Evolution of Interdependencies between Education and the Labor Market in the View of Sustainable Development and Investment in the Educational System

Abstract

Education represents the basic pillar of preparing individuals for integration into the labor market, but also is a crucial component of ensuring sustainable development. The purpose of this research was to identify the type of influences existing between education and the labor market in EU member countries in the context of different levels of investment in the educational system. Cluster analysis and the ordinary least squared method were used to identify the type of influences between the indicators characterizing the level of education and the labor market between 2000 and 2021. The empirical results showed that there was a significant negative correlation of the educational dropout rate with the level of employee compensation, number of hours worked by each employee, and their labor productivity, in the countries with the poorest level of educational investment. In the countries with significant investments in education, getting a graduate diploma and participating in vocational training programs led to a better compensation of employees and a higher employee productivity while the financial aid given by the state for pupils and students reduced the number of worked hours, brought down unemployment amongst people with primary and secondary education and, last but not least, increased the employment rate for higher education graduates. An average level of educational investment led to negative influences between early-stages and employees’ payment level and real labor productivity, while becoming involved in educational activities and participating in vocational training programs increased their rates of remuneration and real productivity. A significant impact of higher education graduates on both increasing unemployment rates and falling employment rates was noticed as has been identified in other studies.

1. Introduction

The challenge of correlating educational systems with the labor market requirements has become not only a national priority but also a European one and even a global one [1]. The financial crisis followed by the pandemic crises have both generated big problems at the European level in terms of inactivity and unemployment rates, especially when it comes to young people [2]. A qualified workforce and a good educational system in line with the existence of academic and institutional structures that ensure the use of knowledge are the prerequisite for the guarantee of sustainable development [3]. From the perspective of sustainable development, both the Sustainable Development Goal SDG 4 “Quality education” and SDG 8 “Decent work and economic growth” can be regarded as accelerating goals because implementing them is the key for their success. At the same time, the two goals are intensely interrelated. The role of SDG 4 “Quality education” in implementing SDG 8 “Decent work and economic growth” means a smooth transition from one goal to another based on their full symbiosis. The SDG 4 is a resource for the development of SDG 8, offering quality education. At the same time, SDG 8 provides appropriate material resources for the functioning of SDG 4 [4].

Both education and training are the basis of any effort to improve the level of productivity of a country and to increase people’s chances not only in finding a good workplace but in getting good jobs. The educational level and the set of skills of the workforce make a clear impact both at individual as well as national level. Education can influence various labor market outcomes such as: employability or unemployment, wages, worker productivity, working conditions, worker health, etc. The mechanisms by which these effects are produced are different, namely: number of years of schooling, level of education, investment in education, certain accreditations, etc. [5].

Ionescu [5] observed that an increase in investment in education leads mainly to positive effects on the labor market, in the sense of increasing employability and income, and less to the reduction of negative effects, i.e., unemployment. Therefore, the higher the level of education, the higher the chances of getting a job or keeping it in times of crisis.

Hien [6] stated that public investment in education and training has many positive benefits and impacts on the economy and society by improving people’s qualification and intellectual level, and in this way reduces unemployment and poverty in a country.

In general, higher levels of education will lead to more opportunities in the labor market to get a higher position and have better chances to avoid unemployment [7]. If there are people with higher education degrees who are unemployed, this is because they refuse to content themselves with poor jobs that are inadequate for their level of qualification.

The effect of education can be observed not only when it comes to the employment rates but also with respect to the working conditions [8]. Even overqualified people (those whose level of qualification exceeds the demands of their jobs) earn more, in general, than those who are hired for the same job but have a lower level of qualification [9]. A more recent paper has tested whether the students’ expectations regarding wages are realistic and if there is a connection between the future income and the type of school they attend. In this work, it was proved that students’ expectations regarding wages were higher than the employers’ realistic offer, that students do not make a distinction between their wage expectations depending on the school they attend [10], and they do not consider things such as the type of contract or the arrangements regarding working time. The educational level influences other essential aspects of the working conditions as workers with a higher level of education are in a better position to negotiate more satisfying working conditions.

A high level of education leads to an increase in worker productivity, but also improves the quality of life by ensuring better jobs and increased earnings for the individual [11]. Boissiere [12] reported a positive relationship between education and income in developing countries, taking into account the importance of public investments in primary education.

In developing countries, a dilemma has been identified with respect to the stimulation of or restraint on access to education. Consequently, on the one hand, the level of education must increase to determine economic growth while, on the other hand, a higher level of education involves negative consequences for the employment rate and salaries because the economic structure of developing countries does not have the capacity to absorb more graduates [13].

Although the need to achieve a better match between the skills acquired by young people at school and those needed in the labor market has been recognized, there are studies [14,15] that believe that this is not enough, and that economic and social policies are needed to generate jobs or additional work for young people.

The main aim of the paper was to identify the type of influences existing between the relevant indicators regarding the educational system and those regarding the employment market in EU member states within the current macroeconomic context, taking into account the different levels of investment in the educational system allocated by each country. The importance of the studied topic relates to the correlation of the economy’s needs with the problems generated by educational level in developed countries, with implications on how budgetary resources are oriented to make more efficient the relationship between the two analyzed systems.

The scientific originality of the research derives from conducting a cluster analysis based on several indicators concerning both government expenditure on education and the level of development expressed by the real GDP per capita as well as from conducting several correlations analyses between the main indicators at the level of the identified clusters. Hence, in the previous literature, although there are studies which have conducted a cluster analysis, they have not been continued with an econometric research program, while the studies which have conducted an econometric analysis of the relevant indicators with regard to the employment market and the educational system did not divide the analysis to consider the development level of the states and the investment level in education, nor did they consider a great number of indicators.

The paper is structured as follows: Section 2 presents the literature review. In Section 3, methodology and data are presented. The empirical results and discussions are provided in Section 4. Section 5 presents the study’s main conclusions, along with some policy recommendations, limitations and directions for future work.

2. Literature Review

For a long period of time, the interest in education and the labor market was oriented towards the effects generated by education on the capital accumulation as part of the model of economic growth. Having analyzed different cultures and political regimes Becker [16] noticed that economic gains were positively correlated with education and the level of qualification, while education negatively influenced unemployment. Moreover, in the same period, Selowski [17] analyzed education as a way to increase the quality of the workforce. He analyzed the positive relationship between labor productivity and incorporated education from an empirical perspective, and at the same time considering the states’ expenditure on education. Another contribution was brought by Mincer [18] who tried to determine the statistical relationship between salaries, education and training on the job and his research was continued in many studies. More recent papers have demonstrated that the different levels of education (primary, secondary, tertiary) and the different types of education (general and vocational) can explain the variations of salaries between individuals [19,20].

