Youth unemployment and active labor market policies in Europe

Caliendo, Marco; Schmidl, Ricarda Article Youth unemployment and active labor market policies in Europe IZA Journal of Labor Policy Provided in Cooperation with: IZA – Institute of Labor Economics Suggested Citation: Caliendo, Marco; Schmidl, Ricarda (2016) : Youth unemployment and active labor market policies in Europe, IZA Journal of Labor Policy, ISSN 2193-9004, Springer, Heidelberg, Vol. 5, Iss. 1, pp. 1-30, https://doi.org/10.1186/s40173-016-0057-x This Version is available at: https://hdl.handle.net/10419/154723 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. http://creativecommons.org/licenses/by/4.0/ Caliendo and Schmidl IZA Journal of Labor Policy (2016) 5:1 DOI 10.1186/s40173-015-0056-3 ORIGINAL ARTICLE Open Access Youth unemployment and active labor market policies in Europe Marco Caliendo1,3* and Ricarda Schmidl2 *Correspondence: [email protected] 1University of Potsdam, IZA Bonn, DIW Berlin, IAB Nuremberg, Potsdam, Germany 3University of Potsdam, Chair of Empirical Economics, August-Bebel-Str. 89, 14482 Potsdam, Germany Full list of author information is available at the end of the article Abstract Since the economic crisis in 2008, European youth unemployment rates have been persistently high at around 20% on average. The majority of European countries spends significant resources each year on active labor market programs (ALMP) with the aim of improving the integration prospects of struggling youths. Among the most common programs used are training courses, job search assistance and monitoring, subsidized employment, and public work programs. For policy makers, it is of upmost importance to know which of these programs work and which are able to achieve the intended goals – may it be the integration into the first labor market or further education. Based on a detailed assessment of the particularities of the youth labor market situation, we discuss the pros and cons of different ALMP types. We then provide a comprehensive survey of the recent evidence on the effectiveness of these ALMP for youth in Europe, highlighting factors that seem to promote or impede their effectiveness in practice. Overall, the findings with respect to employment outcomes are only partly promising. While job search assistance (with and without monitoring) results in overwhelmingly positive effects, we find more mixed effects for training and wage subsidies, whereas the effects for public work programs are clearly negative. The evidence on the impact of ALMP on furthering education participation as well as employment quality is scarce, requiring additional research and allowing only limited conclusions so far. JEL codes: J13, J68, J64 Keywords: Youth unemployment, Active labor market policies, Evaluation, Training, Job search 1 Introduction Young individuals entering the labor market are generally considered to be an at-risk population. They face a higher risk of unemployment than older workers, are more likely to switch between states of joblessness, training and working, and are more likely to enter temporary or precarious types of employment (see, e.g., Quintini et al. 2007). One reason for the lower labor market attachment of youth is their initially low labor market experience. During the school-to-work transition period, labor market entrants tend to learn about their abilities and preferences by “job-shopping” (Topel and Ward 1992), resulting in higher rates of turn-over and more frequent periods of non-employment. At the same time, firms commonly face higher costs of investment and lower costs of termination when employing young workers, making the youth labor market situation more sensitive to demand-side fluctuations, which was recently demonstrated in the aftermath of © 2016 Caliendo and Schmidl. Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. Caliendo and Schmidl IZA Journal of Labor Policy (2016) 5:1 Page 2 of 30 the financial crisis 2007/2008 (Bell and Blanchflower 2010; Choudhry et al. 2012; Verick 2011). Between 2008 and 2009 youth unemployment rates increased by about five percentage points to a 20% average, then until 2013 they further climbed to 24%; during the same time period, adult unemployment rate changed from 6% to 10%.1 High and persistent levels of youth unemployment give rise to concern, as the negative consequences of extended spells of unemployment early in the career are well documented. On the one hand, joblessness is directly associated with psychological distress and financial hardship for the affected youth (Goldsmith et al. 1997). On the other hand, early unemployment spells may have negative effects on later-life outcomes such as lower wages (Arulampalam 2001; Gregg and Tominey 2005; Kahn 2010; Skans 2004), lower labor market attachment (Gregg 2001), lower well-being (Daly and Delaney 2013), and a higher propensity to engage in criminal activities (Bell et al. 2014; Fougère et al. 2009). The immediate monetary costs of youth unemployment include direct costs for unemployment benefits and social assistance, as well as indirect costs of foregone tax payments and social security contributions. Calculations by Eurofund (2011) suggest that the immediate public costs accruing from youth neither in education nor in employment (NEET) in the EU-26 amounted to 120 billion euros (1% of GDP) in 2008, and to 153 billion euros (1.2% of GDP) in 2011, not accounting for discounted future costs and foregone payments due to the lowered labor market attachment in the long run. In light of demographic change and increasing old-age dependency ratios, the burden of these costs are expected to further increase. Raising employment levels is seen as the most effective strategy with which countries can prepare for population ageing (European Commission 2008), making the youth unemployment problem all the more urgent. To smooth the school-to-work transition process, to prevent extended spells of unemployment or the complete withdrawal from the labor market, and to promote entry into stable employment relationships, policy makers in many countries resort to active labor market programs (ALMP). Prominent national initiatives include the “New Deal for Young People (NDYP)” in the UK, “Jugend mit Perspektive (JUMP)” in Germany and the “Youth Unemployment Program (YUP)” in Denmark. More recently, the “Youth Guarantee (YG)” adopted by the European Union in 2013 called all member states to set up ALMP programs to ensure that unemployed youth were offered high quality employment or education