partners such as labor unions have codecision power, they can only codecide if the government allows them to.
The study’s second, more important contribution is that it enhances our knowledge of the conditions under which governments pursue activation. In this vein, this study addresses the scientifically and socially relevant question of when governments take up the advice of the EU and the OECD and increase spending on ALMPs. Despite its relevance, we have only scant evidence of the conditions leading to changes in spending on ALMPs (examples include Armingeon, Reference Armingeon2007; Gaston and Rajaguru, Reference Gaston and Rajaguru2008; Vis, Reference Vis2009b), as most studies focus on the level of ALMPs (e.g. Franzese and Hays, Reference Franzese and Hays2006; Rueda, Reference Rueda2006, Reference Rueda2007; Bonoli, Reference Bonoli2008; Huo et al., Reference Huo, Nelson and Stephens2008). This study’s new methodological and theoretical angles offer insights into the degree to which governments change their labor market policies.
The study’s structure is as follows. The next section develops the theoretical argument and presents the fuzzy-set hypotheses. Subsequently, I elaborate the operationalization of the outcome and the causal conditions. Then, I present and discuss the results of the fsQCA analysis. The final section offers some concluding remarks.
Accounting for the cross-government variation in activation
Over recent years, we have witnessed a steady influx of studies addressing ALMPs. Comparative political economists have typically focused on the economic effects of ALMPs, assessing, in particular, if they attain their intended goals such as lower unemployment or increased opportunities for individuals to find a job (for an overview of ALMPs’ macroeconomic effects, see Kluve Reference Kluve2006; for ALMPs’ effects on individuals, see Martin, Reference Martin2000: 89ff). Although there are exceptions, these studies typically find that the positive effect of ALMPs is weak at best. In a recent contribution, Anderson (Reference Anderson2009) argues that there is more to ALMPs than economic outcomes, as ALMPs enhance people’s opportunities and motives for interacting with others and thereby foster social cohesion. By doing so, Anderson (Reference Anderson2009: 342) argues, ALMPs ‘help to sustain a healthier, happier, and socially connected society’. The hierarchical analyses of the effect of ALMPs on individuals’ social ties in 17 countries in 2002–03 support Anderson’s hypothesis and offer comforting evidence to those pushing for increased spending on ALMPs. Another interesting contribution is Martin (Reference Martin2004). Martin argues that employers are more likely to support, and participate in, active labor market programs when they consider the program of good quality (like in Denmark). When they doubt the quality of the ALMP, employers are less likely to participate in it (such as in the United Kingdom). Martin offers important insights into the conditions under which employers are willing to participate in an ALMP scheme once in place.
This article intends to contribute to a related, but distinct literature that focuses not so much on the effects of ALMPs or on when firms participate, but on underlying conditions. This body of work addresses questions such as: why do some countries spend more on ALMPs than others? Why does this change over time? Franzese and Hays (Reference Franzese and Hays2006), for example, put forward an intriguing yet not very plausible answer. They argue that governments’ policymaking is strategic; governments look at what their neighbors are doing and adjust their own behavior accordingly. Focusing on training program expenditure in 15 European countries between 1987 and 1998, Franzese and Hays show that increasing expenditure in one country did indeed result in lower expenditure in its neighbors. However, there are some important problems with Franzese and Hays’ study. One is the assumption that ALMPs in one country (say France) offer the unemployed who live near the border of a neighboring country (say Belgium) the opportunity to obtain training in France, and return – more employable – to Belgium (Franzese and Hays, Reference Franzese and Hays2006: 173). The authors leave unmentioned the fact that participating in an ALMP scheme in another country has huge consequences for an unemployed individual. In most countries, it would mean having to move to the country offering the ALMP program, as residency in the country is required for such participation. It seems highly unlikely that individuals who already find themselves in a precarious situation – being without a job – will move to another country just to participate in a scheme that may, or may not, enhance their opportunities of finding a job. Therefore, I see more merit in four other explanatory factors typically identified in the literature: partisanship, corporatism, economic openness, and the socio-economic situation. I will argue that the latter condition is the key one, as it is a blossoming socio-economic situation that fosters increases in spending on ALMPs. The other conditions are relevant too since a good socio-economic situation is only sufficient to bring about such increases when combined with one or more other conditions.
