Open Access

Active labour market policies in a recession

IZA Journal of Labor Policy20121:7

https://doi.org/10.1186/2193-9004-1-7

Received: 4 September 2012

Accepted: 7 October 2012

Published: 29 November 2012

Abstract

A major labour market challenge following the Great Recession is to avoid an increase in long-term unemployment and thus the structural unemployment rate. Active labour market policies may play an important role in this respect. We consider these issues with outset in Denmark which is a country relying extensively on active labour market policies. It is also characterised by high flows into and out of employment, implying that most unemployment spells are rather short and the unemployment rate is relative low. These features have been maintained despite the crisis, and in this sense the Danish flexicurity model has coped reasonably well with the crisis. The Danish active labour market policies have been continuously changed in light of experience, new knowledge and changes in the labour market situation. This paper reviews and discusses these changes in light of the recession and considers how the labour market has coped with the recession.

Keywords

Active labour market policy Long-term unemployment Great recession J60 J38 E32

1. Introduction

The great recession has had a large labour market impact causing a steep decrease in employment rates. For a given macroeconomic background and policy setting the overriding task for labour market policies is to reduce the extent to which the rise in unemployment translates into an increase in long-term unemployment and thus the structural unemployment rate. Experience from the high unemployment period in the 1970s and 1980s shows both how steep increases in unemployment can translate into increases in the structural unemployment and how difficult it may be to bring down the structural unemployment rate subsequently.

Active labour market policies may play an important role in this context. However, how should active labour market policies (ALMP) be designed so as to reach this objective in a very slack labour market? This raises several issues. First, there is a volume effect. For given ALMP policies an increase in unemployment causes a larger inflow into the ALMP programmes. This raises both a funding and a capacity issue. Can such an increased level of activity be funded? How fast can the supply capacity in ALMP programmes be expanded. A particular concern is that the quality or effectiveness may be decreasing in the overall activity level e.g. because private job training (one of the instruments which is usually found to be most effective) may be in shorter supply. Second, the composition of the pool of unemployed may change. In a situation with low unemployment, the group is dominated by individuals who face more serious barriers for employment, while in a severe slump more core workers enter. Does this call for a different level or type of ALMP programmes? Thirdly, is the effect of ALMP policies business cycle dependent? There are two sides to this question since ALMP serves both to strengthen search incentives and to improve qualifications to enhance job finding possibilities. A slump may at the same time reduce the importance of the incentive problem and increase the importance of the qualification problem. The former because unemployment benefits may be less distortionary in a slump and hence the need to maintain search incentives via ALMP is accordingly smaller. The latter because a deep recession is associated with structural changes and therefore some core workers may find that their human capital to some degree becomes obsolete. Finally, the political support for ALMP may be business cycle dependent. It may seem more reasonable to have tight ALMP conditions when unemployment is low and job finding rates are high, and the opposite with high unemployment and low job finding rates.

We consider some of these questions with outset in the Danish experiences and policy responses to the great recession. The Danish case is interesting due to the role ALMP plays in the so-called flexicurity model (see e.g. Andersen and Svarer (2007)). The Danish labour market has flexible hiring and firing rules, relatively generous unemployment insurance benefits and a strong focus on ALMP (highest among OECD countries 1). At the same time it has delivered a rather low unemployment rate. How well has this model coped with the great recession, and how have ALMP policies been adapted to the crisis?

Denmark is also an example of how ALMP has been continuously changed. Even though the overall level measured in terms of expenditures as a share of GDP has been relatively constant over the years, the specific design of ALMP has changed significantly over the years in response to the labour market situation, new evidence on the effects of ALMP and changes in political prioritizations.

Clearly it is not possible from one observation (and even a relatively short observation period) to make a reliable inference on how a recession influences the effectiveness of ALMP. The aim of this paper is to highlight some of the insights which can be inferred so far from the Danish case. The paper provides a discussion of how ALMP has been changed in light of the recession as well as new evidence and insights on how it works. The focus will thus be on some policy issues and lessons.

This paper is organized as follows: The developments in the Danish labour market since the onset of the Great Recession are mapped in Section 2, followed by a more detailed account of active labour market policies in Section 3. Active labour market policies have been continuously changed in light of experience and new evidence, and in Section 4 we discuss some recent randomized experiments, their findings and influence on labour market policies. The management structure of labour market policies have recently been changed according to a one-shop principle at the municipal level, and the issues this has raised are discussed in Section 5. A few concluding remarks are given in Section 6.

2. Labour market response to the crisis

Prior to the Great Recession the Danish economy was booming and the unemployment rate reached record low levels, cf. Figure 1. However, at the eve of the crisis there were signs of a turn in the business cycle situation, and the Great Recession contributed strongly to deepen this. The Danish economy has thus experienced a boom-bust pattern. Although unemployment has increased steeply in recent years 2, it is noteworthy that unemployment remains below the OECD average.
Figure 1

Unemployment, Denmark and OECD, 1991 2011.

A particular feature of the Danish flexicurity model is a very high level of churning in the labour market; that is, both the level of inflows and outflows from jobs are high. As a consequence job tenure is relatively low in international comparison, and most unemployment spells are rather short.