Oancea et al. [21] analyzed the causality relationship between economic growth and higher education in Romania and the Czech Republic for the 1980–2013 period and they noticed a significant positive impact of university education on economic growth. In turn, Jovović et al. [22], in a study conducted on the labor market in Montenegro, concluded that the inadvertencies between the needs of the educational system and the needs of the labor market can significantly influence economic growth and economic development. Another paper, which analyzed the effects of the financing of education in Croatia using “the hypothesis of growth driven by education”, tested the fact that Croatia could flourish following increased financing of higher education [23]. At the same time, another analysis confirmed the existence of a positive impact of vocational training costs on economic growth, allowing the authors to conclude that there was a need to develop professional training programs for staff as one of the drivers of macroeconomic development [24].

Within China, a recent study that used panel data for the 1997–2020 period, analyzed the impact of investments on economic growth for different levels of education and concluded that investment in general education had a negative effect on economic growth, while investment in advanced education could lead to economic growth [25].

Nevertheless, in a segmented labor market where occasional work and temporary contracts are common and formal permanent contracts are not that common, human capital could be changed for job security. In such environments, workers with higher skills than the necessary ones for their position are most likely to be found in permanent rather than temporary jobs [26]. Consequently, education can offer protection to a certain extent for some vulnerable jobs. One study found that the proportion of young people who were vulnerable in terms of finding a job because of a primary level of education was higher than that of young people with a superior level of education [27].

Another topic which was tackled in the analysis of the relationship between the labor market and education is the one placing emphasis on the link between education, job opportunities and unemployment. Hence, McKenna [28] drew the conclusion that schooling extends job opportunities because educated workers are productive in all workplaces while unqualified workers are productive only in certain workplaces. Many empirical studies have shown that a high level of education ensures a relatively low risk of unemployment [24,29,30,31,32]. In some papers, a bidirectional relationship was identified between schooling and employment: on the one hand, education ensures a drop in the unemployment rate [33], on the other hand, unemployment raises the demand for education which in turn determines people to invest in their own instruction [34]. Masarova et al.’s research findings revealed that the educational attainment of the active and employed population is increasing and that the educational level of the unemployed rises and falls irregularly depending on how the labor market evolves [35].

Another analyzed issue regarding education and the labor market is the one concerning the inadvertencies between people’s level of training and their jobs. Hence, we can talk about over-education or under-education, but also about the differences between the graduates’ field of study and the fields generating the real demand on the employment market [36,37]. McGuiness and Sloane [38] studied the discrepancies in the labor market in the United Kingdom and drew the conclusion that over-education and over-qualification imply a wage penalty, but over-qualification has stronger negative effects than over-education on job satisfaction. In one analysis carried out in the US, Tsai [39] observed that over-education does not generate lower income as it has been noticed in other studies. Spatial mismatches have also been analyzed and they appear when the quantity of human capital with a certain educational potential has a different spatial distribution compared to the demand for qualified workers. Eliminating these mismatches requires changes of the educational policy in a certain region (possible correspondence for a long period of time) or the relocation of the workforce [40].

One important observation referred to the fact that mismatched skills lead frequently to migration, because educated workers decide to leave their country of origin and migrate to other countries where they can find better job opportunities [41,42]. In the literature, an emphasis has been placed on the importance of graduates’ skills when they enter the labor market. Hence, the phrase “skills of the 21st century” has been coined and it regards the general necessary skills besides the competences that are specific to a job and they refer to collaboration, communication, critical thinking, etc. [43,44,45]. Consequently, an analysis was carried out with respect to the students’ soft skills that have proved to be crucial when entering the labor market [46]. These skills have been classified into seven general groups: communication, arithmetical, technological, learning, problem-solving, teamwork, and specialized skills [47].

All institutions, both educational and those in the labor market, must make a collaborative effort to train students for them to acquire necessary skills and to increase their chances in the labor market [48].

The gaps in literature regarding the possibility of some combined effects between education and the labor market have led us to undertake this research. We aimed to identify the type of influences between several dimensions of education, such as the demand for education, supply of education, and the results of the educational system, and the labor market outputs, such as the employment rate, unemployment rate, and percentage of young people neither educated nor on the labor market, in the context of a different levels of investment in the educational system.

3. Methodology and Data Used

In the first phase of the analysis, our goal was to group the EU member countries into clusters and then to continue the econometrical analysis within each identified cluster. The k grouping is a method of vectorial quantization, originating from the processing of the signal whose purpose is to divide n observations into k clusters where each observation belongs to the cluster with the closest mean (cluster centers or centroid cluster), serving as a prototype for the cluster. In k-means cluster analysis, k represents the number of clusters we define when starting the algorithm [49]. The algorithm involves alternating between two steps: the assignment step, in which each observation is assigned to the group with the closest mean, i.e., the one with the closest Euclidean distance small square [50], and the updating step, in which the averages (centroids) are recalculated for the observations assigned to each group. The algorithm converges when the assignments no longer change.

We used three indicators for the clusterization method of the 27 EU member states. Two of the indicators used refer to investments in education and they are: General government expenditure on education as a percentage of GDP (GGE) and public expenditure on education as a percentage of GDP (PEE). Using these indicators is very significant because it gives a hint regarding the importance given to education by governments. We also used the real GDP per capita expressed as a ratio between the real GDP and the average population of a certain year. The GDP measures the value of the total production of goods and services in an economy during a time period, so we used it to express the economic development of a country.

We took 2019 as the reference year for the three variables because this was the last year unaffected by the pandemic crisis and we considered that it was important that the indicators used in defining clusters should derive from a period with no economic turmoil.

In

Table 1. Indicators used in forming the clusters for the 27 EU member states (%, 2019).

GEO	PEE	GGE	GDP
Austria	4.71	4.8	1.1
Belgium	6.25	6.1	1.7
Bulgaria	4.20	3.8	4.8
Croatia	3.92	4.7	4.0
Cypress	5.24	5.2	4.1
Czech Republic	4.50	4.9	2.6
Denmark	6.36	6.3	1.1
Estonia	4.85	6.1	4.4
Finland	6.06	5.6	1.1
France	5.35	5.2	1.4
Germany	4.70	4.3	0.8
Greece	3.59	4.0	2.0
Hungary	3.90	4.7	4.9
Ireland	3.77	3.1	4.0
Italy	4.10	3.9	0.7
Latvia	4.43	5.7	3.3
Lithuania	3.80	4.6	4.9
Luxembourg	3.72	4.8	0.2
Malta	4.59	5.1	1.8
Netherlands	5.16	5.0	1.3
Poland	4.67	5.0	4.5
Portugal	4.61	4.5	2.7
Romania	3.16	3.6	4.3
Slovakia	3.98	4.3	2.4
Slovenia	4.61	5.5	2.6
Spain	4.08	4.0	1.2
Sweden	7.06	6.9	1.0

Source: calculations based on EUROSTAT data [51].

we list the indicators based on which we started forming the clusters.