Socio-economic situation
The first condition, the socioe-conomic situation, is the central one for accounting for the cross-government variation in increases in ALMP spending. Specifically, this study hypothesizes that an improving socio-economic situation is the necessary condition for such an increase. Governments’ room to maneuver within their budgets is higher under better economic conditions, allowing them to prioritize ALMPs. It thus may seem obvious that good socio-economic conditions lead to higher spending on ALMPs. However, some authors have found the opposite, namely that ALMP spending increases when unemployment rises. Swank and Martin (Reference Swank and Martin2001), for instance, show a positive effect of the average unemployment rate over the previous 3 years on ALMP spending as a percentage of GDP. While significant, the impact is very low, with a one-unit increase in unemployment leading to an increase in spending on ALMPs of about 0.01% of GDP. Moreover, Rueda (Reference Rueda2005) finds that increases in unemployment combined with a leftist government have a positive effect on spending on ALMPs (see below for Rueda’s full argument). Other scholars find no significant relationship between socio-economic performance and ALMP spending (Martin and Swank, Reference Martin and Swank2004; Gaston and Rajaguru, Reference Gaston and Rajaguru2008). Despite these latter findings, it is more plausible that a blossoming socio-economic state induces governments to increase their spending on ALMPs (cf. Armingeon, Reference Armingeon2007; Vis, Reference Vis2009b). Socio-economic changes, such as de-industrialization and demographic change, produce the ‘need’ for such reforms. Specifically, the shift toward a postindustrial welfare state has produced new social risks (NSRs) in that different groups are hit by them (e.g. women, the young, the low skilled) (Armingeon and Bonoli, Reference Armingeon and Bonoli2006). Consequently, socio-economic change likely induces governments to take measures to deal with these new risks, for example, by increasing spending on ALMPs. However, since ALMPs benefit labor market outsiders (e.g. the unemployed and temporary workers) most, the electoral hurdle for governments to pursue them is rather high. Outsiders are less active politically and – more importantly for this study – less important electorally for mainstream parties (Rueda, Reference Rueda2005). Mainstream parties typically making up the government are unlikely to reap electoral gains through ALMPs. Existing studies show that mainstream parties cater to the median voter (e.g. Adams et al., Reference Adams, Clark, Ezrow and Glasgow2004; Adams et al., Reference Adams, Haupt and Stoll2009), who is hardly favorable toward higher spending on ALMPs (Eurobarometer 56.1, 2001, see Vis, Reference Vis2010). Since ALMPs are very expensive, at least in the short run, this limited possibility for reaping electoral gains puts a high price on pursuing ALMPs. Even though (some kinds of) ALMPs fit the policy preferences of most mainstream parties, the conditions for actually pursuing this preference need to be right. When the socio-economic situation is improving, parties’ room to maneuver increases, both politically and fiscally. Therefore, I propose that governments increase spending on ALMPs only when the socio-economic situation is improving.
Partisanship
An improving socio-economic situation is not sufficient for higher spending on ALMPs; we need one or more other conditions as well. A first condition that seems a plausible candidate is leftist partisanship. Most scholars agree that the objectives of leftist and rightist parties vary with respect to socio-economic policies. Regarding ALMPs, leftist parties favor a larger role for labor market policy in general than rightist ones. Leftist, social democratic parties pursue policies to keep individuals in the labor market once they have entered it, for example, by using social policy to protect individuals’ investments in specific skills (Iversen, Reference Iversen2005), or by retraining them when needed. Given social democracy’s interest in labor market participation (e.g. Huber and Stephens, Reference Huber and Stephens2001; Huo et al., Reference Huo, Nelson and Stephens2008), policies that intend to uphold and boost this participation fit these parties’ ideology well. Authors typically find empirical support for the leftist partisanship hypothesis. Some authors show that the positive effect of leftist partisanship works only in combination with another condition, namely economic openness (Bonoli, Reference Bonoli2008). The argument here is that an open economy ties leftist parties’ hands as regards adopting strong proworker policies, because such measures hurt international competitiveness. ALMPs offer leftist parties a way to cater to their constituency while simultaneously upholding competitiveness.
The largest part of the existing literature thus expects, and finds, a (individual) positive effect of leftist partisanship or an interaction between openness and leftist partisanship. There are some deviant voices, though, the most well known of which is Rueda (Reference Rueda2006, Reference Rueda2007). Rueda proposes that ALMPs foremost benefit labor market outsiders while potentially hurting labor market insiders (those with steady employment) through tax effects or wage competition. Since insiders are social democracy’s actual constituency, social democracy will oppose increases in ALMPs. Unlike previous studies, Rueda finds a negative relationship between leftist partisanship and spending on ALMPs (see also Franzese and Hays, Reference Franzese and Hays2006; Gaston and Rajaguru, Reference Gaston and Rajaguru2008). However, when the level of unemployment increases, making insiders more vulnerable to unemployment and hence more like outsiders, the relationship between leftist partisanship and increases in ALMPs turns positive. Moreover, when examining the interaction between leftist partisanship and corporatism, Rueda finds that in those countries with high levels of corporatism leftist governments are associated with lower levels of ALMPs. In countries with a low level of corporatism, the interaction is insignificant. Without corporatist arrangements to protect insiders, these insiders resemble outsiders and therefore become more interested in ALMPs. For insiders in corporatist countries, the opposite holds. Since insiders are protected, they are less interested in ALMPs.
Rather than assuming that the disagreement in the literature means that some scholars are wrong and others are right, this study allows for the possibility that both strands of work may be correct, albeit under different circumstances. This latter view allows the strategic character of leftist parties to be taken into account. As Scharpf (Reference Scharpf1991 [1987]) already noted over two decades ago, the left faces tough choices between its preferred policies and power (office and/or votes) ambitions. Ever tighter government budgets intensify the strategic choices to be made (see also Kitschelt, Reference Kitschelt1999). Against this backdrop, it seems plausible to hypothesize that leftist partisanship is an INUS condition, that is, an insufficient but non-redundant part of an unnecessary but sufficient (combination of) condition(s) (Mahoney and Goertz, Reference Mahoney and Goertz2006: 232), for increased spending on ALMPs. If Bonoli (Reference Bonoli2008) were right, one route toward higher ALMP spending would be leftist partisanship combined with openness. If Rueda (Reference Rueda2006, Reference Rueda2007) were correct, one route would be leftist partisanship combined with low corporatism, and another leftist partisanship combined with increased unemployment.Footnote 1
Economic openness
Regarding the second condition, economic openness, this study also expects to find an INUS condition. Overall, the literature is unclear as to the precise relationship between openness and governments’ spending on ALMPs. Most findings suggest that there is a relationship (such as Bonoli (Reference Bonoli2008) discussed above), but there is insufficient evidence that openness as an individual condition will be necessary or sufficient for increased ALMP spending. The idea that economic openness relates to ALMP expenditures goes back to at least Katzenstein (Reference Katzenstein1985), who showed that it was precisely those small, open economies that adopted corporatist practices. They did so, he argued, because the ideological social partnerships that corporatist arrangements entail shield these economies from perceived vulnerabilities, whether economic or other types. Following this line of argument, one would expect a positive relation between openness and spending on ALMPs because such policies serve as domestic compensation for the arising vulnerabilities (see also Cameron, Reference Cameron1978).
