Is labour market policy heading for social investment in European countries?

1 Introduction

Labour market policy (LMP) in European countries has long been more than just income support in the case of unemployment, i.e. passive LMP. Bonoli (2012: 188) distinguishes three phases of active labour market policies (ALMP) after World War II, each dealing in a new way with novel economic conditions. After a phase of “upskilling” job seekers and then a phase of just keeping them “busy”, he observes a particular emphasis on “employment assistance” in the latest phase, i.e. interventions aimed at (re-) integrating the unemployed and other non-working persons rapidly into paid employment of any kind. This phase starts in the mid-1990s with the so-called “activation turn”: at European level, it was promoted by the European Employment Strategy of 1997 and the Lisbon Agenda of 2000 (van Vliet/Koster 2011: 232). However, activation is but one part of the new, overarching social investment approach (Morel et al. 2012) that became prominent in the European policy discourse since the Lisbon Agenda (Vandenbroucke/Vleminckx 2011), the other being investment in human capital. According to De Deken (2012), “[t]he return on investment that is hoped for is an increased labour market participation of the population as well as a higher productivity of those participating” (2).

The present article provides an analysis of LMP spending trends in European countries in recent years. These years were marked above all by the Great Recession and its impacts on both labour markets and state budgets. These macroeconomic conditions clash with the social investment approach that takes a preventive stance to labour market problems and requires substantial monetary resources. Still, some years into the crisis, the European Commission (2013) encouraged EU Member States to further pursue the social investment track.^[1] How does European countries’ LMP spending evolve in this situation, marked by the need to react quickly to pressing social and labour market problems, the request to keep up social investment, and the necessity to keep public finances under control? Does LMP now differ from what it was before the crisis? Does employment assistance remain the main orientation of LMP? What about expenditure on training, the measure most closely connected to the human capital aspect of social investment? Finally, how does the relationship between active and passive interventions evolve – is there any sign of increased “resource competition”? We chose a descriptive method to answer to these questions; neither causality concerning the genesis of policy nor the effects of policy are dealt with in this article.^[2]

The following section discusses the relationship between social investment and different categories of LMP spending and dicusses contemporary research on LMP spending trends. section 3 introduces the data and method of our analysis. section 4 comparatively analyses the evolution of active and passive LMP, while section 5 is dedicated to developments within the ALMP spending block. The final section of the paper summarises our findings and provides our conclusions.^[3]

2 Labour market policy from a social investment perspective: the state of knowledge

In comparing other fields of social policy, like family-, education-, or old age policy, labour market policy is unique; it makes up a relatively small fraction of total social spending (cp. Nikolai 2012), but it is crucial for a social investment strategy (De Deken 2012). The OECD and Eurostat, as providers of cross-country harmonised spending data, register LMP spending as expenditure specifically aimed at three groups: unemployed persons, persons at risk of unemployment, and inactive persons who would to be employed (Eurostat 2013: 7-8). Total LMP expenditure can be subdivided into three categories (13): service, measures and supports. Services comprise public employment services (PES), including interactions with job seekers and the administration of measures and supports. Measures include training, employment incentives, supported or sheltered employment and rehabilitation, direct job creation and start-up incentives.^[4]Supports include out-of-work income maintenance and early retirement benefits.

In line with established terminology, measures will be referred to as active labour market policy (ALMP) in the analysis. They are designed to bring programme participants into employment. Supports will be referred to as passive labour market policy (PLMP). They “compensate individuals for disadvantage caused by labour market circumstance” (8). This distinction^[5] echoes the difference between social protection and social investment: the latter approach advocates that social policy should help persons succeed in the market, rather than just compensating unwanted market outcomes: preparing instead of repairing (Morel et al. 2012: 9; Nikolai 2012: 92). Yet, the match is not as clear as it may seem: on the one hand, not all ALMP is equally social-investment friendly, and on the other, PLMP can also have a forward-looking character: it supports workers’ capacity to invest in themselves and their families in times of hardship (Nicaise/Schepers 2013: 213), and compensatory payments also prevent the loss of human capital by allowing workers to wait for a job that matches their acquired qualification. De Deken (2012) points out the importance of the duration of payments; however, if social benefits allow a long-term exit from the labour force, this can incite workers to hold onto outdated skills, and it conflicts with the aim of social investment to increase employment.