It is seen from Figure 2 that the boom period preceding the Great Recession was accompanied by a steep increase in the employment inflow rate, but also in the outflow rate (though there was net job creation). Shortage of labour caused intense bidding for workers and induced higher job mobility as well as accelerated wage increases. The Great Recession was on impact associated with a sharp drop in the inflow rate, and it has later returned to its historical level. The outflow rate remained high and has later fallen to its historical level. The increase in unemployment has thus been associated with both a high outflow rate and a low inflow rate. It is noteworthy that the levels of inflow and outflow rates are back at their historical levels. Hence, even in the midst of the recession a very high level of job turnover remains a key feature of the Danish labour market implying that there are jobs to search.
Figure 2

Inflow and outflow rates from employment, 1999 2011.

It is a consequence of the high turnover rates that most unemployment spells are rather short, cf. Figure 3. The average duration of an unemployment spell for a person receiving unemployment benefits is about 13 weeks. In the boom years the average duration decreased, and it has later increased, but it remains at a rather low level. It is a consequence that the ratio of persons experiencing unemployment during the year relative to the number of unemployed in full-time equivalents is rather high, cf. Figure 3. Interestingly this ratio has followed an inverse U-pattern suggesting a substantial reallocation of labour during the recession.
Figure 3

Open unemployment numbers affected and average duration of unemployment spells, 2004 2012.

The high rate of job turnover in the Danish labour market can be interpreted as providing implicit work sharing 3. Rather than having unemployment concentrated on a smaller group, many are affected, but most of them only for a short period. This “rotation” ensures that most unemployed maintain contact with the labour market avoiding depreciation of human capital. As such it should contribute to lower the risk that the higher unemployment rate translates into an increase in long-term unemployment. Although long-term unemployment has increased, the increase is relatively small compared to the increase in unemployment in general, see Figure 4).
Figure 4

Long - term unemployed , 2007 2012.

A further crucial implication of the high level of job turnover is that it also facilitates entry into the labour market for youth. The level of job-openings is relatively high, even in the midst of the recession, and that is beneficial for youth. Youth unemployment has increased in Denmark (for the age group 15–24 it has increased from about 6% to about 14%), but it is significantly below the OECD average.

The initial employment decline was concentrated in the private sector, particularly building and industry. As a consequence males where more affected than females. The shifting gender composition is rather striking, c.f. Figure 5. Later the private sector has recovered somewhat while the public sector has experienced some labour adjustment, and this explains the convergence between males and females. Due to the lower level of hiring youth, unemployment increased.
Figure 5

New unemployment spells, 2004 2012.

3. Active labour market policies

To explain active labour market policies in Denmark a few institutional details are in order. Unemployment insurance 4 is a voluntary scheme based on membership fees and tax-financed subsidies. Persons not eligible for UIB are eligible for Social Assistance (means tested on a family basis).

All unemployed are categorized in so-called match groups based on an overall assessment (qualifications, experience, social situation etc.) of the potential for the individual to be employed. Earlier the system had five groups, now it has three match-groups 5. The system applies to all on unemployment benefits, social assistance, sickness-payment, flex-job etc.

  • Group I (job-ready): Individuals with no problems except unemployment. A person who is available for a job and who can become self-supportive within 3 months.

  • Group II (ready for activity): Individual who is not at present ready to start working, but is capable of participating in a programme activity aiming at later employment.

  • Group III(temporarily passive): A person who is neither ready for a job nor for participation in a programme activity aiming at later employment.

Individuals on UIB are automatically in group I. The rules for ALMP reviewed below apply to all individuals (whether on UIB, SA or other forms of transfer) in match group I. In the following we distinguish between unemployed entitled to unemployment insurance benefits – UI unemployed – and unemployed entitled to social assistance and in match group I – SA unemployed. The sum of UI and SA unemployed is denoted gross unemployment.

While the overall focus on ALMP has been strong since the mid-1990s, the specific design has been under more or less continuous change. The current system has two key elements (contact and activation):

  • Contact: An unemployed (whether on UIB or SA) has to make a CV available on the job centre website within the first four weeks as unemployed, and he/she must participate in interviews on job search and labour market availability at the job centre at least every 3rd month.

  • Unemployed have a right and duty to activation after 9 months of unemployment (duration variable, see below), and thereafter every 6 months.

  • For youths there are special rules:
    • Below the age of 30: First interview within one month, thereafter every 3 months.

    • Below the age of 30: First activation (right and duty) after 3 months (duration at least 6 months). If the person does not have an education, the activation programme should be aiming at education in the ordinary educational system.

    • Below the age of 25 without education and dependent children: mandatory education; if not immediately suited for enrolment in the ordinary education system, activation should focus on improving the pre-conditions making this possible at a later stage. For individuals with a labour market relevant education, the activation should focus on enhancing the scope for ordinary work.

    • Immediate activation of the very young (age-group 18+19). Some municipalities have introduced immediate and full-time activation for very young individuals claiming SA.