Once the clusters were formed, we continued the analysis to identify the type of influences existing between education and the labor market by outlining the model used in the econometrical analysis. We decided to use a multiple linear regression model because for each dependent variable we identified several explanatory variables.

The mathematical equation for the multiple linear regression model which links a y variable to k x variables is written in the following form:

$$y_i={\beta }_0+{\beta }_1{x}_{i,1}+{\beta }_2{x}_{i,2}+\dots +{\beta }_k{x}_{i,k}+{\epsilon }_{i }$$

(1)

where:

-

y is considered to be the answer, the result or the dependent variable

-

x is considered to be the predictive, explanatory or independent variable

-

β₀ is the free term of the equation and expresses the value of y when all the other parameters are programmed to 0)

-

β_1..k represents the coefficients associated with the independent variables and expresses the change of y per unit of increase in the predictor variable associated when all the other predictors remain constant.

-

ε_i represents the standard error of the model and expresses how much variation is in our estimation for the dependent variable y.

We adapted the regression model to our analysis, and we obtained the following equation:

$$Y_i={\beta }_0+{\beta }_{1}ACTIV{02}_{i,1}+{\beta }_{2}ACTIV{34}_{i,2}+{\beta }_{3}ACTIV{58}_{i,3}+{\beta }_{4}ELE{T}_{i,4}+{\beta }_{5}FA{D}_{i,5}+{\beta }_{6}PARTE{T}_{i,6}+{\beta }_{7}AT{58}_{i,7}+{\beta }_{8}YOUN{G}_{i,8}+{\epsilon }_{i }$$

(2)

where Y_i will be represented one after the other by the following dependent variables: COMP, HRW, PROD, UNEMP02, UNEMP34, UNEMP58, EMP02, EMP34 and EMP58.

We used the Ordinary Least Squares regression—OLS—to estimate the coefficients of the linear regression equations which describe the relationship between several independent quantitative variables and a dependent variable (multiple linear regression). The OLS method aims at minimizing the sum of the square differences between the observed values and the predicted values. Hence, β vector of coefficients can be estimated through the following formula:

β = (X’DX) − 1X’Dy

(3)

where: X is the matrix of explanatory variables preceded by a vector of 1 s, D is a matrix with the weights w_i across its diagonal and y the vector of the n values observed of the dependent variable.

The vector of the predicted values can be written as follows:

y* = Xβ = X(X’DX) − 1X’Dy

(4)

We can even determine the variance σ² of the random error ε through the following formula:

$${\sigma }^2=\frac{1}{\left(W-p^{*}\right){{\displaystyle \sum }}_{i=1}^n{w}_i\left(y_i-y_i^{*}\right)}$$

(5)

where p* is the number of explanatory variables at which we add 1 if the intercept is not fixed, w_i is the weight of the i observation, W is the sum of the weights w_i, y is the vector of the observed values and y* is the vector of the predicted values.

To determine the type of influences between education and the labor market, we started by identifying the necessary variables in accordance with both the relevance for the two fields, but also the availability of data. The time frame of the analysis was 2000–2021. Data were collected and processed from the official website of the European Commission—EUROSTAT. We grouped the variables into two categories: dependent variables—in this respect we used indicators that characterize the labor market and independent variables—represented by the indicators of education (see

Table 2. Variables used in the cluster analysis.

Name	Acronym	Unit of Measurement
Dependent variables
Compensation of employees per hour worked	COMP	Euro
Employee hours worked	HRW	Hours worked
Real labor productivity per person employed	PROD	Index, 2015 = 100
Unemployment rates of persons aged between 15 and 74 with less than primary education, primary education, and lower secondary education (levels 0–2)	UNEMP02	Percentages
Unemployment rates of persons aged between 15 and 74 with upper secondary education and post-secondary non-tertiary education (levels 3–4)	UNEMP34	Percentage points
Unemployment rates of persons aged between 15 and 74 with higher education (levels 5–8)	UNEMP58	Percentage points
Unemployment rates of persons aged between 15 and 74 with less than primary education, primary education, and lower secondary education (levels 0–2)	EMP02	Percentage points
Unemployment rates of persons aged between 15 and 74 with upper secondary education and post-secondary non-tertiary education (levels 3–4)	EMP34	Percentage points
Unemployment rates of persons aged between 15 and 74 with higher education (levels 5–8)	EMP58	Percentage points
Independent variables
Activity rates of persons aged between 15 and 74 with less than primary education, primary education, and lower secondary education (levels 0–2)	ACTIV02	Percentage points
Activity rates of persons aged between 15 and 74 with upper secondary education and post-secondary non-tertiary education (levels 3–4)	ACTIV34	Percentage points
Activity rates of persons aged between 15 and 74 with higher education (levels 5–8)	ACTIV38	Percentage points
Early educational dropout (persons between 18 and 24 years old)	ELET	Percentage points
Financial aid for pupils and students as percentage of total public expenditure on education	FAD	Percentage points
Rate of participation in education and training (last 4 weeks—population aged between 25 and 64 years old)	PARTET	Percentage points
Higher education graduates (population aged between 25 and 34 years old)	AT58	Percentage points
Young people who are neither employed, nor enrolled in an educational program (15–24 years old)	YOUNG	Percentage points

Source: calculations made by the authors based on EUROSTAT data.

).

Further on, we will present the descriptive statistics for the variables selected for the analysis. It is very important to analyze the values registered by Skewness and Kurtosis indicators to determine the normality of the distribution of data series used. In Appendix A, we present the descriptive statistics for the 17 variables used in analyzing the type of influences between the labor market and education for each formed cluster.

For Cluster 1, the values of the Skewness indicator showed a normal distribution for 14 variables and a long tailed distribution to the right for three variables (COMP; FAD and YOUNG). Concerning the flattening degree, there were two series of mesokurtic data (PROD and UNEMP58), three series of leptokurtic data (COMP, FAD and YOUNG) and the rest of the series were platykurtic. All data series in Cluster 2 had a normal distribution with a mesokurtic flattening degree for three of the variables (PROD, UNEMP58 and AT58), a leptokurtic one for the YOUNG variable, while the rest of them had a Kurtosis platykurtic. In the 3rd cluster, we found a more diversified distribution of the data series, with long tails to the left (PROD) and to the right (UNEMP02, UNEMP34, UNEMP58, ELET), but also normal distributions of the data series. The flattening degree was normal for three of the variables (EMP02, FAD, YOUNG), arch-like for seven variables and flat for another seven variables.

The Jarque–Bera test is used to verify if a series of data is distributed normally. The null hypothesis of this test is the normal distribution of the series. If the p probability associated with the test is lower than 0.05 (we can take into consideration also values up to maximum 0.10) then the null hypothesis of a normal distribution of the series will be rejected. According to the results of the Jarque–Bera test in Appendix A, the following series of data had a normal distribution: HRW, EMP58 and ELET for Cluster 1; COMP, UNEMP02, UNEMP34, UNEMP58, ELET and AT58 for Cluster 2; ACTIV02 and ACTIV 04 for Cluster 3; the remainder of the data series do not have normal distributions.