opportunities within four months of entering unemployment (European Commission 2014). The types of programs most commonly used can be divided into four categories, namely, labor market training, job search assistance and monitoring, wage subsidies and public sector work programs. Given the relatively high rates of youth unemployment, it is not surprising that large shares of the active youth population participate in ALMP, as can be seen in Fig. 1. The average youth unemployment rate in the presented countries was at 23% in 2012, and, on average, 9% of the active youth participated in ALMP. Unfortunately, however, the quantitative importance attributed to the use of ALMP to fighting youth unemployment stands in stark contrast to the low level of knowledge we have regarding their effectiveness. Grubb (1999) provides a descriptive summary of the early evidence (up to 2000) on the effectiveness of specific ALMP for disadvantaged youth in the US. He concludes that training and education measures are largely unsuccessful, but points to promising combinations of training with a close labor market link and additional support schemes. Reproducing the negative findings on training programs for Caliendo and Schmidl IZA Journal of Labor Policy (2016) 5:1 Page 3 of 30 Fig. 1 Youth unemployment rates and stock of ALMP participants relative to the active working population in selected European countries, 2012(a).Source: Eurostat, 2012. Note: Dark gray bars depict youth unemployment rates, light gray bars the ALMP participation among all active youth. (a)Only countries that provide information on ALMP participant numbers for the youth population. ALMP participants numbers refer to the stock of participants in ALMP types 2 to 7 (Training, Employment Incentives, Supported Employment and Rehabilitation, Direct Job Creation and Start-up Incentives) are included OECD countries, Martin and Grubb (2001) suggest that wage subsidies are among the most promising programs for youths. Heckman et al. (1999) study the evidence of ALMP in the US and Europe, reaching the conclusion that none of the program types substantially benefit unemployed youth. A meta-analysis of European ALMP evaluation studies up to the year 2000 is provided by Kluve et al. (2002) who find that youth tend to benefit less from ALMP participation than adults. Similar results emerge in the more recent and comprehensive meta-analyses by Kluve (2010) and Card et al. (2010) showing that programs targeted at youths, and in particular those involving training, are less effective in increasing employment levels than programs targeted at the general population of unemployed. The particularity of the youth labor market situation and the results from the metaanalyses suggest that assessments of the effectiveness of ALMP for adults are most likely not valid for youth. So far, no consensus exists on the effectiveness of the different ALMP programs for this age group. With this study we aim to fill this gap. Aggregating the evidence of a substantial number of recent studies assessing the effectiveness of ALMP for youth, we provide a comprehensive re-assessment of the questions revolving around which policies work for whom and under which conditions. Unlike the recent metaanalyses we thereby pay special attention to the specificity of the youth labor market situation, highlighting the particularities that might promote or mitigate the effects of ALMP programs for young individuals. In line with the previous literature, we focus on the micro-evidence of the effectiveness of ALMP, distinguishing between the four types of programs that are most commonly employed to promote a direct labor market entry: labor market training, job search assistance and monitoring, wage subsidies and public sector work programs. Besides focusing on the effect of ALMP on employment take-up, Caliendo and Schmidl IZA Journal of Labor Policy (2016) 5:1 Page 4 of 30 we also examine the evidence available on education participation, and the findings on the quality of the accepted employment whenever available. The paper is organized as follows. Section 2 gives a quick overview of the labor market situation of youth in Europe and discusses the role of ALMP. Section 3 presents the collected evidence on the effects of the different policies, before Section 4 concludes. 2 The labor market for youth in Europe 2.1 A quick glance at the European situation Table 1 depicts various indicators describing the labor market situation of youth for different European countries in 2013. Column (1) shows that there exists substantial variation across countries in the unemployment rates2for youth below 25 years, ranging from 8% in Germany to over 40% in Italy, Spain and Greece. Similarly, the share of youth in long-term unemployment in column (3) ranges from 5% to 7% in Finland and Sweden to over 50% in Italy, Greece and the Slovak Republic. As these cross-country differences are linked to differences in economic performance or the institutional set-up, youth-specific labor market patterns are highlighted when relating youth labor market outcomes to Table 1 Absolute and relative labour market indicators for youth in selected European countries, 2013 Unemployment Unemployment >1 year Inactive all NEET(a) ALMP(b) Country youth youth/adult youth youth/adult youth youth/adult youth youth youth/adult (1) (2) (3) (4) (5) (6) (7) (8) (9) Austria 9.2 2.0 14.8 0.6 40.7 3.6 10.6 1.03 1.5 Belgium 23.7 3.2 30.8 0.6 69.0 4.7 18.7 0.75 0.8 Czech Republic 19.0 3.1 32.7 0.7 68.5 6.3 14.2 - - Denmark 13.0 2.1 10.1 0.3 38.3 3.1 13.4 0.28 0.3 Estonia 18.8 2.3 34.8 0.8 60.2 4.9 16.8 0.06 0.6 Finland 19.6 3.0 5.3 0.2 48.2 3.6 15.5 0.32 0.5 France 24.0 2.8 27.3 0.6 62.7 5.4 19.4 0.91 2.3 Germany 7.8 1.6 23.0 0.5 49.2 4.0 10.3 1.29 2.5 Greece 58.3 2.2 52.0 0.8 71.6 4.4 33.1 - - Hungary 27.2 3.0 33.0 0.6 72.8 4.3 26.1 1.09 1.8 Ireland 26.8 2.2 41.2 0.6 60.3 3.1 22.0 0.27 0.9 Italy 40.0 3.5 53.3 0.9 72.8 3.2 33.7 0.58 1.3 Luxembourg 15.8 3.0 23.0 0.8 74.1 5.9 8.2 0.52 0.3 Netherlands 11.0 1.9 17.0 0.4 30.0 2.4 9.5 0.09 0.1 Norway 9.1 3.1 11.0 0.4 43.0 3.2 10.9 0.34 0.4 Poland 27.3 3.0 31.7 0.7 66.7 4.3 20.2 0.16 0.4 Portugal 38.1 2.5 36.3 0.6 65.0 5.6 22.0 0.34 1.8 Slovak Republic 33.7 2.6 61.3 0.9 69.2 5.4 21.0 0.20 1.0 Slovenia 21.7 2.2 39.4 0.8 66.2 7.1 13.7 0.09 0.6 Spain 55.5 2.3 39.4 0.8 62.2 4.8 32.4 0.27 0.5 Sweden 23.5 3.9 6.9 0.3 45.5 5.0 12.9 0.20 0.2 Switzerland 8.5 2.1 16.9 0.5 32.4 3.2 10.2 - - United Kingdom 20.7 3.6 28.8 0.7 41.6 2.9 19.1 - - Source: Eurostat, 2013 Youth unemployment rates