Training programmes play a crucial role in supporting the reskilling and upskilling of jobseekers, enabling them to secure good jobs. This chapter examines the impact of Slovenia’s training programmes for unemployed individuals on their labour market outcomes through a counterfactual impact evaluation. It focuses on the main training programmes (Preparation for National Vocational Qualifications, Institutional Training and Non-Formal Education and Training) offered between 2015 and 2018, analysing a wide range of outcomes. These include the probability of finding employment, remaining unemployed or becoming inactive, receipt of unemployment benefits, earnings, total days worked, occupational mobility and the probability of migrating abroad.
Impact Evaluation of Wage Subsidies and Training for the Unemployed in Slovenia
Abstract
Training programmes are a key element of active labour market policies (ALMPs), as they equip unemployed individuals with new skills or enhance their existing ones, thereby supporting improved labour market outcomes. Indeed, creating effective upskilling and reskilling opportunities is increasingly important in light of ageing population and the twin digital and green transitions, which are reshaping labour markets and leading to skill and labour shortages in specific sectors. However, these programmes can be expensive, and not all are equally effective in achieving their objectives. While participants in training programmes often show higher employment rates following the training completion compared to non-participants, it is not always clear whether these differences are caused by the programme itself or by other characteristics of the participants. To study the causal impact of training programmes in Slovenia, this chapter presents a counterfactual impact evaluation (CIE) of training programmes available to unemployed individuals in Slovenia, using the methodology and data presented in Chapter 3.
The CIE shows that Slovenia’s training programmes are effective across a number of outcomes. The findings indicate that these programmes significantly increase the probability of finding employment while reducing the probability of remaining registered as unemployed or becoming inactive, with little impact on the likelihood of receiving unemployment benefits. In addition, the training programmes are effective in increasing cumulative days worked and cumulative earnings, while reducing occupational mobility. These programmes support participants’ labour market outcomes in the short, medium, and long term, except for the initial months when participants are actively engaged in training activities. The estimated effects exceed most international estimates of the employment effects of training programmes. Moreover, the training programmes are effective in supporting a wide range of jobseekers into employment, across age, gender and duration of unemployment.
The rest of this chapter is laid out as follows. Section 4.2 examines the effect of Slovenia’s main training programmes (considered as one single programme) for unemployed people on the probability of employment, registration as unemployed, inactivity, the probability of receiving unemployment benefits, as well as the effect of training programmes on cumulative earnings, cumulative days in employment, and occupational mobility. It also provides distinct estimates of the effect of each of the five training programmes on the probability of finding employment. Section 4.3 looks at effectiveness of training programmes across various jobseeker characteristics: gender, age, education level, nationality and unemployment duration. Finally, Section 4.4 compares the findings of this report with international studies and related evidence on the effectiveness of Slovenia’s training programmes.
Studying the causal effect of training requires understanding what would have happened to participants had they not participated in the programme – the counterfactual. As discussed in Chapter 3, the counterfactual is not directly observable in the data because individuals can only be observed in one state at a time, either as participants or non-participants. The primary goal of an impact evaluation, therefore, is to use existing data to construct a credible counterfactual. This report employs propensity score matching to estimate the counterfactual, which involves comparing the outcomes of training participants with those of comparable – along multiple characteristics – non-participants. This section outlines the results of this analysis which makes use of linked administrative data from the Statistical Office of the Republic of Slovenia (SORS) and the Employment Services of Slovenia (ESS), as described in Chapter 3.
As highlighted in Chapter 3, the various training programmes share certain similarities while also exhibiting some differences. While the main analysis in Chapter 4.2 examines the overall effect of all training programmes collectively, Chapter 4.3 disaggregates this effect by individual programme. Training programmes increase the probability of finding employment.
Figure 4.1 presents the main findings of the propensity score matching analysis for the likelihood of being employed. Panel A shows the percentage of individuals who find employment over time (measured in months since the start of training) for both the treatment group (training participants) and the matched control group (non-participants). Panel B reports the net effect of the training programmes, calculated as the difference in employment rates between the two groups shown in Panel A.
The analysis indicates that training programmes are effective in increasing the probability of unemployed individuals finding employment. Twelve months after the start of the programme, training participants are 10 percentage points more likely to find employment compared to the matched control group (Figure 4.1, Panel B). This effect diminishes only slightly to approximately 9 percentage points after 30 months but remains otherwise consistent and stable over the medium to long term. The employment data provided for this report covers the period up to 2022, allowing the employment effect to be tracked for a maximum of 48 months from the start of training for all participants. As shown in the accompanying technical report (OECD, 2025[1]), this effect remains consistent throughout the full 48‑month tracking period available in the data.
