Mixing apples with oranges? Partisanship and active labour market policies in Europe

Inappropriate inclusion of programmes

First, public employment services and administration (PES) include placement and related services and the administration of benefits. Spending on PES therefore includes the cost of employing people to administer benefits and organizing the placement services. It is entirely unclear whether this benefits the unemployed or whether PES is used to monitor benefit recipients more closely. For instance, the 2001 plan of help to return to employment made it compulsory for the unemployed in France to ‘take an ‘acceptable’ job’ (Barbier, 2009: 178). The impact of this programme on unemployment and employment is also contested. Estevão (2003: 15) finds that spending on PES is associated with lower employment rates.

Second, ALMPs include spending on job rotation and job sharing. This is the sort of programme that Germany’s Kurzarbeit schemes during the recent economic crisis involved. This is a way to prevent redundancies rather than to reduce unemployment or increase employment rates. Third, expenditures on start-up incentives entail helping the unemployed to start their own business as self-employed workers. Again, the promotion of self-employment has little to do with the interests of labour or with worker–employer relationships. Note that spending on job rotation and start-up incentives represent in any case a very small share of aggregate ALMPs.

Direct job creation

‘Direct job creation’ is a fourth type of ALMP. This sort of programme has a much longer history than ALMPs as an integrated conceptual category. Germany was implementing ‘national job creation policies’ as early as the 1920s (Janoski, 1990: 63). In the 1970s, Sweden expanded public sector employment and used ALMPs to provide an occupation for unemployed workers. This included ‘temporary jobs arranged mostly in the public sector’ (Bonoli, 2010: 18). These job creation schemes were therefore classic interventions on the demand side of the labour market. In 1979, the Danish left created a job offer scheme guaranteeing a 7-month job to the long-term unemployed (Huo, 2009: 105). The Dutch PvdA’s left party also offered government subsidized jobs in the public sector. For instance, the so-called ‘Melkert job schemes’ directly created jobs (Huo, 2009: 124, 125).

In France, the left-wing government introduced ‘collective utility work’ in 1984 (Lødemel and Trickey, 2000). Similarly, it was the socialists who in 1997 introduced the “Nouveaux Services Emplois Jeunes” providing 18–30-year-olds with 5 years of full time employment (Daguerre, 2007: 116). The impact of this initiative on spending on direct job creation in France, where the Socialists were in power from 1988 to 1992 and 1997 to 2002, can be seen in Figure 1. In France, the underlying public rationale of these schemes was both to create jobs in the context of large unemployment and to address ‘unmet needs in the public sector’ (Lødemel and Trickey, 2000: 60). Governments that initiated these schemes did so with the official objective to deal with mass unemployment through demand side programmes to ‘keep jobless people busy’ (Bonoli, 2010: 441).

OPEN IN VIEWER

Figure 1.

Spending on direct job creation at a percentage of GDP in France.

Thus, direct job creation involves the use of public funds to directly create employment. Most often these jobs are created in the public or non-commercial sector. This measure therefore directly reduces unemployment. There is evidence that direct job creation was effective in increasing employment in the 1990s (Estevão, 2003: 15). By reducing unemployment, direct job creation may therefore have a positive impact on wages. As a result, this measure is consistent with the interest of both the employed and unemployed workers. Social democratic parties can therefore be expected to support direct job creation.

Employment incentives and rehabilitation

Spending on ‘employment incentives’ constitutes a fifth type of ALMPs. This includes both recruitment incentives and employment maintenance incentives. This is part of a broader agenda that reinforces incentives for the unemployed to take up jobs. Economists (Phelps, 1997; Snower, 1997) have stressed the role of (targeted) employment subsidies in reducing unemployment and making low wage workers better off.

However, this type of ALMP may also put downward pressures on wages in private sector employment. This occurs through two mechanisms. First, by subsidizing low wage work in the private sector, this makes it more interesting for employers to offer low wage jobs. Such a substitution effect is consistent with some of the empirical literature (Calmfors et al., 2001). Second, this programme rewards the acceptance of any jobs by the unemployed. This makes it more likely that the unemployed would take up jobs that they would otherwise not accept.