As for ALMP, several scholars have pointed to the crucial difference between enabling and constraining measures, in the words of Bonoli (2012: 183) “those which are about improving human capital, and those which use essentially negative incentives to move people from social assistance into employment” (cp. also Vlandas 2013: 5). As both traits are not mutually exclusive, Bonoli proposes to consider them as dimensions by which ALMP measures can be classified. He terms the first dimension “human capital investment” and the second “pro-market employment orientation”.^[6] Bonoli chooses to consider only those ALMPs that score on both dimensions, i.e. they boost human capital and are connected to regular jobs, as genuine examples of social investment. Such measures reduce the burden of unemployment both for the concerned person and society, without substituting unemployment for jobs of low quality and low productivity. Among the four types of ALMP that Bonoli discusses, he argues that two qualify as genuine examples of social investment: “employment assistance” (facilitating contacts between job seekers and potential employers) and “upskilling” (investing in human capital). “Incentive reinforcement” (mere activation) and “occupation” (keeping jobless people busy) are not, or much less, accepted as examples of social investment (Bonoli 2012: 185-186).

Each of Bonoli’s types of ALMP comprises one or several concrete ALMP spending categories. Among the ones for which data is available (see above), we can attribute training to the “upskilling” type, direct job-creation and supported or sheltered employment and rehabilitation to the “occupation” type, as well as employment incentives and start-up incentives to the “employment assistance” type. Bonoli (2012: 187) also includes expenditure on PES and administration in the “employment assistance” ALMP type, but this spending category also includes the cost of administering supports, among other things, so further disaggregation is required. As for “incentive reinforcement”, it is a type of ALMP that cannot be adequately expressed by spending data.^[7]

A severe drawback of any classification of ALMP spending lies in their internal heterogeneity, caused by unobserved design features. For example, Vlandas (2013: 6) points out that in the case of spending on services, it is “entirely unclear whether this benefits the unemployed or whether PES is used to monitor benefit recipients more closely”. Similarly, Clasen et al. (2016: 30) highlight that spending on public job-creation schemes can mean substantially different things according to whether they are well-paid or not: social participation and preservation of human capital, or just a test of availability and “workfare” (Nelson 2013: 259). Even training can be reduced to a mere way of keeping unemployed persons “busy” if the content and presentation of courses is poor or if acquired competences cannot be put to use due to mass unemployment (Bonoli 2012: 199). Clasen et al. (2016: 30) view these ambiguities as the reason why “controversies over the determinants of ALMP effort have proved rather difficult to resolve”. In the case of descriptive analyses like the present one, these ambiguities impact the validity of comparisons between countries over time.

3 Data and method

The basis of this investigation is a database on LMP expenditure provided by Eurostat and the EU-Commission’s DG Employment, Social Affairs & Inclusion (DG EMPL).^[8] Our sample includes 27 countries.^[9] All are members of the EU except Norway; some EU countries (Croatia, United Kingdom) are not included due to a lack of data in some years. Our observation period reaches from 2006, the eve of the Great Recession, to 2015, the year with the most recent data available at the time of writing. We will compare the beginning of our observation period with the end. The timing of our study has the advantage that we do not compare spending between “normal” times and “extraordinary” times: we consider that no conclusions on spending trends and welfare state recalibration can really be drawn from taking the peak of the Great Recession as a reference point. Instead, we compare the situation before 2008 with a moment when the grip of the crisis on European countries began to loosen.^[10]

4 Active and passive expenditure – does one crowd out the other?

We begin our analysis with a comparison of spending on active and passive labour market policies; types of ALMP will be distinguished only in section 5. The main question we will analyse here is whether active expenditure crowds out passive expenditure under the impression of the social investment discourse, or whether passive expenditure gets even more important as a means to repair the social damage done by the crisis.

