Allowance for School Graduate Practice Performance in Slovakia: Impact Evaluation of the Intervention

Abstract

This study aims to evaluate the impact of a selected active labour market policy measure that has been applied in Slovakia—Allowance for school graduate practice performance—on the employability of young jobseekers and their sustainability in the labour market, and thus, it will also empirically contribute to the field of relevant literature. The policy targets unemployed school graduates, and it enables them to acquire professional skills and practical experience that corresponds with their level of education, work habits, and possible direct contact with potential employers. At the same time, this measure addresses a long-standing gap in the Slovakian education system, namely, the insufficient linkages between the educational process, the practices in the field, and the requirements of the labour market. Using fiscal resources to finance this policy, it provides a natural and logical platform to investigate the relevance of the outcome of this measure in the context of its proclaimed objectives. In light of this, we employed a counterfactual approach to compare the results of the participants who were affected the measure (recipients; treated group) and non-participants, as their counterparts (comparison/control group), using an instrumental variable to mitigate self-selection and selection-bias problems. Our findings show that this policy intervention has a short- or medium-term impact on the employability of unemployed school graduates and the sustainability of their careers. In addition, a positive impact on their monthly wages was observed. We also came to the conclusion that, assuming the measure is linked to other labour market policy interventions, which is aimed at employers that are willing to hire young unemployed people, it would be possible to improve the functionality and effectiveness of support for the unemployed through indirect measures.

1. Introduction

Unemployment is a serious economic, social, and political problem that negatively affects national economies, society, and unemployed individuals. Moreover, there is a significant risk that the unemployed will lose work habits, faith in their abilities, or will even build a passive attitude toward the need to be employed. This situation is even more acute in the case of school graduates who have little experience with the real labour market, are in the phase of building working habits and gaining professional skills, and have little working experience. School graduates looking for their first job after leaving school, are generally among the most vulnerable groups in the labour market [1].

Due to the objective seriousness of the unemployment problem (in general), national governments, including the Slovakian government, are attempting to keep the national unemployment rate as low as possible (towards the natural rate of unemployment) through various measures aimed at either the supply or the demand side of the labour market, by supporting unemployed people with the aim of successfully integrating them into the labour market and preventing them from being excluded from the labour market. The Ministry of Labour, Social Affairs and Family of the Slovak Republic is responsible for labour market policy conditions in Slovakia. The goals and priorities of the labour market policy are determined based on economic and social policy intentions, which are specified in the program statement of the government and other documents approved by the National Council of the Slovak Republic. The Central Office of Labour, Social Affairs and Family (hereinafter, “COLSAF”) performs the executive tasks in this area through the local labour offices and the Social Insurance Agency.

Labour market policy measures are generally divided into active and passive measures. Although passive policies and their measures are historically older (created during the last century, approximately during the 1920s) and primarily aim at providing unemployment benefits [2], active labour market policy interventions (hereinafter, “ALMP”), which also include the measures under analysis, aim to increase employment opportunities for jobseekers and improve matching between vacancies and workers [3]. The specific measures of ALMP are usually provided using four fundamental pillars—vocational training, assistance in the process of job searching, public work programs, and support to self-employed persons. Both types of policies are the subject of discussions and studies, particularly regarding the effectiveness of spending, benefits for their beneficiaries and society, and so on. According to [4,5], both labour market policies should be seen as two essential parts of a broader national social protection system.

Unemployment of school graduates (i.e., youth unemployment) is a dangerous phenomenon in all countries of the world, including Slovakia. At present, the risk of transforming this group of unemployed people into the long-term unemployed is even more pronounced due to the ongoing COVID-19 health crisis and its effects on the labour markets. For instance, authors of the studies [1,6,7] state that it can be expected that young people transitioning from education to work will find it more difficult to find employment due to increased competition for jobs and the declining availability of jobs precisely because of the continuing negative impacts of the COVID-19 crisis on the labour market worldwide [8]. In terms of the labour markets recovering from the COVID-19 crisis (which can currently be observed in several countries of the world), it is expected or it is already confirmed that vacancies are preferentially filled by people who already have work experience [9]. For example, upon examination of Slovakia, COLSAF registered 6518 unemployed school graduates in May 2019, whereas in May 2020 (the period of the first wave of the health crisis), it registered up to 8150 with a registered unemployment rate of 7.20%. In April 2021 (the end of the second wave of the COVID-19 pandemic), COLSAF registered 9597 unemployed school graduates, which was approximately 4.10% of the total number of jobseekers for a given period with a registered unemployment rate of 8%. Currently, in February 2022, the registered unemployment rate reached only 6.86% (what we can consider as being a gradual recovery of the labour market); however, the share of unemployed school graduates out of the total number of unemployed has not decreased; indeed, it was 4.60%. This is an even worse situation than in April 2021 [10,11,12].

Helping young people enter and remain in the labour market is an important part of policies, as they help to promote economic growth and obtain better living conditions. To support employment of school graduates in Slovakia, the Slovakian government operates with various measures suitable for this specific group of young unemployed people, including the Allowance for school graduate practice performance (hereinafter “Graduate practice”). According to Slovakian legislation, a school graduate is a person, a citizen of Slovakia or other countries of the EU, or a third-country national who has been granted residence in Slovakia with the status of being a long-term resident of the EU, under the age of 26, who has completed continuous full-time vocational training less than two years ago with the relevant level of education and has not had a regularly paid job since its completion (Sections 2 (2) and 8 (1) letter (a) of the Act no. 5/2004 Coll. on employment services and on amending certain laws) (hereinafter “Act on employment services”) [13]. The measure specifically targets school graduates as a group of young unemployed people. Its main aim is to enable them to first acquire professional skills and practical experience corresponding to their level of education, and support creating work habits. At the same time, it is a measure that indirectly addresses a long-standing gap in the Slovakian education system—the insufficient linkages within the educational process, practice in the field, and the labour market requirements.

The main aim of the study is to evaluate the results of the Graduate practice ex post in the context of its defined objectives, i.e., the effects of the measure on involved individuals (beneficiaries, participants; usually called “treated” individuals) in terms of their employability and sustainability in the labour market. The evaluation is made by a counterfactual approach, producing highly rigorous results by applying quantitative econometrics methods. As the measure under analysis is of a voluntary nature, the main contribution is the application of the instrumental variable method, trying to mitigate the self-selection, and selection-bias problems resulting from the voluntariness of the measure.

The originality of this study lies in the rigorous evaluation of the policy measure and quantifying the impact of the measure on its beneficiaries in comparison to other members of their cohort who did not participate in the measure (non-treated, non-involved individuals; control group). As a key to profiling this measure in the future, we consider (1) how to ensure its better functioning (from the view of participants) and (2) efficient spending of public funds (from the point of view of the state budget and EU funds, respectively), since all institutions of the public administration of Slovakia when using fiscal resources and EU funds are required to preserve the economy, efficiency, and effectiveness of their use according to the Act no. 523/2004 Coll. the Act on budgetary rules of the public administration and on amending certain laws [14].

Literature Review

Not a single commodity can be created without using a certain amount of human labour, and that is why the labour market is the market with the rarest factor of production. Human resources (known as “labour”) are thereby one of the most important factors that contribute to socioeconomic development [15]; however, in practice, the labour market does not have the features of a perfect market, and the situation is getting riskier, especially in the case of a high unemployment rate. Without doubt, unemployment not only has a significant impact on the economy, but it may also have an impact on many other aspects of life in general, such as the health and social life of individuals in society [1,2,16].

Additionally, one serious problem is the unemployment of young people, and thus, the group of the population that is in the life stage of developing work habits and acquiring professional and practical skills [17,18,19,20,21,22]. The author of [23] emphasises that the transition of young people from school to work is fraught with challenges, which mostly result in relatively low employment rates, high unemployment rates, or creating a group of young people who are neither employed nor in education or training (NEETs). The successful employment of a young person is not only beneficial for national welfare, but also directly for them. For young people who have completed their education and found a suitable job, a rapid establishment of social contacts and an improvement to their mental state has been observed.

From the “traditional” point of view—youth unemployment as the unemployment of frictional nature was analysed (e.g., by [24,25,26]). The conclusions of the authors of the studies [27,28,29] point to the link between youth unemployment and insufficient training systems within the process of education, resulting in a lack of working experience. The shortage of skilled (young) workers is also one of the greatest strategic challenges for most companies in Slovakia [30].