All unemployed eligible for unemployment benefits have, at their own discretion, a right to 6 weeks education (retaining UI benefits during participation). The type of education is an individual choice but the possibility set depends on the previous education record. Most educational activities are within the system of “labour market educations” (AMU) offering skill-specific short-term courses. The educational activity should be in the first 9 months (persons above age 25) or 6 months (persons below age 25) of an unemployment spell. Of the total number of persons eligible for unemployment benefits in 2010, 13% used this education option 6. This programme has not yet been evaluated.

The degree of activation (number of people in programmes relative to the relevant target group) and the average duration are shown in Figure 5 for both UIB- and SA-unemployed. It is slightly above 20% for the UIB-unemployed and 35-40% for the SA-unemployed. The average duration of programme activities is slightly longer (14–15 weeks) for UIB-unemployed than for SA-unemployed (10–11 weeks). It is noteworthy that the level of activation has been maintained despite the increase in unemployment (Figure 6).
Figure 6

Activation rate and average duration of activation activities UI and SA unemployed.

The programme types are illustrated in Figure 7. Since the activation rates have been maintained and unemployment has increased, it follows for all types of programmes that there has been an absolute increase in the number of participants. Counselling and qualification activities as well as employment with wage subsidies (for UIB-unemployed) have decreased in relative importance, while work practice, which is short-term affiliations with private or public-sector firms, has increased in relative importance.
Figure 7

Activation types for UIB - and SA - Unemployed.

It is noteworthy that the level of ALMP measured by activation rates for the unemployed has been maintained despite the crisis and the increase in unemployment. This is reflected in a clear counter-cyclical pattern for the expenditures on ALMP, cf. Figure 8.
Figure 8

Unemployment and ALMP expenditures, 2004 2010.

4. Changes in ALMP-instruments

ALMP is continuously adjusted in light of the labour market situation and new evidence. In recent years ALMP has shifted more towards focus on job search, counselling and sanctions. This reflects an adaptation to new research results – some based on natural experiments. We will account for the changes in ALMP and the empirical research motivating these changes.

In recent years there has been a remarkable focus on implementing evidence based labour market policies in Denmark. In 2005 the Danish Labour Market Board initiated the first randomized experiment to test the effectiveness of conducting an intensified programme for newly unemployed. The initiative – which is described in more detailed below – not only led to changes in the active labour market policy, but also started a sequence of new controlled experiments. Since 2005 more than six controlled experiments have been implemented and they have all to a smaller or larger extent induced changes in the design of the active labour market policies.

In this section we go through a couple of the experiments and report their main findings and discuss how they have affected policy.

Quick back to work – 2005–2006

The first random control experiment was conducted in two counties in Denmark during the winter of 2005–6. All individuals in the two counties who became unemployed during this period were randomized into a treatment group or a control group based on their birthday. Those born on the 1st to the 15th were given the treatment, and those born on the 16th to the 31st were not. The treatment consisted of a dramatic intensification of labour market policies, involving information, very early mandatory participation in job search assistance programmes, frequent meetings with employment officers, and full-time programme participation in an active labour market programme for at least three months if they had not found employment before 18 weeks of unemployment. The unemployed were all members of UI-funds and therefore they had a close connection to the labour market and did not in general have other problems besides being unemployed. Figure 9 shows the difference in the transition rates from unemployment for the treatment and the control group.
Figure 9

Transition rates out of unemployment, QBW.

It is clear from the figure that the exit rate is higher for the treatment group in the first 20 weeks of the experiment. Several authors have evaluated the experiment further. Graversen & Van Ours (2008) use duration models to address issues of dynamic selection and estimate the effect of the activation programme on exit rates to work. They find that the re-employment rate increases about 30%, and this effect is the same across age and gender. Rosholm (2008) finds similar results when estimating the effects of the activation programme separately for both counties. Graversen & Van Ours (2008), Rosholm (2008) and Vikström et al. (2011) all investigate which elements of the activation programme are most effective. Graversen & Van Ours (2008) find that the threat effect and job search assistance are most effective. A similar conclusion is drawn by Vikström et al. (2011), who construct nonparametric bounds. Also Rosholm (2008) finds substantial effects, but he also shows that when controlling for time-varying indicators of treatment all positive effects vanish and some even become negative, the so-called locking-in effect. He finds that the estimated risks of meetings and being activated drive the difference in the job finding rates between treated and non-treated Individuals. Additional evidence for threat effects is provided by Graversen & Van Ours (2011). They show that the effect of the activation programme is largest for individuals with the longest travel time to the programme location. Finally, Blasco & Rosholm (2011) analyse long-term effects on post-unemployment employment stability in terms of duration of self-support after leaving the unemployment pool. They find that treatment increases the post-unemployment self-support duration by 10% for men, while treated women show no post-unemployment stability effects. Decomposing the effect, they show that 20–25% is due to lagged duration dependence.

Although the intensified treatment seems to have been effective for the treated individuals, it might be at the expense of the control group. As case workers meet more often with members of the treatment group, they may suggest job openings to them rather than to members of the control group. To investigate this congestion effect, that is, whether the control persons in the QBW regions had a more difficult time to locate jobs than unemployed in other regions of Denmark, Figure 10 compares the survival rates for the treatment and control groups in the two regions that participated in QBW to the survival rate in unemployment for unemployed in the remaining parts of Denmark.
Figure 10

Survival rates from unemployment, QBW.