are based on individuals below 25 years. The youth-adult ratio is based on yearly unemployment rates of youth (15–24 years) and adult (25–54 years) unemployment levels in 2013 (a)NEET (Neither in Employment nor Education) rates among all inactive youth are depicted for the 20 to 24 year-old’s only (b)Participation shares in ALMP refer to the year 2012 and are calculated as the number of ALMP participants over the stock of unemployed and may be larger than 1 due to multiple participation. The youth-adult ALMP participation ratio divides the share of ALMP participants among youth by the share of ALMP participants among adults Caliendo and Schmidl IZA Journal of Labor Policy (2016) 5:1 Page 5 of 30 those of more senior workers. In particular, this reveals that the youth labor market situation across countries shares several common features. The youth/adult unemployment ratio in column (2) shows that youth are, on average, two to three times more likely to be unemployed than adult workers – with the largest differences observed in Sweden, the UK, and Italy. At the same time, the probability of becoming long-term unemployed is commonly substantially lower for youths than for adults. The relative shares of longterm unemployed between youth and adults depicted in column (4) show that youth long-term unemployment risk is, on average, only two thirds the value of adult long-term unemployment. Part of the differences in the youth-adult (long-term) unemployment risk can be explained by the higher job mobility and the higher likelihood to enter inactivity among youth. As depicted in columns (5) and (6) of Table 1, the inactivity rate3among youth ranges between below 30% (Netherlands, Switzerland and Iceland) and 70% (Greece, Hungary, Luxemburg and Italy). On average, they are hence about four times more likely to be inactive than adults. While part of this higher inactivity is driven by entry into formal education – note that countries with an extensive apprenticeship-based secondary education system in which youth are counted as working, such as Germany, Austria, Denmark and Switzerland, exhibit a much lower inactivity rate – youth are generally also less attached to the labor market, and hence more likely to withdraw from labor force. An indicator capturing the size of this risk group is the share of youth neither in education nor in employment (NEET). Column (7) shows that about one third of countries exhibit NEET shares above 20% among the age group of 20 to 24-year-olds, with the Southern European countries (Italy, Portugal, Spain and Greece), some Eastern European countries (Poland, Slovak Republic and Hungary) and Ireland being affected most heavily. The labor market transition of youth is generally highly correlated with the levels of educational attainment (Quintini et al. 2007). Averaged European unemployment rates from 2013 suggest that youth who have obtained at most a schooling degree at the lower secondary level are 1.7-times more likely to be unemployed than youth who have at most obtained an upper secondary degree, and they about two times more likely to be unemployed than youth with a tertiary education degree. In the Southern European countries that were hit particularly hard during the economic crisis – Greece, Italy, Portugal and Spain – the unemployment rates of lower secondary and upper secondary school graduates only differ by a factor of 1.1 in favor of the higher educated. In the other European countries, the respective unemployment rates differed by a factor of 1.8 (Eurostat 2013).4 Differences in economic conditions, education and labor market policies may result in systematic cross-country differences in the chances of youth to enter the labor market. As outlined above, youth from countries that were hit hardest by the economic crisis (Southern European countries, Ireland) and that are hence plagued with structurally low labor demand, are more likely to be (long-term) unemployed, are more likely to enter inactivity, and to experience lower returns to education. Several studies underline the close relation between labor market institutions, e.g., hiring policies, minimum wages, and youth unemployment levels (Addison and Teixeira 2003; Bertola et al. 2007; Jimeno and Rodríguez-Palenzuela 2003). The close link between the schooling system and labor market entry of youth is documented most intensely. On the one hand, the general schooling system may affect the tendency of youth to leave school early, i.e. to finishing their Caliendo and Schmidl IZA Journal of Labor Policy (2016) 5:1 Page 6 of 30 education with a degree at the lower secondary or lower level, with the consequence of experiencing more difficulties entering the labor market. On the other hand, the structure of post-mandatory, professional education options affects the youth labor market transition. In countries with established vocational education tracks or an extensive apprenticeship system, youth participating in vocational education tend to experience a faster entry into the labor market than youth entering the labor market after participating in mostly general education (Hanushek et al. 2011; Neuman and Ziderman 1999; Riphahn and Zibrowius 2015; Ryan 2001; Winkelmann 1996).5The reason for this faster entry due to vocational education is attributable to an improved short-run alignment of the supply and demand for skills. In particular, skill or qualification mismatch is likely to reduce the quality and stability of initially accepted jobs (Wolbers 2003), with potentially long-lasting negative effects of later-life labor market outcomes (Liu et al. 2012). At the same time, the increasing phenomenon of over-education (Quintini 2011) may result in decreased chances of low-skilled workers finding employment due to crowding-out (Borghans et al. 2000; Dolado et al. 2000). According to Quintini (2011), the Netherlands, Greece, Spain, Sweden, Luxembourg and Portugal are amongst the countries with highest levels of over-qualification. 