From a macroeconomic point of view, the impact of a crisis or economic downturn on youth unemployment was analysed (e.g., by [31,32,33,34,35]), and the effects were seen negatively and mostly as long-term. For example, with regard to the financial crisis in 2008, the results of the study [35] show that the crisis affected the unemployment of young people for five years after the onset of the crises; however, most adverse effects were found in the second and third years after the financial crisis. During the same crisis, the authors of [36] conclude that youth unemployment was not uniform, since some countries, mainly those affected by sovereign debt crises or construction bubbles (e.g., the Baltic States, Greece, Ireland, and Spain), recorded substantial increases to the rate of youth unemployment. In connection with the impact of the COVID-19 health crisis that has had a devastating impact on the economies of countries and on the Slovakian labour market, Ref. [37] states that the less time school graduates were on the labour market at the beginning of the pandemic, the greater the impact of the crisis on them. Moreover, the author in [37] emphasises that school graduates graduating during the COVID-19 crisis are in the worst position. In [38] the authors draw attention to the fact that youth unemployment is not only a problem for emerging nations or developing countries but also well-developed ones. This fact is in connection with the ongoing COVID-19 crisis and has been confirmed (e.g., in [1,39]). The results of [39] indicate that although young people between 16 and 28 years make up less than a quarter of the American labour force, from February to April 2020, they accounted for a third (approximately) of the increase in the overall unemployment rate of the USA. The authors in [22] present an opinion (based on the experience from the great recession due to the financial crisis) that a decrease in the GDP of EU countries by 1% due to COVID-19 crisis will be reflected in a decrease in youth employment of approximately 1.77%. The ILO estimates that more than one in six young people have stopped working due to the pandemic. As for the education and training opportunities for youth, the ILO states that these opportunities have been interrupted, creating long-term implications for post-pandemic recovery [40].

As for policy/program interventions targeting the young unemployed and their ex post evaluation, several researchers have evaluated these programs; although, as the authors in [41] state, these evaluations should be carried out to a much greater extent, including at the EU level. The European Commission has been putting pressure on applying evaluation methods that are empirically able to test the impact of the EU cohesion policy [42]. The authors of [36] state that the evaluation of ALMP measures within EU countries is supported by the European Commission, as the EU often participates in funding these measures. According to the recommendations of the European Commission formulated in [43], the implementation of impact assessments is mandatory for EU countries, and it can be expected that the evaluation will be emphasised further in connection with the labour market measures taken under the recovery plan for EU countries.

Most common measures aimed at young jobseekers within European countries include training courses, job search assistance and monitoring, subsidised employment, and public work programs. As for training courses, there is generally a consensus on the heterogeneous effects of this type of measure—the effectiveness varies among different types of courses and groups of participants [20,44,45,46,47]. The authors in [20] state that measures aimed at acquiring practical skills among employers in the case of young unemployed people in Slovakia are more effective than measures aimed at further education, which is often a more expensive form of support. The effects of educational programs usually do not have an immediate positive effect on increasing employment among young people (up to 1 year); however, they have a clear impact in the medium term (1 to 2 years) and mainly on the group of young jobseekers with a university degree. The authors of [48] offer almost the same conclusions. The results show that the effect of target training activities on the chances of re-employment financed by the European Social Fund in Italy is heterogeneous between types of courses and the unemployed. Based on the results, the authors suggest profiling the unemployed in order to address each individual to the training course with the highest employment impact according to their characteristics [49]. They also present other useful findings, mainly from a practical point of view, including the functional side of the processes of the measures. The study analyses selected active measures in 31 countries (a total of 107 interventions) aimed at young jobseekers. They found that investing in youth through active measures might pay off since skills training and entrepreneurship promotion interventions appeared to yield positive results on average; however, the impact of the measures were more significant in magnitude in low- or middle-income countries than in high-income countries. The results show that being unemployed or unskilled in a high-income country (where the labour demand is skill-intensive) puts youth at a distinct disadvantage in comparison to well-educated cohorts. Interventions aimed at young jobseekers in low- and middle-income countries impacted both employment and earnings outcomes. The finding is very important since it points to the potential benefits of combining supply and demand interventions to support youth in the labour market.

As far as Slovakia is concerned, one of the first studies analysing the effects of the measures of the ALMP was the study presented by [50]. The authors analysed the effects of the measures on unemployment in 1991–1994 and compared them with the situation in neighbouring Czechia. In both countries, a positive correlation between the ALMP measures and the number of vacancies was revealed. The measures also had a significant positive effect on the placement process of jobseekers. Some studies (e.g., [51,52,53]), also evaluated the measure under our analysis. The authors found out that participants in the Graduate practice achieved better outcomes than unemployed individuals who did not participate in the measure. In [18,51] the authors emphasise inter alia the institutional aspect of implementing this measure, since implementation plays an important role in determining its net effects. They found out that the implementation of the measure among responsible institutions (labour offices) was not uniform, which was reflected in the great variability of the evaluated effects. Regarding the evaluation of the impact of the measure on public finance, [54] states that the participation in the measure saved on average 20% of the government spending spent on each young person that was officially registered as unemployed in 2014, and more than 70% in 2015. Despite this study and others, we must state that, in general, there are not many comprehensive impact assessments of ALMP measures in Slovakia, let alone the use of counterfactual impact evaluation.

This is also a reason for carrying out this empirical study. The obtained results will serve as the foundation for evaluating the functionality of the measure under consideration in the context of its defined objectives based on historical data, as well as the foundation for reassessing its procedural functioning in the future. We expect this measure to play a leading role in supporting the employment of school graduates in the current crisis and the post-COVID-19 period given the conditions of the Slovakian economy and the Slovakian labour market. From a theoretical point of view, the benefit of the study lies in the connection between the theoretical mathematical–statistical basis of ex post counterfactual analysis with economic practice.

2. Materials and Methods

2.1. Aim of the Study

The main aim of the study is to evaluate the effects of the measure complexly—“Allowance for school graduate practice performance” (the legal definition of the measure)—in the context of the objectives of the measure (i.e., the effects of the measure on its participants (young jobseekers, school graduates)) in terms of their employability and sustainability in the labour market and their earning outcomes. The secondary aim is to analyse the state of the labour market and the area of the ALMP measures through appropriately selected indicators and the legal conditions of the Graduate practice in Slovakia. Knowledge of the situation is a prerequisite for an adequate evaluation of the measure under analysis.

We expect this measure to play a main role in supporting the employment of school graduates in the current COVID-19 situation and the post-COVID-19 period given the conditions of the Slovakian economy and the Slovakian labour market. In terms of methodology, another aim of the study also seeks to point out the relevance of the use of counterfactual evaluation in Slovak economic practice.

2.2. Method of the Evaluation

The evaluation of the Graduate practice is an ex post analysis. The analysis is guided by a counterfactual approach, bringing highly rigorous results through the application of quantitative econometrics methods. Furthermore, as the measure under analysis is of a voluntary nature, the evaluation is based on the application of the instrumental variable method, trying to mitigate the self-selection and selection-bias problems resulting from the voluntariness of the measure.

According to [55], the impact evaluation of any measure is an analysis of cause–effect relationships (causality), and the evaluation aims to answer the question of whether the participation in the measure has the desired effect on the outcome variables; therefore, the evaluation of the measure analyses to what extent the changes in the treated participant’s outcome variable can be attributed to their participation in the measure. Put simply, the impact of the measure can be obtained as the difference between the value of the outcome variable after participating in the measure and the value that this variable would have obtained in the absence of the measure (net difference) [56]. In the literature, this situation is most often indicated as “what would have happened in the absence of the intervention” and indicates the counterfactual results (values of the outcome variables) of treated individuals in case they would not have participated in the measure.

Let us mark the state associated with participation in the intervention as “1” and, conversely, the state associated with non-participation in the intervention with the value “0”. Intervention participation is indicated by the variable $D_i$. Thus, if an individual $i$ is included in the program, then his value of the variable is $D_i=1$ and vice versa; if he is not the recipient of the intervention, his value is $D_i=0$. The aim of the evaluation is to compare the values of outcome variables in cases where he participated in the program (i.e., if he completed the intervention ($D_i=1$)), with those cases where he did not complete it ($D_i=0$) [57,58].

Moreover, let $Y$ be the outcome variable by which we quantify the effects of the intervention. Then, to estimate the effect of the treatment, we need to calculate the difference in the means of the outcome variable

$$E\left(Y|X, D=1\right)-E\left(Y|X, D=0\right)$$

(1)

Then, let $Y_{ti0}$ be the value of outcome variable for individual $i$ in period $t$ after the intervention for a non-participant, and $Y_{ti1}$, for a participant of the intervention. Such a label makes it possible to directly formulate the causal effect of the program on the individual’s outcomes as the difference of its values $Y_{ti1}-Y_{ti0}$, which expresses the effect of the treatment at the individual level. For simplicity, we will label the outcome variable for the untamed individual as $Y_0$ for the non-participant and $Y_1$ for the intervention participant. The impact of the intervention will simply be $Y_{i1}-Y_{i0}$ at the individual level, or just $Y_1-Y_0$ at the population level [59].