Comparing the survival curve for the control group with the comparison group of unemployed in other regions of Denmark indicates that in the first 20 weeks of the experiment the control group members in the QBW regions stay unemployed a little bit longer, although the difference is quite small. Gautier et al. (2012) have taken a closer look at the negative externality of the experiment on the control group members and find that the job-finding rate for this group is somewhat lower and hence that the effects reported in the first set of papers that evaluated QBW are slightly overrated.

All in all, the micro econometric evaluations of Quick Back to Work (QBW) suggest that the intensified policy worked both in the short term when focusing on exit rates from unemployment and in the longer term when focusing on return to employment. In addition, a cost-benefit analysis conducted by the Danish Economic Council (2007) demonstrated large net gains from QBW. What is more unclear is what caused these effects of the different instruments in QBW. To evaluate this further a new experiment was initiated in 2008. The new random control experiment was called Quick Back to Work 2 and will be described in the next subsection.

The findings from QBW also affected labour market policies. First, the lessons that the policy makers drew from the experiments (of course combined with other types of input) were that early intervention was beneficial for newly unemployed, and second, that meetings were presumably a more cost effective tool compared to other types of interventions. As a consequence activation was moved forward from 12 months to 9 months (as described in Section 2) for unemployed above 30 years and to 3 months for unemployed younger than 30. In addition, the first meeting for unemployed under 30 was moved forward from 3 months to after 1 month of unemployment. Although these changes took place before the onset of the Great Recession the policies are still in place, and although there might be reasons to soften the requirements this has not happened so far.

Quick back to work 2 – 2008

As mentioned above it was somewhat unclear what caused the effects of QBW. To evaluate this further, a new experiment was initiated in 2008. The new random control experiment was called Quick Back to Work 2. It comprised four different experiments. These experiments were designed in such a way that they would yield estimates of the effects of single elements of the QBW package. The treatments consist of weekly group meetings between a group of unemployed workers and 1–2 case workers (A), bi-weekly individual meetings between one unemployed worker and one case worker (B), early programme participation (C), and a combination of group meetings and early programme participation (D).

Again, the participants were newly unemployed who were members of UI-funds. The subjects of the experiments were individuals becoming unemployed during weeks 8–29 in 2008. The individuals were randomized into treatment or control groups based on their date of birth. Individuals born on the 16th-31st were assigned to the treatment groups, while those born on the 1st to the 15th were assigned to the control groups.

QBW2 has been evaluated by Pedersen et al. (2012). They find that especially the individual meetings in experiment B were effective. Figure 11 shows the accumulated difference in the employment rate between the treatment and control groups in experiment B. There were around 400 individuals in the treatment and control groups for both genders.
Figure 11

The employment effect of individual meetings ( experiment B ).

Pedersen et al. (2012) find that the meetings between newly unemployed workers and caseworkers on average increase employment rates over the next two years by 10% corresponding to 5 weeks. In addition, a cost-benefit analysis shows that the surplus per unemployed is around 4725 Euros.

The group meetings in experiment A were clearly the cheap option, and although the effect of these was quite small, the cost benefit analysis showed a minor positive return to these meetings.

The early activation experiment (C) showed positive effects for men, especially young men, and especially during more favourable macroeconomic conditions, while for women there was actually a negative effect of early activation due to locking-in effects. The findings suggest that the effect for men arises due to a threat effect (ex-ante effect) and more stable subsequent employment. This evidence on threat effects of early activation for men but not for women corresponds closely to results found by Rosholm & Svarer (2008) based on register data. In addition, the cost-benefit analysis also reports a surplus from the early activation experiment.

The final experiment that combined group meetings and early programme participation did not give a statistical significant effect and the cost-benefit analysis showed a deficit from this experiment.

In conclusion, QBW2 confirmed that both meetings and early activation were effective – although it also revealed that the early activation programme involved increased costs and increased the risk of prolonging unemployment spells due to the risk of locking-in effects.

Both QBW and QBW2 were targeted at newly unemployed. To test if the meetings were also effective for less employable unemployed, two additional experiments were implemented. One targeted at long-term recipients of social assistance and one at sickness benefits recipients.

Experiments for social assistance recipients and sickness benefits recipients

The experiment for unemployed on social assistance was implemented in 2008 and consisted of weekly meetings for 26 weeks. The target population was unemployed on social assistance who had been out of work for at least 26 weeks. Again, the treatment was randomly given to only a part of the unemployed in the job centres that participated in the experiment. The group of unemployed that participated and the control group turned out to be unemployed with an average unemployment spell of around 4 years. The group therefore consisted of unemployed with – on average – a very loose connection to the labour market. The experiment has been evaluated by Rosholm & Svarer (2009), and the evaluation showed that the meetings did not lead to increased exits to employment for the treatment group. In fact, what happened was that the intensified meeting propensity led to increased activation for the group, and it appeared that in terms of increasing employment the activation for this group was not effective either. In total, the experiment implied a deficit of little more than 1,000 Euros per participant.