4. Results and Discussion

In order to define the centers of the three clusters, we started from three points chosen randomly; in our case, we opted for the minimum, maximum and mean values of the PEE indicator. Thereafter, we repeatedly used the approach that was presented previously [49] until the elements of the group stopped modifying. The results obtained by the cluster analysis led to the identification of the clusters identified in Figure 1.

The composition of the three clusters that were established depending on the level of educational investment and the level of economic development of EU member states is presented in

Table 3. Cluster composition for the 27 EU member states.

Cluster 1. 10 Countries Low Educational Investment Level	Cluster 2. 4 Countries High Educational Investment Level	Cluster 3. 13 Countries Average Educational Investment Level
Bulgaria	Belgium	Austria
Croatia	Denmark	Czech Republic
Cypress	Finland	France
Estonia	Sweden	Germany
Hungary		Greece
Ireland		Italia
Latvia		Luxembourg
Lithuania		Malta
Poland		Netherlands
Romania		Portugal
		Slovakia
		Slovenia
		Spain

Source: calculations made by the authors in Microsoft Excel.

. Only four member states (cluster 2) made big investments in education, that is, 6–7% of GDP; in 13 states (cluster 3), there was an average level of educational investment; while in 10 EU member states (Cluster 1), the level of investment in education was low.

We found an analysis of the characteristics of both educational systems and the labor market in the countries forming the clusters to be very useful to identify the common elements specific to educational systems.

Hence, cluster 1 was grouped into two categories of countries. First, was a category with the Baltic countries together with Ireland and Poland that were characterized by a high participation rate in education in general, a high participation rate in tertiary education, and a quite low employment rate of the youth generated by the high number of young people who were interested in getting a higher education degree. The second country category included Romania and Bulgaria, which was characterized by a low participation rate in education in general, both regarding preschool education and higher education. Simultaneously, the rate of early education dropout and the NEET rate were both high.

Cluster 2 was composed of especially the northern countries with a high interest in education and a high participation rate in tertiary education. Noteworthy here were aspects such as the very good results when it comes to integrating the young into the labor market, the lowest NEET rates in the EU, and a high employment rate for young people aged between 15 and 24 years old.

Cluster 3 consisted of the largest group of countries which were grouped approximately in the same way in several studies analyzing similar aspects. Concerning the participation in education in these countries, preschool education and higher education both enjoy rates above the European mean. At the same time, the rate for leaving school at an early age was low, the NEET rate was low as well, but the employment rate of the young was quite different between the countries.

The econometrical analysis initially implies the identification of a possible self-correlation between the selected variables. This was done with the help of a correlation matrix for each of the three formed clusters. Moreover, through this matrix we could check if there were any positive or negative correlations between the variables used in the analysis and their intensity level.

The Pearson correlation coefficient varied from −1 to 1. An absolute value of exactly 1 implies that a linear equation perfectly describes the relationship between X and Y, with all data points situated on a line. The correlation sign is determined by the regression: a value of +1 implies that all data points are situated on a straight line for which Y increases as X increases and conversely for −1. A value equal to 0 implies that there is no linear dependency between the variables.

Table 4. Pearson correlation matrix for Cluster 1. Dependent and Independent variables are defined in Table 2.

	Dependent Variables	COMP	HRW	PROD	UNEMP02	UNEMP34	UNEMP58	EMP02	EMP34	EMP58
Independent Variables
ACTIV02		0.48	−0.20	−0.10	−0.54	−0.24	0.06	0.97	0.53	0.21
		0.00	0.01	0.21	0.00	0.00	0.40	0.00	0.00	0.01
ACTIV34		0.36	0.08	−0.45	−0.13	0.06	0.15	0.43	0.66	0.31
		0.00	0.27	0.00	0.10	0.44	0.05	0.00	0.00	0.00
ACTIV58		0.36	0.30	0.06	−0.19	−0.04	0.01	0.30	0.34	0.87
		0.00	0.00	0.44	0.01	0.59	0.87	0.00	0.00	0.00
ELET		−0.23	−0.34	−0.41	−0.28	−0.27	−0.21	0.50	0.42	0.02
		0.00	0.00	0.00	0.00	0.00	0.01	0.00	0.00	0.79
FAD		0.14	−0.58	0.02	0.02	−0.27	−0.16	0.16	0.32	0.04
		0.07	0.00	0.75	0.83	0.00	0.04	0.04	0.00	0.61
PARTET		0.47	0.32	0.13	0.09	0.08	0.18	0.09	0.32	0.24
		0.00	0.00	0.08	0.25	0.30	0.02	0.22	0.00	0.00
AT58		0.63	−0.04	0.41	0.12	0.06	0.22	0.02	0.12	0.41
		0.00	0.58	0.00	0.11	0.45	0.00	0.75	0.13	0.00
YOUNG		−0.16	−0.54	−0.34	0.26	0.38	0.39	0.03	−0.37	−0.58
		0.04	0.00	0.00	0.00	0.00	0.00	0.67	0.00	0.00

Source: calculations made by the authors based on Eviews results.

,

Table 5. Pearson correlation matrix for Cluster 2. Dependent and Independent variables are defined in Table 2.

	Dependent Variables	COMP	HRW	PROD	UNEMP02	UNEMP34	UNEMP58	EMP02	EMP34	EMP58
Independent Variables
ACTIV02		−0.28	−0.46	−0.37	−0.44	−0.55	−0.24	0.97	0.87	0.73
		0.02	0.00	0.00	0.00	0.00	0.04	0.00	0.00	0.00
ACTIV34		−0.64	0.08	−0.25	−0.02	−0.21	−0.11	0.68	0.96	0.68
		0.00	0.51	0.03	0.84	0.07	0.37	0.00	0.00	0.00
ACTIV58		−0.55	−0.08	−0.44	−0.24	−0.38	−0.33	0.71	0.78	0.96
		0.00	0.50	0.00	0.04	0.00	0.00	0.00	0.00	0.00
ELET		0.11	−0.21	−0.33	−0.38	0.01	−0.26	−0.06	−0.42	0.03
		0.37	0.08	0.00	0.00	0.91	0.03	0.62	0.00	0.83
PARTET		−0.15	−0.20	0.25	0.07	−0.23	0.13	0.53	0.70	0.24
		0.21	0.09	0.03	0.55	0.05	0.29	0.00	0.00	0.04
FAD		−0.07	−0.66	−0.35	−0.64	−0.68	−0.25	0.92	0.73	0.58
		0.56	0.00	0.00	0.00	0.00	0.03	0.00	0.00	0.00
AT58		0.61	0.05	0.72	0.52	0.08	0.36	−0.44	−0.32	−0.39
		0.00	0.66	0.00	0.00	0.48	0.00	0.00	0.01	0.00
YOUNG		0.08	0.33	−0.09	0.22	0.48	0.13	−0.68	−0.75	−0.39
		0.48	0.00	0.47	0.06	0.00	0.27	0.00	0.00	0.00

Source: calculations made by the authors based on Eviews results.

and

Table 6. Pearson correlation matrix for Cluster 3. Dependent and Independent variables are defined in Table 2.