Supporting this argument, some authors find a positive effect of economic openness on ALMP spending (e.g. Swank and Martin, Reference Swank and Martin2001; Franzese and Hays, Reference Franzese and Hays2006). Others, however, reveal a negative effect between openness and spending on ALMPs (e.g. Martin and Swank, Reference Martin and Swank2004; Armingeon, Reference Armingeon2007). According to Armingeon (Reference Armingeon2007: 924), the latter result from open economies not being able to ‘afford ‘luxurious’ social spending’. Interestingly, this goes against Bonoli’s (Reference Bonoli2008) hypothesis that in an open economy, social democracy’s only option to deal with the vulnerabilities of an open economy is turning to ALMPs. Conversely, Gaston and Rajaguru (Reference Gaston and Rajaguru2008) find no effect of globalization (openness) on ALMPs, which in their view results from the importance of ALMPs’ wage effects, which increase job competition. This wage effect may stem from immigration having a negative impact on the wages of native workers. Therefore, it is likely that openness as an individual condition is hardly relevant for changes in spending on ALMPs. Still, it may very well have an effect when combined with one or more other conditions.
Corporatism
The literature also pays ample attention to the effect of the degree of corporatism on ALMPs, as in Rueda’s (Reference Rueda2006, Reference Rueda2007) work discussed earlier. In his revisiting of Small States, Katzenstein (Reference Katzenstein2003) argues that small states have a large(r) capacity to learn and can therefore disregard threats and avoid past mistakes. With regard to ALMPs, this would imply that governments in a corporatist country might well adopt ‘fashionable’ policies such as ALMPs more quickly. Although most contemporary scholars expect that corporatism relates to spending on ALMPs, they are not conclusive about its effects. One of the reasons for this indistinctness is that the theoretical expectations about the preferences of trade union members on ALMPs are unclear (Nelson, Reference Nelson2009). Traditionally, trade union members support generous social policies and therefore are likely to favor ALMPs. The higher the degree of job insecurity they face, the higher this support might be, in order to safeguard the trade union member’s own position or that of fellow workers for whom he or she feels responsible. Job insecurity rises because of, for instance, high levels of unemployment, low levels of employment protection or deindustrialization. However, as discussed above, according to insider–outsider theory (for an application see e.g. Rueda, Reference Rueda2006, Reference Rueda2007), trade union members will oppose ALMPs. Trade union members are the insiders of the labor market and wish to maintain their advantageous position vis-à-vis the outsiders. Since ALMPs improve the competitiveness of labor market outsiders, who will typically be participants in activation schemes, trade union members will oppose such policies. This effect will be higher, not lower, when job insecurity increases. Nelson (Reference Nelson2009) shows that neither of these theoretical positions finds full empirical support. Consequently, it seems plausible that corporatism as an individual condition is hardly relevant for changes in spending on ALMPs, yet that it may very well work in combination with one or more other conditions.Footnote 2
Operationalization
Outcome
Following Armingeon (Reference Armingeon2007) and OECD (2003), this study’s primary measure of the outcome is active spending per unemployed, that is the share of GDP spent on ALMPs per 1% standardized unemployment. Or in formula
![\[--><$$>{{\rm Active}} \ {{\rm spending}} \ {{\rm per}} \ {{\rm unemployed}} = [{{\rm spending}} \ {{\rm on}} \ \frac{{{{\rm ALMPs}}}}{{{{\rm GDP}}}}\times 100]/{{\rm st}}{{. \ \rm unemployment}}\eqno<$$><!--\]](https://static.cambridge.org/binary/version/id/urn:cambridge.org:id:binary:20160920234526606-0837:S1755773910000378:S1755773910000378_eqnU1.gif?pub-status=live){kind=small}
This is a better measure than the typically used active spending as a share of GDP because it is less sensitive to the state of the economy (e.g. recession). Specifically, a recession leads – all other things being equal – to higher spending on ALMPs as a share of GDP (GDP falls and thus ALMP/GDP rises, even when the government does not do anything). This is where the problem of many studies lies. However, by dividing ALMP/GDP by the level of standardized unemployment, one corrects – at least partly – for this problem. The larger number of unemployed in a recession means that the (higher) ALMP spending as a share of GDP needs to be divided among a larger number of unemployed. For activation to appear using this study’s measure, the government would thus need to (substantially) boost ALMP expenditure.Footnote 3
The first step to raise the validity of active spending per unemployed as an indicator for activation is to remove those governments displaying less than one percentage point change. While such a small change could result from a political decision, it is more likely that it results from measurement error.Footnote 4 Table 1 lists the resulting cabinets, their period in office and the countries in which they govern; Appendix B in the web appendix (available on the author’s website: www.barbaravis.nl) provides the argumentation for the exclusion or combination of some governments. The exact period depends on when a new cabinet entered office (either in or after 1985) and when a cabinet finished its period in office (around 2003). Table 1 also presents an overview of the percentage point change during the cabinet period in active spending per unemployed per cabinet. The changes range from −23.5 percentage points under Carlsson 3 (Sweden, 1990–91) to +29.0 percentage points under Carlsson 2 and 1 (1986–90). The average change per cabinet (either ±) is 4.5 percentage points. Interestingly, both increases and reductions are not limited to one period; both occur in the 1980s, 1990s and early 2000s alike. Moreover, and related, spending on ALMPs did not ‘automatically’ rise after the OECD’s and EU’s recommendations in the mid 1990s, with some governments even displaying reductions in this period (e.g. Kohl 4 in Germany and Bolger 3 & Shipley 1 in New Zealand).