Unfortunately, and this is at the heart of the problem of carrying out program evaluations, the effectiveness of the program can be estimated with considerable reliability at best, but it cannot be measured with certainty, as we can never realistically simultaneously observe the individual’s potential outcomes. $Y_{i1}$ if he was a participant in an intervention, and the outcome of the same individual is $Y_{i0}$ if he was not a participant in an intervention. Each individual can either participate in the program or not; thus, only one of these two values of the outcome variables can be observed for every individual [57]. In the literature, this fact is called the “fundamental problem of evaluation” or, more generally, the “fundamental problem of causal inference” [60].

For those individuals who participate in the program, we can track their outcomes $Y_1$, whereas the outcome $Y_0$ is a counterfactual outcome for them, which is not observable. On the other hand, in the non-intervention population, it is possible to observe only the values of their outcome variable $Y_0$, whereas the counterfactual outcomes $Y_1$ are not available for them. Participation in the program $D_i$, decides which of the outcomes, $Y_{i1}$ or $Y_{i0}$ will be observed for a given individual.

The average effect of the program (i.e., the total average treatment effect (ATE)) can be expressed as the average value of the differences between treated and non-treated groups:

$$\begin{array}{ll}ATE& =E\left(Y_1-Y_0\right)\\ & =\left\{\pi E\left(Y_1| D=1\right)+\left(1-\pi \right)E\left(Y_1| D=0\right)\right\}\\ & -\left\{\pi E\left(Y_0| D=1\right)+\left(1-\pi \right)E\left(Y_0| D=0\right)\right\},\end{array}$$

(2)

where $\pi =P\left(D=1\right)$ represents the probability of an intervention, and thus, the additional probability of no intervention is $\left(1-\pi \right)=P\left(D=0\right)$. This term expresses the expected impact of the intervention on the outcomes of all individuals in the population (treated and non-treated) [59]. For the validity of (2), it is necessary to assume that the so-called Conditional independence assumption (CIA) holds that there are no significant differences in unobservable characteristics between participants and non-participants that would affect the outcome variable $Y$. CIA is also called “selection on observables” and can be written

$$D\coprod \left(Y_0,Y_1\right)|X$$

(3)

for the effect on the population ATE [61].

The average effect of treatment on the treated (ATT) is given by

$$ATT=E\left(Y_1-Y_0| D=1\right)=E\left(Y_1| D=1\right)-E\left(Y_0| D=1\right)$$

(4)

with the CIA assumption for the effect on the treated

$$D\coprod Y_0|X.$$

(5)

When a CIA holds, instead of the mean difference (1), it is possible to estimate the effect on the population by (2) and the effect on the treated by (4) [61].

Similarly, as any other population parameter, the mean value of $E\left(Y_1│D=1\right)$ is unknown and must therefore be estimated using a sample as tan approximation of the true value of a population parameter [57]. On the other hand, although this parameter can, in principle, be estimated from the available data on program participants, it is not even hypothetically possible to estimate the mean value $E \left(Y_0│D=1\right)$. This is because the value of $Y_0$ for program participants cannot be measured because program participants “produce” only the values of $Y_1$ [60]. This means that only the expected outcome $E\left(Y_1│D=1\right)$ can be estimated from the observable data on the intervened individuals, but not $E\left(Y_0│D=1\right)$ [57]. It should therefore be emphasized that in the previous formulas (2) and (4) for ATE and ATT, the values of $E\left(Y_1│D=0\right)$ and $E \left(Y_0│D=1\right)$, are unobservable in reality [59].

Here lies the basic problem of the evaluation: the problem of finding a suitable identification assumption to obtain the value $E \left(Y_0│D=1\right)$ in the definition of ATT (4) is that it can be replaced by an estimate that can be obtained from observable data. The mean value $E\left(Y_0│D=1\right)$ is thus counterfactual and expresses what would have happened to the participants on average if they had not participated in the program. This problem of non-measurability of the counterfactual situation cannot be solved by collecting more data or measuring suitable variables. It can only be solved by finding a suitable comparison group [57]; therefore, the control group of not involved individuals (counterparts, called “non-treated”) compensates for the missing data. For the individuals in the control group, we have data on the outcome variables in case they do not participate in the measure. Provided this control group is sufficiently similar to the group of treated individuals, this will create a credible counterfactual situation. In this situation, participants of the measure and their counterparts are very similar in their characteristics, so any difference in their outcomes can be ascribed to participation in the measure.

The creation of the most accurate counterfactual situation is realised by matching individuals based on their pre-measure characteristics. There are various matching methods, the most accurate of which is the exact matching method, but this method raises the problem of dimensionality and a low degree of exact match of values of the variables [62]. The most widely used method is propensity score matching, which solves this dimensionality problem by matching based on the propensity score instead of exact values. The score represents the probability of the individual’s participation in the measure, estimated by the logistic regression model. Every treated and non-treated individual is assigned an estimated propensity score based on their pre-interventional characteristics. This score is then used for matching treated participants of non-treated individuals instead of exact values of their characteristics, thus solving the problem of the high dimensionality of matching. The basic identification assumption for the creation of samples in the matching methods according to [61] is

$$E\left(Y_0|D=1, x\right)=E\left(Y_0|D=0, x\right),$$

(6)

thus, unobservable (or missing) values of $E \left(Y_0│D=1, x\right)$ can be replaced by observable values of $E \left(Y_0│D=0, x\right).$

However, the so-called selection bias problem should also be considered in the impact evaluation of the measures. A control group is used as a substitute for a counterfactual situation of the measure participants’ results in selection bias, resulting from an individual’s decision to participate in the measure (self-selection problem), influenced by many factors. At the same time, the impact of unobservable characteristics of individuals must be considered, such as talent, motivation, skill, and so on. All these characteristics impact the outcomes of the individual, and because of their effect, the individual’s better performance may be incorrectly attributed to the measure. This may result in bias concerning estimates of the effects of the measures, which are then overestimated.

Moreover, if participation in the intervention program is affected not only by the observable characteristics of individuals but also by their unobservable characteristics, the CIA conditions (3) and (5) do not hold, and the ATE and ATT estimates given by (2) and (4) will be skewed (overvalued or undervalued) [44,61]. The variable $D_i$, indicating the individual’s participation in the intervention, becomes endogenous, correlated with the error term in the equation, thus preventing ordinary least squares (OLS) from producing consistent estimates of regression parameters, including ATE (2) and ATT (4).

More specifically, in the case of the measure under analysis—Graduate practice—the participation in the measure is voluntary, hence it results from the individual decision of each young jobseeker; therefore, it is true that the effect of the measure on the meaning of sustainability and employability of young jobseekers can be influenced by the self-selection of participation of individuals in the measure (i.e., by their motivation—that is, a variable that evaluators cannot measure and quantify); however, this motivation may also affect their results in the post-intervention period. A more motivated jobseeker can also be more motivated to find a job and keep it, and vice versa. As [61] writes, econometric methods based on the CIA are no longer appropriate for estimating the actual program effect on the outcome variables.

One solution to this problem is using the “exogenous variation” in $D$ caused by the variable $\delta$, where $\delta$ does not affect $Y$ directly, so “$\delta \to D\to Y$” holds; $\delta$ affects $Y$ only indirectly through $D$ [63]. Consider a linear model for the outcome variable

$$Y={\beta }_0+{\beta }_1{X}_1+{\beta }_2{X}_2+\dots +{\beta }_K{X}_K+{\beta }_{D}D+U$$

(7)

where $E\left(U\right)=0$ and $cov\left(X_j,U\right)=0$ for $j=1,2,\dots ,K$. The explanatory variables $X_1,\dots , X_K$ in (7) are exogenous, but $cov\left(D,U\right)\ne 0$, i.e., $D$ is a potentially endogenous variable in Equation (7) due to the influence of unobservable characteristics (let us say, motivation) in $\epsilon$ and $U$, but $cov\left(Z,U\right)=0$ [61,62,63]. As listed in [62], if $cov\left(D,U\right)\ne 0$, OLS estimation of Equation (7), it generally results in inconsistent estimators of all parameters ${\beta }_j$. Without more information, we cannot consistently estimate any of the parameters in (7). A general solution for this problem of an endogenous explanatory variable is the method of instrumental variables (IV).

For practical application of the IV method, we need an observable variable $Z$, not in Equation (7), that meets two conditions. First, $Z$ must be uncorrelated with $U$, that is

$$cov\left(Z,U\right)=0$$

(8)

that is, $Z$ is exogenous in (7) similarly as other variables $X_1,\dots , X_K$. The second condition involves the relationship between the instrument $Z$ and the endogenous variable $D$ given by the linear projection of $D$ onto all the exogenous variables, given by

$$D={\alpha }_0+{\alpha }_1{X}_1+{\alpha }_2{X}_2+\dots +{\alpha }_K{X}_K+{\alpha }_{Z}Z+\epsilon$$

(9)

with $E\left(\epsilon \right)=0$, where $\epsilon$ is uncorrelated with $X_1,\dots , X_K$ and $Z$. The key assumption on the linear projection (9) is that the coefficient on $Z$ is non-zero [60]:

$${\alpha }_Z\ne 0.$$

(10)

When $Z$ satisfies both conditions (8) and (10), it is an IV candidate for $D$. In (8), $Z$ is an instrumental variable for $D$ (or simply an instrument) and IV estimator can be applied. We assume that this instrument does not directly affect the outcome variable [61]. This condition is called the “exclusion restriction” and prevents any direct effect of the instrumental variable $Z$ on the outcome variable $Y$, defining only the indirect effect via $D$ [55].