	Dependent Variables	COMP	HRW	PROD	UNEMP02	UNEMP34	UNEMP58	EMP02	EMP34	EMP58
Independent Variables
ACTIV02		0.29	−0.09	0.17	−0.57	0.01	0.19	0.96	0.13	0.24
		0.00	0.19	0.01	0.00	0.92	0.00	0.00	0.05	0.00
ACTIV34		−0.02	−0.28	−0.25	0.08	−0.17	−0.24	0.18	0.73	0.36
		0.72	0.00	0.00	0.24	0.01	0.00	0.01	0.00	0.00
ACTIV58		−0.19	0.23	−0.15	−0.26	−0.01	0.03	0.55	0.16	0.67
		0.01	0.00	0.02	0.00	0.92	0.62	0.00	0.01	0.00
ELET		−0.18	0.31	−0.05	−0.25	0.11	0.23	0.61	−0.11	0.15
		0.01	0.00	0.42	0.00	0.10	0.00	0.00	0.10	0.02
FAD		0.24	−0.49	−0.10	−0.19	−0.36	−0.43	0.20	0.58	0.32
		0.00	0.00	0.13	0.00	0.00	0.00	0.00	0.00	0.00
PARTET		0.59	−0.53	0.14	−0.31	−0.31	−0.26	0.23	0.37	0.28
		0.00	0.00	0.03	0.00	0.00	0.00	0.00	0.00	0.00
AT58		0.55	−0.35	0.32	−0.11	0.13	0.24	0.16	−0.14	−0.14
		0.00	0.00	0.00	0.10	0.05	0.00	0.02	0.04	0.03
YOUNG		−0.41	0.50	−0.08	0.48	0.70	0.62	−0.29	−0.61	−0.55
		0.00	0.00	0.24	0.00	0.00	0.00	0.00	0.00	0.00

Source: calculations made by the authors based on Eviews results.

describe the correlation matrixes for the three clusters. The first row of the matrix for any variable, e.g., ACTIV02, lists the values of the correlation coefficient and on the second row we have the values of the associated probabilities.

In the first cluster we observed a strong positive correlation between the active persons with primary education (ACTIV02) and their employment rate (EMP02) of 0.97, which was statistically significant. We also observed a significant positive correlation in the case of active persons who have a higher education degree and their employment rate, which was again statistically significant. Noteworthy, were the negative correlations between the number of people who at an early stage either dropped out of school or gave up the vocational training programs and the remuneration level, the number of worked hours, the labor productivity, and the unemployment rates.

For Cluster 2, we identified four strong positive correlations between the active persons with primary education (ACTIV02) and their employment rate (EMP02) of 0.97, and the employment rate of persons with secondary education, respectively (EMP34) of 0.87, between the active persons with secondary education (ACTIV34) and their employment rate (EMP34) of 0.96, as well as between the active persons with tertiary education (ACTIV58) and their employment rate (EMP58) of 0.96, all of them statistically significant.

Another major positive correlation was found between the financial aid given to pupils and students and the employment rate of persons with primary education (EMP02) of 0.92, which was also statistically significant. Among these results we also discovered a significant negative correlation between the young unemployed persons with no education and the employment rate of persons with secondary education.

As for Cluster 3, we found a strong positive correlation between the active persons with primary education and their employment rate of 0.96, which was statistically significant. Generally, for the three clusters, the most significant correlations between the selected variables were found in cluster 2 where the level of investment in education accounted for a significant percentage of the total public expenditure on education. Regarding the other two clusters, although there were both statistically significant positive and negative correlations, they had a lower intensity.

The results obtained following the use of the OLS method in determining the type of influences existing between education and labor market for each of the three clusters, are shown in

Table 7. OLS regression results for Cluster 1. Dependent and Independent variables are defined in Table 2.

	Dependent Variables	COMP	HRW	PROD	UNEMP02	UNEMP34	UNEMP58	EMP02	EMP34	EMP58
Independent Variables
ACTIV02		0.4403 ***	−2.7945 ***	0.1244	−0.6685 ***	−0.1446 ***	0.0285	1.0759 ***	0.1004 ***	−0.0256
ACTIV34		0.3513 ***	3.2273	−1.4914 ***	0.0443	0.1552	0.0909 *	0.0013	0.8273 ***	−0.0584
ACTIV58		−0.4019 ***	25.8221 ***	2.3062 ***	0.1376	−0.0717	−0.0983 **	0.0070	0.1087 *	1.0699 ***
ELET		−0.5230 ***	−9.7249 ***	−0.8673 ***	−0.1511	−0.3368 ***	−0.1953 ***	0.0885 **	0.1996 ***	0.1481 ***
FAD		0.0507	−6.1953 ***	0.6194 **	0.0595	−0.3265 ***	−0.1466 ***	−0.0239	0.2356 ***	0.1220 ***
PARTET		−0.1841	19.7965 ***	0.4931	1.4490 ***	0.6399 ***	0.2522 **	−0.5895 ***	−0.4353 ***	−0.2075 ***
AT58		0.2598 ***	−9.1320 ***	−0.0098	0.0390	0.0164	0.0493 **	−0.0036	−0.0257	−0.0475 ***
YOUNG		−0.0059	−0.9958	0.1430	1.2743 ***	0.9018 ***	0.4399 ***	−0.4542 ***	−0.5924 ***	−0.3461 ***
R2		0.69	0.65	0.19	0.62	0.60	0.58	0.97	0.75	0.89
Adjusted R2		0.67	0.63	0.16	0.61	0.58	0.57	0.97	0.74	0.88

*, **, and *** represents the significance at 10, 5 or 1% level. Source: calculations made by the authors based on Eviews results.

,

Table 8. OLS regression results for Cluster 2. Dependent and Independent variables are defined in Table 2.