The activation rate for this group of unemployed was in fact relatively high already, and the findings from this experiment together with other Danish studies have led to a significant decrease in the activation rate for unemployed with long unemployment spells and in many cases other problems than unemployment. As a result, the persons in group 3 7, who are labelled as temporarily passive, are no longer required to participate in activation.

The experiment for sickness benefit recipients was implemented in 2009 and targeted at all newly registered recipients of sickness benefits. Again, the setup was a randomized experiment and the treatment was intensified meetings and participation in activation or other health related activities. The experiment was evaluated by Rosholm & Svarer (2010) and the conclusion followed to a large extent the conclusion from the experiment for social assistance recipients. The intensified treatments did not lead to higher return rates to employment, and since the activation rate went up, the costs of the experiment exceeded the returns and the cost benefit analysis reported a minor deficit.

In 2009 the government allowed the municipalities to use activation for recipients of sickness benefits on the same terms as ordinary unemployed and as a consequence to be compensated for active periods. This possibility was partly redrawn in 2011 when it was required that only sick-listed who could not be expected to find jobs within the next 8 weeks could be activated on these terms.

Youth experiment

As in most countries youths have been disproportionally affected by the financial crisis and the unemployment rate has risen more sharply for this group compared to older persons. This has raised demand for more active policies towards youth unemployment. Since many of the unemployed youth lack labour market relevant education, a dual strategy seems appropriate. For young people with a qualifying education, the goal is to bring them back into employment relatively fast, whereas for uneducated youths it might be a better long-term strategy to encourage enrolment in relevant education if they possess the relevant qualifications.

Currently the policies toward youths are already quite intensive as discussed earlier. To test if it would be helpful with even more intensive guidance and a stronger focus on education for uneducated youths, a randomized controlled experiment was implemented in November 2009.

The experiment had two branches; one for unemployed below 30 with some kind of further education and one for young people without further education. Both consisted of intensified counselling. The educated group had individual weekly or biweekly meetings in the first 13 weeks of the experiment, whereas the uneducated group had individual weekly meetings for the first 32 weeks of the experiment. There was also an increased focus on activation, especially job training, but since the group already participated in activation quite early in the unemployment spell, there was no major difference between the activation rate in the treatment group and the control group.

The major impact of the experiments was counselling, and the evaluation (Ramböll, 2011) showed that there was a positive effect on employment rates for the educated group and a positive effect on the exit to education for the uneducated group. For the uneducated group the increased exit to education happened at the expense of exit to employment. A cost-benefit analysis of the treatment for the educated group showed a surplus of around 700 Euros per unemployed. Again, this experiment confirmed that intensive counselling is effective – even when the labour market is in a slump.

The conclusion so far from the experiments is that counselling is effective – especially for newly unemployed. In addition, early activation is in some cases effective since it leads to increased search activities before actual participation. It is, however, also clear that activation has a locking-in effect and therefore tends to prolong unemployment spells for unemployed who participate in activation. On top of this are the costs of running activation programmes. To decrease both the locking-in effects and the costs of programmes there has been a shift from more expensive programmes like class-room training to less expensive programmes like work practice, where the unemployed has an internship at a private or public firm for on average 4 weeks on unemployment benefits. Although it has been the intention to implement a more work-related activation policy, it has not been implemented in the experiments in a sufficiently clean way to allow for a test of the effectiveness of the two training schemes against each other.

Finally as a general comment on most assessments of the effects of ALMPs it should be noted that they do not consider general equilibrium effects arising via job creation induced by changes in job creation rates due to changed search behaviour or wage effects, cf. Andersen and Svarer (2012). The wage effect is potentially important since ALMP affects not only unemployed and participants in programmes, but also employed facing an unemployment risk.

Business cycle dependent activation strategies

An important policy question is how active labour market policies should be adapted to the business cycle situation. Although unemployment has increased in Denmark, the active labour market policy has basically remained unchanged since the onset of the crises in the sense that activation rates have been maintained. Clearly there has been a lot of political debate on the issue of which instruments are most appropriate and whether the active focus makes sense in a slack labour market.

The guidance from the economic literature on this issue is not that clear cut. There is not much work on how labour market policies in general should be adapted to the business cycle situation 8, and even less on active labour market policies. Andersen and Svarer (2009) consider the issue of whether active labour market policies (either the propensity or the duration) should be made business cycle dependent in a standard search framework. The key issue is thus the balancing of the search incentives relative to the insurance provided, and it is found that there is a welfare case for having active labour market policies to be pro-cyclical; that is, the conditions should be more demanding (high incentives, duration) in a boom compared to a slump. However, this analysis considers only the direct search effect (via the threat effect and locking-in) of activation as it has no effects on qualifications. If qualifications are decreasing in the duration of an unemployment spell, there is an argument for more intensive active labour market policies during a slump to prevent this decline. This may be reinforced to the extent that a severe business cycle downturn is associated with structural shifts, implying that the market value of certain qualifications is depreciated significantly. Well-targeted activation measures can thus help prevent that the rise in unemployment translates into an increase in the structural unemployment rate. The difficult policy issue is how to match the content of activation programmes to the characteristics of and barriers for employment faced by the individual.

There is a small but growing literature that has investigated whether activation is more or less efficient when unemployment increases and, in addition, which types of activation are most efficient in economic downturns 9.