This method of IV tries to mitigate the influence of the self-selection effect and reduce the selection bias problem by introducing an IV into the regression model. The ATE (2) and ATT (4) can be quantified if it is possible to identify and measure a variable that affects the selection of individuals into the measure, but does not affect the values of the outcome variable or other unobserved variables in the impact period [55].

In previous studies, many different variables have been used as appropriate instrumental variables. These are usually context variables, as characteristics of the environment in which jobseekers live, work, and so on. In this study, we apply one instrumental variable that was preferred in other studies (e.g., [64,65,66]), the distance of the permanent residence of the jobseeker from the provider of the Graduate practice (place of practice). We think that a possible benefit that may encourage young people to participate in Graduate practice, is a short distance to the practice provider. Conversely, a long distance from the office could discourage young people from taking part in the intervention. Moreover, we assume that this distance could affect an individual’s decision to participate in the measure, but does not affect the individual’s outcomes in employment and keeping a (real) job position after participating in the Graduate practice in the impact period after the termination of the measure. In our study, the impact period is 24 months after the successful termination.

For the instrumental variable used in this study, i.e., the distance from the permanent residence from practice provider, the correlation between the instrumental variable and participation in the intervention is $corr\left(Z,D\right)=0.203$ with p-value of the test of its significance < 0.05. This means a relatively weak but statistically significant correlation. Moreover, the correlation coefficient between the instrument and both outcome variables shows a very weak dependence; $corr\left(Z,wage\right)=0.04$ and $corr\left(Z,employment\right)=0.08$. Thus, we consider this instrument to be relevant for this analysis, as it influences an individual’s decision to participate in the program, but is not directly related to the probability of finding employment and to the wages of the participants in the future. This variable is also not associated with the individual’s motivation. There is no obvious reason to believe that the distance between the jobseeker’s permanent residence and the practice provider is associated with the individual’s motivation or probability of finding a job.

Practically, by the regression (9) of $D$ on $\left(X_1,\dots ,X_K,Z\right)$ we get the estimates $\left({\widehat{\alpha }}_0,{\widehat{\alpha }}_1,\dots ,{\widehat{\alpha }}_K,{\widehat{\alpha }}_Z\right)$ and then by the regression (7) of $Y$ on $\left(X_1,\dots ,X_K,\widehat{D}\right)$, where $\widehat{D}={\widehat{\alpha }}_0+{\widehat{\alpha }}_1{X}_1+\dots +{\widehat{\alpha }}_K{X}_K+{\widehat{\alpha }}_{\delta }Z$ we can estimate the parameter ${\beta }_D$ consistently. This is the principle of the two-stage regression model. The coefficient of the variable ${\widehat{D}}_i$ in this second regression represents the ATE estimate (2) [63].

Another way of finding ${\beta }_D$ is by the regression of $D$ on $\left(1,\delta \right)$ to get the estimate ${\widehat{\alpha }}_{\delta }$ and then $Y$ on $\left(1,\delta \right)$ to get ${\widehat{\gamma }}_D$. Finally, ${\beta }_D={\widehat{\gamma }}_D/{\widehat{\alpha }}_{\delta }$. Here, ${\gamma }_D={\alpha }_{\delta }{\beta }_D$ is the slope of $\delta$, which is the product of two effects: ${\alpha }_{\delta }$ for $\delta$ on $D$ and ${\beta }_D$ for $D$ on $Y$. These two found that the instrumental variable estimates are numerically the same [61,64].

In [60] the author further lists that the abovementioned two-stage approach has an important shortcoming. The consistency of the estimates obtained depends very much on the correctly specified model in the first step of the procedure. If this assumption is not met, deviations are transferred from the first model to the second model, disrupting the resulting ATE estimate. To solve this problem, [61] proposes using a three-step model as a combination of a probit model and a two-stage least squares method.

Therefore, in this study, we apply the three-stage least-square instrumental variable method to estimate the effect of the Graduate practice. The estimated effect of the measure is then as follows:

• logit model for $D$ on all $X$ and $Z$ from which we predict probabilities for the participation of individuals in the Graduate practice ${\widehat{p}}_1\left(D_i\right)$;
• ordinary least squares regression model of $D$ on all $X$ and ${\widehat{p}}_1\left(D_i\right)$; the predicted values from this model are ${\widehat{p}}_2\left(D_i\right)$;
• second ordinary least squares regression model of $Y$ on all $X$ and ${\widehat{p}}_2\left(D_i\right)$; the regression coefficient of the variable ${\widehat{p}}_2\left(D_i\right)$ in this third regression represents the estimate of the average treatment effect.

According to [62], the preference of this approach ensures that the resulting standard deviations of estimates do not need to be adjusted to use the predicted values from the first two models as the consistency assumption required in such cases is met.

2.3. Data and Period of the Analysis

The analysis relies on administrative data concerning young jobseekers that meet the legal conditions of the Graduate practice as defined in Section 51 of the Act on employment services [13]. The data comes from the database of jobseekers governed by the COLSAF and the database of Social Insurance Agency (SIA) policyholders. All the data refers to Slovakia and covers the years 2015–2018 (i.e., we work with the data on those treated school graduates, who started the intervention on 1 October 2015 at the earliest, and at the same time, the impact period under review expired on 31 December 2018). Unfortunately, at the time of preparing the study, we did not have access to newer data in terms of precise information on individual participants of the measure, and other young jobseekers as their counterparts, because in Slovakia, these data are usually collected at the end of each programming period and require considerable time for checking and “cleaning”.

Both databases were used to create a treated group of school graduates (participants of the Graduate practice) and a control group of non-treated young jobseekers who were eligible but decided not to participate. All sample eligibility checks, as well as logical checks (such as date sequence, exclusion of individuals outside the period of evaluation, elimination of duplicate registrations in the database of jobseekers, elimination due to an individual’s departure abroad, or due to death, etc.), were performed manually. Thus, after all controls and restrictions, we had a database of 96,860 young jobseekers (12,953 participants in the Graduate practice, 83,907 counterparts). As is obvious, the control group is six times more numerous.

Each school graduate who was actively looking for a job in the analysed period or participating in the Graduate practice was characterised by the variables presented in

Table 1. Variables used in the evaluation. Source: authors.

Variable Name	Description
ID	Anonymized identification code of each individual
Intervention	Attending the Graduate practice
School code	Last finished school according to the list of study and vocational fields
Age at registration	Age of each individual on the date of entry into the database of jobseekers
Previous evidence	Sum of days of all previous registrations of the individual in the database of jobseekers
Profession before registration	Last profession before the individual’s registration according to the international classification of occupations ISCO-08
Start of registration	Date of entry into the database of jobseekers
End of registration	Date of termination from the database of jobseekers
Evidence length	Number of days of the last evidence in the database of jobseekers
Start of intervention	Date of start of the participation in the Graduate practice (if)
End of intervention	Date of termination of the participation in the Graduate practice (if)
Duration of intervention	Number of days of participation in the Graduate practice (if)
Grant	Sum of allowances paid during the Graduate practice
Age shifted	Age of the individual shifted from the start of the registration in the database of jobseekers to the date of the start of the participation in the Graduate practice (for the treated participants) or to 1st October 2015 (for non-treated jobseekers)
Gender	Gender of each individual
Marital status	Marital status at the time of the individual’s entry in the database of jobseekers (single, married, divorced, widow/widowed, or not specified)
Level of education	Highest degree of education at the time of the individual´s entry into the database of jobseekers (10 levels of education from unfinished primary education to the highest level of higher education, plus the category not specified)
Region of residence	Region of the permanent residence at the time of the individual´s entry in the database of jobseekers (Bratislava, Trnava, Trencin, Nitra, Zilina, Banska Bystrica, Presov, Kosice, plus category not specified)
Disadvantages	Disadvantage of jobseekers under Section 8 of Act on employment services [13] (other than the status “school graduate”, i.e., a long-term unemployed individual who has completed education lower than the secondary vocational education, individual living as a single adult with one or more people dependent on their care or caring for at least one child before the end of compulsory schooling, disabled person, plus category no disadvantage)
Last school	Type of the last (attended) school of the individual at the time of the entry into the database of jobseekers
Driving license group 1	Ownership of the driving license for motorcycles (categories AM, A1 and A; binary type of variable—0 if the individual does not own the driving license of a given category, 1 if the individual owns the driving license of given category)
Driving license group 2	Ownership of the driving license for cars and small lorries (categories B1, B and B + E; binary type of variable—0 if the individual does not own the driving license of a given category, 1 if the individual owns the driving license of given category)
Driving license group 3	Ownership of the driving license for trucks (categories C1, C1 + E, C and C + E; binary type of variable—0 if the individual does not own the driving license of a given category, 1 if the individual owns the driving license of given category)
Driving license group 4	Ownership of the driving license for lorries and buses (categories D1, D1 + E, D and D + E; binary type of variable—0 if the individual does not own the driving license of a given category, 1 if the individual owns the driving license of given category)
Driving license group 5	Ownership of the driving license for tractors (category T; binary type of variable—0 if the individual does not own the driving license of a given category, 1 if the individual owns the driving license of given category)

.