	Dependent Variables	COMP	HRW	PROD	UNEMP02	UNEMP34	UNEMP58	EMP02	EMP34	EMP58
Independent Variables
ACTIV02		−0.3742 ***	−6.3319 ***	−0.9759 ***	0.0396	−0.0591	−0.0367	0.7988 ***	0.0153	0.0184
ACTIV34		−0.7205 ***	11.275 ***	−0.0795	0.3445 **	0.1612	0.0365	−0.1882 **	0.8083 ***	−0.0336
ACTIV58		0.4419 **	17.974 ***	1.2193 ***	−0.1599	0.0185	−0.0131	0.1956 **	0.0592	1.0163 ***
ELET		1.2820 ***	−11.842 ***	1.1102 ***	−0.5808 **	−0.1034	−0.0456	0.3320 ***	0.1024	0.0480
FAD		0.8631 ***	−14.100 ***	0.2187	−0.6507 ***	−0.2621 ***	0.0385	0.3392 ***	0.2108 ***	−0.0282
PARTET		0.1822 ***	−1.9876 **	0.4766 ***	0.0509	0.0267	0.0355 **	−0.0026	−0.0115	−0.0254 *
AT58		1.0103 ***	−3.3871 **	0.7585 ***	0.2119 **	−0.0615	0.0550 *	−0.1046 **	0.0490	−0.0412 *
YOUNG		−0.3605 *	−0.6147	−0.8455 ***	0.2401	0.2000 *	0.1415 **	−0.1676 **	−0.1538 *	−0.1193 ***
R2		0.84	0.87	0.74	0.77	0.55	0.26	0.98	0.96	0.94
Adjusted R2		0.82	0.86	0.71	0.74	0.50	0.19	0.98	0.96	0.94

*, **, and *** represents the significance at 10, 5 or 1% level. Source: calculations made by the authors based on Eviews results.

and

Table 9. OLS regression results for Cluster 3. Dependent and Independent variables are defined in Table 2.

	Dependent Variables	COMP	HRW	PROD	UNEMP02	UNEMP34	UNEMP58	EMP02	EMP34	EMP58
Independent Variables
ACTIV02		0.3233 ***	−4.8194 ***	0.2573 ***	−0.4701 ***	0.1965 ***	0.2097 ***	0.9582 ***	−0.1380 ***	−0.1728 ***
ACTIV34		0.1554	−8.4046 ***	0.0572	0.8665 ***	−0.0597	−0.0543	−0.2452 ***	0.9610 ***	0.0485
ACTIV58		−0.2125	30.8758 ***	1.0025 ***	−0.4034 ***	−0.0473	−0.0498	0.2190 ***	0.0961 **	1.0379 ***
ELET		−0.1781 **	3.1599 **	−0.4262 ***	−0.0883	−0.2330 ***	−0.1739 ***	0.0562 **	0.1546 ***	0.1333 ***
FAD		0.0521	−8.5293 **	−0.3045	−0.3625 **	−0.2202 **	−0.2709 ***	0.1552 **	0.1771 ***	0.2224 ***
PARTET		0.5910 ***	−6.7987 **	−0.1804	−0.3465 ***	−0.1828 ***	−0.1423 ***	0.0716	0.1170 ***	0.1111 ***
AT58		0.2990 ***	−3.0894 ***	0.1894 ***	0.2046 ***	0.1356 ***	0.1059 ***	−0.1017 ***	−0.0873 ***	−0.0855 ***
YOUNG		−0.0409	12.317 ***	0.2880 **	0.7865 ***	0.9340 ***	0.5419 ***	−0.3895 ***	−0.6136 ***	−0.4305 ***
R2		0.49	0.57	−0.04	0.57	0.66	0.71	0.95	0.83	0.84
Adjusted R2		0.48	0.56	−0.08	0.56	0.65	0.70	0.95	0.82	0.83

**, and *** represents the significance at 5 or 1% level. Source: calculations made by the authors based on Eviews results.

.

From Table 7, we discuss only those variables whose influence was statistically significant (marked with *, ** or *** for probabilities of 1, 5 or 10%). Consequently, we see positive influences on the level of compensation for employees (COMP) because of the increase by a unit in the percentage of active population with primary and secondary education (ACTIV02 and ACTIV34), and of persons with tertiary education (AT58). On the other hand, we have negative influences coming from the active population with tertiary education (ACTIV58) and from persons aged between 18 and 24 who gave up school at an early age (ELET). Regarding the number of hours worked per employee (HRW), we identified both positive influences coming from the active persons with secondary and tertiary education and as a result of the increase in the rate of participation in other vocational training courses (PARTET), but also negative influences generated by the increase by a unit in ACTIV02, in the rate of persons who dropped out of school at an early age, in the rate of higher education graduates (AT58), but also due to the rise in the financial aid given by the state to pupils and students (FAD). The real labor productivity per employee (PROD) will go up only under the influence of the positive change of two of the selected variables (ACTIV58 and FAD) and down as a result in the increase by a unit in ACTIV34 and ELET.

Concerning the evolution of the unemployment rates, statistically significant influences were present at the level of each of the three types of unemployment rates. Hence, we identified positive influences following the increase by a unit in the variables ACTIV34, PARTET, AT58 and YOUNG, but also negative influences generated by ACTIV02, ACTIV58, ELET and FAD. The employment rates were influenced by the same indicators but in the opposite direction of the influence generated on the unemployment rates, the only exception being the variable ACTIV34, which even in this case had a positive influence.

It is important to also analyze the values of R² and adjusted R², which indicate the extent the selected variables within each model explain the change in the dependent variable. As shown in Table 7, employment rates (especially EMP02) were best explained by the selected independent variables in percentage terms of over 75%. The rest of the dependent variables were explained in percentage terms of approximately 60–65%, except for the real labor productivity per employee, which dependent on the evolution of the variables selected in proportion of only 19%.

This shows a negative impact of educational dropout on the employees’ level of compensation, on the number of hours worked by each employee and, finally, on labor productivity. At the same time, having graduated with a Bachelor’s degree positively influenced the level of compensation, consequently, graduates with a Bachelor’s degree were a lot better paid than those with an average education [4].

The state’s intervention in education by giving financial aid to pupils and students generated a reduction in the number of worked hours. This fact was explained by the option of many employees and employers to sign part-time contracts, especially so for students. We also observed that people who drop out at an early stage, generally do so because they want to get a job. The regression results show the positive impact on employment rates and the opposite effect on unemployment rates of people who received a primary and secondary education. On the other hand, both the increase in the number of participants to vocational training courses and the rise in the number of young people who were neither employed nor enrolled in the educational system had a negative impact on the employment rates, and consequently they push up the unemployment rates in these countries. The explanation for the effects of participating in vocational training courses on these rates resides in the fact that to gain new knowledge and skills people leave their jobs during the training period.

Consequently, although in these countries the objective of sustainable development Quality education does not seem to be targeted, given the low levels of investment in education, positive effects are still possible in these conditions through measures to limit school dropouts. Simultaneously, through measures to encourage tertiary education, progress can be made in relation to the sustainable development objective 8—Decent work and economic growth—by improving the productivity of workers, by orienting towards productive activities, towards entrepreneurship.

The results in Table 8 show that most of the influences are different in cluster 2, where investments in education account for a significant share in the public expenditure, compared with the 1st cluster, where we find the lowest percentage points of investments in education.

