Mitigating Long-term Unemployment in Europe

LTU remained a key challenge in some European countries, especially for those who were hit the hardest during the GFC. As of 2018/2019, the LTU rate of Greece was the highest, where almost one in five working-age population were LTU. Along with Spain, the LTU rate in Greece increased about fourfold after the GFC, compared with the pre-crisis period. Nonetheless, there were significant improvements in the LTU rates in some countries, many of which were the new member states (NMS).

NMS includes Bulgaria, Croatia, Czech Republic, Estonia, Hungary, Latvia, Lithuania, Poland, Romania, Slovakia, and Slovenia.

In some countries, the youth LTU rates remained high historically and increased significantly after the GFC. Countries with high LTU rates tend to have very high long-term youth-unemployment rates. These include Greece, Italy, Spain, Croatia, and Slovakia. Between pre- and post-GFC, the most significant increases in LTU rates among young people were seen in Greece (>15 percentage points), Italy, and Spain (about 10 percentage points increases). Overall, the youth LTU rates declined significantly in Germany and in many NMS, particularly in Poland, the Czech Republic, Hungary, and Slovakia.

Despite significant improvement in the LTU rate, the incidence of LTU remained elevated in the EU. After the GFC, the EU average of the incidence of LTU slightly increased to 46%, compared to 43% during 2005–2008. Greece had the highest level of incidence of LTU after the crisis, where almost three-fourths of the unemployed were long-term, followed by Slovakia, Italy, and Bulgaria. Generally, when the LTU rate improved, the incidence of LTU also declined. The exceptions include Bulgaria and Belgium, where LTU rates declined after the GFC but the incidence of LTU increased, suggesting that LTU could be attributed to structural factors.

The incidence of LTU varied across demographic groups. Across age groups, the incidence of LTU in the EU was the highest in people with the age group of ≥ 50 years: 60% of the unemployed in that cohort were out of job for >12 months. By gender, the incidence of LTU appeared to be similar for male and female. By education levels, the incidence of LTU was the highest in those with low skill levels. Depending on the structure of the labor force and the LTU risk, either the low-skilled or those with medium skill levels represented the largest group among the unemployed. The risk of falling into LTU among those with an intermediate skill level was still high in several EU countries, including Greece, Spain, Croatia, Portugal, Slovakia, and Ireland.

Another interesting aspect is to assess factor contributing to the incidence of LTU at a regional level (Elhorst, 2003). Nonetheless, our study is bounded by data limitation, particularly on ALMP spending at a regional level.

Addressing LTU would alleviate poverty and social exclusion, and would be beneficial for many EU countries, particularly those with demographic challenges. LTU has adverse consequences for individuals and the economy as the long-term unemployed tend to suffer from low social mobility and face higher risk of poverty. Even those who are reemployed tend to earn less than in their previous jobs and are demoted from their past careers (Nichols et al., 2013). Getting jobs for the long-term unemployed would bolster income and alleviate poverty. For countries facing demographic challenges, integrating the long-term unemployed into the labor market (and also discouraged persons) would help to alleviate the issue as they are a potential source of productive labor. LTU also has negative effects on youths’ prospects and their risk of social exclusion. Existing studies show that experiencing a protracted unemployment not only strongly affects all dimensions of a young person's psychological wellbeing and quality of life but also weakens their future employment outcomes as well as trust in institutions (Eurofound, 2017; Duell et al., 2016).

To explore the potential drivers of LTU, we consider the following panel regression model: LTUit=γ1Xit+γ2Zit+ηi+vt+εit,i=1,…,N;t=1,…,T,\matrix{{{LTU}_{it} = {\gamma_1}{X_{it}} + {\gamma_2}{Z_{it}} + {\eta_i} + {v_t} + {\varepsilon_{it}},} \hfill & {i = 1, \ldots ,N;t = 1, \ldots ,T,} \hfill \cr} where LTU_it denotes the incidence of LTU of country i in year t, X_it is a vector of macroeconomic variables, and Z_it is a vector of policy variables. We have fixed effects for country (η_i) and time (ν_i) to account for unobserved country-specific heterogeneity and the global common factor, and ɛ_it is a random error term.

Regressors comprise macroeconomic and labor market variables. We include lagged GDP growth, inflation rate, skill mismatches, and estimated labor market matching efficiency. Based on the Okun's law and the Philips’ curve, economic growth and inflation rate are inversely associated with unemployment and, to some extent, LTU. Moreover, LTU could be a result of the unemployed having skills that do not match demand and low matching efficiency (Pissarides, 2000). As the incidence of LTU is known to be persistent, we also include the lagged incidence of LTU (one lag). For policy variables, our model includes both passive labor market policies (PLMPs) and ALMPs as well as a tax wedge. The literature shows that certain types of ALMPs are effective in reducing LTU (Card et al., 2010, 2018; Kluve et al., 2019). Moreover, unemployment benefits and a tax wedge also play some roles in reducing LTU (Bassanini and Duval, 2006; OECD, 2006).