After controlling input data on young jobseekers, we used a sample of 12,953 participants in the Graduate practice and 83,907 non-treated individuals. As the sample in the analysis period was dominated by young people under the age of 26 (as determined by the eligibility criterion), their marital status was predominantly single (more than 99% in both groups). In the sample, complete secondary vocation education prevailed (45.7% in the treated group, 54% in the control group). 62% of jobseekers in the control group and 35.3% of participants in the Graduate practice were not disadvantaged jobseekers as defined in Section 8 (1), except for the letter (a) of the Act on employment services [13]. If they stated that they were disadvantaged jobseekers, they mostly mentioned the status “long-term unemployed” (24.5% in the control group, 45.4% in the treatment group).

Regarding gender, women predominated in the group of participants (64.8%), while in the group of non-participants, men predominated (58.6%). In both samples, most individuals came from the eastern parts of Slovakia, namely from the Presov region (18.9% in the control group, 22.8% in the treatment group) and the Kosice region (16.3% in the control group, 13.9% in the treatment group). Approximately every third treated (but also non-treated) young jobseeker owned a license for motorcycles, but also cars and small trucks. Overall, the average age of the participants was 22.09 years and 23.19 in the case of their counterparts.

Through the quantitative research, we found out that the average duration of the Graduate practice in the observed period was almost five months, considering a month with a time fund of 22 working days and four hours per working day (i.e., 338 h). The average hourly value of the allowance in the observed period was EUR 1.468 for the given time fund. The average length of registration in the database of jobseekers before the school graduate became a participant in the measure was 8.17 months.

According to the selected results of the qualitative research (a questionnaire survey on a sample of 280 treated individuals who participated in the Graduate practice), we state that almost 80% of participants would undergo the practice again. Furthermore, the measure participants almost unequivocally considered the procedural acts preceding the performance of the Graduate practice as the most negative aspect of the measure (the procedural side of the administrative support of the measure).

All the characteristics of the samples used in this study are in

Table A1. Descriptive characteristics of the sample. Source: authors.

Variable	Total	Non-Treated	Treated
ageatregistration	22.36	22.48	21.62
previousevidence	460.64	493.91	245.12
evidencelength	311.28	287.92	462.59
ageshifted	23.04	23.19	22.09
Sharesin%
men	0.55	0.59	0.35
maritalstatus
married	0.05	0.05	0.06
divorced	0.20	0.20	0.20
NA	0.61	0.70	0.00
levelofeducation
primary	8.48	9.70	0.60
lowersecondary	0.65	0.70	0.30
secondaryvocational	15.76	17.00	7.70
completesecondaryvocational	46.81	45.70	54.00
uppersecondaryvocational	7.13	6.70	9.90
highervocational	0.20	0.20	0.20
university1st	1.75	1.70	2.10
university2nd	5.27	2.80	21.30
NA	13.08	14.50	3.90
disadvantages
school-leaver	14.19	13.40	19.30
longtimeunemployed	27.29	24.50	45.40
loweducation	<0.01	<0.01	<0.01
noregularpaidjob	0.09	0.10	0.00
childcare	<0.01	<0.01	<0.01
drivinglicensegroup1	32.10	37.00	0.34
drivinglicensegroup2	0.41	0.37	0.67
drivinglicensegroup3	0.02	0.02	0.01
drivinglicensegroup4	0.20	0.20	0.20
drivinglicensegroup5	0.03	0.03	0.02
regionofpermanentresidence
Bratislava	5.90	6.30	3.30
Trnava	9.13	9.10	9.30
Trencin	10.13	10.30	9.00
Nitra	12.33	12.30	12.50
Zilina	13.74	13.50	15.30
BanskaBystrica	13.27	13.20	13.70
Presov	19.42	18.90	22.80
Kosice	15.98	16.30	13.90

in the Appendix A. We present the average values and the frequencies of the whole sample included in the study, as well as the treated and non-treated groups.

The impact period of each participant was considered 24 months and started immediately after the termination of participation in the Graduate practice. During this period, the employment of participants and their counterparts was monitored, and all needed information was taken from the database of the SIA.

Based on the information on the start of the registration, termination of the registration, type of the registration, and officially registered monthly assessment base (earnings outcomes), we have been able to determine the employment of each individual in the labour market and the amount of his/her monthly wage; therefore, the outcome variables used as the variables identifying the employability and sustainability of young jobseekers and their earnings outcomes in the labour market in the case of our study logically are:

• wage—the average monthly wage of individuals (in EUR) over a 24-month impact period;
• employment—the period of registration of individuals in the SIA as a full-time employed person or self-employed person (in days).

3. Results

3.1. Basic Characteristics of the Labour Market and Labour Market Policy Measures in Slovakia

Figure 1 shows the development of the amount of funds spent in Slovakia on labour market policy measures (grey dotted line) in 2004–2020 and ALMP instruments (black line). It also shows the development of the registered unemployment rate (grey column) and the youth unemployment rate as a percentage of the youth labour force in Slovakia (framed column).

It is obvious that until 2019, the unemployment rate in Slovakia was mostly dropping, except for the years 2008–2012 when the Slovakian labour market was hit by the consequences of the global financial crisis; however, the opposite trend is currently being observed due to the COVID-19 crisis. Currently (June 2021), the unemployment rate in Slovakia has reached 7.8% [67]. As far as youth unemployment is concerned, its rate for each depicted year was more than two times higher and copied the development of the overall unemployment rate. According to the data from [67], the youth unemployment rate in Slovakia reached 23.4% in May 2021, which is the worst monthly value in the last 5 years. With regard to long-term unemployment, we still consider it a challenge for young people, since this rate is still above the EU average and is even accentuated by regional disparities in Slovakia. The worst situation is in areas that generally suffer from unemployment (the east and south-east parts of Slovakia, mainly municipalities located in the Banska Bystrica region, Presov region, and Kosice region) [10]. Finding a job for young people from these parts usually means migrating for work to more economically developed parts of Slovakia or abroad.

According to the data provided by the OECD [68] and COLSAF [69], in 2019, Slovakia used approximately EUR 470 million for labour market policy tools. Of this amount, a total of EUR 185.5 million was spent on ALMP measures (Categories 2–7); for the measure under analysis EUR 1.98 million was spent. Approximately EUR 156 million came from the European Social Fund from the Operation program “Human resources 2014–2020”. Expenditures on ALMP in 2020 increased by 16.7%; however, the funds used directly for the analysed measure reached the level of EUR 1.56 million (i.e., they dropped by 21.2% as did the number of participants of the measure).

For the purposes of international comparison, the expenditures on labour market policy measures accounted for 0.55% of the Slovakian GDP, which ranks Slovakia among the EU Member States with relatively low expenditures. The expenditures on ALMP measures accounted for 0.198% of the Slovak GDP in 2019, and they increased by 0.4 percentage points to 0.238% of the Slovak GDP in 2020. Overall, and for all the years observed, it is clearly true that the expenditures on ALMP measures lag behind the spending on passive labour market support [13,69].

From the analysis of the available data, it generally follows that the Graduate practice is one of the most widely used ALMP measures in Slovakia. Although the number of its participants is decreasing (in comparison to its “best years”), this intervention is still one of the most interesting ways of acquiring job skills and contacts with “real working life” in the labour market in case of school graduates. The greatest interest in this measure was recorded in the period of the financial crisis and the post-crisis period (2009–2012), and the number of participants ranged from 12,000 to 22,000. In the last six years (2015–2020), including the years under analysis, the number of participants dropped (it was 7398; 5689; 5439; 4154; 3575; and 2402 participants, respectively). The funds fixed for the measure were continuously dropping, logically, regarding the decreasing number of participants, but the legal conditions of the measure remained unchanged. They reached the volume, namely of EUR 4.48 million; EUR 3.45 million; EUR 2.91 million; EUR 2.22 million; EUR 1.98 million and EUR 1.56 million. The average expenditure per participant in these years ranged from EUR 534.4 (in 2018) to EUR 649.5 (in 2020). The decline of treated school graduates is related mainly due to the generally improving situation in the labour market and the existence of other measures aimed at young jobseekers (we are referring to the period until the outbreak of the COVID-19 crisis); however, the COVID-19 period and the expected post-COVID-19 period in particular, in our opinion, will bring new challenges for this measure.