Looking at average spending amounts across our country sample, there are three observations to make. First, expenditure on passive LMP dominates active expenditure in all observed years, while labour market services are the smallest spending category (ETUI 2018: 40). This is shown in Figure 1, which depicts relative spending, thus the mean share of countries’ wealth dedicated to LMP. On the left is aggregate expenditure, on the right need-adjusted expenditure, i.e. deflated by the share of the population wanting to work. Aggregate PLMP spending is almost exactly double ALMP spending both in 2006 and 2015, at the mean across our sample countries. We see a small increase of ALMP relative to PLMP per person wanting to work (from 49.3% in 2006 to 53.7% in 2015). A sustained shift in the relationship between average active and passive spending is not visible in either direction in the observation period. We take this as a sign that there is no sharpened “resource competition” (Vandenbroucke/Vleminckx 2011) between active and passive spending.

Figure 1

Relative LMP spending, European* mean, 2006–2015

Source: Eurostat/DG EMPL, own calculations. *includes AT BE BG CY CZ DE DK EE EL ES FI FR HU IE IT LT LU LV MT NL NO PL PT RO SE SI SK

However, our second observation is that the predominant short-term reaction to the crisis was in terms of supports rather than measures: for the years 2009 and 2010, we see a much stronger bulge in PLMP spending as compared to ALMP. The largest overhang is observed in 2009, where passive spending amounts to approximately 2.5 times the active spending volume. A plausible explanation lies in the distinction between “mandatory” and “discretionary” spending (Streeck/Mertens 2011): job seekers are legally entitled to a certain level of supports for a certain period of time, while training programmes, employment subsidies etc. are not automatically offered by the PES.

Our third observation is that aggregate expenditure rises a bit in the course of the observation period, while spending per person wanting to work declines. On the aggregate level, PLMP and to some extent also ALMP expenditure are greater at the end of the observation period than at its beginning, whereas need-weighted expenditure (right side of the graph) is smaller.^[14] This means that LMP becomes heavier fiscally, while the value of measures and supports that reaches the individual job seeker shrinks. Depending on the country, the latter phenomenon can be explained by policy reform (“retrenchment”), but also by a lack of reform (“policy drift”, cp. ETUI 2018: 41; Streeck/Thelen 2005: 24–26), e.g. in a situation where statutory entitlements run out due to sustained labour market strain.

The averages just reported are unweighted, small countries thus weigh more than the size of their population or their GDP would suggest. In the following, we will look at spending trends at the country level.^[15]

4.1 Spending trends at the country level

There is a bulge of LMP spending in most countries during the crisis years from 2009 on, but the steep increases are soon reversed and spending returns to its pre-crisis level or trajectory in many countries. To determine the longer-term direction of LMP spending, we abstract from the special situation in the midst of our observation period and compare amounts spent in the years 2006 and 2007, the eve of the crisis, with amounts in 2014 and 2015.^[16]

Table 1 reports trends of absolute expenditure by country, distinguishing between ALMP and PLMP. Aggregate expenditure is shown in the left half of the table, need-adjusted expenditure in the right half. For aggregate expenditure, a rising trend is evident: 19 out of 27 observed countries spend at least 5% more on ALMP at the end of the reference period, and 16 out of 27 raise their PLMP expenditure by at least 5%. In 10 countries, both ALMP and PLMP expenditure increase. At the opposite side, there are only two countries with decreased ALMP and PLMP expenditure on aggregate, Germany and Romania. No country shows overall stability of aggregate expenditure.

Table 1

Trends in absolute LMP spending (PPS) on aggregate, from 2006/2007 to 2014/2015

PLMP	on aggregate ALMP	per person wanting to work ALMP
	rising	Stable	falling	rising	stable	falling
rising	AT CY EE FI FR, IE LT LU SI SK	NO	BG ES IT LV NL	CZ EE, LV SI	LT	BG DE IE IT NO
stable	CZ EL MT PT	–	–	AT SK	FI	–
falling	BE DK HU PL SE	–	DE RO	HU MT PL SE	–	BE CY DK EL ES FRLU NL PT RO

1. Source: Eurostat/DG EMPL, own calculation. Stability defined as no more than 5% difference between initial and final amount.