2.2 The role of active labor market policies The primary objective of active labor market programs for youth is to integrate unemployed youth into the labor market, stabilize their career entry, and/or to promote the take-up of vocational training as an intermediate step to labor market entry. Column (8) in Table 1 shows that countries exhibit substantial differences with respect to participation in ALMP. While in Slovenia, Estonia and the Netherlands the share of youth in ALMP is below 10% of the unemployed youth population, it exceeds 100% in Austria, Germany and Hungary (meaning that individuals participate in more than one program on average). Column (9) relates the share of unemployed youth participating in ALMP to the share of unemployed adults participating in ALMP. An indicator larger than 1 shows that youth are over-represented in ALMP compared to adults (e.g., in Austria, France and Germany), while an indicator below 1 indicates that they are under-represented (e.g., in Denmark, Estonia and Finland). The previous Section points to three areas of intervention depending on the source of the labor market problem: the stimulation of labor demand, the avoidance of long-term unemployment, and the elimination of educational mismatch. In case educational mismatch is at the root of the unemployment problem, ALMP training programs serve to align the skill-level of the unemployed youth to that of labor demand. As youth are more prone to engage in further formal education, it is sensible to distinguish shortand longterm training measures. While short-run programs tend to be of remedial nature, aiming to overcome minor and specific skill deficits, long-run training programs aim to overcome more structural skill deficits such as gaps in general education. As the latter types of programs stand in direct competition to formal education, their benefits need to be evaluated relative to regular schooling or training options and are thus likely to be less strong than programs for adults who would otherwise not engage in any education. Also, depending on the types of professional education options available, ALMP training may not be similarly valued as formal education by employers, and participation in them may entail negative stigmatization effects. A smaller part of training ALMP are preparatory training Caliendo and Schmidl IZA Journal of Labor Policy (2016) 5:1 Page 7 of 30 programs that promote the take-up of regular formal education, such as the continuation of general schooling, or participation in apprenticeship-based vocational education. Since the returns to upgrading skills may be low if the overall demand for youth labor is low – as outlined above – a second issue to be addressed by ALMP programs is the stimulation of youth labor demand. There are several reasons why the demand for youth labor may be low. First, youth are commonly the ones most affected by economic downturn, as firms may be less willing or able to let go of workers with longer tenure. Second, even under normal economic conditions, employers may prefer hiring more experienced workers, in particular, if previous work credentials or colleague referrals allow employers to discriminate better between low and high ability workers (compare Montgomery, 1991, for a theoretical analysis of employee referrals). Third, in the presence of job-specific human capital, firms may be less willing to hire youth and invest in costly training if there is a high probability that youth leave the firm without redeeming this initial investment (see Wolter and Ryan, 2011, for an extensive summary of firm’s training involvement). Hence, if low workor job-specific experience is a barrier to initial labor market entry, the provision of financial incentives for employers to hire and train young people may constitute an effective tool to improve labor market integration. As youth gain more experience and firms are better able to observe their ability, it is intended that youth are offered regular work contracts following the initial subsidy period. Furthermore, Cockx and Picchio (2013) suggest that stigmatization rather than the depreciation of human capital may be the source of state-dependence in long-term unemployment among youth. Following this line of reasoning, wage subsidies promoting the take-up of “real” employment, albeit subsidized, may help youth signal their employability. A third issue to be addressed by ALMP is the avoidance of long-term unemployment spells, which were found to have particularly detrimental long-term effects for youth (see Section 1). Job search programs and tight monitoring schemes aim to achieve a fast activation of youth early in the unemployment spell. As youth are more likely to enter inactivity as a response to longer spells in unemployment, these measures may also prevent complete withdrawal of youth from the labor market in the long-run. A potential downside of these approaches is that they may result in a direct withdrawal from the labor market when monitoring and sanction are imposed too fiercely. Studies for adult unemployed find that monitoring and sanctioning increase take-up of employment, but may also result in withdrawal from the labor market (see, e.g., Arni et al. 2013; van den Berg et al. 2013). Because of the higher risk of youth of entering inactivity, higher intensity activation schemes may result in stronger drop-out responses among youths compared to adults. In a similar vein, but targeted at more disadvantaged youth, public employment schemes are used to keep youth in the labor market by offering a work-like environment including low levels of payment. While these types of programs are usually ineffective for adults, they could provide a stepping-stone for youth when combined with training for regular jobs. A general issue arising from the previous Section is that youth tend to experience higher labor market mobility than adults, the consequences of which on labor market outcomes are ambiguous. On the one hand, job changes may allow an upward movement in the careerand wage-ladder and allow youth to become aware of their skills and preferences (Topel and Ward 1992). On the other hand, it has been found that job-changes promote further job-changes (occurrence dependence) and that early job instability significantly Caliendo and Schmidl IZA Journal of Labor Policy (2016) 5:1 Page 8 of 30 reduces later wages (see, e.g., Doiron and Gørgens 2008; Neumark and Wascher 2006). Consequently, ALMP programs should also be measured by their ability to improve the job match quality and stability of accepted employment. If participation in ALMP helps youth to learn about their preferences, career or schooling opportunities, or allow to signal their abilities to employers, this is likely to improve choices and stability of subsequent employment. Against this background it is particularly interesting to assess the long-run effects of, e.g., training measures or financial incentives to employers, as the benefits of these measures may lie in an increased long-run stability of employment. 