3.2. Legal Analysis of the Graduate Practice

In Slovakia, a jobseeker is generally understood to be a citizen who can work, wants to work, is looking for a job, and is kept in the Register of jobseekers. If we mention jobseekers in connection with the measure, which is the subject of the analysis, we speak of so-called unemployed school graduates.

According to Sections 2 (2) and 8 (1), letter (a) of the Act on employment services [13], a school graduate is a person, a citizen of Slovakia, other countries of the EU or a third-country national who has been granted residence in Slovakia with the status of a long-term resident of the EU, who is under the age of 26, who has completed continuous full-time vocational training less than two years ago with the relevant level of education and has not had a regularly paid job since its completion. At the same time, the jobseeker defined in this way falls into the legally defined category of so-called disadvantaged jobseekers.

The Graduate practice is primarily governed by the already mentioned Act on employment services [70]; the conditions of the measure are directly defined in Section 51. The measure is of a voluntary nature and is provided by the COLSAF through 46 local labour offices, following the same section. The general aim of the measure is to reduce the unemployment rate among young jobseekers under the age of 26. The specific aim of the measure is to support the competitive ability of school graduates, and thus, their employability and long-term sustainability in the labour market through the acquisition of working skills, professional experience, and working habits. It is therefore a means of building their working careers.

The Graduate practice was put into practice in 2004. Since then, it has undergone several adjustments to reflect the needs of the labour market and its changing conditions. The major legal changes in the measure are summarized in

Table 2. Legal changes of the conditions of the Graduate practice. Source: authors based on the amendments to the Act on employment services.

Period of Validity	Conditions of the Intervention
	Age of the Jobseeker/Conditions of the Evidence	Duration of the Intervention	Monthly Allowance
April 2004–April 2008	<25 years	up to 6 months, possible repetition 1 year after the previous Graduate practice	56,43 EUR (fixed amount)
May 2008–December2010
January 2011–June 2011	<25 years	3–6 months, 20 h a week, repetition is not possible	based on the statutory amount of the subsistence minimum
July 2011–April 2013	<26 years
	registered in the Register of jobseekers for at least three months
from May 2013	<26 years
	registered in the Register of jobseekers for at least one month

.

The period analysed in this study corresponds to the current wording of the Act on employment services [13], regulating the conditions of this measure. The measure is intended for a school graduate who meets the following conditions:

• is a person not older than 26 years;
• has left the continuous preparation for a job with relevant education stage in the full-time study less than two years ago;
• at the same time, did not have a regular paid job for longer than six consecutive months before the official registration in the Register of jobseekers;
• is registered in the Register of jobseekers for at least one month.

The funding of the measure is subject to the signing of an agreement between the labour office and the participant (school graduate) and, at the same time, between the labour office and the employer (provider of the Graduate practice). The duration of the measure is from 3 months to 6 months, for 20 h weekly, without any possibility of repetition and extension of the participation in the measure for the same school graduate. During the participation in the measure, the participant remains in the Register of jobseekers and cannot do any other job.

During the Graduate practice, the relevant labour office pays the participant a monthly allowance. From 2011, the allowance is fixed at 65% of the statutory amount of the subsistence minimum provided to one adult person in a given year. As the subsistence minimum is changing according to Act no. 601/2003 Coll. on subsistence minimum and on amending certain laws [70], the allowance given to the Graduate practice is changing according to the amendments to this Act. The amount of these allowances are in

Table 3. Monthly allowance of Graduate practice intervention. Source: authors based on the Act no. 601/2003 Coll. on subsistence minimum and the Act. on employment services.

Period	2013 to 2nd Quarter of 2017	3rd Quarter 2017 to 2nd Quarter 2018	3rd Quarter 2018 to 2nd Quarter 2019	3rd Quarter 2019 to 2nd Quarter 2020	3rd Quarter 2020 to 2nd Quarter 2021	3rd Quarter 2021 to 2nd Quarter 2022
Subsistence minimum (1 adult person) in EUR	198.09	199.48	205.07	210.20	214.83	2018.06
Allowance in EUR	128.75	129.66	133.30	136.63	139.64	141.74

.

Among other things, after working for two calendar months, each participant in the Graduate practice is entitled to time off in the range of 10 days. It is also obvious that by incorporating this right, this measure aims to approximate as closely as possible the employment duties and claims of participants in the Graduate practice to a “normal” employment relationship and a “normal” job position.

3.3. Results of the Counterfactual Evaluation of the Measure

According to the counterfactual analysis and its results, to alleviate the already mentioned problem of the existence of unobserved non-measurable factors and the self-selection of participants into the measure, resulting from the voluntary basis of the measure, we have used the method of instrumental variables and three-step approach during its process.

As we have already mentioned, we consider the distance of the jobseeker’s permanent residence to the provider of the Graduate practice as a suitable instrumental variable that may influence the decision to participate in the measure. In the use of the instrumental variable (as suggested in already mentioned studies), we have based on the parallel between benefits and costs that the measure generates for the potential participant and, at the same time, the fact that the Graduate practice is temporary and the fact that despite the effort to approximate it, it is not a real job with respect to wages and social conditions resulting from “classic” job positions and employment relationships. As we have mentioned before, we assume that this distance could affect an individual’s decision to participate in the measure but does not affect his/her outcomes in employment and work career.

The coefficient of the variable $ste{p}_2$ that represents the predicted values of participation in the measure from the second step of the procedure quantifies the effect of the measure by applying the three-step least-square approach. Regression models for both outcome variables are in the Appendix A (

Table A2. Regression model for the outcome variable wage, step 3. Source: Authors.