From the need-adjusted perspective (right side of Table 1), the picture is more balanced, but the tendency points downward: there are 10 countries with rising and 14 with falling PLMP expenditure per person wanting to work, and there are 10 countries with rising and 15 with falling expenditure on ALMP. The largest group, 10 countries, show both falling ALMP and falling PLMP expenditure. Finland has stable expenditure within our definition of stability as less than 5% deviation from the initial value.

Cyprus, France and Luxembourg are highlighted in bold letters in Table 1, they feature a combination of rising aggregate and declining need-adjusted expenditure for both active and passive ALMP. Our third observation above applies to these three countries in particular.

Table 1 also illustrates the alignment or dissociation of ALMP and PLMP in the individual countries: in many cases, ALMP and PLMP either rise jointly or fall jointly, or both remain stable (12 in the aggregate, 15 in the need-adjusted view). In contrast, there are (10, respectively 9) countries where ALMP and PLMP per person wanting to work take opposite directions. Yet, due to the approximately equal number of countries on either side, our observation period does not seem a time of rising active interventions at the expense of passive interventions in Europe, or vice versa. This confirms Kuitto’s (2016: 453) “mixed” findings (see above).

Table 1 highlights in italics the countries where the mix of active and passive LMP does change, both in aggregate and in need-adjusted terms: Hungary, Poland and Sweden feature rising ALMP while PLMP spending declines, while in Bulgaria and in Italy, more is spent on PLMP while ALMP expenditure shrinks.

Figure 2

Absolute changes in relative need-adjusted ALMP and PLMP expenditure from 2006/2007 to 2014/2015, by country

Source: Eurostat/DG EMPL, own calculations

An important aspect not covered by Table 1 is the magnitude of changes. We therefore complement it by Figure 2 and Figure 3, which plot absolute, respectively relative changes of relative expenditure,^[17] with PLMP on the horizontal and ALMP on the vertical axis. While most countries cluster around the origin, we see strong decreases in real spending in Denmark, Belgium and Luxembourg (Figure 2). In the former two countries, decreases in ALMP are not nearly as important as decreases in PLMP: these countries (with large relative spending levels and large absolute changes, even if they are small countries) have a sizeable influence on average absolute changes of spending in Europe (Figure 1).

In Figure 3, the graph with relative changes,^[18] four countries stand out: Hungary and the Czech Republic strongly increase their active expenditure, compared to their initial spending, while PLMP remains rather stable. Estonia boosts both active and passive expenditure. Bulgaria considerably increases its PLMP spending, relatively, while decreasing ALMP. These countries pull average relative changes of spending upwards in Europe. Relative changes are very sensitive to the starting level, which is quite low in the for aforementioned Central and Eastern European countries. The heterogeneity of countries in our sample requires European averages to be interpreted very prudently.

Figure 3

Relative changes in relative ALMP and PLMP expenditure from 2006/2007 to 2014/2015, by country

Source: Eurostat/DG EMPL, own calculations

It has become clear that our observation period is marked by quite heterogeneous LMP spending trends in our 27 sample countries. Let us examine, for illustrative reasons, two country cases with contrasting developments concerning the active/passive LMP mix: Italy and Hungary. Italy (Figure 4) is a prominent example of weak and further declining social investment. Kazepov and Ranci (2017: 98) describe Italy as an “adverse case” of social investment, “a country unfriendly to SI”. In our observation period, passive expenditure increased tremendously in Italy with is a particular jump in the year 2009, and sustained for some years afterward, while active expenditure experiences a steady decline. In the last year observed, however, both trends are reversed.

Figure 4

LMP expenditure trends in Italy, 2006–2015

Source: Eurostat/DG EMPL, own calc. *Amounts calculated at 2011 prices, divided by 2011 PPP.

In contrast, Hungary (Figure 5) is a rather spectacular case of rising ALMP, its policy arrangement changes from a dominance of passive to a dominance of active expenditure. As we will see below (section 5), this is due to a boost in public job-creation. Simultaneously, reforms to lower unemployment benefits are also reported (Bouget et al. 2015: 38).

Figure 5

LMP expenditure trends in Hungary, 2006–2015

Source: Eurostat/DG EMPL, own calc. *Amounts calculated at 2011 prices, divided by 2011 PPP.