Based on a meta-study of around 100 micro econometric evaluation studies, Kluve (2010) has investigated whether the effect of activation on labour market outcomes is related to the unemployment rate. Kluve (2010) finds that the propensity to find a positive effect of activation is increasing in the unemployment rate and, in addition, that the propensity to find a negative effect is decreasing in the unemployment rate. In fact, this finding is most pronounced when he focusses on class-room training. The drawback of Kluve (2010) is that he does not correct for possible changes in either the composition of unemployed or the composition of activation choices across business cycles. To address these issues, micro econometric analyses based on rich data sets are needed.

In Lechner & Wunsch (2009) a German micro data set covering the period from 1984 to 2003 is used. The data has rich information on both characteristics of the unemployed and on types of activation. Lechner & Wunsch (2009) focus on class-room training, which - like in the Danish case - is the most used activation type in the German active labour market system. The study applies a matching estimator that enables control for both the composition of unemployed and activation types and finds that the long-run effect of activation is positive in terms of improving employment rates for the unemployed. In the short run activation decreases employment rates due to the locking-in effect of class-room training. To shed light on the sensitivity of these results to business cycle indicators, the effects are correlated with the unemployment rate at the time the activation period is initiated. The article finds that the locking-in effect decreases with the unemployment rate, whereas the positive programme effect increases. The first finding suggests that the detrimental locking-in effect of activation is reduced when unemployment is low. Since the job finding rate is lower in economic downturns, it appears intuitive that the costs of participating in activation are lower when unemployment is high. It is less obvious what drives the positive programme effect.

In Forslund et al. (2011) it is investigated how the relative effect of work practice and vocational class-room training is affected by the aggregate unemployment rate. The study uses Swedish data from 1999 to 2005, and like Lechner & Wunch (2009>) they use a matching estimator that enables control for both the composition of unemployed and activation types. They find a negative locking-in effect and positive programme effects for both types of activation and that both effects are larger for the vocational class-room training. This implies that in the short run work practice is more effective, whereas vocational class-room training is more effective in the longer run. When the effects are correlated with the unemployment rate the article finds that the relative effectiveness of vocational class-room training increases in the unemployment rate. Like in Lechner & Wunch (2009) this is related to a decreasing locking-in effect when unemployment increases.

On Danish data two studies provide indications on the effects of activation across the business cycle, although they are not explicitly focusing on these aspects as was the case in the two articles mentioned above.

Staghøj et al. (2007) analyse the effect of four different kinds of activation and let the locking-in- and programme effect vary with the unemployment rate. For class-room training they find that the locking-in effect increases in the unemployment rate and that the programme effect decreases. The same is true for public sector job training, whereas they find no business cycle dependency in the effects for private sector job training. Lauzadyte & Rosholm (2011a) estimate the effect of activation for the period 1999 to 2004. They find that the effects of activation decrease with the unemployment rate and that this is due to a drop in the programme effects in particular. In a Norwegian study by Røed & Raum (2006) it is found that the effect of activation in general is less positive when unemployment increases.

Although the two first – and also most detailed – studies suggest that it is more effective to engage in class-room training when unemployment increases, there is also a cost side to it. A Danish cost-benefit analysis by Jespersen et al. (2008) shows that class-room training is the most expensive activation tool, and with rapidly increasing levels of unemployment the cost side can of course question the optimal policy mix in a situation where public funds are already rather limited. Presumably for this reason the choice of activation type in the Danish activation policy has remained rather constant since the onset of the economic crisis.

5. New management system

Since 2009 the municipalities have been responsible for governing the labour market policy towards all types of unemployed and sick-listed. This indicated a change from a system where the insured unemployed were governed by the state. The new system implies that all unemployed are handled by one authority – the so-called one-stop shops. This system has the apparent advantage of being simple and it benefits from economies of scale, but it also implies that unemployed eligible for unemployment insurance are handled by the same authority that provides assistance to unemployed on social assistance.

The new job centres are governed by the municipalities and the labour market policy is decided by the local city councils. This basically implies 98 possibly different labour market policy strategies, and it also runs the risk that each local job centre has a strong focus on the local labour markets and local jobs at the cost of less focus on geographical mobility.

The two major issues related to this therefore concern both how financial incentives affect the policies of the municipalities and whether geographically organized job centres can provide the most relevant counselling for job seekers.

Regarding financial incentives the state has several means for affecting the actual policy conducted in the 98 different municipalities in Denmark.

The most direct instrument is compensation to the municipalities of a fraction of the benefit payments to different categories of unemployed or sick-listed. This compensation is higher if the municipalities activate the unemployed. The most recent reform implied that only certain types of activation lead to increased compensation. To induce municipalities to switch from different kinds of class-room training to more work-related activation, the compensation rate was made dependent on the type of activation. This implied that job training in public or private firms and work practice have a higher compensation rate than if the unemployed is enrolled in e.g. a job search course. As shown in Figure 7 this seems to have influenced activation choices.