Parameter	B	Std. Error	t	Sig.	95% Confidence Interval
					Lower Bound	Upper Bound
Intercept	−826.157	68.217	−12.111	0.000	−959.861	−692.452
[gender_man]	71.117	40.157	1.771	0.077	−7.591	149.825
[gender_woman]	54.267	40.156	1.351	0.177	−24.439	132.973
[marital_status_single ]	67.002	8.654	7.743	0.000	50.041	83.963
[marital_status_married]	69.916	9.090	7.692	0.000	52.100	87.733
[marital_status_divorced]	40.453	17.147	2.359	0.018	6.845	74.061
[marital_status_widow]	−68.011	76.316	−0.891	0.373	−217.589	81.567
[level_of_education_not finished]	38.585	14.711	2.623	0.009	9.751	67.419
[level_of_education_primary]	55.567	13.383	4.152	0.000	29.336	81.797
[level_of_education_lower secondary]	70.235	15.174	4.629	0.000	40.494	99.976
[level_of_education_secondary vocational]	71.265	12.984	5.488	0.000	45.815	96.714
[level_of_education_complete secondary vocational]	65.378	12.926	5.058	0.000	40.044	90.713
[level_of_education_upper secondary vocational]	5.157	14.860	0.347	0.729	−23.969	34.283
[level_of_education_higher]	30.986	18.486	1.676	0.094	−5.248	67.219
[level_of_education_university 1st]	−4.618	36.049	−0.128	0.898	−75.273	66.038
[level_of_education_university 2nd]	0.698	36.026	0.019	0.985	−69.912	71.309
[level_of_education_university 3rd]	−63.899	75.987	−0.841	0.400	−212.834	85.035
[disadvantages_no]	−26.941	46.131	−0.584	0.559	−117.356	63.475
[disadvantages_school-leaver]	−28.967	46.127	−0.628	0.530	−119.374	61.441
[disadvantages_long time unemployed]	−28.341	46.097	−0.615	0.539	−118.691	62.010
[disadvantages_low education]	75.772	69.375	1.092	0.275	−60.202	211.745
[disadvantages_no regular paid job]	−22.515	53.835	−0.418	0.676	−128.030	83.000
[disadvantages_child care]	30.527	70.706	0.432	0.666	−108.056	169.109
[driving license_group1_no]	−3.953	11.308	−0.350	0.727	−26.117	18.210
[driving license_group2_no]	21.128	11.326	1.865	0.062	−1.071	43.326
[driving license_group3_no]	−28.934	9.421	−3.071	0.002	−47.399	−10.469
[driving license_group4_no]	13.459	14.144	0.952	0.341	−14.263	41.181
[driving license_group5_no]	4.162	8.073	0.515	0.606	−11.661	19.984
[region_Bratislava]	168.495	18.452	9.131	0.000	132.329	204.662
[region_Trnava]	167.341	18.385	9.102	0.000	131.306	203.376
[region_Trencin]	153.800	18.378	8.369	0.000	117.781	189.820
[region_Nitra]	139.816	18.354	7.618	0.000	103.842	175.791
[region_Zilina]	141.493	18.343	7.714	0.000	105.541	177.444
[region_Banska Bystrica]	125.672	18.347	6.850	0.000	89.712	161.631
[region_Presov]	117.204	18.324	6.396	0.000	81.289	153.118
[region_Kosice]	132.310	18.333	7.217	0.000	96.378	168.242
[school_1]	−40.347	13.386	−3.014	0.003	−66.583	−14.111
[school_2]	58.183	18.888	3.080	0.002	21.162	95.204
[school_3]	40.541	20.335	1.994	0.046	0.684	80.398
[school_4]	30.391	17.372	1.749	0.080	−3.659	64.440
[school_5]	16.707	16.172	1.033	0.302	−14.990	48.403
[school_6]	−1.082	29.275	−0.037	0.971	−58.462	56.297
[school_7]	31.478	16.028	1.964	0.050	0.063	62.893
[school_8]	−6.986	47.122	−0.148	0.882	−99.344	85.372
[school_9]	81.595	30.567	2.669	0.008	21.685	141.505
[school_10]	17.261	17.970	0.961	0.337	−17.959	52.482
[school_11]	0.464	35.533	0.013	0.990	−69.181	70.109
[school_12]	62.308	24.203	2.574	0.010	14.870	109.746
[school_13]	51.315	15.806	3.247	0.001	20.336	82.294
[school_14]	28.298	15.864	1.784	0.074	−2.795	59.390
[school_15]	−56.627	101.547	−0.558	0.577	−255.658	142.405
[school_16]	5.064	83.397	0.061	0.952	−158.393	168.522
[school_17]	−91.687	21.849	−4.196	0.000	−134.510	−48.865
[school_18]	−189.438	35.828	−5.287	0.000	−259.660	−119.216
[school_19]	−3.973	19.787	−0.201	0.841	−42.755	34.809
[school_20]	42.552	15.988	2.662	0.008	11.217	73.888
[school_21]	22.919	16.446	1.394	0.163	−9.315	55.152
[school_22]	5.896	18.100	0.326	0.745	−29.580	41.372
[school_23]	46.937	16.663	2.817	0.005	14.278	79.596
[school_24]	78.390	115.830	0.677	0.499	−148.635	305.415
[school_25]	85.485	27.919	3.062	0.002	30.765	140.206
[school_26]	−35.728	37.628	−0.950	0.342	−109.478	38.021
[school_27]	−44.091	42.487	−1.038	0.299	−127.366	39.184
[school_28]	34.412	37.712	0.913	0.362	−39.503	108.327
[school_29]	10.975	40.458	0.271	0.786	−68.322	90.271
[school_30]	21.377	37.754	0.566	0.571	−52.620	95.374
[school_31]	1.265	38.374	0.033	0.974	−73.947	76.478
[school_32]	−11.596	41.047	−0.283	0.778	−92.047	68.855
[school_33]	−9.776	38.562	−0.254	0.800	−85.357	65.805
[school_34]	27.574	38.906	0.709	0.478	−48.680	103.829
[school_35]	17.305	38.444	0.450	0.653	−58.044	92.655
[school_36]	41.112	38.703	1.062	0.288	−34.746	116.970
[school_37]	−20.469	42.500	−0.482	0.630	−103.769	62.830
[school_38]	−95.066	51.466	−1.847	0.065	−195.938	5.806
[school_39]	37.513	44.333	0.846	0.397	−49.378	124.405
[school_40]	−26.530	38.524	−0.689	0.491	−102.037	48.978
[school_41]	14.967	16.005	0.935	0.350	−16.403	46.336
age_at_registration	−75.672	1.704	−44.420	0.000	−79.011	−72.333
evidence_length	−0.264	0.004	−59.522	0.000	−0.273	−0.255
previous_evidence_records	−0.108	0.009	−12.416	0.000	−0.126	−0.091
age_shifted	103.376	1.656	62.413	0.000	100.129	106.622
step2	278.185	4.540	61.280	0.000	269.287	287.082

for wage and

Table A3. Regression model for the outcome variable employment, step 3. Source: Authors.

Parameter	B	Std. Error	t	Sig.	95% Confidence Interval
					Lower Bound	Upper Bound
Intercept	−422.068	35.536	−11.877	0.000	−491.718	−352.417
[gender_man]	26.993	20.919	1.290	0.197	−14.008	67.993
[gender_woman]	30.689	20.918	1.467	0.142	−10.311	71.688
[marital_status_single ]	34.807	4.508	7.721	0.000	25.972	43.643
[marital_status_married]	40.509	4.735	8.555	0.000	31.228	49.790
[marital_status_divorced]	44.160	8.932	4.944	0.000	26.653	61.668
[marital_status_widow]	12.023	39.755	0.302	0.762	−65.897	89.942
[level_of_education_not finished]	19.318	7.664	2.521	0.012	4.298	34.339
[level_of_education_primary]	28.069	6.972	4.026	0.000	14.405	41.733
[level_of_education_lower secondary]	36.165	7.905	4.575	0.000	20.672	51.658
[level_of_education_secondary vocational]	40.389	6.764	5.971	0.000	27.132	53.646
[level_of_education_complete secondary vocational]	32.092	6.733	4.766	0.000	18.894	45.290
[level_of_education_upper secondary vocational]	−3.999	7.741	−0.517	0.605	−19.171	11.173
[level_of_education_higher]	11.406	9.630	1.184	0.236	−7.468	30.281
[level_of_education_university 1st]	8.356	18.779	0.445	0.656	−28.450	45.162
[level_of_education_university 2nd]	−7.704	18.767	−0.411	0.681	−44.487	29.079
[level_of_education_university 3rd]	−47.247	39.584	−1.194	0.233	−124.831	30.337
[disadvantages_no]	−15.545	24.031	−0.647	0.518	−62.645	31.555
[disadvantages_school-leaver]	−24.566	24.029	−1.022	0.307	−71.662	22.530
[disadvantages_long time unemployed]	−24.774	24.013	−1.032	0.302	−71.839	22.292
[disadvantages_low education]	26.475	36.139	0.733	0.464	−44.357	97.307
[disadvantages_no regular paid job]	−12.384	28.044	−0.442	0.659	−67.350	42.582
[disadvantages_child care]	27.312	36.832	0.742	0.458	−44.879	99.503
[driving license_group1_no]	−2.088	5.891	−0.354	0.723	−13.633	9.458
[driving license_group2_no]	9.891	5.900	1.676	0.094	−1.673	21.455
[driving license_group3_no]	−22.070	4.908	−4.497	0.000	−31.689	−12.451
[driving license_group4_no]	6.524	7.368	0.886	0.376	−7.917	20.965
[driving license_group5_no]	5.632	4.205	1.339	0.180	−2.610	13.875
[region_Bratislava]	81.686	9.612	8.498	0.000	62.845	100.526
[region_Trnava]	85.582	9.577	8.936	0.000	66.811	104.354
[region_Trencin]	82.983	9.573	8.668	0.000	64.219	101.747
[region_Nitra]	76.428	9.561	7.994	0.000	57.689	95.168
[region_Zilina]	74.467	9.555	7.793	0.000	55.739	93.195
[region_Banska Bystrica]	66.437	9.557	6.951	0.000	47.705	85.169
[region_Presov]	64.158	9.545	6.721	0.000	45.450	82.867
[region_Kosice]	72.452	9.550	7.587	0.000	53.734	91.170
[school_1]	−25.684	6.973	−3.683	0.000	−39.351	−12.017
[school_2]	30.506	9.839	3.100	0.002	11.221	49.792
[school_3]	25.088	10.593	2.368	0.018	4.325	45.850
[school_4]	13.333	9.050	1.473	0.141	−4.404	31.070
[school_5]	12.734	8.424	1.512	0.131	−3.778	29.246
[school_6]	−0.932	15.250	−0.061	0.951	−30.822	28.959
[school_7]	13.762	8.350	1.648	0.099	−2.603	30.127
[school_8]	16.521	24.547	0.673	0.501	−31.590	64.633
[school_9]	61.592	15.923	3.868	0.000	30.384	92.801
[school_10]	10.516	9.361	1.123	0.261	−7.831	28.863
[school_11]	0.339	18.510	0.018	0.985	−35.941	36.619
[school_12]	24.289	12.608	1.926	0.054	−0.423	49.000
[school_13]	24.016	8.234	2.917	0.004	7.879	40.154
[school_14]	15.684	8.264	1.898	0.058	−0.513	31.881
[school_15]	−0.794	52.899	−0.015	0.988	−104.475	102.887
[school_16]	41.456	43.444	0.954	0.340	−43.693	126.606
[school_17]	−55.766	11.381	−4.900	0.000	−78.074	−33.459
[school_18]	−90.171	18.664	−4.831	0.000	−126.752	−53.591
[school_19]	2.489	10.307	0.241	0.809	−17.714	22.691
[school_20]	15.820	8.328	1.899	0.058	−0.504	32.143
[school_21]	11.379	8.567	1.328	0.184	−5.412	28.171
[school_22]	−1.783	9.429	−0.189	0.850	−20.263	16.698
[school_23]	23.500	8.680	2.707	0.007	6.487	40.513
[school_24]	81.193	60.339	1.346	0.178	−37.070	199.457
[school_25]	42.132	14.544	2.897	0.004	13.626	70.637
[school_26]	−30.943	19.601	−1.579	0.114	−69.361	7.476
[school_27]	−48.867	22.133	−2.208	0.027	−92.247	−5.487
[school_28]	−14.570	19.645	−0.742	0.458	−53.074	23.935
[school_29]	−24.319	21.075	−1.154	0.249	−65.627	16.988
[school_30]	−16.326	19.667	−0.830	0.406	−54.873	22.221
[school_31]	−18.975	19.990	−0.949	0.343	−58.155	20.205
[school_32]	−23.652	21.382	−1.106	0.269	−65.561	18.257
[school_33]	−23.117	20.088	−1.151	0.250	−62.489	16.256
[school_34]	−18.945	20.267	−0.935	0.350	−58.668	20.777
[school_35]	−12.171	20.026	−0.608	0.543	−51.423	27.080
[school_36]	−13.195	20.161	−0.654	0.513	−52.711	26.322
[school_37]	−23.817	22.139	−1.076	0.282	−67.210	19.576
[school_38]	−54.849	26.810	−2.046	0.041	−107.396	−2.302
[school_39]	−2.892	23.094	−0.125	0.900	−48.157	42.372
[school_40]	−35.351	20.068	−1.762	0.078	−74.685	3.982
[school_41]	9.627	8.337	1.155	0.248	−6.715	25.968
age_at_registration	−42.152	0.887	−47.498	0.000	−43.891	−40.412
evidence_length	−0.144	0.002	−62.430	0.000	−0.149	−0.140
previous_evidence_records	−0.059	0.005	−12.877	0.000	−0.067	−0.050
age_shifted	56.925	0.863	65.976	0.000	55.234	58.616
step2	138.742	2.365	58.670	0.000	134.107	143.377