Thus, employment incentives may promote low-wage work and make employers substitute non- subsidized labour by subsidized labour. This means that ALMPs may become ‘financial subsidies that firms exploit for hiring cheap labour’ (Huo, 2009: 111; Martin and Swank, 2004). This is not likely to be popular with core social democratic voters. This concern of a potentially detrimental effect of employment incentives on the type of employment has for instance been voiced by French trade unions (Naton, 2009). Similarly in Sweden, the social democratic position was that ‘the state should not subsidize or encourage low wage employment’ (Huo, 2009: 116).

Liberals as well as Conservatives have supported the reinforcement of incentives (Bonoli, 2010). This type of programme promotes market mechanisms and reduces unemployment by raising incentives, which is consistent with Liberal and Conservative ideology. For instance, in 1990, the Danish centre-right government introduced a scheme that promoted the young unemployed to participate in activation (Huo, 2009: 104). The impact of this initiative on spending on employment incentives in Denmark, where a Liberal Conservative coalition ruled from 1982 to 1993 and after 2001, can be seen in Figure 2. Similarly, the centre-right government in the Netherlands introduced a programme, the Loonsuppletie, which granted the unemployed a temporary wage supplement. This was only awarded where the wage of the new job was inferior to that of the previous job (Huo, 2009: 123).

OPEN IN VIEWER

Figure 2.

Spending on employment incentives as a percentage of GDP in Denmark.

The programme ‘supported employment and rehabilitation’ promotes mobility on the part of the unemployed to get into employment. This is done for instance by providing mobility grants to unemployed workers who accept moving region to seize an employment opportunity. It also consists of ‘subsidies for the productive employment of persons with … a long-term … reduced capacity to work’ (Organisation for Economic Co-operation and Developement, 2010). This programme makes it more likely that a job seeker in a given region would move to another region.

Thus, supported employment and rehabilitation has similar effect to employment incentives. Most often the stated aim of these programmes is to promote re-entry into the labour market (Bonoli, 2010). Both measures incentivize the unemployed to take jobs, thereby potentially putting downward pressures on wages. Rueda (2007: 74) recognized this possibility when he contends that ALMPs ‘promote entry into the labour market of outsiders who will underbid insiders’ wage demands’. I argue that this is only the case for the aforementioned programmes and not for the other types of ALMPs.

Training schemes

Training schemes in ALMPs have the aim of raising human capital. This was, for example, the main reason for the promotion of ALMPs by Swedish social democrats in the early 1950s. The Rehn Meidner model involved a solidaristic wage system that priced out low productivity industries. The resulting unemployed could then be retrained and incorporated into high productivity industries (Huo, 2009). Thus, contrary to measures that incentivizes unemployed people to take up jobs, training schemes aim to enable unemployed people to re-skill thereby increasing their chances of successfully attaining their preferred employment position.

It is precisely because training ALMPs raise human capital that Boix (1998) argues that the Left would support these programmes. By raising the productivity of unemployed people, this allows social democrats to raise both economic growth and equality. Conversely, the conservatives may see publically funded training as unnecessary. For instance, when the centre-right party took power in Sweden in 1991, they reduced spending on ‘skill and competence development’ in ALMPs (Huo, 2009: 113). And it was a left government that introduced the ‘vocational training programme’ in Norway (Huo, 2009: 120).

However, some historical evidence partly contests Boix’s argument. In 1963, the Gaullist party in France attempted to introduce training schemes in the unemployment benefit system. This was partly opposed by unions who resented additional state involvement in unemployment insurance (Bonoli, 2010; Clegg, 2005). The promotion of vocational training to address unemployment also occurred around the same time in Germany. This took the form of the 1969 employment promotion act which was proposed by the CDU–SPD coalition (Bonoli, 2010).

As Bonoli (2010: 17) concludes, training was supported by very different political parties: ‘Swedish Social democrats, the French Gaullists, Italian Christian democrats and a coalition government in Germany’. In addition, training may not be relevant for unemployment or employment levels. This echoes Estevao’s (2003: 15) conclusion that ‘training programmes for unemployed … adults seemed irrelevant’ for employment.