We use annual data of 25 European countries over the period 2000–2018 (see Appendix A). Data for all variables (including underlying data for estimated variables) come from the Eurostat database. Labor market policy variables include both expenditures on ALMPs and PLMPs. They are defined as public spending on labor market policies (LMP) per unemployed and expressed in percentage of GDP per capita, which is often used in the empirical research on unemployment (Arpaia et al., 2014). Skill mismatch is calculated as a sum square of the differences between the shares of population and employment by education attainment (see Section 5). Matching efficiency is computed using the matching function approach (see Section 5 and Appendix B). Table 1 presents the descriptive statistics and pairwise correlations for the incidence of LTU and its potential drivers.

Table 1 shows that the correlations between the incidence of LTU and its potential drivers are of expected signs. The incidence of LTU is negatively correlated with the lagged GDP growth rate and the inflation rate as suggested by the Okun's law and the Phillips curve. A positive correlation (0.41) between the incidence of LTU and skill mismatch is as expected. In line with the prediction of the search and matching model (Pissarides, 2000), the incidence of LTU and labor market matching efficiency are negatively correlated with the correlation of −0.37. Furthermore, the incidence of LTU is also positively correlated with the youths who are not in employment, education, or training (NEET). Finally, there is a negative relationship between the incidence of LTU and expenditures in LMP.

The main results are broadly robust. A significant degree of persistence in LTU implies that our estimators could be biased as the lagged dependent variable is correlated with the error term. To account for dynamic effects and address the endogeneity problem, we also perform dynamic panel data estimation as in Arellano and Bond (1991). The results are shown in column (2) of Table 4. Similar to the fixed-effect estimation, the incidence of LTU is associated with skill mismatches, matching efficiencies, and some types of ALMPs. For robustness check, we also perform OLS estimation on the pooled data [Table 4, column (3)]. The results are similar to those of the benchmark and dynamic panel data estimation, although training is not statistically significant.

The incidence of LTU appears to be more sensitive to economic growth, matching efficiency, and ALMPs in the NMS. We now apply the fixed-effect panel data estimation but limit the sample to the NMS. Column (4) of Table 4 shows that while the qualitative results remain broadly unchanged, the size of coefficients on of economic growth, matching efficiency, and ALMPs become larger for the NMS.

Employment protection is not found to be associated with the incidence of LTU. The role of employment protection legislation (EPL) on unemployment has been discussed in the literature. Its impact on LTU remains mixed in empirical studies. For example, OECD (2004) finds that strict employment protection increases the incidence of LTU, while Heckman and Pagés-Serra (2000) show that EPL has no effect on LTU. By using the indicator of EPL developed by OECD, we find that EPL does not have statistically significant impact on the incidence of LTU.

Labor market policy interventions cover labor market services, PLMPs, and ALMPs. The Eurostat database breaks down LMP intervention into three main categories: (1) labor market services, (2) PLMPs, and (3) ALMPs. Labor market services cover all services of the PES and any other publicly funded services for jobseekers. It generally includes “services and sanctions” and aims at enhancing job search efficiency (see Section 4.3). Second, PLMPs cover financial supports that aim to compensate jobseekers for loss of income during unemployment. They include unemployment benefits and incomes that facilitate early retirement. Finally, ALMPs cover policies that activate jobseekers to find employment. They comprise training, employment incentives, startup incentives, supported employment and rehabilitation, and direct job creation.

The levels and types of spending in labor market policy intervention varied considerably across Europe. These reflected diverse labor market characteristics, challenges, and government's priorities across countries. In 2017, the average LMP spending for the whole EU was at 1.3% of GDP but differed significantly across countries. In general, LMP spending in the advanced economies was twice as large as those in the NMS. By country, spending on LMP interventions ranged from just 0.1% of GDP in Romania to 3% of GDP in Denmark. In many countries, spending on LMP interventions peaked during the GFC and continued to decline to the levels slightly below the pre-crisis level. These countries included Denmark, Germany, Ireland, France, the Netherlands, Portugal, Romania, Sweden, and Finland.

Most countries allocated the largest share of LMP resources on PLMPs, followed by ALMPs and PES, respectively. In 2017, Cyprus and Bulgaria allocated more than three-quarters of its LMP resources to PLMPs, in contrast with Hungary and Malta, where the shares of spending on PLMPs were about 20% and 15% of the total labor market expenditure, respectively. For the EU average, about half of the LMP interventions spending was allocated to PLMPs (mostly in unemployment benefits), and one-third was spent on ALMPs. Some countries, however, put spending priorities on ALMPs. They included Czech Republic, Denmark, Croatia, Lithuania, Luxembourg, Hungary, and Poland. Malta was the only country among the EU that spent the largest share of LMP on labor market services (i.e., PES).