The analysis also highlights the importance of comparing outcomes for similar individuals when assessing the impact of the programme. Simply considering the employment rate of training participants 12 months after the programme – 44.8% (Figure 4.1, Panel A) – would be misleading. This is because only part of this effect can be attributed to the programme, as some participants would have found employment anyway. Employing a matched comparison (or control) group provides a credible counterfactual: Panel A shows that approximately 34.2% of the control group were employed 12 months after comparable individuals in the treatment group started the training. The difference between the two lines in Panel A represents the effect of the training on the probability of finding employment at different intervals from the start of the training. At the 12‑month mark, this effect amounts to 10.6 percentage points.
The impact of training on the likelihood of finding employment takes time to materialise. As discussed in Chapter 3, the median duration of the training programmes analysed in this report is just under two months. During this period, participants have less time available to work or search for employment. As a result, the probability of finding employment during the training is higher for the control group than for the treatment group, leading to a temporary negative effect of training in these initial months. In the literature, these short- to medium-term effects are referred to as “lock-in” effects (Card, Kluve and Weber, 2018[2]).
Share of individuals in employment and percentage point effect on employment probability
Note: Panel A shows the percentage of employed persons among the treatment and matched control group. Panel B plots the treatment effect (i.e. the difference between the treatment and control group shown in Panel A). The analysis presents nearest-neighbour propensity score matching results which matches individuals based on several characteristics: each individual’s employment history (earnings, occupation), unemployment information (unemployment spell duration, unemployment benefit level and remaining duration), employability rating (based on ESS counsellor judgement), disability status, demographic characteristics (age, education, gender, nationality, marital status), and location. For every individual in the treatment group, the matching is conducted based on the values of these characteristics in the calendar month when the individual enters the programme. The control group is comprised of individuals with similar characteristics not entering active labour market programmes in that same calendar month. For paired individuals in the treatment and control groups, this calendar month is then the reference point after which outcomes are measured. Some individuals in the treatment group are dropped as they do not have a corresponding match. The confidence intervals in Panel B are shown at the 5% level of significance and represented by the whiskers delimiting the dotted lines on the charts.
Source: OECD calculations based on data from the Statistical Office of the Republic of Slovenia and the Employment Service of Slovenia.
4.2.1. Training programmes improve a wide range of outcomes
While employment is a key outcome that training programmes are likely to affect, it is crucial to consider other outcomes to gain a comprehensive understanding of the impact of training. The rich administrative data shared by Slovenia for the CIE enable the analysis of outcomes beyond employment probabilities. These additional outcomes include the likelihood of remaining unemployed or becoming inactive, the probability of receiving unemployment benefits, cumulative earnings, cumulative days worked, occupational mobility and the probability of migrating from Slovenia.
Figure 4.2 shows the estimated effects of training programmes on various labour force status, including employment, unemployment, benefit receipt, and inactivity. These outcomes are based on administrative data rather than surveys. Therefore, caution is advised when interpreting these results due to the potential disparities between registered and survey-based unemployment. For completeness, the impact on employment has been carried over from Figure 4.1, Panel B. To simplify the presentation, the figure displays only the treatment effects, without showing the outcomes for the treatment and control groups as in Figure 4.1. Overall, the findings indicate that training programmes:
Reduce registered unemployment: During the initial “lock-in” phase, participants are more likely to remain registered as unemployed, but this trend reverses 4 to 5 months after the start of the training programme. After 12 months, training participants are approximately 5.3 percentage points less likely to be registered as unemployed compared to matched non-participants, a difference that is statistically significant. This effect also persists in the long term, although its magnitude diminishes slightly to around 4 percentage points.
Leave benefit receipt unaffected: The impact of training on unemployment benefit receipt remains negligible in the short, medium, and long term, with most estimates being statistically non-significant. At first glance, this may seem counterintuitive: if training participants are more likely to enter employment than comparable non-participants, should they not also be less likely to receive unemployment benefits? In reality, entering employment has two opposing effects on benefit receipt. On one hand, employed individuals are no longer registered as unemployed and are therefore not entitled to unemployment benefits, reducing the probability of receiving them. On the other hand, employed individuals may work for a period long enough to qualify for unemployment benefits if they subsequently return to unemployment, increasing the probability of receiving benefits. In Slovenia, these two effects appear to balance each other out, resulting in an overall impact of training on unemployment benefit receipt that is close to zero.