As a result, we can perceive the importance given to education through the positive influences generated by the rise in the percentage of active population with tertiary education (ACTIV58) and in the total percentage of higher education graduates (AT58), and last but not least, by the increase in the rate of participation in training courses (PARTET). The financial aid on behalf of the public institutions given to pupils and students also exerts a positive influence in the rise in the employees’ compensation level.

The only explanatory variables which negatively influence the compensation of employees were the active persons with primary and secondary education and the young unemployed persons aged between 18 and 24 who were not enrolled in any type of educational program. The average number of hours worked per employee (HRW) increased in line with the increase in the percentage of active persons with secondary and tertiary education and dropped in line with the rise by a unit in any of the other variable.

Another thing that significantly influences the labor productivity is the level of education of the population, as proven by the negative impact generated by the increase in the percentage of active population with primary education and the drop in percentage terms of young people (18–24 years old) who were unemployed and were not enrolled in any form of educational program. On the other hand, there was a major positive impact of having graduated from higher education and other vocational training programs (ACTIV58 and AT58) on the real labor productivity.

The importance of high investments in education was confirmed within our equation through the impact of the rising financial aid given to pupils and students (FAD) on the falling unemployment rates of persons with primary and secondary education. Another thing observed in the case of Cluster 2 was the positive impact of both the variable FAD on the increasing employment rates of persons with primary and secondary education but also of the variable ACTIV58 on the rising employment rates of the population with primary and tertiary education.

The values of R² and adjusted R² indicate the relevance of the model mainly in the case of the employment rates where values were over 0.94 for these indicators. Consequently, we considered that 94% of the change in the variables EMP02, EMP34 and EMP58 was explained by the independent variables included in the proposed model. We also noticed an improvement of indicators for the explanatory model of COMP and HRW of over 84% and for the explanatory model of the evolution of PROD of over 74%.

For the countries with significant investments in education and which were grouped into cluster 2, we concluded that graduating from tertiary education and participating in vocational training programs led to a better compensation of employees, just like it was shown by Akay et al. and Caglayan-Akay and Komuryakan [52,53], and to a higher labor productivity, hence confirming other past studies [54]. As in the case of the states in cluster 1, the financial aid given by the state for pupils and students reduced the number of worked hours, cut the unemployment rate for people with primary and secondary education, and finally, it raised the employment rate for people with tertiary education. Noticeable for cluster 2 was the fact that the variables selected do not significantly influence either the unemployment rates nor the employment rates of people with higher education, but only those of the people with primary education, which highlights the importance given by employers to degrees and qualifications of potential employees.

We also identified positive influences of the active population with tertiary education on the employment rates but also negative ones on the unemployment rate. Simultaneously, we discovered negative influences on the active population with primary and secondary education of the employment rates and positive ones of the unemployment rate, which is in accordance with the results obtained by other researchers [24,41].

In the case of the countries that formed cluster 2 in the results of the econometric analysis, there was a trend towards creating jobs for people with tertiary education, improving working conditions, income from work, giving greater importance to education, both through the prism of investments in education, and by the importance given to diplomas and professional training, so that these countries were much more oriented towards achieving the sustainable development goals related to education and work.

The results in Table 9 indicate positive influences of the participate rate in education and vocational training (PARTET) and of graduates with tertiary education (AT58). Dropping out at an early stage led to a decrease in the employees’ compensation level. Regarding the number of hours worked per employee, we noticed a positive influence of the variable ACTIV58 and of the explanatory variables ELET and YOUNG, unlike the negative influences indicated in the case of the other two clusters. Regarding real labor productivity, there was a single significant negative influence of the variable ELET; the remaining variables with statistically significant results had a positive influence.

The unemployment rates were negatively influenced by most variables apart from the increase in the number of higher education graduates and of young persons who were neither employed not enrolled in a vocational training program. The employment rates were influenced by the same indicators but opposite to the influence exerted on the unemployment rates.

The proposed model explained the evolution of the employment rates in percentage terms of over 83%, the evolution of unemployment rates and the number of hours worked per employee in percentage terms of over 57% and in smaller proportions the other dependent variables according to the values of R² and adjusted R².

The regression results for cluster 3, consisting of the countries with average investments in education, show that dropping out of the educational system at an early stage had a bad influence on both the employees’ remuneration level and the real labor productivity, and on the contrary, participating in education and vocational training programs made both the employees’ remuneration and their real productivity go up. Another aspect that can be emphasized refers to the major impact of higher education graduates on the rising unemployment rates and on the falling employment rates, a conclusion reached also by other studies [9]. This fact can also be explained by the migration of those who are university-educated towards more developed countries where specialists in all fields are much better paid than in the states in which they acquired their qualification. Another explanation could be the one given by the theory of appropriate jobs, in that students will get a full picture of their own productive capacities only after they have started working. Hence, integrating graduates into the labor market can imply searching for a suitable job which leads to changing jobs many times with adverse effects on both the employment and unemployment rates. However, by analyzing the impact of the total active population with tertiary education on the employment rates, we observe significant positive influences that confirm other studies’ results [27,28,29].

Moreover, in the case of these countries, to ensure sustainable development, measures are needed to reduce school dropout and encourage participation in tertiary education to ensure a better quality of life, by increasing incomes and productivity. Measures are also needed to limit the departure of higher education graduates from these countries in favor of others that offer better remuneration by correlating the salary level with those in other countries, and by offering other advantages (housing, access to a series of services in advantageous conditions, etc.).

Moreover, we shall identify the possible causality relationships between education and the labor market through the variables analyzed previously with the help of the causality test, Granger. The correlation does not necessarily imply the existence of causality. The Granger [55] method to find out if x causes y determines how much of the actual y can be explained through the past values of x and then it discovers if adding the later values of x can improve the explanation. They say that y is caused Granger by x if x helps to predict la y or equivalently, if the coefficients of late x-s are statistically significant. We must also take into account the fact that the bidirectional causality is often encountered: x Granger causes y and y Granger causes x. It is also important to mention that the assertion “x Granger causes y” does not imply that y is the effect or result of x. The Granger causality measures the precedence and content of information but does not indicate through itself the causality in the more common use of the term.

To identify the causality relationships between education and the labor market we used the bivariate regressions:

$$y_t={\alpha }_0+{\alpha }_1{y}_{t-1}+\dots +{\alpha }_l{y}_{t-l}+{\beta }_1{x}_{t-1}+\dots +{\beta }_l{x}_{-l}+{\epsilon }_{t }$$

(6)

$$x_t={\alpha }_0+{\alpha }_1{x}_{t-1}+\dots +{\alpha }_l{x}_{t-l}+{\beta }_1{y}_{t-1}+\dots +{\beta }_l{y}_{-l}+u_{t }$$

(7)

for all possible pairs of series (x, y) from the group. The F statistics reported are the Wald statistics for the common hypothesis:

$${\beta }_1={\beta }_2=\dots ={\beta }_l=0$$

(8)

For each equation. The null hypothesis is that x does not Granger-cause y in the first regression and that y does not Granger-cause x in the second regression.