Table 1 Changes in spending on ALMPs
aActive spending per unemployed is the share of GDP spend on active labor market policies (ALMPs) per 1% standardized unemployment.
bActive spending in total labor market policies (LMP) spending is spending on ALMPs as a percentage of GDP divided by spending on ALMPs and spending on the passive LMP (unemployment compensation and early retirement) as a percentage of GDP.
Sources: Armingeon et al. (Reference Armingeon, Gerber, Leimgruber and Beyeler2008); Woldendorp et al. (Reference Woldendorp, Keman and Budge2000); recent years collected by author from different sources.
Notes: Changes in percentage points. Only those years are included in which the government was in office for more than half a year. For coding decisions of cabinets removed or combined with other cabinets, see Appendix 2 in the web appendix.
For a ‘truly’ active orientation, spending on ALMPs as a share of total labor market spending (spending on ALMPs and spending on passive labor market policies, unemployment compensation, and early retirement, all as a percentage of GDP) should be high as well (OECD, 2003; Armingeon, Reference Armingeon2007). This second indicator, see formula (2) below, is to some extent influenced by the state of the economy since passive expenditures increase automatically when the level of unemployment rises, resulting ceteris paribus in a lower share of active spending in total labor market spending. Still, the need for ALMPs rises when the level of unemployment increases, suggesting that spending on ALMPs may go up too. For the latter, contrary to the former, one usually needs a government decision. This makes the change in active spending as a share of total spending on labor market policies a substantively interesting measure of these – at least somewhat – opposing forces. Table 1 (last column) displays the percentage point change in active spending as a share of total spending on ALMPs.
![\[--><$$>\eqalignno{\qquad\qquad{\rm{Active}} \ {\rm{spending}} \ {\rm{in}} \ {\rm{total}} \ {\rm{labor}} \ {\rm{market}} \ {\rm{spending }} \cr \qquad{{\rm = spending}} \ {\rm{on}}\ \frac{{{\rm{ALMPs/GDP}}}}{{\displaystyle\left. {\left[ {\frac{{{\rm{ALMPs}}}}{{{\rm{GDP}}}}{\rm{ + }}\frac{{{\rm{PLMPs}}}}{{{\rm{GDP}}}}} \right.} \right]}}\times 100 <$$><!--\]](https://static.cambridge.org/binary/version/id/urn:cambridge.org:id:binary:20160920234526606-0837:S1755773910000378:S1755773910000378_eqnU2.gif?pub-status=live)
Table 1 shows that for most governments, the direction of both activation indicators is the same. For several cabinets the directions of the indicators conflict, suggesting that is unclear whether the change in spending on ALMPs is positive or negative. If the change (either positive or negative) in the indicator active spending as a share of total spending is less than one percentage point, active spending per unemployed is more relevant as the former change could easily have resulted from measurement error. These governments thus remain in the analysis. If, however, the two indicators differ in sign and both of them indicate a change larger than one percentage point, I remove the government because the direction of the change is too unclear. This leaves us with 53 governments. The changes in active spending as a share of total spending range from −12.2 percentage points under Aho 1 (Finland, 1991–95) to +16.5 percentage points under Delamaruz (Switzerland, 1995–99). The average change per cabinet (either ±) is somewhat higher than for the first indicator, namely 6.2 percentage points. Moreover, Table 1 shows that for both activation indicators the cross-government variation is substantial.
The data presented thus far are raw data that need to be transformed into fuzzy sets (calibrated, see Ragin, Reference Ragin2008: Ch. 4, 5) to make fuzzy-set qualitative comparative analysis possible. Fuzzy sets are ‘finegrained, [pseudo] continuous measures that have been carefully calibrated using substantive and theoretical knowledge relevant to set membership’ (Ragin, Reference Ragin2000: 7). A fuzzy set includes three qualitative breakpoints, 1, 0 and 0.5, relating to the situations that a government is ‘fully in’, ‘fully out’ and ‘neither in nor out’ of a set (here: high spending on ALMPs, whereby I label the first indicator ALMP per unemployed and the second active spending as a share of total spending). As in fuzzy-set applications establishing these breakpoints and the inbetween scores is important (Ragin, Reference Ragin2000, Reference Ragin2008), let me discuss the procedure. Table 1 helps assign the fuzzy-set scores. Based on substantive knowledge of the cases, derived from others Cox (Reference Cox1998), Huber and Stephens (Reference Huber and Stephens2001), Clasen (Reference Clasen2005), and the OECD Employment Outlooks (various years), I place the qualitative breakpoints 0 and 1 at −15 and +15. Such a reduction (increase) means a change of 0.15% of GDP per percent standardized unemployment. If the unemployment rate is 6%, the share of GDP spent on activation would then reduce (increase) by 0.9% during the cabinet period – a lot given that total social expenditure does not generally exceed 30%. I set the qualitative breakpoint 0.5, the so-called crossover point, at 0. After having established the three qualitative breakpoints, the following procedure calibrates the raw data. First, all raw data below or above the qualitative breakpoints, that is, <15 and >15, are recoded as follows (see Ragin, Reference Ragin2006): lowest through −15, new value −15; 15 through highest, new value 15. The new minimum and maximum are −15 and +15. Then, the fuzzy set is computed by taking these transformed raw data and subtracting the lower limit (here: −15) from each score and then dividing the result by the upper limit minus the lower limit, here: 15 to −15 = 30. In formula: fuzzy-set score=[transformed raw data − lower limit]/[upper limit − lower limit].Footnote 5
In most countries, the fuzzy-set data for active spending per unemployed display substantial within country variation. The Netherlands is the only country where no cabinet reduces active spending. In some other countries, the variation disappears somewhere in the 1990s. Specifically, no Belgian cabinets from Dehaene 1 (1992) onwards display activation. The same holds for the New Zealand cabinets since Bolger 3 (1996). Conversely, the Danish cabinets since Schlüter 5 (1990) all increase spending on ALMPs, as do the Swiss cabinets since Felber (1991). Nevertheless, there is a lot of cross-government variation in changes in spending on ALMPs to explain.