Reduce inactivity: From the start of the programme, training participants are less likely to be inactive (i.e. not in employment, not unemployment and not having migrated abroad) than comparable non-participants. Twelve months after starting the programme, this effect amounts to 4.7 percentage points and persists in the long term, fluctuating between 4 and 6 percentage points. In the early months, this reduction in inactivity is primarily due to a higher likelihood of being registered as unemployed, whereas in the longer term, it is driven by an increased likelihood of being employed.
Increase employment (as discussed in Section 0 above).
The accompanying technical report (OECD, 2025[1]) shows that these effects are statistically significant except for the effect on unemployment benefit receipt.
Percentage point effect on being in different labour market status
Note: The analysis presents nearest-neighbour propensity score matching results which matches individuals based on several characteristics: each individual’s employment history (earnings, occupation), unemployment information (unemployment spell duration, unemployment benefit level and remaining duration), employability rating (based on ESS counsellor judgement), disability status, demographic characteristics (age, education, gender, nationality, marital status), and location. For every individual in the treatment group, the matching is conducted based on the values of these characteristics in the calendar month when the individual enters the programme. The control group is comprised of individuals with similar characteristics not entering active labour market programmes in that same calendar month. For paired individuals in the treatment and control groups, this calendar month is then the reference point after which outcomes are measured. Some individuals in the treatment group are dropped as they do not have a corresponding match.
Source: OECD calculations based on data from the Statistical Office of the Republic of Slovenia and the Employment Service of Slovenia.
In addition to analysing the effects of training programmes on labour force status, it is important to examine their impact on a broader set of labour market outcomes. Figure 4.3 offers insights into the quality of employment by assessing the impact of training programmes on cumulative days in employment, cumulative earnings and occupational mobility since the start of the programme. The findings indicate that participation in training programmes results in (as shown by the confidence intervals reported in Figure 4.3, all effects become statistically significant after the initial months of training):
Increased cumulative days in employment: 3 months after the start of the programme, training participants begin to accumulate more days in employment compared to the control group. By the 3‑year mark, training participants have, on average, worked 100 more days than comparable non-participants.
Increased cumulative earnings: Cumulative earnings follow a similar pattern to cumulative days in employment. Beginning 3 months after the start of the programme, cumulative earnings for the treatment group start to exceed those of the control group. These effects build steadily over time, and by three years after the start of training, participants have earned an additional EUR 4 500 compared to the control group (adjusted to 2022 prices).
Decreased mobility towards higher-paying occupations: Following the initial training period, individuals in the treatment group exhibit a lower occupational index (details on the index can be found in Chapter 3) than those in the control group. This suggests that, conditional on finding employment, treated individuals are more likely to find employment in lower-paying occupations compared to their counterparts in the control group who also found employment. However, the magnitude of this effect, at 2.5%, is small.
The effects on days in employment and cumulative earnings exhibit a linear trend, showing consistent growth. This aligns with the earlier finding of a steady increase in employment each month. However, conditional on finding employment, training participants appear to enter lower-paying occupations compared to non-participants, as shown in Figure 4.3, Panel C.
A comparison of occupations in which training participants are employed 12 months after starting the training with those of matched non-participants reveals that training participants are overrepresented in the 4‑digit ISCO‑08 occupational groups Bus and Tram Drivers (2.27% of training participants versus 0.07% of matched non-participants), Heavy Truck and Lorry Drivers (3.44% versus 0.87%), and Security Guards (9.74% versus 0.74%). Although these occupations have a lower occupational index than the labour market average, they have consistently been identified by the ESS as occupations experiencing labour shortages (Employment Service of Slovenia, 2018[3]; 2024[4]). This partly explains why, despite being more likely to enter lower-paying occupations than the control group, training participants also tend to secure more stable employment relationships, characterised by more days worked and higher cumulative earnings than those of comparable non-participants.
Training programmes can influence individual outcomes beyond labour force status and labour market outcomes, including the decision to migrate abroad. These programmes aim to reduce barriers to labour market entry in Slovenia, enhance employment prospects, and thereby potentially decrease the incentive to seek employment abroad.