3 Active labor market policies and their effects In this Section, we summarize the findings of 37 evaluation studies assessing the impact of ALMP measures for youth in Europe (see Table 3 in the Appendix for a full list of studies). We include evaluation studies that consider programs specifically targeted at unemployed youth, as well as evaluation studies assessing the performance of general ALMP programs including a subgroup-specific analysis for youth. Tables 4 to 7 in the Appendix provide a brief overview of the studies included in the analysis, classified by the type of program under investigation. In line with the most common types of ALMP encountered in practice, four broad types of programs are distinguished: labor market training, job search assistance and monitoring, wage subsidies, and public sector work programs. We restrict our overview to studies that convincingly address the problem of nonrandom selection into treatment, e.g., by conducting randomized field experiments, exploiting exogenous variation in program access, controlling for a large set of informative control variables, or using other state-of-the-art econometric evaluation techniques. The bulk of evaluation studies use the conditional independence or “unconfoundedness” assumption for identification, thus assuming that controlling for observed characteristics is sufficient to capture selection into the different treatment options. The ALMP evaluation literature offers some guidelines regarding the most informative characteristics to be included to control for non-random selection (e.g., Caliendo et al. 2014; Lechner and Wunsch 2013), and we preselected studies adhering to these guidelines. However, it is clear that by the short previous labor market history of youth and the stronger variability in terms of initial labor market attachment, there is a risk that unobserved heterogeneity plays an even stronger role for the youth than for the adult working population. Next to considering the employment effect of the ALMP programs, we also assess, where available, program impact estimates on education participation and the quality of accepted employment. On the one hand, participation in education constitutes a relevant outcome of interest as youth may wish to continue further formal education once exposed to the treatment rather than entering the labor market directly. Under the assumption that higher levels of human capital result in a better long-term labor market prospects, this may also be a desirable outcome of ALMP programs for youth. On the other hand, as youth tend to consider participation in formal education as an alternative to remaining unemployed, short-run positive employment effects of policies may arise due to the control group entering education rather than remaining unemployed. As it is not clear how the returns to labor market policy programs compare to that of formal education participation, this substitution effect should be kept in mind when evaluating the effect of ALMP for youth. Caliendo and Schmidl IZA Journal of Labor Policy (2016) 5:1 Page 15 of 30 Table 2 Overview - Number of studies finding positive, zero or negative effects of ALMP for different outcomes Impact Employment Education Quality Labor market +11 2 5 Training 0 12 2 3 (Studies: 19/ - 6 7 1 Effects: 47) (n/a) (1) (11) (12) Job search +12 0 0 Assistance 0 8 3 2 (Studies: 16/ - 1 0 1 Effects: 27) (n/a) (0) (12) (12) Wage subsidies +411 (Studies: 8/ 0 4 1 1 Effects: 13) - 0 1 0 (n/a) (1) (5) (6) Public work +111 Programmes 0 8 2 0 (Studies: 8/ - 5 1 1 Effects: 20) (n/a) (0) (6) (7) Note: A detailed list of the used studies can be found in Tables 4 to 7 in the Appendix. The number of effects includes effects for different sub-groups within a study however, a further comparison in terms of education outcomes and employment stability is not possible due to the lack of evidence. Unfortunately, most of the evaluation studies under consideration do not address the issue of costs-effectiveness of the respective programs. It is interesting, however, to look at approximate calculations8based on the program specific expenditure information drawn from the Eurostat Database. This suggests that the average per-capita expenditure is highest for wage subsidies (about 3,200 euros), followed by training programs (2,500 euros), job creation schemes (2,200) and job search services as the least expensive (1,200 euros). The relatively high costs of job creation schemes stand in contrast to their effectiveness, and are likely to be driven by having the disadvantaged youths as focus group. Based on the calculations by Eurofund (2011, p. 62ff.) of the country-specific costs of NEET arising due to yearly foregone earnings in 2008, the per-capita costs incurred by these ALMP programs can be compared with the costs incurred by not having youths in the labor force. This can provide a more tangible assessment of the importance of the costs of maintaining ALMP. This exercise shows that the per-capita cost of wage subsidies amount to about 40% of foregone earnings, 30% of foregone earnings in case of training programs, about 35% for job creation schemes and 10% for job search programs.9The costs of wage subsidies are quite sizeable, but given their above average effectiveness, they stand out relative to similarly costly job creation schemes that are often found to result in zero or negative employment effects. In contrast, the low relative costs of job search measures in combination with their relatively high effectiveness suggest that they are commonly cost-effective. 4Conclusions Reducing youth unemployment is a key challenge in many European countries, especially in the aftermath of the most recent economic slowdown. Youth are a population at risk Caliendo and Schmidl IZA Journal of Labor Policy (2016) 5:1 Page 16 of 30 and exhibit a much higher risk of becoming unemployed than the adult population. The youth-adult unemployment ratio ranged from 1.6 to 3.6 in 2013, but even in relatively calm economic circumstances, as in the pre-crisis years 2005 to 2007, the youth-adult unemployment ratio was between 2 and 3 on average. Keeping in mind that unemployment at early stages can have severe long-term consequences on labor market outcomes as well as on other socio-economic factors, policy-makers use active labor market policies as a potential solution and spend significant resources on them. Most commonly, four broad types of programs are distinguished: labor market training, job search assistance and monitoring, wage subsidies, and public sector work programs. We review the existing