That the design of the compensation scheme affects behaviour is more thoroughly investigated in a report by Slotsholm (2010) evaluating a reform of the compensation scheme. In 2006 the state changed the compensation for social assistance. Prior to the reform the municipality was compensated directly by 50% for the social assistance payments. This was changed so that the rate of compensation was 65% when the unemployed were activated and 35% in passive periods. Figure 12 shows the activation rate in a sample of municipalities before and after the reform date.
Figure 12

Activation rate for unemployed on social assistance before and after a reform in the compensation system.

As the figure suggests the activation rate increased. In the report a micro econometric analysis suggests that activation increased on average by 8 per cent as a consequence of the reform.

6. Concluding remarks

Despite the increase in unemployment induced by the Great Recession, in- and outflows to employment remain high in the Danish labour market. Implicitly the burden of unemployment is thus shared in the sense that many are affected by unemployment, but most unemployment spells are short. Although the labour market has been adversely affected by the crisis the unemployment rate remains below the OECD average. The concern that the unemployment increase would be particular steep in a country with relatively lax employment protection and high job turnover rates has so far proven unjustified. In this sense one may say that the Danish flexicurity labour market has coped relatively well with the crisis.

Active labour market policies play an important role in this endeavour. There is a long tradition of extensive use of active labour market policies in Denmark, and active labour market policies have been maintained during the Great Recession. With the increase in unemployment this means that there has been a volume increase in active labour market policies. The flipside of this is an expenditure increase, but so far it has been possible to accommodate this since there has been some fiscal space. This points to the importance of creating fiscal space in “good times” to be able to accommodate negative shocks. While it in the current situation may be difficult to find policy measures which significantly can boost job creation, it is important to maintain the active focus of labour market policies to prevent an increase in long-term and thus structural unemployment.

Active labour market policies have been changed repeatedly often based on new empirical insights from natural experiments. Important lessons have been the effects of early intervention with a focus on job search. To reduce locking-in effects and programme costs there has been some shift away from expensive programmes like class-room training to less expensive programmes like job training. Another important lesson is that late intervention for groups marginalized in the labour market has small, if any, effects. Whether this is due to wrong instrument choice or the difficulty of overcoming barriers for this group cannot be settled from the empirical evidence. Irrespective of the reason this underscores the importance of early intervention to prevent that passive unemployment spells become too long. An overall lesson on active labour market policies from the Danish case it that the specific details of the content and structuring of labour market programmes is crucial, and therefore there is a continuous learning process in adjusting policies in light of experience, research and the labour market situation.

On both theoretical and empirical grounds it is not clear whether there is a case for strengthening or weakening ALMP with the business cycle situation. Job search may be less distorted in a slump, and this implies that the role of ALMP in countering disincentive effects from e.g. unemployment insurance is less strong. On the other hand depreciation of human capital with the duration of unemployment spells calls for intervention to restrain this process which eventually leads to an increase in the structural unemployment rate.

7. Endnotes

1According to OECD data, Denmark has the highest expenditures on active labour market programmes relative to GDP. In 2010 expenditures constituted 1.9% of GDP, and before the crisis about 1.3% of GDP cf. www.oecd-ilibrary.org.

2The decrease in employment is roughly double the increase in the number of unemployed, cf. Andersen (2012). The difference is accounted for by migrant workers, a decrease in the number of multiple job-holders, and a decrease in the labour force (increase in education).

3There are two explicit ways of adapting labour input to variations in need. One is via a “time account” which stipulates that the average number of working hours should be 37 hours per week over a certain period. This allows variation in hours without necessarily inducing overtime pay. About 2/3 of the workers in the main DA/LO-area is on a contract with a “time account”. In addition there is a “work rotation” scheme where working hours can be reduced and the workers can receive unemployment benefits for the reduced hours. The standard period for a job rotation arrangement is 13 weeks. There was some expansion of the use of this scheme in 2008, but it has subsequently levelled off (http://www.jobindsats.dk).

4The maximal benefit level is regulated by i) the fact that the replacement rate relative to past wage income cannot exceed 90%, and ii) a nominal cap (indexed to wage development). As a consequence the effective replacement rate is declining in the wage and is on average about 65%. Benefit duration has been shortened over the years and since 2011 it has been 2 years (2 ½ years). To remain eligible for unemployment benefits there is a work requirement equal to 12 months within the last 36 months.

5https://www.retsinformation.dk/Forms/R0710.aspx?id=139870.

6 http://www.ams.dk/Publications/2011/13-12_Evaluering-af-6ugers-selvalgt/Evaluering-af-6-ugers-selvvalgt-uddannelse-sep-2011pdf.pdf

7As mentioned earlier the Danish labour market system allocates unemployed into three different groups.

8There is some recent work on business cycle dependent unemployment insurance, see e.g. Andersen and Svarer (2011).

9In a recent cross-country study of the effects of structural reforms Bouis et al. (2012) find that expenditures on ALMP going to either training or employment incentives have no effects on the aggregate employment rate under favourable economic conditions (boom), they both have a significant effect in a normal business cycle situation, and only employment incentives work under less favourable economic conditions (slump).

Declarations

Acknowledgements

We gratefully acknowledge comments and suggestions made by an anonymous referee.

Responsible editor: Juan Francisco Jimeno.