Cumulative days worked, cumulative earnings, and occupational mobility for those who found a job
Note: The analysis presents nearest-neighbour propensity score matching results which matches individuals based on several characteristics: each individual’s employment history (earnings, occupation), unemployment information (unemployment spell duration, unemployment benefit level and remaining duration), employability rating (based on ESS counsellor judgement), disability status, demographic characteristics (age, education, gender, nationality, marital status), and location. For every individual in the treatment group, the matching is conducted based on the values of these characteristics in the calendar month when the individual enters the programme. The control group is comprised of individuals with similar characteristics not entering active labour market programmes in that same calendar month. For paired individuals in the treatment and control groups, this calendar month is then the reference point after which outcomes are measured. Some individuals in the treatment group are dropped as they do not have a corresponding match.
Source: OECD calculations based on data from the Statistical Office of the Republic of Slovenia and the Employment Service of Slovenia.
Figure 4.4 illustrates the effect of training programmes on the likelihood of migrating from Slovenia. Not surprisingly, during the initial months of training, participants are less likely than comparable non-participants to migrate abroad. Interestingly, however, this effect persists even after the training is completed. Twelve months after the start of the training, participants are 0.40 percentage points less likely than matched non-participants to migrate abroad. While modest in absolute terms, this represents a significant reduction in relative terms: by 12 months, 1.04% of individuals in the control group have migrated abroad, compared to just 0.65% of individuals in the treatment group – a reduction of 37.5%. Although this effect remains stable until 27 months after the programme’s start, it becomes insignificant thereafter. One potential factor contributing to this is the return of Slovenians from abroad at the onset of the COVID‑19 crisis in March 2020.
Probability of migrating from Slovenia
Note: The analysis presents nearest-neighbour propensity score matching results which matches individuals based on several characteristics: each individual’s employment history (earnings, occupation), unemployment information (unemployment spell duration, unemployment benefit level and remaining duration), employability rating (based on ESS counsellor judgement), disability status, demographic characteristics (age, education, gender, nationality, marital status), and location. For every individual in the treatment group, the matching is conducted based on the values of these characteristics in the calendar month when the individual enters the programme. The control group is comprised of individuals with similar characteristics not entering active labour market programmes in that same calendar month. For paired individuals in the treatment and control groups, this calendar month is then the reference point after which outcomes are measured. Some individuals in the treatment group are dropped as they do not have a corresponding match.
Source: OECD calculations based on data from the Statistical Office of the Republic of Slovenia and the Employment Service of Slovenia.
4.2.2. Certain limitations of the counterfactual impact evaluation should be taken into account when interpreting the results
When interpreting the findings of this report, it is important to consider one key limitation. Omitted variable bias resulting from unobserved characteristics – such as motivation – that influence both the decision to participate in training and labour market outcomes (e.g. the likelihood of finding employment or earnings) may lead to an upward bias, causing the results to overstate the true effect of training.
Omitted variable bias is a common concern in observational studies. Participants often self-select into training programmes based on individual characteristics (see Chapter 3), leading to systematic differences between participants and non-participants, which may partially drive variations in outcomes. The propensity score matching methodology allows to control for differences between training participants and non-participants with respect to observed characteristics. However, it relies on the assumption that all factors influencing individuals’ selection into training are observed and accounted for. In other words, it assumes that participants and non-participants do not differ in unobserved characteristics that simultaneously affect their labour market outcomes (Rosenbaum and Rubin, 1983[5]).
Unfortunately, it is not possible to guarantee that all relevant characteristics are controlled for. Certain attributes, such as motivation and ability, are typically unobserved but may plausibly influence both the decision to participate in training and subsequent labour market outcomes. If this is the case, training participants might have been more likely to find employment even without the training. This could result in an overestimation of the effectiveness of training. The accompanying technical report (OECD, 2025[1]) addresses some of these concerns. Specifically, it presents evidence that the treatment and control groups do not differ in terms of pre‑treatment outcomes, which mitigates concerns about omitted variable bias.
One way to avoid omitted variables bias is to conduct a randomised controlled trial (RCT) instead of an observational study. In RCTs, individuals are randomly assigned to the training, thereby ensuring that the treatment and control groups are similar in terms of both observables and non-observable characteristics. In the future, Slovenia could consider using RCTs to test the effectiveness of ALMPs.