evidence on the effectiveness of these different ALMP programs from evaluation studies in Europe addressing the potential problem of selection bias in a convincing way. What becomes immediately apparent is that the overwhelming majority of studies only assess the effects on employment take-up. While this is understandable given the main objective (and also comparable to the adult population), it might be too shortsighted for the youth population. Here, additional outcome dimensions – such as the take-up up of higher formal education or the job quality – should also be considered. The presented evidence shows differences in the results across but also within different ALMP types. Job search assistance with and without monitoring results in overwhelmingly positive effects. These programs are usually not very cost intensive and seem to work in many different circumstances. However, it also seems that “too much” job search activation can result in zero or even negative employment effects, suggesting that they do not provide the optimal solution for all youth. Similarly, the question of whether job search and monitoring also lead to “better” jobs has not been addressed sufficiently by the literature. The few studies considering employment quality find rather mixed effects, suggesting, however, that harsh monitoring and sanctioning schemes result in negative employment outcomes. Labor market training programs are most commonly used for youths and have been studied most extensively. The overall results are somewhat mixed: for less than half of the programs/sub-groups, positive effects are found; and for the majority, we find insignificant or even negative effects. The negative findings are predominantly driven by the performance of combinations of classroom and practical training, as well as classroombased training programs under adverse economic conditions in which an upgrading of skills may not be sufficient to improve labor market integration chances in the shortto medium-run. Excluding these studies, the evidence on classroom-based training is similarly positive compared to that of job search assistance. A particularity of classroom-based training ALMP is that it may reduce the take up of formal education. This crowding out of formal training is clearly an issue to be addressed in further research. The hopes one might have had with public work programs are not met. Very similar to the results for the adult population, only one of the reviewed studies shows positive effects. Instead of providing a bridge to regular employment, they seem to entail lockingin or even stigmatization effects. It is thus questionable whether they should be used in the current (or any) situation. In contrast, the effects of wage subsidies on employment takeup are always positive (or insignificant), suggesting that subsidized “real” work experience often provides a stepping stone in regular employment. Wage subsidies evaluated under bad economic conditions were found to perform less well, suggesting that they are no Caliendo and Schmidl IZA Journal of Labor Policy (2016) 5:1 Page 17 of 30 panacea for stimulating labor demand. Unfortunately, little is known about the effects on job quality or stability. Overall, the aggregate evidence of the effectiveness of ALMP is somewhat discouraging, suggesting that some – but not all – elements of ALMP programs can be a solution for the youth unemployment problem. The evidence also raises the question of whether the money spent on ALMP for the young unemployed is well invested or whether the money should be used to tackle potential problems earlier. While investing in early education is definitively favorable (Heckman 2006), these investments have a long-term horizon and will not be a solution for all youth at risk. In particular, it remains an important challenge for policy-makers to devise activation schemes that benefit youth with very low skills and motivation and/or socially disadvantaged backgrounds. While job search and monitoring programs seem to be a very promising start for many youth, neither of the programs under study were targeted to the benefit of the most disadvantaged. Similarly, training programs with practical elements that tend to be targeted at youth with low initial schooling are found not to work very well. Promising strategies for disadvantaged youth may comprise intensive mentoring programs (see Rodriguez-Planas 2012 for evidence on the U.S.) but also programs that take into account social or health-related factors that may be at the source of a difficult labor market entry (currently ongoing projects include Crépon et al. 2015; van den Berg et al. 2015). Similarly, more policy emphasis should be given to promoting the take-up of stable employment relationships. Helping youth to learn about their preference by personalized coaching, information about career opportunities, or structured labor market entry program could help improve the career choices made by youth. Saniter and Siedler (2014) show that the introduction of occupational counseling centers in Germany resulted in smoother labor market entry among affected school leavers. Similarly, it is expected that training programs might also bring the largest return in terms of human capital investment in the long-run. Unfortunately, most of the studies summarized here were not able to identify long-term effects. The observation period of most examined studies ranges from several months to two years (with very few exceptions), thus hardly qualifying as long-term. The review hence also highlights the shortcomings in current research with respect to the long-term consequences of policy measures. Further research should put a stronger focus on education and job quality outcomes in the long-term perspective. Endnotes 1Based on unemployment rates for youths (aged 15 and 24) and adults (aged 25 and 54) in 2008 and 2013 in the EU-27; provided by Eurostat. 2The European Labor Force Survey defines unemployment as “without work during the reference week, currently available for work and either actively seeking work in the last four weeks, or expecting to start a job within the next three months.” The unemployment period is defined as the duration of job search or the time since the last job was held. 3Inactivity is defined as not working and not searching for work and/or unavailable for work. The concept of inactivity is hence not related to receipt of unemployment benefits, which would affect youth disproportionately as they had less time to build up unemployment benefit claims. 