Causal conditions
Based on the hypotheses presented above, I construct five causal conditions: unemployment (UNEM), economic growth (GROWTH), partisanship (RIGHT), corporatism (CORP), and openness (OPEN). As for the outcome, the raw data need to be transformed into fuzzy sets. Let me discuss the procedure per condition. For the first two conditions that capture the socio-economic situation a government faces – growth and unemployment – I use the change in the level of economic growth and change in the level of unemployment during the cabinet period. The first qualitative breakpoint 0 (fully outside the set of GROWTH or UNEM) is placed at −5. Substantive knowledge about developed democracies indicates that a reduction of economic growth or unemployment by 5% is both unusual and has a substantial impact on the possibilities for and necessity of socio-economic policymaking. For similar reasons, I place the second qualitative breakpoint 1 (fully in the set of GROWTH or UNEM) at +5. The inbetween scores are established based on the coding scheme in Table 2.
Table 2 Coding scheme fuzzy-sets GROWTH and UNEM
Sources: Armingeon et al. (Reference Armingeon, Gerber, Leimgruber and Beyeler2008); from 2001 onwards: OECD labour force statistics (2008); author’s calculations.
To establish the fuzzy set RIGHT, I focus on leftist cabinet composition, calculated as social democratic and other leftist parties as a percentage of total cabinet posts (Armingeon et al., Reference Armingeon, Gerber, Leimgruber and Beyeler2008). Measuring the complexion of a cabinet by means of the share of leftist parties is conventional in the literature (e.g. Huber and Stephens, Reference Huber and Stephens2001; Allan and Scruggs, Reference Allan and Scruggs2004). I set the qualitative breakpoints 0 (fully outside the set of RIGHT) and 1 (fully inside the set of RIGHT) at 100 and 0, as these scores correspond with the hegemony of social democratic and other leftist parties and the hegemony of rightwing parties. The crossover point is placed at 0.50, that is, where leftist and rightist parties hold the same percentage of cabinet posts. To differentiate between more leftwing and more rightwing cabinets, I code these cases as either 0.55 (when the rightist party or parties receive most of the votes) or 0.45 (when the leftist party or parties receive most of the votes). Table 3 displays the coding procedure, including how I establish the scores inbetween the qualitative breakpoints. Table A1 in the web appendix shows the fuzzy-set scores per condition.
Table 3 Coding scheme fuzzy-set RIGHT
aBased on the Schmidt index, see Armingeon et al. (Reference Armingeon, Gerber, Leimgruber and Beyeler2008).
Source: Armingeon et al. (Reference Armingeon, Gerber, Leimgruber and Beyeler2008).
The resulting scores indicate that both cabinets including Christian democrats (such as Kohl 1–4) as well as secular conservative cabinets (such as Thatcher 1–3) are coded as rightwing. This coding actually makes sense based on the emphasis on leftist and rightist issues in the manifestos of the parties in government (see Vis, Reference Vis2009a). The LeftRight scale constructed by the Comparative Manifesto Project taps the policy orientation of a party by means of the percentage of references to rightist issues, such as freedom and economic incentives, and leftist issues, such as democracy and labor groups (for all categories, see Budge et al., Reference Budge, Klingemann, Volkens, Bara and Tanenbaum2001: Table 1.1). The scale ranges from −100 (when the entire program is devoted to leftist issues) to +100 (when the entire program is devoted to rightist issues). The policy orientation of the government can be calculated as follows: [∑(absolute number of seats of party i in government × party i’s LeftRight score)]/(total number of seats for the cabinet). Using the LeftRight score per cabinet, the secular conservative Thatcher governments and the German and Dutch cabinets including the Christian democrats (Kohl and Lubbers) are not that far apart ideologically. Thatcher 2 and Lubbers 1 hardly differ, with both scoring well into the rightist part of the scale (29.0 vs. 28.33). Moreover, the Kohl cabinets are overwhelmingly rightist, with Kohl 1 even scoring higher than Thatcher 1 (26.76 vs. 24.4).