Having established the effectiveness of Slovenia’s training programmes across various outcomes, the next question is how these effects vary based on the type of training and the characteristics of participants. This section delves into the aggregate results by first examining the impact of each of the five different training programmes and then analysing their effects on various groups of participants.
4.3.1. All five training programmes have positive employment effects, but the magnitude of these effects varies
The effect shown in Figure 4.1 represents an average across all five training programmes. However, the effectiveness of each programme varies significantly (Figure 4.5). While all five programmes evaluated improve employability in Slovenia, some are more effective than others. The national vocational qualifications (NVQ) preparation programme and the local Non-Formal Education and Training programmes attain the highest effectiveness, increasing the probability of finding employment 12 months after programme start by approximately 17 percentage points. The effect is slightly lower for the regular Non-Formal Education and Training programme (12 percentage points) which targets older jobseekers and long-term unemployed people or with low education level and its variant for young people (14 percentage points). In contrast, Institutional Training programmes show a much smaller effect, increasing the probability of employment by only 5 percentage points 12 months after the programme’s start.
The reasons for these differences across programmes are likely to be manifold. First, the programmes differ in their purpose (see Chapter 3). The NVQ preparation programme focuses on helping participants prepare for examinations to obtain a national certificate that validates their skills. Their goal is to promote the recognition of prior learning, enabling participants to certify they possess the skills needed to work in certain occupations (or part of these skills). This directly enhances participants’ employability, which may explain the stronger employment effects observed for these programmes. Local Non-Formal Education and Training programmes offer targeted opportunities for reskilling and upskilling. Their local nature enables them to be tailored to the specific needs of the local labour market, which may help explain the stronger employment effects observed.
Programme costs also vary, as reported in Table 3.1. Interestingly, the most effective programmes – the NVQ preparation programme and local Non-Formal Education and Training programmes – are also the most expensive, whereas the least effective (i.e. Institutional Training programmes) are the least costly. This suggests a positive correlation between programme costs and effectiveness, although a direct comparison is not possible due to differences in the composition of participants across programmes.
However, even within specific training programmes, there is considerable variation in training content, covering diverse fields such as business, IT, construction, and production technologies. Additional variation arises from the fact that courses within the same programme may be delivered by different external providers. Currently, providers are primarily selected based on cost, largely due to a lack of systematic measures of provider quality, which results in variation in training quality across providers. While outside the scope of this report, detailed information on the attributes of specific courses attended by participants (e.g. a Slovene language course) and the characteristics of training providers would enable a more granular analysis of the impact of training programmes across different fields and providers.
Second, as discussed in Chapter 3, the different programmes target different groups of jobseekers. For example, the regular Non-Formal Education and Training programme primarily targets jobseekers aged 50 and above, as well as long-term unemployed individuals of all ages. Meanwhile, the Non-Formal Education and Training programme for youth targets young jobseekers under 30. These groups of jobseekers may experience different returns to training, which in turn could contribute to the observed differences in programme effectiveness.
Effect of training on employment probability across five training programmes
Note: The analysis presents nearest-neighbour propensity score matching results which matches individuals based on several characteristics: each individual’s employment history (earnings, occupation), unemployment information (duration, unemployment benefit level and remaining duration), employability rating (based on ESS counsellor judgement), disability status, demographic characteristics (age, education, gender, nationality, marital status), and location. For every individual in the treatment group, the matching is conducted based on the values of these characteristics in the calendar month when the individual enters the programme. The control group is comprised of individuals with similar characteristics not entering active labour market programmes in that same calendar month. For paired individuals in the treatment and control groups, this calendar month is then the reference point after which outcomes are measured. Some individuals in the treatment group are dropped as they do not have a corresponding match.
Source: OECD calculations based on data from the Statistical Office of the Republic of Slovenia and the Employment Service of Slovenia.
Figure 4.5 shows the effect of different training programmes on the probability of finding employment. However, how do these programmes compare in terms of their impact on other outcomes, such as participants’ earnings and occupational mobility? In the medium and long term, all five programmes lead to significant increases in cumulative earnings, though the differences in programme effectiveness observed in their impact on employment probability are also mirrored in their effect on earnings. Indeed, the NVQ preparation programme and the local Non-Formal Education and Training programmes increase cumulative earnings by up to EUR 10 000 at 36 months after the programme’s start. In comparison, the increase is lower for the regular Non-Formal Education and Training programme and its variant for young people, at EUR 8 000 and EUR 7 000, respectively. Once again, the Institutional Training programme proves to be the least effective, raising cumulative earnings by only around EUR 3 500 at the 36‑month mark.