4The country-average is based on Belgium, Czech Republic, Denmark, Germany, Estonia, Ireland, Greece, Spain, France, Italy, Luxemburg, Hungary, the Netherlands, Austria, Poland, Portugal, Slovenia, Finland, Sweden and the UK. 5European countries with the most developed vocational education and apprenticeship systems are, e.g., Germany, Austria, Switzerland, Denmark and partially Caliendo and Schmidl IZA Journal of Labor Policy (2016) 5:1 Page 18 of 30 also the Netherlands. See Eichhorst et al. (2015) for a more detailed definition and assessment of the role of vocational education in industrialized countries. 6The Swedish unemployment rates more than tripled during the time period of investigation (from 3% to over 9%). In Norway, the unemployment rate during the period of investigation was at an all-time high at 5%. 7Depending on the specification; unfortunately the authors do not provide guidance of which of the two estimates is more reliable. 8According to the share of young participants in the respective program types, total expenditure information for youths were calculated for each country and program-type assuming that the program cost for youth and adult participants are similar. Price-differences across countries are accounted for by dividing country-specific costs by the purchasing power parities for government services. To increase reliability we only included countries that provided expenditures on at least two types of programs, and averaged expenditure information available between 2010 and 2013 whenever possible, otherwise the latest information available was used. Countries included are BE, DE, EE, ES,FR,LV,LT,AT,PL,PT,SI,SK,FI,SE. 9The increase in relative costs of job creation schemes suggests that countries with low foregone earnings invest relatively more on job creation schemes than on other measures. Appendix Table 3 Evaluation literature by Countries Country Author Austria Winter-Ebmer (2006) Belgium Cocky et al. (2013) Denmark (1) Graversen and van Ours (2008) Denmark (2) Jensen et al. (2003) Denmark (3) Maibom et al. (2014) Denmark (4) van den Berg et al. (2012) Finland Hämäläinen and Ollikainen (2004) France (1) Bonnal et al. (1997) France (2) Brodaty et al. (2011) France (3) Cavaco et al. (2004) France (4) Crépon et al. (2013) France (5) Fougère et al. (2009) Germany (1) Achatz et al. (2012) Germany (2) Bernhard and Kruppe (2012) Germany (3) Bernhard and Wolff (2008) Germany (4) Caliendo et al. (2008) Germany (5) Caliendo et al. (2011) Germany (6) Hartig et al. (2008) Germany (7) Hohmeyer and Wolff (2007) Germany (8) van den Berg et al. (2014) Germany (9) Wolff and Jozwiak (2007) Germany (10) Wolff et al. (2010) Hungary Micklewright and Nagy (2010) Norway Hardoy (2005) Portugal Centeno et al. (2009) Sweden (1) Bennmarker et al. (2013) Sweden (2) Carling and Larrson (2005) Sweden (3) Costa Dias et al. (2013) Sweden (4) Engström et al. (2012) Sweden (5) Forslund and Skans (2006) Sweden (6) Larrson (2003) Sweden (7) Andrén and Gustafsson (2004) Sweden (8) Hägglund (2014) UK (1) Bell et al. (1999) UK (2) Blundell et al. (2004) UK (3) Dorsett (2006) UK (4) Petrongolo (2009) UK (5) van den Berg et al. (2014) Caliendo and Schmidl IZA Journal of Labor Policy (2016) 5:1 Page 19 of 30 Table 4 Evaluation studies on labor market training Author Country Treatment Sample Horizon Estimation Employment Unemployment Education Quality/Wages Winter-Ebmer (2006) Austria School-based, full-time retraining courses and coaching Unemployed ≤26, lay-offs at Steel Factory Five years after treatment Tobit IV 0 n/a n/a + Jensen et al. (2003) Denmark (1) Introduction of vocational education program with 50% reduction of UB during participation Unemployed 16 to 24 years, long-term unemployed, low education levels. Up to 10 months after introduction Competing risk duration model 0n/a+0 Hämäläinen and Ollikainen (2004) Finland Labor Market Training school-based, various durations Youth Practical Training: paid training within a firm First-time unemployed youth, between 16 and 30 years 5 years after entry Propensity Score Matching LMT: + YPT: 0 LMT 0 YPT: 0 LMT - YPT: 0 LMT + YPT: 0 Bonnal et al. (1997) France (1) Workplace training on a temporary contract Young men who were less than 26 years, low or high levels of education Directly after participation, 6–12 months and >12 months after participation Multiproportional hazard model with unobserved heterogeneity Low educ: + high educ: 0/- Low educ: - high educ: 0/+ n/a +(probability of employment over 1 year) Brodaty et al. (2011) France (2) General or job-oriented practical training Low-skilled unemployed below 27 years Up to 6 month after treatment Propensity Score Matching with PS derived from a competing-risk duration model 0 (relative to JCS) - relative to fixed-term contracts n/a n/a n/a Cavaco et al. (2004) France (3) Retraining and job seeking assistance for 6months Recently displaced workers with min 2 years experience in the firm, <25 years Up to 3 years after program entry Generalized Tobit Model with multiple selectivity criteria + n/a n/a n/a Caliendo and Schmidl IZA Journal of Labor Policy (2016) 5:1 Page 20 of 30 Table 4 Evaluation studies on labor market training (Continued) Achatz et al. (2012) Germany (1) Short coaching or classroom training 18–30 year old recipients of means-tested UB benefits entering treatment within 9 weeks of start of receipt Up to 30 months Propensity Score Matching +n/a n/a n/a Bernhard and Kruppe (2012) Germany (2) Mediumto long-term labor market training based on vouchers Unemployed between 15 to 24 Up to 28 month after program entry Propensity Score Matching +0 n/a n/a Caliendo et al. (2011) Germany (5) Classroom based training programs Unemployment entrants up to 25 years, East and West separately, high/low previous education (lE,hE,lW,hW) Up to 5 years after treatment entry Propensity Score Weighting lE: 0 hE: + lW: +hW: + n/a lE: - hE: - lW: - hW: - n/a Hartig et al. (2008) Germany (6) Short coaching or classroom training Unemployed between 15 to 25 years, receiving means-tested benefits, by East/West/ Male/Female (Wm, Wf, Em, Ef) Up to 25 months after treatment entry Propensity Score Matching Wm: + Wf: 0 Em: 0 Ef: 0 Wm: 0 Wf: 0 Em: 0 Ef: 0 n/a n/a Wolff and Jozwiak (2007) Germany (9) Short classroom training Unemployed 15–24 receiving means-tested unemployment benefits Up to 20 months after program start Propensity Score Matching Men: + women: 0 Men: 0 women: + n/a n/a Hardoy (2005) Norway Mix of classroom and firm-based training, max 6 months with potential subsequent program