Partisanship and different ALMPs

This section has shown that there are important differences between ALMPs. I have identified three distinct types of ALMPs: employment incentives and rehabilitation, direct job creation and training schemes. From this discussion, my argument is that Social democratic parties, all other things being equal, support direct job creation but do not support employment incentives and rehabilitation. This is because direct job creation benefits the unemployed without putting pressure on employed workers, whereas employment incentives and rehabilitation may have adverse consequences for employed workers.

Historically, both social democratic and conservative parties may support training schemes. Theoretically, this is because training also matters to employers and hence these schemes are more likely to be driven by the coordination regime than partisanship. I therefore derive the following hypothesis and three observable implications.

H1 The control of the government by Social democratic parties is: (a) positively related to spending on direct job creation; (b) negatively related to spending on employment incentives and rehabilitation; and (c) not significantly related to spending on training.

The potential impact of welfare regimes and varieties of capitalism

Welfare regimes may affect the preferences of political parties for different ALMPs. There are three sets of reasons why welfare regimes affect political parties’ preferences for labour market policies. First, there are enduring historical differences in the sort of problems different regimes have faced. Long-term unemployment was traditionally much higher in continental Europe than in the Nordic cluster (Esping-Andersen, 1990: 152). For example, Norway and Sweden were among the few countries to achieve unemployment rates of about 2 to 3 percent during the post-war era (Esping-Andersen, 1990: 163). More generally, Scandinavian countries have much lower poverty and inequality rates than other European countries (Häusermann and Palier, 2008).

The second reason is that the ability to undertake policies may also be regime-dependent. Political parties also choose policies in the context of existing policy tools, which may differ significantly in different regimes. Scandinavian social democratic parties can expand public sector employment more than on the continent. Their tax revenues are larger, which allows them to spend much more on all welfare state policies. Similarly, the introduction of ALMPs in the 1960s in Scandinavia was made easier by the expansion of the welfare state at that time. Today, the long history of ALMPs in Sweden makes them an easy policy tool to expand. Later retrenchment may be prevented by the popular support these programmes have created. This is what Armingeon (2007: 913) calls the ‘regime specific predisposition of expanding ALMP’. The logic of the welfare state becomes partly independent of ‘temporal political power distribution’ (Armingeon, 2007: 914).

Third, there are different coalitions and ideology behind left-wing parties in different welfare regimes. For instance, the Scandinavian left drew its strength from a coalition between labour and the peasant movement. Subsidies for farmers were granted in exchange for a ‘full employment welfare state’ (Esping-Andersen, 1990: 30). The labour movement was therefore much more encompassing in Scandinavia than in continental Europe. The labour movement may also be more divided because of religious cleavages, as in the Netherlands, or between different left-wing parties, as in France (Clegg et al., 2010). This could imply that left-wing parties may promote different types of ALMPs in continental European countries from those in Scandinavia.

The main contender to the power resource approach, the Varieties of Capitalism literature (Hall and Soskice, 2001), underscores the importance of the type of capitalism for the sort of social policy and social protection that emerges (Estevez-Abe et al., 2001). Recall that the previous section suggested that there were no clear partisan drivers of training. This does not imply that training is irrelevant in other respects. Given the importance of skills for the industry, spending on this type of ALMP is likely to be driven more by the type of capitalism than by partisanship (Estevez-Abe et al., 2001; Hall and Soskice, 2001). If employers need workers to acquire specific skills, political parties may support training schemes for the unemployed.

For instance, coordinated market economies (CMEs) have ‘high levels of vocational educational and … industry specific and firm specific skills’ (Iversen and Stephens, 2008: 31). This is consistent with historical evidence that most political parties in CMEs supported training measures for the unemployed. There is also evidence that employers in Sweden had a strong interest in the development of training schemes (Swenson, 2002). Similarly, Danish employers were heavily involved in the extension of training schemes (Martin and Thelen, 2007: 24). Indeed, the drastic expansion of training programmes in the post-Second World War period was at least partly driven by a need to address important skill shortages (Bonoli, 2010). In line with this theory, continental and Scandinavian welfare regimes may have less different preferences or needs for training because both these regimes have broadly similar coordination structures.