Across different types of ALMPs, most countries allocated resources to employment incentives, followed by training programs. On average, EU countries spent one-third of ALMPs on employment incentives in 2017, followed by training (about 30%), rehabilitation, and direct job creation, respectively. Startup incentives generally accounted for a very small share in most countries (<5%), with exceptions for Spain, Poland, Sweden, and Croatia. There were significant country-level variations of ALMP spending. Austria, Germany, France, Croatia, Latvia, Portugal, and Finland prioritized their spending on the training programs, whereas Bulgaria, Greece, Hungary, and Slovenia spent the largest share of their ALMP budgets on direct job creation.

Policy implications: Existing literature, successful country experiences, and our findings point to the following implications for designing ALMPs.

Skill mismatches are defined as imbalances between skills demanded for labor and skills available in labor supply. Ideally, the proper way to measure skill mismatch would require data on vacancies and unemployment separately for different skill levels, proxied by education. The higher is the discrepancy between vacancies and unemployment within a particular skill compared with that prevailing throughout the whole economy, the higher the associated degree of mismatch. As vacancies by education level are not available, we follow Estevao and Tsounta (2011) and construct a mismatch indicator as differences between employment and working-age population by education groups.

The degree of skill mismatches varied greatly across the EU and across time. Skill mismatches appeared to be high in Belgium, Ireland, and Bulgaria, while they were low in the Netherlands, Portugal, and the United Kingdom. Relative to pre-GFC, about half of the EU member states experienced an improvement in skill matching. The largest improvements occurred in many NMS. In contrast, skill mismatches deteriorated after the GFC in many advanced countries, including Ireland, Greece, Spain, and Portugal.

Digital skills are a prerequisite for the current and future of work (Figure 2). Changes in skill demand and supply have resulted in difficulty for firms to find employees with the right skills (Cedefop, 2018a). In light of a growing digitalized economy, the skills that will be increasingly in high demand include digital skills. Based on Cedefop's skills online vacancy analysis tool, almost half of the vacancies were related to Information and Communication Technologies (ICT), healthcare, science, and business and retail sectors and they are expected to increase in the future (Cedefop, 2018b). Many of these vacancies require some knowledge of digital skills. According to the ESJS in 2014, >85% of the EU employers required at least basic ICT skills to perform the job. Going forward, future structural transformation of the EU labor markets will be accompanied by a high demand for digital skills. The experience of the COVID-19 lockdown will further increase the premium of digital skills.

Figure 2

Despite the growing importance of digital skills in the EU labor market, digital skill gaps—measured by the difference between the percentage of employees who need at least the basic level of ICT skills to work and the percentage of individuals who have at least basic and overall digital skills—remained high in the EU. All countries except the Netherlands experienced skill gaps of at least 15%, where digital skill gaps in Bulgaria, Croatia, and Italy were as high as 50%.

Policy implications: Based on the experiences of the world's best performers in the area of digital skills, digital literacy can be strengthened through policies aimed at creating a conducive environment for the development of digital skills, and sectoral policies focusing on education and training.

Labor market matching efficiency in EU generally declined after the GFC and improved slightly in recent years. Most EU countries experienced declines in their matching efficiency immediately after the financial crisis (Figure A1 in Appendix B). However, the recovery process in the post-crisis period differed across countries. For example, after deteriorating during the GFC, the matching efficiency improved to the pre-crisis level in the Czech Republic and Estonia. On the other hand, the matching efficiency in Bulgaria, Greece, Ireland, Lithuania, Romania, the United Kingdom, Spain, and Slovenia continued to decline.

Strengthening the role of PES could enhance matching efficiency. Improving labor market matching efficiency requires policies beyond those aimed at stimulating aggregate demand, since it relates to structural factors, including institutional inefficiencies, which dissuade job seekers from accepting a job (Bova et al., 2016). PES play an important role in contributing to a well-functioning labor market, as they facilitate the process of matching the jobseekers and employers. Based on the EU's peer-to-peer review and G20 case studies, successful programs offered by PES that bring about sustainable employment for the LTU include the following factors:

Targeted and personalized programs that combine several interventions: Initial profiling is an effective tool to categorize jobseekers before designing other interventions. Germany has developed a comprehensive skills assessment profiling tool, so-called “Kodiak,” and linked the profiles to regional labor market vacancies. It consists of a self-assessment questionnaire, aptitude test, interview to assess achievement motivation, assessment of social and communication skills required in selected occupations, and technical standards for the analysis of personal skills. Denmark's approaches for interventions depend on matching groups categorized by the Employability Profiling Toolbox. Beyond initial profiling, labor market interventions for the LTU generally require a step-by-step approach, starting from strengthening basic skills and coaching, followed by workplace-oriented training, vocational training, and job search assistance. Austria, Belgium, Portugal, and the United Kingdom encourage employers to offer internships to the LTU by providing allowances for those registered with PES.