4.3.2. Groups of jobseekers such as the short-term unemployed disproportionally benefit from training
In addition to variation across programmes, the impact of training can also vary across different types of jobseekers. Figure 4.6 reports the employment effects of training for groups of jobseekers defined by gender, age, education, unemployment duration, and nationality. These groups may have distinct needs and, as a result, may experience varying returns from training. Although the employment effect of training programmes varies across jobseeker groups, it is important to note that the impact is positive for all groups.
Interestingly, the effect is lower for female participants compared to male participants. The difference between the two groups is substantial, amounting to 5.5 percentage points: the effect for women is 7 percentage points, while for men it is 13.5 percentage points. This difference is present across all age groups, although its size varies. Among participants aged 30‑50, the gap between women and men is only 1.5 percentage points, but it widens considerably for younger participants under 30 and older participants over 50, reaching 9 and 11 percentage points, respectively, although for participants over 50 the large confidence intervals result in imprecise estimates.
Percentage point effect on employment probability at 36 months after starting training
Note: The analysis presents nearest-neighbour propensity score matching results which matches individuals based on several characteristics: each individual’s employment history (earnings, occupation), unemployment information (duration, unemployment benefit level and remaining duration), employability rating (based on ESS counsellor judgement), disability status, demographic characteristics (age, education, gender, nationality, marital status), and location. For every individual in the treatment group, the matching is conducted based on the values of these characteristics in the calendar month when the individual enters the programme. The control group is comprised of individuals with similar characteristics not entering active labour market programmes in that same calendar month. For paired individuals in the treatment and control groups, this calendar month is then the reference point after which outcomes are measured. Some individuals in the treatment group are dropped as they do not have a corresponding match. Unemployment duration is measured either as the duration before ALMP entry (for ALMP participants) or entry into the control group (for their matched controls).
Source: OECD calculations based on data from the Statistical Office of the Republic of Slovenia and the Employment Service of Slovenia.
When considering the heterogeneity of effects across gender and age groups, the point estimates for both women and men over 50 are higher than those for their younger counterparts under 30, although the confidence intervals largely overlap. This pattern is even more pronounced for earnings (see Annex Figure 4.A.1. ), where the effects are largest among participants over 50 compared to all other age groups, suggesting that training is particularly beneficial for older jobseekers. However, similar to employment, the earnings effects are generally greater for men than for women. The difference in the impact of training programmes between women and men can be partly attributed to the fact that men are overrepresented in the more effective programmes, specifically NVQ preparation programmes and local Non-Formal Education and Training programmes, while women are more likely to participate in programmes with smaller effects, particularly in institutional training.
Individuals with secondary-level education appear to benefit more from training. The fact that some programmes are designed for individuals with secondary-level qualifications could partly drive this result. Indeed, a secondary-level diploma is a prerequisite for participation in some programmes. The impact is slightly lower for registered jobseekers with tertiary education and significantly lower for registered jobseekers with only primary-level education or less, although the confidence intervals for these two groups largely overlap. For individuals with primary education, the lower effect could be driven by the limited availability of programmes that focus on providing basic skills and are specifically targeted at individuals without secondary-level qualifications. This is also reflected in the lower uptake of this group of jobseekers (see Chapter 3). Indeed, since 2011, ESS has no longer been mandated to provide primary education for adults. Ensuring effective collaboration with institutions offering these programmes is essential to provide reskilling and upskilling opportunities for this group. In this context, Micro-credentials – smaller, more targeted and more flexible type of qualification – have the capacity to rapidly equip individuals with skills that are demanded in changing job markets (OECD, 2023[6]).
When examining effect heterogeneity across jobseekers with varying unemployment duration, the impact of training programmes appears to be greater for individuals who have been unemployed for over 12 months. A potential reason is that these programmes, particularly Non-Formal Education and Training, are targeted at the long-term unemployed.
Finally, the analysis by nationality reveals that the impact of training programmes is greater for EU nationals (including Slovene nationals) than for non-EU nationals. This difference may partly be explained by differences in language proficiency, which is often a prerequisite for participation in training programmes, as well as by the differential selection of these two groups into specific programmes.