Mapping ALMPs in different welfare regimes

Ferrera (1996) distinguishes between four types of welfare regimes: Scandinavian, continental, Liberal and Southern. For simplicity, Liberal and Southern regimes can be put together under the label ‘minimalist labour market policies’ welfare regimes. Their welfare regimes are smaller and less decommodifying than in the rest of the continent. Thus, the Scandinavian welfare regime comprises Sweden, Denmark and Finland. The continental cluster includes France, Germany, Austria, Belgium and the Netherlands. The minimalist category includes the UK and Ireland, Spain, Portugal, Greece and Italy.

Social democratic parties in Scandinavia are stronger than in other regimes. We should therefore expect the difference between spending on employment incentives and rehabilitation under a left-wing government and under a right-wing government to be larger in Scandinavia. In addition, if left-wing parties in continental Europe are less inclusive, they may care less about the adverse effects of employment incentives and rehabilitation. Workers in precarious employment may also be less important to social democrats in continental Europe than in Scandinavia.

The ability to expand public sector employment in Scandinavia is higher than in the other two regimes. If Social democrats are able to expand standard permanent public employment, they may not promote direct job creation. This is because employment secured by direct job creation is more temporary and precarious than standard public sector employment. The more inclusive nature of the labour movement in Scandinavia means that Social democrats may want to offer standard public sector jobs to the unemployed. The recent opposition of unions and social democrats towards a work scheme introduced by the centre right government in Sweden best illustrates that the Left may be strongly opposed to certain types of ALMPs (Kullander and Bjornberg, 2011). This is ultimately an empirical matter.

As argued in the previous section, the effect of partisanship on training is historically unclear and theoretically less important than the type of capitalism in which government policy making takes place. This is because training is particularly important to employers and economic performance more generally in CMEs (Hall and Soskice, 2001; Swenson, 2002). Thus, one can expect training spending to be higher in the two welfare regimes that encompass more coordinated market economies, such as the Scandinavian and, to a lesser extent, the continental welfare regime. From this discussion, the following hypothesis is posited.

H2 (a) The negative correlation between left parties in government and employment incentives and rehabilitation will be stronger in Scandinavia; (b) the control of government by left parties is positively correlated with direct job creation in continental and minimalist welfare regimes; and (c) training spending is higher in non-minimalist welfare regimes.

Description of data

Throughout, I rely on panel data for 14 European countries (EU15 minus Luxembourg) over the period 1990 to 2007, though data availability varies depending on variables and countries. The lower bound results from data availability, but also from the fact that most countries outside Scandinavia did not undertake significant ALMPs prior to the 1990s. The upper bound is determined by the onset of the crisis, which may introduce a bias after 2007.

With regard to my dependent variables, I rely on the OECD statistics database. The OECD provides annual data on spending as a percentage of GDP on my three types of ALMP. My first dependent variable is constructed by summing employment incentives and supported employment and rehabilitation. The second dependent variable is training measures in the database, and the third concerns the ‘direct job creation’ category. More details on the definitions and average values for each country of the dependent variables can be found in the appendix (Tables A1 and A2).

Given that ALMPs aimed to address problems that were driven by labour market and macroeconomic developments (see Bonoli, 2010), it is particularly important to control for the performance of the labour market and the economy. To control for the former, the analysis includes annual harmonized unemployment rates defined as the number of unemployed people as a percentage of the civilian labour force. The state of the economy is controlled by including annual GDP growth in percentages because higher growth of GDP may affect both the cyclical and the discretionary component of policies. Further, controlling for the degree of deindustrialization or trade openness did not alter the results. ²

The measure of the impact of partisanship is an updated version of the Schmidt index taken from the Comparative Political Data Set III, 1990–2009 (Armingeon et al., 2011). This calculates the political composition of the Cabinet. The original coding is from (1) hegemony of right-wing (and centre) parties through to (5) hegemony of social-democratic and other left parties. I have rescaled this variable to take values from −2 to +2.