We show the results obtained following the causality Granger test in Figure 2, Figure 3 and Figure 4 in which we find causality directions from the labor market to education and from education to the labor market for each of the three formed clusters.

In Figure 2, we can observe three causality relationships from the employees’ compensation level to education, one in Cluster 2 and two in Cluster 3. In the opposite direction, we observe nine causality directions from education to the employees’ compensation level, but only as it concerns clusters 1 and 3. We can state that the values from the analyzed time frame of the variables ACTIV02, ACTIV34, ACTIV58 and ELET explain the actual value of the variable COMP in the case of cluster 3 while the variables ACTIV34, ACTIV58, PARTET, AT58 and YOUNG explain the actual value of COMP variable in the case of the 1st Cluster.

Concerning the causality links from HRW to education, we identified six such relationships, the majority belonging to Cluster 2, to PARTET, AT58, YOUNG and ELET. In the opposite direction, from education to the number of hours worked per employee, we identified five causality relationships, three in the case of cluster 3 from ACTIV34, FAD and YOUNG, and two regarding cluster 1 (ELET and YOUNG).

The causality relationships from real labor productivity to education show that they exceeded the number of those from education to labor productivity. Regarding cluster 1, we observe the influence of labor productivity on the percentage of active people with primary education, of graduates with tertiary education but also of unemployed youngsters who were not enrolled in either an educational or vocational training program. The evolution of real labor productivity explained the actual value of the variables ACTIV34 and PARTET for cluster 2, while for cluster 3, it explained the actual value of the variables PARTET and YOUNG. Only three relationships of causality were identified from education to labor productivity, that is, from ACTIV34 and AT58 for cluster 2, and from ACTIV02 for cluster 3.

We came across many causality relationships between unemployment rates and the selected indicators concerning education, but the most numerous ones were those from unemployment rates to education, especially in the case of clusters 2 and 3 where the relationships were approximately identical. Figure 3 shows a series of bidirectional relationships: between UNEMP02 and ACTIV58 (cluster 1 and cluster 3), UNEMP34 and ACTIV02 and FAD (cluster 2), UNEMP34 and ACTIV58 (cluster 3), UNEMP58 and ACTIV34 and YOUNG (cluster 1), UNEMP58 and YOUNG (cluster 3).

The causality relationships from education to unemployment rates were determined mainly by indicators such as the percentage of active population with tertiary education, the percentage of persons who dropped out of school at an early stage, and the unemployed youngsters who were not enrolled in any type of educational program. Notably, in the case of cluster 2, there was no causality link between the selected indicators concerning education and unemployment rates of persons with primary education.

In Figure 4, just like in the previous case, most causality relationships came from the employment rates to education and with several bidirectional relationships: between EMP34 and ACTIV34 (cluster 1), between EMP34 and YOUNG (cluster 3), and between EMP58 and YOUNG (cluster 3). We can underline the fact that the smallest number of causality relationships between employment rates and education were found in cluster 2, which is the one with the highest educational investment level. We also noticed that the number of causality relationships became higher in line with the type of employment rates. If only eight causality relationships could be found at the level of employment rates of people with primary education, then regarding the employment rates of people with tertiary education, we arrive at a number of 14 causality relationships.

5. Conclusions

The aim of the research was to investigate the types of influences existing between education and the labor market using a cluster analysis based on different levels of investments in the educational system. Three indicators were used to cluster the 27 EU member states. Two of the indicators referred to investments in education, namely: General government expenditure on education as a percentage of GDP (GGE) and public expenditure on education as a percentage of GDP (PEE). We also used the real GDP per capita expressed as a ratio between the real GDP and the average population of a certain year. As the reference year for the three variables, we opted for the year 2019 given that it was the last year unaffected by the pandemic crisis. We thought that it was important that the indicators we used in defining clusters should belong to times with no economic turmoil. Three clusters were formed composed of the following: four member states (cluster 2) with big investments in education, that is, more than 6–7% of the GDP; 13 states (cluster 3) with an average level of educational investment; and 10 EU member states (cluster 1) with a small level of educational investment.

This division into clusters was necessary to continue the econometric analysis for each cluster separately because we considered that results will be clearer if the analysis refers to countries with similar characteristics, as was the case in other studies [29,56]. The Pearson correlation matrix showed that the majority of the existing, strong and positive correlations between the labor market and education took place between: active persons who received a primary education and the employment rate of people with primary education for each of the clusters; between the active population who got a higher education and the employment rate of people with higher education for both countries with the lowest and the highest investments in education;, and between the financial aid given by the state for pupils and students and the employment rate of the population with a primary education. We also identified some negative correlations, the most significant one being between the young unemployed who were not registered under any form of education and the employment rate of people who got a secondary education. This could be explained mainly by the migration abroad of seasonal workers of the labor force with a secondary education.

The empirical results show that there was a significant negative impact of educational dropout on the level of employee compensation, number of hours worked by each employee, and their labor productivity in the countries with the poorest level of educational investment. It the countries with significant investments in education, getting a graduate diploma and participating in vocational training programs led to a better compensation of employees and a higher employee productivity, while the financial aid given by the state for pupils and students reduced the number of hours worked, brought down unemployment amongst people with primary and secondary education, and last but not least, increased the employment rate for higher education graduates. An average level of educational investment led to negative influences between early-stages and employees’ payment level and the real labor productivity, while becoming involved in educational activities and participating in vocational training programs increased their rates of remuneration and their real productivity. The significant impact of higher education graduates on both increasing unemployment rates and falling employment rates was also noticed, as has been pointed out in other studies.

Limitations and Prospects for Further Study

Nevertheless, this study has its own limits, such as the lack of data for all countries for several indicators characterizing both education and the functioning of the educational system, which could probably lead to even more illustrative results. Consequently, the future developments of the study could include indicators linked with social protection and results of the educational system, but only if all countries will supply a lot more available data to the public.

Taking into account the war crises near to the European Union, future research will focus on this topic, which surely will affect the labor market and the entire economy. It is obvious that migration will have negative effects on the economy of Ukraine, such as brain drain, and a decrease of high-skilled specialists. The biggest problem for this country can be losing the young generation, mostly students, who do not have strong family connections and usually are more open to changes. According to a survey conducted by Work.ua, 59% of Ukrainian refugees were looking for a job in the country where they were staying temporarily.

Furthermore, remittances from abroad increased to the pre-war level, which means that people started to work and send money home to family members. This is good for consumption and positively influences the exchange rate, which dropped recently. Moreover, Ukrainian migrants are spending their savings abroad, negatively influencing the resources of the National Bank of Ukraine. It is of course important that people return, but if they cannot find work in Ukraine, it can be an option for them to work in Europe.