Beyond diverging trends, the two preceding graphs also show that levels of ALMP and PLMP spending can diverge between countries. In the following section, we will again examine all our sample countries: are there other cases beside Hungary where more is spent on active than on passive LMP? Can we distinguish between ALMP- and PLMP-countries in Europe?

4.2 Are there “ALMP-” and “PLMP-countries” in Europe?

Figure 6 and Figure 7 compare spending structures synoptically for our sample at the beginning and at the end of our observation period. The solid diagonal line marks all points where ALMP and PLMP spending are equal in size. All dots above it stand for higher ALMP, while all dots below the line stand for higher PLMP expenditure. There are two dashed lines above and two below the diagonal. The two upper dashed lines mark an ALMP expenditure 50% respectively 100% higher than PLMP spending, and vice versa.^[19]

Clearly, a majority of countries are below the diagonal line, both at the beginning and at the end of the observation period. In 2006/2007 (Figure 6), most countries are situated in the area where PLMP is between 50% and 100% higher than ALMP, thus between the two lower-most dashed lines. Three countries go beyond this and reach into the area where PLMP is more than double ALMP expenditure: Belgium, Cyprus, Malta, and also Portugal. Some years after the Great Recession, in 2014/2015 (Figure 7), there are numerous countries with a PLMP expenditure more than twice as high as ALMP expenditure: Italy, Spain, Germany, and the Netherlands have joined, and also Estonia and France.

Figure 6

Active and passive relative need-adjusted LMP expenditure, mean of 2006–2007

Source: Eurostat/DG EMPL, own calculations

Norway, the only country with at least medium-sized spending leaning toward ALMP in 2006/2007, balances its expenditure between ALMP and PLMP in 2014/2015. Lithuania and Bulgaria, also leaning toward ALMP in 2006/2007 (though at a much lower spending level), have split by 2014/2015: Lithuania stays ALMP-leaning, while Bulgaria switches to the PLMP side. Denmark also has balanced spending at the end of the observation period, having increased their active share of LMP. Sweden and Hungary stand out for having distinguished themselves with high levels of ALMP spending, as they are far above the diagonal in 2014/2015. This also holds for Poland and the Czech Republic; however, their spending volume is much smaller. In sum, we observe a certain fanning out during the observation period, as distinctive ALMP and PLMP profiles emerge.

Figure 7

Active and passive relative need-adjusted LMP expenditure, mean of 2014–2015

Source: Eurostat/DG EMPL, own calculations

Another observation to make from the above two graphs is that the big spenders on PLMP reduce their spending significantly. Yet, the distinction between strong and weak LMP remains important, maybe more important than the distinction between ALMP- and PLMP-countries: we can see this in Figure 7 if we focus on the dotted lines; they mark median spending levels and divide the area into four quadrants.^[20]

In 2014/2015, there are countries clearly below (above) the median in ALMP spending, but above (below) in PLMP: Hungary and the Czech Republic on the active side and Italy and Spain on the passive side. This makes that the two countries in the ALMP quadrant are Eastern and the two in the PLMP quadrant are Southern European. Yet, the 2nd (high ALMP and high PLMP) and 4th (low ALMP and low PLMP) quadrants are by far the most crowded. Our observations thus confirm for the field of labour market policy the findings of Kuitto (2016: 454) that “rather than a trade-off, there is (still) a distinction between those countries spending less on welfare policies in general (most of the CEE countries) and those spending more in both compensating and social investment policies and offering more generous benefits” (in addition, cf. Bruno/Rovelli 2010: 670). In line with Bonoli’s findings (see above), welfare regimes are partly reflected here, with the biggest spenders being Nordic and Continental countries, and the smaller spenders being Southern and Eastern European. There are, however, considerable differences within these groups.

Summing up this section, there is no shift towards either active or passive expenditure in Europe as a whole between 2006 and 2015, though policy arrangements change in some individual countries and spending profiles get a bit more polarised between active and passive spending. Some Eastern European countries strongly increase their spending on ALMP (and partly PLMP) during the observation period, while some Western European countries strongly decrease their PLMP (and partly ALMP) spending. Spending on labour market policy remains drastically unequal across Europe. Moreover, we observe a reduction of spending per person wanting to work in most countries even though aggregate LMP spending increases, which has to do with rising labour market challenges.