Empirical model and estimation strategy

I construct two dichotomous variables to capture the impact of welfare regimes. The dummy ‘MIN’ is equal to 1 when the country has a minimalist welfare regime, and 0 otherwise. Similarly, the dummy ‘CONT’ is equal to 1 when the country is a continental welfare regime, and 0 otherwise. When both dummy variables are zero, the intercept then captures the impact of the Scandinavian welfare regime on the dependent variables.

The mediating effect of welfare regimes on the impact of partisanship on different ALMPs is captured by introducing an interaction term between my measure of partisanship and the set of without s, i.e.: regime dichotomous variables. Panel data regression analysis of the three dependent variables is used to test my hypotheses. More specifically, the general regression model that is tested is as follows:

y it = β 0 + β 1 PARTY i , t + β 2 PARTY i , t * MIN i , t + β 3 PARTY i , t * CONT i , t + β 4 MIN i , t + β 5 CONT i , t + β 6 HU i , t – 1 + β 7 GDP i , t – 1 + α t + ε i , t

where y_it is the dependent variable in country i at time t. There are three dependent variables expressed in levels: ³ spending on direct job creation; employment incentives and rehabilitation; and training as a percentage of GDP. With regard to the explanatory variables, PARTY is the index measure of partisanship that was explained earlier. MIN and CONT are dummy variables measuring the intercept effect of belonging to minimalist and continental welfare regimes, respectively. In addition, PARTY*CONT and PARTY*MIN capture how welfare regimes influence the impact of partisanship on the dependent variable. For instance, to assess the effect of left-wing power in continental Europe one should look at β₁+ β₃. Unemployment (HU) and GDP growth (GDP) are lagged one period. Last but not least, the α_t’s are t−1 year dummies and ϵ_i,t is the residual. Time dummies are included to capture time effects but statistical tests suggest that country dummies are not required. ⁴

The various diagnosis tests ⁵ have identified that there is heteroskedasticity, contemporaneously cross-sectionally correlated and auto-correlated errors. The appropriate estimation method in such a case is to carry out ordinary least square (OLS) with panel-corrected standard errors and Prais–Winston transformation (Hoechle, 2007: table 1, p. 4).

OPEN IN VIEWER

Table 1

The determinants of spending on employment incentives and rehabilitation (as % of GDP)

Columns	1	2	3
Government partisanship (from −2, right wing, to +2, left wing)	−0.0098*	−0.0097*	−0.0324**
Dummy variable for minimal welfare regime		−0.4459**	−0.4310**
Dummy variable for continental welfare regime		−0.3010**	−0.2892**
Minimal welfare regime*Partisanship			0.0312*
Continental welfare regime*Partisanship			0.0303*
Harmonized unemployment rate (lagged one period)	−0.0013	0.0021	0.006
GDP growth (lagged one period)	−0.001	−0.0006	0.001
Constant	0.2580*	0.5417**	0.5275**
Observations	242	242	242
R-squared	0.1438	0.2743	0.3029

Notes: *p<0.05; **p<0.01

The determinants of employment incentives and rehabilitation

The results for employment incentives and rehabilitation are presented in Table 1. My results suggest that the Left and spending on employment incentives and rehabilitation are negatively related. This is in line with the qualitative evidence and with the hypothesis presented earlier. This contradicts the empirical results of the power resource approach that has analysed the determinants of ALMPs. It also contradicts Boix’s (1998) contention that left-leaning governments necessarily undertake a supply side strategy, whatever its nature or the domain in which it is applied. This finding is consistent with the notion that left-wing parties will not want to spend more on employment incentives and rehabilitation because this may be neither beneficial for workers nor for the unemployed. This result is stable and significant across specification. Note that including employment protection legislation, union density, an index of wage coordination, or spending on passive labour market policies did not affect this result.

Second, the coefficients of both regimes dummies are negative and significant (columns 2 and 3). This is consistent with the notion that both Minimalist and continental welfare regimes spend less on employment incentives and rehabilitation than the Scandinavian regime. Omitting regimes might spuriously attribute the higher spending to partisanship. This is because Scandinavian countries have on average been ruled by social democrats more than in the rest of Europe.