participation Unemployment entrants between 16 to 25 2 years after unemployment entry Structural discrete choice model with selection -+(also ALMP participation) -n/a Caliendo and Schmidl IZA Journal of Labor Policy (2016) 5:1 Page 21 of 30 Table 4 Evaluation studies on labor market training (Continued) Hardoy (2005) Norway Classroom training for work-related skills, up to 5 months Unemployment entrants between 16 to 25 2 years after unemployment entry Structural discrete choice model with selection -+(also ALMP participation) -s n/a Carling and Larrson (2005) Sweden (2) Youth Guarantee (workplace training, classroom trainig, mix of the two) within 100 days of unemployment entry 20 to 24 year old unemployment entrants 18 months years after unemployment entry DIDID, using variation in time, age and municipality dimensions n/a - n/a n/a Costa Dias et al. (2013) Sweden (3) Practical labor market training (mostly work) Unemployment men 20 to 24 years 1 and 2 years after unemployment registrations Propensity Score Matching with IV-correction exploiting age-eligibility cut-offs -++(upon request) n/a Forslund and Skans (2006) Sweden (5) Youth practice vs. classroom training Unemployed youth between 20 and 24 Up to two years after program entry Propensity Score Matching, Regression +0n/a+ Larrson (2003) Sweden (6) Labor market training of different durations First-time unemployed youth 20 to 24 years One and two years after program start Propensity Score Matching 0n/a0Andrén and Gustafsson (2004) Sweden (7) Labor market training Unemployed entering training 20–25, different cohorts: T1(84/85), T2(87/88), T3(90/91) Average of three years after treatment entry Selection model with random coefficient n/a n/a n/a T1: 0 T2: + T3: +/0 Dorsett (2006) UK (3) Full-time education (FTET) relative to other NDYP-options Unemployed male UB benefit claimants for longer than 6 months, 18 to 24 years Up to 1.5 years Propensity Score Matching -/0 (relative to JCS, WS, and extended job search) +(rel. to WS), -/0 (relative to JCS and extended job search) n/a n/a Caliendo and Schmidl IZA Journal of Labor Policy (2016) 5:1 Page 22 of 30 Table 5 Evaluation studies on job search assistance and monitoring Author Country Treatment Sample Horizon Estimation Employment Unemployment Education Quality/Wages Graversen and van Ours (2008) Denmark (1) Intensive contact with caseworkers and job search program with potential subsequent training participation Unemployment entrants <30 years 10 to 15 weeks after unemployment entry (before training entry) Randomized Controlled Trial +n/a n/a n/a Maibom et al. (2014) Denmark (3) Information letter, and intensified caseworker meeting, and mentoring, training participation. Shortor long-term unemployed youth below 30 years, with or without qualifying education Up to 3 years after entry into program Randomized Controlled Trial Uneducated: - educated: 0 n/a Uneducated: 0 educated: 0 n/a van den Berg et al. (2012) Denmark (4) Meeting with the caseworker (counseling or monitoring) Unemployed men and women with UB receipt, below 30 years (Instantaneous effect) Timing-of-events +n/a n/a n/a Crépon et al. (2013) France (4) Job search assistance by by private providers University graduates (under 30 years), at least 6 months in unemployment 8 month Randomized Controlled Trial +(partially due to displacement) n/a n/a n/a Fougère et al. (2009) France (5) Vacancies information by the public employment service Registered unemployed job seekers Max. 1.5 years Structural partial job search model +n/a n/a n/a Bernhard and Wolff (2008) Germany (3) Contracting out placement services to private providers Recipients of social welfare, 15 - 24, East and West Germany, males and females (Wm, Wf, Em, Ef) Up to 20 month Propensity Score Matching Em: +Ef: + Wm: 0 Wf: 0 Em: 0 Ef: + Wm: +Wf: + n/a n/a van den Berg et al. (2015) Germany (8) Partial or complete withdrawal of social benefits (sanctions) for up to 3 months Unemployed social welfare recipients 18 to 24 years Immediate effect of sanction Timing-of-events +- (only for single households) n/a - Caliendo and Schmidl IZA Journal of Labor Policy (2016) 5:1 Page 23 of 30 Table 5 Evaluation studies on job search assistance and monitoring (Continued) Micklewright and Nagy (2010) Hungary Increased job search monitoring for 4 months New UI claimants with 75–179 days of UI entitlement Up to 3 months Randomized Controlled Trial 0 n/a n/a n/a Centeno et al. (2009) Portugal Early intervention, including job search assistance, counseling, monitoring and training, threat of sanctions Youths below the age of 25 who were unemployed less than 6 months 12 and 24 months after participation Difference-inDifference Matching 0 n/a 0 n/a Bennmarker et al. (2013) Sweden (1) Assignment to private placement providers, up to 6 months Unemployed less than 25 years, with unemployment duration >3 months Up to 12 months after assignment Randomized Controlled Trial 0n/an/a0 Engström et al. (2012) Sweden (4) Increased threat of monitoring of job search activity Unemployed less than 30 years 4 to 8 months Randomized Controlled Trial 0 n/a n/a n/a Hägglund (2014) Sweden (8) Increased intensity of job search activities and monitoring Regular and longterm unemployed between 18 to 24 years Up to 3 years Randomized Controlled Trial regular: + long-term: 0 regular: 0 long-term: 0 n/a regular: + long-term: 0 Blundell et al. (2004) UK (2) Intensified job search assistance/mentoring and wage subsidies, threat of sanctions Unemployed, 18 to 24 years, longer than 6 months unemployed First 4 months after program entry DID using age and regional variation in eligibility Men: +(partially due to wage subsidy) women: 0 n/a n/a n/a Petrongolo (2009) UK (4) Reduction of amount and duration of UB and introduction of job search requirements (diary) Male unemployment entrants 16 to 24 3monthsand6 months DID approach Short-run: + long-run: - +n/a - Caliendo and Schmidl IZA Journal of Labor Policy (2016) 5:1 Page 24 of 30 Table 5 Evaluation studies on job search assistance and monitoring (Continued) Dorsett (2006) UK (3) Extended job search relative to other NDYP options Unemployed male UB benefit claimants for longer than 6 months, 18 to 24 years Up to 1.5 years Propensity Score Matching +(rel. to all other options) 0(rel.toWS); - (rel. to training and JCS) n/a n/a van den Berg et al. (2014) UK (5) Enrollment in NDYP - Gateway period (job search) Unemployed for at least 6 months, 18–24 years One month Kernel hazard regression exploiting random timing of introduction +n/a n/a n/a