For corporatism (CORP), I focus on the degree of coordination of wage setting as measured by the Kenworthy (Reference Kenworthy2001) index. The definition of what corporatism exactly is varies among researchers and has changed over time (see Molina and Rhodes, Reference Molina and Rhodes2002 for an overview). Despite these varying conceptualizations, wage formation is one of the areas in which the involvement of organized interest groups in decision-making has been most extensive and regular (Kenworthy, Reference Kenworthy2002, Reference Kenworthy2003). The degree of wage setting coordination is therefore highly informative as an indicator for the extent of corporatism in a country. An advantage of the Kenworthy index is that it is time varying and covers the countries relevant for this study. The former feature is particularly welcome because the degree of corporatism has changed over time, at least in some countries (Kenworthy, Reference Kenworthy2003; Traxler, Reference Traxler2004). I place the first qualitative breakpoint 0 (fully outside the set of CORP) at 1. Countries that score ‘1’ on the Kenworthy index have fragmented wage coordination, which is confined largely to individual firms or plants, and have no corporatist system. The second qualitative breakpoint 1 (fully in the set CORP) is placed at five. Countries that score ‘5’ on the Kenworthy index have centralized coordinated wage bargaining by peak confederation(s) or government imposition of a wage schedule/freeze, with a peace obligation, which is typically corporatist. The corporatism variable is recoded into a continuous variable, using the same procedure as for the outcome.
Following most studies, I measure openness as total trade in current prices (sum of import and export) as a percentage of GDP (Armingeon et al., Reference Armingeon, Gerber, Leimgruber and Beyeler2008). I set the first qualitative breakpoint 0 (fully outside the set of OPEN) at 0%. An economy scoring 0% on openness is completely closed, having no import or export relations with other countries. The second qualitative breakpoint 1 (fully in the set) is placed at 100%, as that means that a country’s trade relations with other countries are so extensive that they (more than) match that country’s GDP. The inbetween scores are calculated using the same procedure as for CORP and the outcome.Footnote 6
FsQCA analysis
Let me turn to the fuzzy-set qualitative comparative analysis carried out to identify the conditions under which governments increase spending on ALMPs. I use the fsQCA 2.5 software for the analyses. The so called truth table algorithm transforms the fuzzy-set membership scores into a truth table, which lists all logically possible combinations of causal conditions and each configuration’s empirical outcome (Ragin, Reference Ragin2008: Ch. 7). The algorithm uses the direct link between the truth table rows and the corners of the property space, that is the multidimensional space including all logically possible combinations of causal conditions (GROWTH, UNEM, RIGHT, CORP, and OPEN). In this study, the property space has 25 (=32) corners (the configurations). Table A2 in the web appendix displays the truth table. The truth table reveals that the governments displaying the same combination of conditions – and which are thus in the same row – almost always come from different countries, suggesting that the relevant variation is found not so much across countries but across governments. This observation strengthens the argument in favor of taking governments as units of analysis. Moreover, the truth table shows that eight (of the 32) configurations include contradictions – the outliers. I will return to these later on. Finally, the table reveals nine logical remainders, that is, configurations with no empirical observations. This finding means that the variation in the data is limited, as not all possible configurations are observed empirically. However, the degree of limitedness is fairly low since over 70% of the configurations are observed empirically.
After reviewing the truth table, we logically minimize the table using Boolean algebra to reveal the combinations of causal conditions that are sufficient for producing the outcome (Ragin, Reference Ragin2008: Ch. 7). The researcher needs to decide what to do with the logical remainders. The most complex solution arises if no so-called simplifying assumptions are employed, that is, when the positive cases are set as ‘true’ and all other cases are set as ‘false’. Simplifying assumptions are statements about the hypothetical outcome of the logical remainders. The most parsimonious solution of fsQCA is attained if the positive cases are set as ‘true’, the negative cases as ‘false’, and the remainders as ‘don’t care’. I employ the most complex solution, as that is the most conservative approach, and report the results of the most parsimonious solution in a note.
In Boolean logic, logical AND (*) refers to the intersection of sets and logical OR (+) to the union of sets. Moreover, capitals indicate the presence of a condition and lower cases its absence. The fsQCA analysis finds that there are three routes, or causal recipes, for increased spending on ALMPs (ACT): (1) decreasing unemployment combined with openness OR (2) decreasing unemployment combined with a leftist government and the absence of corporatism OR (3) decreasing unemployment combined with a rightist government and the presence of corporatism. In fuzzy-set notation, the analysis’ result isFootnote 7
![\[--><$$>\eqalign{{\rm{unem}} \ast {\rm{OPEN}}\:{\rm{ \,+\, }}\:{\rm{unem}} \ast {\rm{corp}} \ast {\rm{right}}\:{\rm{\, +\, }}\:{\rm{unem}} \ast {\rm{CORP}} \ast {\rm{RIGHT}} \cr\qquad\rightarrow {\rm{ACT \ (coverage:}} \ {\rm{0}}{\rm{.81;}} \ {\rm{consistency:}} \ {\rm{0}}{\rm{.89)}}{\rm{.}}\eqno<$$><!--\]](https://static.cambridge.org/binary/version/id/urn:cambridge.org:id:binary:20160920234526606-0837:S1755773910000378:S1755773910000378_eqnU3.gif?pub-status=live){kind=small}
Consistency indicates the degree to which the solution or result is sufficient for producing the outcome. Coverage gives an indication of the degree to which cases correspond to the (combination of) conditions. Thus, this result suffices to bring about increased spending on ALMPs 89% of the time, covering 81% of the cases.