for employment).

The results are as follows. For the outcome variable wage (brutto), the coefficient of the independent variable $ste{p}_2$ in the regression model is EUR 278.19, which means that the participants of the Graduate practice had an average wage of EUR 278.19 higher in the impact period (24 months) than the non-treated young jobseekers in the control group. According to the p-value of the test of significance of the regression coefficient (p-value < 0.05), it is also evident that the variable $ste{p}_2$ is a statistically significant variable for the dependent variable $wage$ in this regression model. It was also statistically observed that the average monthly wage of women who participated in the Graduate practice was 37.4% higher than the average wage of women who did not participate in the Graduate practice. In the case of men, the difference was even higher, namely 48.9% in favour of the participants. In both groups, we observed that the average monthly wage was above the minimum wage according to [71] in the period under analysis. For the outcome variable, employment regression coefficient of 138.74 was estimated for the variable $ste{p}_2$. This means that the persistence of the participant in the Graduate practise in the employment relationship in the impact period was, in general, more stable. The participants’ employment with one employer or as self-employed people lasted almost 139 days (i.e., 4.6 months longer compared to their counterparts in the control group). This effect is also statistically significant in the model according to the p-value (<0.05). The employability of young jobseekers and their sustainability in the labour market in the impact period is expressed as the percentual share of employed in all unemployed persons in the treatment and control groups. As presented in Figure 2, it was much higher for the participants of the Graduate practice in the first 11 months of the impact period. For instance, in the first month of the observed impact period, only 15% of individuals from the control group (i.e., young jobseekers who were not involved in the Graduate practice, found, and retained a job for one month). On the other hand, 55% of the participants found and retained a job in the first month of the observed impact period.

It is evident that the participants of the Graduate practice had a better starting position in terms of their employability than non-treated individuals; however, the trend of the overall development of employability and sustainability in the labour market (length of employment) of both groups is the same. We have observed the biggest differences in the first ten months of the impact period, favouring the treatment group; however, in the 11th month of this period, the differences between the groups almost disappeared, and during the second year, individuals who did not participate in the Graduate practice reached even slightly better results in terms of their sustainability in the labour market (i.e., 3.2% individuals from the sample of non-treated individuals kept a job position from 13 to 24 months compared with 1.1% participants in the Graduate practice).

The results indicate that the Graduate practice has had a significant effect on the employability and length of employment of the participants in the labour market in the first year after its termination; however, the duration of the effect lasts from the short- to medium-term. Moreover, after the first year, the effect of the Graduate practice is slightly disappearing (the differences between the individuals of both groups in the meaning of their employability and sustainability on the labour market over time).

4. Discussion

As for the ability of the measure to achieve the objectives that are favourable to it, we assessed this through a counterfactual analysis. This is a quantitative way of assessing the impact of policy interventions, that the European Commission is currently promoting. In our case, the basic problem with the above analysis was the considerable inconsistency and error rate in the databases of the responsible institutions we worked with; therefore, we emphasise that if the active labour market measures are to be properly assessed, there is a need to improve communication, information transfer, and control between the responsible institutions. At the same time, it is important to increase the confidence of these institutions in carrying out evaluations.

Generally, and according to the main objectives of the Graduate practice, we consider the measure beneficial for its participants since it allowed them to gain professional skills, working experience, and habits that, based on the quantitative results of the study, played an important role mainly in the first year after the termination of the practice in building a working career.

The counterfactual evaluation of the Graduate practice accepting the self-selection and selection-bias problems resulting from the voluntariness of the measure of individuals who met the legal conditions of the measure revealed that the measure (despite the decreasing number of participants in the last years before the COVID-19 crisis) was beneficial in terms of employment of participants in the labour market, in terms of retaining a job position and, last but not least, in terms of the financial reward for work performed (wage) during the observed impact period of the measure. Compared with their counterparts who were not involved in the Graduate practice, the ability to get a job and keep a job was much higher, especially in the first 11 months of the impact period. Thus, we state that the outcomes of participation in the measure are clearly reflected, especially in the short and medium-term. An important finding is also the fact that the participants of the Graduate practice had higher monthly wages (on average). Thus, it can be assumed that professional skills, working experience, and habits resulting from participation in the measure were positively reflected in the volume of earnings outcomes compared to those who were legally entitled to the measure but did not participate of their own free will.

The evaluation results are beneficial, especially in terms of defining conditions for the operation of the measure in the future. At the same time, the application of the counterfactual approach shows that the proposals for measures aimed at improving the situation in the labour market, especially with regard to the disadvantaged groups of jobseekers, should be promoted as evidence-based, as advocated by the European Commission.

Concerning the limitations of our study, we consider the use of only one evaluation method with one chosen instrumental variable; therefore, we consider it appropriate in the continuation of this study to compare and confirm the results with the application of other counterfactual methods or using another instrumental variable(s). One of the main problems we encountered during the evaluation was that in Slovakia, some variables in the database of jobseekers are recorded statically (i.e., at the time of registration). Some of the variables, such as age, can be updated at the beginning of the intervention. This actualisation was necessary to verify the eligibility criterion for the Graduate practice; however, the problem arises with such types of variables as marital status, level of education, region of permanent residence, and so on. The values of these variables may change over time for each jobseeker, but the evaluator is not able to update them; therefore, it was not possible to carry out the evaluation for smaller target groups of jobseekers and to compare their outcomes achieved during the impact period.

5. Conclusions

The main aim of the study was to evaluate the effects of a selected measure aimed at improving the employment situation of school graduates in Slovakia, namely “Allowance for school graduate practice performance” in the context of the defined objectives of the measure and at the same time, in the context of the current labour market situation. As is already known, the current situation in the labour market, and thus also the issue of employability of school graduates, is largely influenced not only by the graduates’ qualifications and work experience gained in the education process but unfortunately also by the ongoing health crisis.

Employers in Slovakia are interested in employing school graduates but have long pointed out the inconsistency of qualification requirements for school graduates and job vacancies, as well as the lack of practical experience and work habits. Therefore, after the recovery of the situation in the labour market after the health crisis, we expect that this measure will play one of the main tasks of supporting employment of school graduates in the conditions of the Slovak economy and Slovak labour market. However, according to the results of the study, the effectiveness of the measure will need to be supported by simplifying and unifying procedural steps that condition the participation in the measure. At the same time, it will be necessary to link this measure, for example, to other suitable measures for the same group of young jobseekers and try to make the measure more attractive to potential employers.

The topic of this study provides a wide space for discussion and further research. Furthermore, we believe that our results will contribute to the future expansion of the knowledge of the public about this topic, especially in Slovakia.