5 What do governments choose from the ALMP toolbox?

This section examines active labour market policies in a disaggregate way, based on the established spending categories. Our main interest lies in the question whether the trend towards “employment assistance”, identified by Bonoli (2012), continues after the beginning of the crisis in 2008. Alternatively, countries could take recourse to direct job-creation, like they did as of the mid-1970s under the impression of the oil shocks and mass unemployment (188). Equally possible would be a scenario of investment in training, as private-sector jobs became scarcer during the Great Recession and training remains a way to relieve the labour market.

We start again by looking at (unweighted) average spending across our European sample countries. Looking first at aggregate (relative) spending (left part of Figure 8), there is a clear hierarchy: training is most important, employment incentives and the category sheltered and supported employment and rehabilitation share rank number two. Much less is spent on direct job-creation and again much less on start-up incentives. In the per-head view (right part), the advance of training over employment incentives and sheltered and supported employment and rehabilitation is smaller than in the aggregate view, else the relationships between spending categories are similar.

Figure 8

Relative ALMP spending, European* mean, 2006–2015

Source: Eurostat/DG EMPL, own calculations. *includes: AT BE BG CY CZ DE DK EE EL ES FI FR HU IE IT LT LU LV MT NL NO PL PT RO SE SI SK

Let us concentrate for a moment on the change of expenditure as percent of GDP between the first two and the last two years observed: on aggregate, there is rising expenditure for the categories of employment incentives, sheltered and supported employment and rehabilitation, and direct job-creation. There is approximately stable aggregate spending on training and start-up incentives. Deflated by the number of persons wanting to work, the evolution of expenditure appears in a more negative light.^[21] Training in particular declines during the course of the observed period, but also employment incentives and sheltered and supported employment and rehabilitation, in spite of the rise in aggregate spending. Only for direct job-creation – where aggregate expenditure rises most – do we also see a rise in need-adjusted spending, and it mostly occurs after the year 2011.

These trends average out over heterogeneous developments in the individual sample countries, and in some cases they are driven by particularly strong changes in just a few or even just one country.^[22] Above all, this holds for direct job-creation, where the observed rise is almost exclusively due to Hungary. Figure 9 depicts the Hungarian ALMP spending profile: beside the skyrocketing direct job-creation, other ALMP categories continue to play a minor role. This policy shift (connected with declining PLMP, see above) has sparked a lot of critical (“workfare”) and few prudently optimist comments (see Lakner/Tausz 2016; Asztalos Morell 2014; Szikra 2014; Keller et al. 2016). Beside Hungary, direct job-creation is the most significant measure also in Slovenia and Bulgaria, though in Bulgaria it has declined recently.

Figure 9

Spending on ALMP in Hungary, by measure

Source: Eurostat/DG EMPL, own calc. *Amounts calculated at 2011 prices, divided by 2011 PPP.

Other cases where changes in a small number of countries influence the European average are the particularly strong declines of training expenditure in Norway and Germany (aggregate), respectively Luxembourg, Norway and Denmark (need-adjusted). Aggregate spending on employment incentives, rising on average, benefits from strong gains in Luxembourg and Denmark, while Spanish expenditure withers. From a need-adjusted perspective, there are strong declines not only in Spain, but also in Luxembourg and the Netherlands, but they are partly compensated by notable increases in at least seven countries. The spending category sheltered and supported employment and rehabilitation benefits from a favourable trend in Sweden, both on aggregate and per person wanting to work. Another important country here is Denmark, with rising aggregate but declining need-adjusted expenditure. The Netherlands also show a strong decline of need-adjusted spending in this category. As for start-up incentives, the average of European countries is moved most strongly by the marked decline of spending in Germany.

6 Summary and discussion

Our descriptive analysis of European countries’ spending on labour market policies between 2006 and 2015 shows that there is no real trend towards social investment in this field, despite the discursive dominance of the social investment paradigm in Europe since the early 2000s and its reaffirmation by EU-Commission in 2013. This shows both in the relationship between passive and active spending and in the composition of active spending. Concerning levels of spending, labour market policy gets fiscally more demanding on aggregate, yet the amount of resources made available to the average individual job seeker shrinks in recent years. Spending also remains very unequal between European countries.