The fsQCA finding shows that a decreasing level of unemployment is present in each of the three paths toward activation. This suggests that an improving socio-economic situation (decreasing unemployment) is thus – as hypothesized – in most cases a necessary condition for activation. Most of the governments increasing ALMP spending do so under improving socio-economic situations. This condition is only sufficient for activation when combined with other conditions (the presence of openness, or the absence of corporatism and the absence of a rightist government, or the presence of corporatism and the presence of a rightist government). Depending on the path, partisanship does not matter (unem*OPEN), leftist partisanship does (unem*corp*right), or rightist partisanship does (unem*COPR*RIGHT).
Are any of the paths identified more important in the sense that they cover (many) more cases? Based on the so-called unique coverage of the paths, which assesses the relative importance of the different (combinations of) causally important conditions (Ragin, Reference Ragin2008: 658), the answer is yes. Of the total solution’s coverage of 81%, 21% is uniquely due to the first path (unem*OPEN). The second path (unem*corp*right) and the third path (unem*CORP*RIGHT), conversely, uniquely cover only 2%. This means that most of the cases with membership in the latter two paths also have membership in the first one. Stated differently, most of unem*corp*right and unem*CORP*RIGHT is thus unem*corp*right*OPEN and unem*CORP*RIGHT*OPEN.
Do the sufficient paths capture the governments pursuing activation? Table 4 presents the governments’ membership scores in the outcome and the sufficient paths, including only those cabinets that are ‘in’ the set of activation (i.e. a score >0.5) or ‘in’ one or more of the causal paths; Table A3 in the web appendix displays the results of all governments. In 24 of the 30 governments pursuing activation, at least one of the three paths is present. For six cases, however, these paths cannot explain the occurrence of the outcome (Rocard 2&1, Lubbers 3, Guterres 2, Felber, Ogi, and Reagan 2). These are the outliers of this analysis. The presence of outliers indicates that although there are three clear paths toward activation, these are not the only possible ones. This means that it is more apt to describe the paths as ‘almost always sufficient’. A preliminary assessment of the factors that these governments have in common, which would point to an omitted condition, suggests that there is no such factor. The deviating cases, for example, include a surplus coalition and a minimal winning one, as well as a multi-party minority, a single party minority, and a single party majority. Moreover, these governments were in office in the mid-1980s, late 1980s to early 1990s, early 1990s, and 2000s. This observation suggests that in order to understand why these six deviating governments increased spending on ALMPs, in depth case analysis is needed – something that is beyond the scope of this article.
Table 4 Membership in sufficient paths, selected cabinets
Note: The table includes only those cabinets with membership of the outcome (i.e. a score > 0.5) and/or membership of one or more of the paths. Table A3 in the web appendix presents the results for all cabinets. Raw coverage indicates the proportion of the cases covered by that particular path. Unique coverage indicates the proportion of the cases only covered by that particular path (and not also by one or more other paths).
Table 4 also reveals that three governments should have pursued activation, because of their membership in at least one of the paths, but did not (Dehaene 2, Lipponen 2, and Bondevik 1). In this situation one or more unobserved factors prevent reform. Note that these cabinets are only just out of the set of activation, with fuzzy-set scores of 0.41, 0.43, and 0.47.
Discussion of findings
What can we take from these findings? How do they contribute to the discussion on the political economy of labor market reforms? Regarding the effect of partisanship, my findings indicate that whether or not leftist or rightist partisanship relates to increased spending on ALMPs depends on other conditions. As the partisanship hypothesis posits, leftist government is thus neither individually necessary nor sufficient for increases in ALMP spending. For both leftist partisanship and rightist partisanship, decreasing unemployment is central. From this point, the results start to diverge. Although rightist government and corporatism foster activation, it is in the absence of corporatism that leftist governments activate (or when decreasing unemployment coincides with openness, which also holds for rightist governments). This latter finding corroborates Rueda’s (Reference Rueda2007) hypothesis, with one important exception: the almost omnipresent factor of an improving socio-economic situation. This indicates that leftist governments in non-corporatist countries strategically opt for activation only when they have the fiscal room to do so. Rather than opting for higher spending on ALMPs when unemployment rises, as Rueda expects, my fsQCA analysis shows that it is decreasing unemployment that does the trick – as I expected theoretically.
The finding that rightist partisanship and corporatism go hand in hand is an unexpected one. Examining why this is so would be an interesting avenue for future research. One explanation could be that since rightist parties are typically more favorable toward ALMPs than passive ones, trade unions push for more spending on ALMPs when the government is of rightist composition – knowing that higher passive measures, which the unions may actually prefer, are likely a bridge too far. The unions succeed in their attempts only when the rightist government faces a prosperous socio-economic state.
Moreover, the findings on partisanship indicate that, as expected, corporatism alone has no bearing on higher spending on ALMPs, only in conjunction with other conditions. The same applies to economic openness; here we also find a combination of conditions leading toward activation. Specifically, as envisioned by Katzenstein (Reference Katzenstein1985) and supported by Franzese and Hays (Reference Franzese and Hays2006) and Martin and Swank (Reference Martin and Swank2004), many of the governments increasing ALMP expenditure have an open economy. Again, an open economy is not enough; a good socio-economic situation is also required. Furthermore, there are two paths that do not feature openness. Having an open economy is thus not a necessary condition.