Overall, the analysis of the employment effects of training across different groups of jobseekers indicates that some groups actively targeted by training programmes, such as older workers and long-term unemployed individuals, benefit more from training compared to other jobseekers. However, this does not hold true for lower-educated participants, who, despite being prioritised under the programme eligibility criteria, do not appear to benefit as much. Women over 50 also appear to benefit less from training than their male counterparts. Taking into account the specific labour market barriers faced by this group when designing training programmes could enhance their effectiveness.
Numerous training programmes across countries have been the subject of evaluation studies in the international literature. This section compares the findings on the effects of training programmes in Slovenia with those from these studies. The first study, conducted by Card, Kluve and Weber (2018[2]), covered 49 countries and summarised the results of over 200 impact evaluations of ALMPs. Of these, 51 evaluations provide point estimates of the employment effects of training programmes comparable to those assessed in Slovenia. The second meta‑analysis focuses on programmes funded by the EU’s European Social Fund (ESF) and includes estimates from 20 studies examining vocational training, as well as 19 evaluations of mixed interventions combining, for example, vocational training with other types of support (European Commission, 2023[7]).
Figure 4.7 reports the distribution of these estimates on the probability of employment using box plots for the short term (up to 1 year), medium term (1 to 2 years), and long term (more than 2 years). The results from the two meta‑analyses are compared with the findings of this evaluation, which are highlighted in red. Since the meta‑analysis by Card, Kluve and Weber (2018[2]) provides only estimates of the employment effects of training, and the meta‑analysis of ESF programmes (2023[7]) includes only a limited number of estimates on earnings effects, the comparison focuses exclusively on the probability of finding employment.
The comparison shows that, in international comparison, the effects of Slovenia’s training programmes are large. In most cases, the estimated effects in Slovenia rank in the top quarter of the distribution, indicating that these effects are larger than 75% of those reported in the other meta‑analyses. Only in the long term, when compared to the estimates for ESF programmes, Slovenia’s results align with the average.1 In contrast, when compared to the short-term estimates from Card, Kluve and Weber (2018[2]), Slovenia’s results approach the 95th percentile. This suggests that the negative lock-in effects of training programmes in Slovenia are relatively small compared to those observed in other countries, partly due to the relatively short duration of the programmes, which have a median length of approximately two months.
Percentage point effect on employment probability
ESF: European Social Fund.
Note: Short, medium and long-term effects respectively refer to effects up to one year, 1‑2 years, and more than two years after programme completion. For Slovenia, results refer to 12, 24 and 36 months after beginning the programme. As such, the observation periods are similar, but potentially not fully aligned. The box and whisker plots for the other studies refer to the 5th, 25th, 75th and 95th percentiles of the estimates (i.e. the black line at the bottom refers to the 5th percentile and the bottom of the blue box refers to the 25th percentile). Point estimates are included in the chart even if they are statistically insignificant. The studies presented adopt various research designs and econometric techniques – the results for Slovenia use nearest-neighbour propensity score matching (for details, see Chapter 3).
Source: Card, Kluve and Weber (2018[2]), “What Works? A Meta Analysis of Recent Active Labor Market Program Evaluations”, Journal of the European Economic Association, Vol. 16/3, pp. 894‑931, https://doi.org/10.1093/jeea/jvx028; European Commission (2023[7]), Meta‑analysis of the ESF counterfactual impact evaluations, https://data.europa.eu/doi/10.2767/580759; and OECD calculations based on administrative data from Slovenia.
In addition to these international studies, Burger et al. (2021[8]) specifically examine the effects of training programmes in Slovenia. This study uses data on participants in institutional training programmes from 2009 to 2012 and analyses their impact on outcomes similar to those explored in this report, including the probability of finding employment, cumulative days worked, and cumulative earnings. Moreover, the study employs nearest-neighbour matching – the same methodology used in this report.
The results in Burger et al. (2021[8]) indicate that, 12 months after the programme’s start, training participants are approximately 3 percentage points more likely to be employed than non-participants. This effect increases and stabilises at around 5 percentage points 24 months after the programme’s start. These findings align with the estimates presented in this report for Institutional Training programmes (Figure 4.5): 12 months after the programme’s start, training is estimated to increase the probability of employment by 4.6 percentage points, rising to 5.5 percentage points at the 24‑month mark.