More in detail on the relationship between active and passive spending, we find that on average across European countries, passive spending dominates throughout our observation period, and that its advance over active spending is rather stable, apart from an even stronger emphasis on passive spending in the most turbulent crisis years. There is just a minority of countries that spend more on active LMP. In 2014/2015, it is Denmark, Sweden, Hungary, the Czech Republic, and Poland, while the majority of our 27 sample countries are inclined towards passive expenditure. We observe changing – and somewhat sharpening – spending profiles in some countries in recent years, but no common trend toward either active or passive LMP.

The latter would not preclude an increasingly social-investment friendly character of spending if we consider the ambivalent nature of the broad categories of active and passive expenditure with regard to social investment. Yet, we also look at the components of ALMP spending that are strengthened, and weakened, during the observation period. Training is at the heart of social investment. On average across European countries, social investment was clearly the most prominent ALMP spending category at the beginning of our observation period in 2006. Yet, it gradually lost this distinguished status: by 2015, little additional funding is spent on training in comparison with the categories of employment incentives, sheltered and supported employment, and rehabilitation. This is striking provided the EU’s emphasis on social investment and ist objective of fostering more knowledge-based economies.

In contrast, what is paradigmatic for the present phase of labour market policy is the aim of maximising the number of persons in employment, or who are at least economically active. Motivated by the objective to reduce the dependency ratio, this aim has also been discussed as a defining feature of social investment (De Deken 2013). It is thus not surprising that employment incentives become more important in many of our sample countries, relative to other measures. Often, employment incentives consist of temporary wage subsidies paid by the state either to workers (who can thus accept lower wages) or employers (who save money on salaries). We argue that this risks promoting the creation of particular low-productivity jobs connected to a knowledge-based economy only through their role in creating an environment of affordable services for highly-productive knowledge-intensive job holders. The combination of weakening training expenditure and strengthening employment incentives suggests that an increased polarisation of the labour market is politically acceptable as the cost of rising employment rates. The role that activation in the form of benefit conditions and sanctions may play here cannot be verified with spending data. The aspired additional jobs are predominantly dependent employment in the private sector: there is no extension of public job-creation except in some (not all) countries with particular labour market problems, and start-up incentives are at a broad downward trend.

Comparing levels of spending between European countries, we found that the classification of countries with active- versus passive-leaning labour market policies is not the most relevant distinction. Active and passive expenditure are highly correlated, with passive LMP receiving more resources in most countries. Rather, we can discern between countries spending little and countries spending more on LMP. This holds despite some convergence, following from a boost in ALMP in some Eastern European countries with very low initial spending and from strong reductions of PLMP spending in some countries with Continental welfare states and high initial spending.

Notwithstanding, we do observe some sort of “resource competition”: not between types of LMP, but between job seekers. In most countries, the amount of resources granted to the average job seeker shrinks as the crisis unfolds after 2008. In the case of ALMP, this affects the value of training programmes, job subsidies, etc. We explain this with a rising demand for interventions in the face of partly (pre-)fixed budgets. Concerning PLMP, this does not necessarily have to do with welfare reform, even though social spending was curtailed in many European countries to make up for the fiscal losses connected to bank bailouts, stimulus packages and economic slowdown (ILO 2015: 128). Declining passive expenditure per person can also be a consequence of individual benefit entitlements running out in a phase of enduring labour market strain. We conclude that spending per person is not, as Hemerijck et al. (2013: 21) argue, “largely independent of the state of the economy”. Presumably, however, reductions in spending per person would have been effectuated even without the crisis in some countries. According to a recent report to the European Commission (Bouget et al. 2015: 37), cuts sometimes predate the 2008 crisis and do not concentrate on the most affected countries. In our sample, the country with the most important decline in absolute spending per person wanting to work is Luxembourg.

Additional Material: the online version of this article offers additional material (https://doi.org/10.1515/zsr-2018-0018)