Overall, this study’s empirical results offer the strongest corroboration of the hypothesis that improving socio-economic conditions foster spending on ALMPs (cf. Armingeon, Reference Armingeon2007; Keman et al., Reference Keman, Van Kersbergen and Vis2006). Although in being partially supportive to Rueda’s hypotheses, my findings contradict his hypothesis that social democracy supports ALMPs when the economy is deteriorating. I find, instead, that leftist partisanship matters when unemployment rates are decreasing. This result supports my hypothesis that because of their rather expensive nature and because they hardly lead to electoral gains, spending on ALMPs is increased only under improving socio-economic situations.
Cross-validation with the second activation indicator
The first step to raise the validity of active spending per unemployed as an indicator for activation was the removal of governments for which the pursuit of activation was unclear. The second step is to cross validate the findings of the analysis with an fsQCA analysis including the second activation indicator, active spending as a share of total spending, as an outcome.
The fsQCA analysis of active spending on ALMPs as a share of total spending, based on the same coding scheme as used for the first indicator, produces largely similar substantive results. Specifically, the full result isFootnote 8
![\[--><$$>\eqalign{ & {\rm{unem}} \ast {\rm{right}} \ast {\rm{corp}}\:{\rm{ \,+\, }} \:{\rm{unem}} \ast {\rm{right}} \ast {\rm{OPEN}}\:{\rm{ \,+\, }}\:{\rm{unem}} \ast {\rm{OPEN}} \ast {\rm{GROWTH}} \cr & {\rm{ + \,}}\:{\rm{unem}} \ast {\rm{open}} \ast {\rm{CORP}} \ast {\rm{RIGHT}}\:{\rm{\, +\, }}\:{\rm{GROWTH}} \ast {\rm{OPEN}} \ast {\rm{CORP}} \ast {\rm{right}} \cr\rightarrow {\rm{ACT \ (coverage\!:}} \ {\rm{0}}{\rm{.74;}} \ {\rm{consistency:}} \ {\rm{0}}{\rm{.91)}}. <$$><!--\]](https://static.cambridge.org/binary/version/id/urn:cambridge.org:id:binary:20160920234526606-0837:S1755773910000378:S1755773910000378_eqnU4.gif?pub-status=live)
The intermediate solution is – as for the first indicator – the same as the complex one. This finding indicates that here too, most governments increase ALMP spending when socio-economic conditions improve. Differently from the results for the outcome active spending per unemployed, it is either decreasing unemployment, increasing economic growth, or the two combined that does the trick here. Again, an improving socio-economic situation is only sufficient for activation in combination with other conditions. Largely, these combinations of conditions are the same as those in the first fsQCA analysis. Summing up, the cross-validation of the findings of the fsQCA analysis using active spending as a share of total spending as the outcome corroborates the finding of the quasi-necessity, though not sufficiency, of an improving socio-economic situation for activation to be pursued by governments.
Concluding remarks
Since the mid-1990s, the OECD and the EU have been pushing governments to increase their emphasis on ALMPs to improve labor market performance. The literature thus far is inconclusive as regards the conditions under which governments have taken this advice and increased ALMP spending. This study’s contribution to the discussion is twofold. First, based on a new data set of 53 cabinets between 1985 and 2003, it demonstrates that there is substantial variation across governments in the extent to which they increase spending on ALMPs. Second, and more importantly, the fsQCA analysis showed that an improving socio-economic situation, in the form of decreasing unemployment, is almost always necessary for governments to increase their spending on ALMPs. This improving socio-economic situation is sufficient for activation only when combined with (1) openness; (2) the absence of corporatism and a leftist government, or (3) the presence of corporatism and a rightist government.
By presenting data on governments’ labor market policymaking, and by identifying the conditions under which governments increase spending on ALMPs, this study contributes to the existing public policy literature. However, this study also has its limitations. One limitation is that I have not provided an explanation for the deviant cases. The fsQCA analysis of 53 cases conducted here has painted the broad picture by identifying the paths leading to activation about 80% of the time. In order to capture 100%, an additional approach is needed, specifically indepth case analysis. For instance, it would be interesting to examine if the quality of active labor market programs affects governments’ willingness to increase their spending on them. Some countries’ ALMP programs are better than other countries’ programs as they have for example higher success rates in terms of placing participants in jobs. Just as Martin (Reference Martin2004) found that employers are more willing to participate in ALMP programs when they think these programs are of good quality (as in Denmark) than when they consider their quality poor (as in the United Kingdom), governments may also be more eager to use their tight budgets on ALMPs when they expect these programs pay off. Indepth case analysis would also allow for a detailed assessment of the specific spending decisions made by each cabinet, which could serve as an additional cross-validation of the data and the analysis.
Overall, this study’s findings show that ALMPs may indeed be ‘luxuries’ that governments can only afford when the economy is improving. If this is true, the gloomy prediction is that the financial and economic crisis and its aftermath will lead governments to tune down their ALMP spending – despite all the attempts of the OECD and EU to convince them to do otherwise.
Acknowledgements
Earlier versions of this article have been presented at the CSR Colloquium at the VU University Amsterdam, the Netherlands, at the NIG Work Conference 2008, Enschede, the Netherlands, at a lunch seminar of AIAS, Amsterdam, and at the ECPR Joint Sessions of Workshops, Lisbon, Portugal. I would like to thank all the participants for their helpful comments. Also many thanks for the constructive and useful comments of the three anonymous reviewers from the EPSR. Furthermore, I thank Karin Koole for her excellent research assistance. This study is supported by a faculty fellowship of the centre for comparative social sciences (CCSS) of the faculty of social sciences, VU University Amsterdam and a Veni grant from the Netherlands Organisation for Scientific Research (NWO, grant no. 451-08-012).