Third, the interaction terms between partisanship and the type of welfare regime are positive and significant. This suggests that the Left in Scandinavia is indeed more negatively related to employment incentives and rehabilitation than is the case in the continental welfare regime. This finding is also consistent with Jensen’s (2010: 282) argument that ‘in countries that have a tradition of left-wing incumbency … right-wing governments compensate for the distrust of the public because of the popularity of the welfare state and strong vested interests’. As a result, there is a significant positive relation between ‘right-wing governments and social spending in traditionally left-wing countries’ (Jensen, 2010: 282). Last but not least, though significant and positive, the coefficient’s net effect of the Left in minimalist welfare regimes is not very large. This small positive coefficient is consistent with the adherence of the Left to the third way in the UK (Giddens, 1998).

The determinants of direct job creation

The results for the determinants of direct job creation are presented in Table 2. The coefficient for the minimalist welfare regime is negative but not significant, while it is positive but not significant for the continental regime (columns 2 and 3). This cannot conclusively confirm that the continental welfare regime spends more on direct job creation than the other two regimes, regardless of partisanship. However, this would be consistent with earlier qualitative evidence that shows the tendency of the continental regime to reduce labour supply. Early retirement schemes in the 1980s, for example, fulfilled such a role. The fact that the minimalist welfare regime may spend less than the other two regimes would be in line with their smaller welfare states.

Moreover, as shown in columns 1 and 2, the coefficient for partisanship is not significant. However, in the fully specified model (column 3) the power of the Left does have a significant negative coefficient in Scandinavia. This means that left-wing parties in Scandinavian countries are associated with less spending on direct job creation. By contrast, the impact of the Left in Minimalist and continental welfare regimes is significant and positive. Both results are in line with the hypothesis that in the continental welfare regime left-wing parties will favour direct job creation.

It is important to recall that ALMPs are targeting spending at the unemployed. In Scandinavia, a large and expanding public sector may have played the role of direct job creation in the continental regime. Iversen and Cusack (1998) have argued that the large public sector in Scandinavia made it possible for them to achieve the twin objective of employment and equality. Thus, while left-wing parties will not support employment incentives and rehabilitation, these results show that left-wing parties will not oppose all ALMPs in all regimes. Contrary to Rueda’s (2007) findings, the Left does support some ALMPs, provided that this does not hurt employed labour.

The determinants of training schemes

The results for spending on training schemes are presented in Table 3. First and foremost, the coefficient of partisanship is not significantly related to training schemes, regardless of the specification. Thus, spending on training schemes is not driven by the power of left-wing parties. This contradicts Boix’s (1998) argument that left-wing parties will necessarily spend more on training as part of a broad supply side strategy. It is prima facie consistent with Rueda’s (2007) findings that social democratic parties do not matter.

A second reason is that the welfare regime could itself fully determine the amount spent on training. This is partly consistent with our results. Indeed, the coefficient of the dummy variable for the continental welfare regime is negative and significant. Moreover, the coefficient of the dummy variable for the minimalist welfare regime is much more negative than that of the continental welfare regime. Thus, the Scandinavian welfare regime spends, all other things being equal, more than the continental welfare regime, which spends more than minimalist welfare regimes. This is in line with the expectation that training should be higher in CMEs than in liberal market economies.

The differences of effects between Scandinavian and continental welfare regimes require more explanation. The continental regime encompasses economies such as France that are in fact closer to being a mixed market economy (MME) than a fully coordinated market economy. Training may hence be less important in MMEs than in CMEs, which is then captured by differences in the results for Scandinavian and continental regimes. Another possibility is that differences in the way training systems are organized within CMEs has implications for the amount of spending that is channelled towards training the unemployed. While Scandinavian CMEs such as Sweden and Norway rely on vocational colleges, continental CMEs rely instead on the dual apprenticeship system (Estevez-Abe et al., 2001: 170, 171). If much more of the training in continental CMEs is provided privately by the firm, this may explain why public spending on training for the unemployed may be lower than in Scandinavia.