Authors’ Affiliations

(1)
Department of Economics and Business, Aarhus University

References

  1. Andersen TM: A flexicurity labour market in the great recession - the case of Denmark. De Economist, special issue 2012, 160: 117–140.Google Scholar
  2. Andersen TM, Svarer M: Flexicurity -Labour Market Performance in Denmark. CESifo Economic Studies 2007,53(3):389–429. 10.1093/cesifo/ifm015View ArticleGoogle Scholar
  3. Andersen TM, Svarer M: Konjunkturafhængig arbejdsmarkedspolitik (Business cycle dependent labour market policy), Rapport til Arbejdsmarkedskommissionen, København. 2009. http://www.amkom.dkGoogle Scholar
  4. Andersen TM, Svarer M: Risk State dependent unemployment benefits. Journal of Risk and Uncertainty 2011, 78: 325–344.Google Scholar
  5. Andersen TM, Svarer M: The role of workfare in striking a balance between incentives and insurance in the labour market. Economica 2012.Google Scholar
  6. Blasco S, Rosholm M: The Impact of Active Labour Market Policy on Post Unemployment Outcomes. IZA DP 5631, Evidence from a Social Experiment in Denmark; 2011.Google Scholar
  7. Bouis R, et al.: The Short-Term Effects of Structural Reforms: An Empirical Analysis. OECD Economics Department Working Papers, No. 949; 2012.View ArticleGoogle Scholar
  8. Danish Economic Council: The Danish Economy. Spring Report; 2007. http://www.dors.dkGoogle Scholar
  9. Forslund A, Fredriksson P, Vikström J: What active labour market policy works best in a recession? Nordic Economic Policy Review 2011, 171–202.Google Scholar
  10. Gautier P, Muller P, van der Klaauw B, Rosholm M, Svarer M: Estimating Equilibrium Effects of Job Search Assistance, IZA DP 6748. 2012.Google Scholar
  11. Graversen BK, Van Ours JC: How to help unemployed find jobs quickly: Experimental evidence from a mandatory activation program. J Public Econ 2008,92(10–11):2020–2035.View ArticleGoogle Scholar
  12. Graversen BK, Van Ours JC: An activation program as a stick to job finding. Labour 2011,25(2):167–181. 10.1111/j.1467-9914.2011.00521.xView ArticleGoogle Scholar
  13. Jespersen ST, Munch JR, Skipper L: Costs and Benefits of Danish Active Labour Market Programs. Labour Econ 2008,15(5):859–884. 10.1016/j.labeco.2007.07.005View ArticleGoogle Scholar
  14. Kluve J: The effectiveness of European active labour market programs. Labour Econ 2010, 17: 904–918. 10.1016/j.labeco.2010.02.004View ArticleGoogle Scholar
  15. Lauzadyte A, Rosholm M: Cyclical and Regional Variations in the Effectiveness of Active Labour Market Programmes. Aarhus University, Manuscript; 2011.Google Scholar
  16. Lechner M, Wunsch C: Are Training Programmes More Effective When Unemployment is High? J labour Econ 2009, 27: 653–692. 10.1086/644976View ArticleGoogle Scholar
  17. Pedersen JM, Rosholm M, Svarer M: Experimental evidence on the effects of early meetings and activation. Manuscript, Aarhus University; 2012.Google Scholar
  18. Rosholm M: Experimental evidence on the nature of the Danish employment miracle. IZA Discussion Paper, No; 2008. 3620Google Scholar
  19. Rosholm M, Svarer M: The Threat Effect of Active Labour Market Programmes. Scand J Econ 2008,110(2):385–401. 10.1111/j.1467-9442.2008.00544.xView ArticleGoogle Scholar
  20. Rosholm M, Svarer M: Kvantitativ evaluering af Alle i gang. report to the Danish Labour Market Board, Danish; 2009. http://www.ams.dk/Viden/Udvikling%20og%20forsoeg/Alle-i-gang.aspxGoogle Scholar
  21. Rosholm M, Svarer M: Kvantitativ evaluering af Aktive – hurtigere tilbage. report to the Danish Labour Market Board, Danish; 2010. http://www.ams.dk/Viden/Udvikling%20og%20forsoeg/Aktive-hurtigere-tilbage.aspxGoogle Scholar
  22. Røed K, Raum O: Does Labour Market Training Speed Up the Return to Work? Oxf Bull Econ Stat 2006,68(5):541–568. 10.1111/j.1468-0084.2006.00177.xView ArticleGoogle Scholar
  23. Slotsholm : Evaluering af de økonomiske styringsinstrumenter på beskæftigelsesområdet, report (in Danish). 2010. http://www.ams.dk/Aktuelt/Nyheder/2010/11–01%20Oekonomiske%20styringsmekanismer.aspxGoogle Scholar
  24. Staghøj J, Rosholm M, Svarer : A Statistical Programme Assignment Model, IZA Discussion Paper, No 3165. 2007.Google Scholar
  25. Vikström J, Rosholm M, Svarer M: The relative efficiency of active labour market policies: Evidence from a social experiment and non-parametric methods. IZA Discussion Paper, No; 2011. 5596 5596Google Scholar

Copyright

© Andersen and Svarer; licensee Springer. 2012

This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.