Turning to other outcomes, Burger et al. (2021[8]) report that 36 months after the programme’s start, participants have earned approximately EUR 800 more than comparable non-participants. This effect is lower than the estimate presented in this report, which finds an increase of about EUR 2 300.
This chapter has shown that training programmes for unemployed individuals in Slovenia are effective in improving various outcomes. While their effectiveness varies, they benefit jobseekers across a wide range of characteristics. In an international comparison, training programmes in Slovenia appear to be more effective than most similar programmes studied in other countries, even though estimation methods and time periods vary across studies. This suggests that Slovenia’s system of training for unemployed individuals has been successful in designing programmes that equip participants with market-relevant skills, helping them finding employment. The flexibility and dynamic nature of Slovenia’s ALMP system has likely played a key role in developing training programmes that meet the needs of both jobseekers and employers.
References
[8] Burger, A. et al. (2021), “A comprehensive impact evaluation of active labour market programmes in Slovenia”, Empirical Economics, Vol. 62/6, pp. 3015-3039, https://doi.org/10.1007/s00181-021-02111-6.
[2] Card, D., J. Kluve and A. Weber (2018), “What Works? A Meta Analysis of Recent Active Labor Market Program Evaluations”, Journal of the European Economic Association, Vol. 16/3, pp. 894–931, https://doi.org/10.1093/jeea/jvx028.
[4] Employment Service of Slovenia (2024), Poklicni Barometer 2024, https://www.ess.gov.si/fileadmin/user_upload/Trg_dela/Dokumenti_TD/Poklicni_barometer/edited_poklicni_barometer_slo-2024.pdf.
[3] Employment Service of Slovenia (2018), Poklicni Barometer 2018, https://www.ess.gov.si/fileadmin/user_upload/Trg_dela/Dokumenti_TD/Poklicni_barometer/Poklicni-Barometer_2018_SLO.pdf.
[7] European Commission (2023), Meta-analysis of the ESF counterfactual impact evaluations – Final report, Publications Office of the European Union, Directorate-General for Employment, Social Affairs and Inclusion, Pompili, M., Kluve, J., Jessen, J. et al., https://data.europa.eu/doi/10.2767/580759.
[9] Haepp, T. and M. Serrano Alarcon (2024), Meta-analysis of the European Social Fund counterfactual impact evaluations: Brief update with alternative measures, European Commission, Ispra, https://publications.jrc.ec.europa.eu/repository/handle/JRC137571.
[1] OECD (2025), “Technical report: Impact Evaluation of Employment Incentives and Training for the Unemployed in Slovenia”, OECD, Paris, https://www.oecd.org/content/dam/oecd/en/about/projects/technical-reports-and-presentations-dg-reform/slovenia/Technical-Report-Evaluation-of-active-labour-market-policies-in-Slovenia.pdf.
[6] OECD (2023), “Micro-credential policy implementation in Finland, the Slovak Republic, Slovenia and Spain”, OECD Education Policy Perspectives, No. 86, OECD Publishing, Paris, https://doi.org/10.1787/c3daa488-en.
[5] Rosenbaum, P. and D. Rubin (1983), “The central role of the propensity score in observational studies for causal effects”, Biometrika, Vol. 70/1, pp. 41-55, https://doi.org/10.1093/biomet/70.1.41.
Change in cumulative earnings at 36 months after starting training
Note: The analysis presents nearest-neighbour propensity score matching results which matches individuals based on several characteristics: each individual’s employment history (earnings, occupation), unemployment information (duration, unemployment benefit level and remaining duration), employability rating (based on ESS counsellor judgement), disability status, demographic characteristics (age, education, gender, nationality, marital status), and location. For every individual in the treatment group, the matching is conducted based on the values of these characteristics in the calendar month when the individual enters the programme. The control group is comprised of individuals with similar characteristics not entering active labour market programmes in that same calendar month. For paired individuals in the treatment and control groups, this calendar month is then the reference point after which outcomes are measured. Some individuals in the treatment group are dropped as they do not have a corresponding match. Unemployment duration is measured either as the duration before ALMP entry (for ALMP participants) or entry into the control group (for their matched controls).
Source: OECD calculations based on data from the Statistical Office of the Republic of Slovenia and the Employment Service of Slovenia.
← 1. The meta‑analyses of the ESF-funded studies refers to results that includes more than one result per study. Including only one result per study, the “balanced” average effects for training are slightly lower (Haepp and Serrano Alarcon, 2024[9]).