The Hard Worker, the Hard Earner, the Young and the Educated: Empirical Study on Economic Growth across 11 CEE Countries

Abstract

Economic growth is an important metric for the sustainable development of any region or country. Central and Eastern Europe members of the European Union are important players of the single market, which implements regional policies to mitigate socio-economic differences between its newer and established members. The present study examines the factors that shape the phenomenon of economic growth across 62 NUTS 2 regions from 11 countries in Central and Eastern Europe during the period 2011–2020. The study investigates determinants related to education level, involvement of young people in the labor market, household net income, high-speed internet facilities and overall hours spent at work during a year. Three panel data models estimated with first-differenced generalized method of moments showed that regional economic growth was significantly influenced mainly by income, the rate of young employees and educational attainment level. Relevant insights and policy implications for regions in CEE countries are addressed.

1. Introduction

The European Union (EU), currently numbering 27 states, promotes regional policies that aim at reducing socio-economic disparities between its members and at speeding up the structural convergence of regions. The rationality behind regional policies is straightforward: once socio-economic gaps are bridged, the EU community block can strengthen its single market in relation to other regional trade blocks (e.g., United States-Mexico-Canada Agreement, Association of Southeast Asian Nations, Common Market of Eastern and Southern Africa, Southern Common Market) and increase the competitiveness of goods and services produced in its realm.

Considering the structure and diversity of EU members, relevant comparisons and the tracking of macroeconomic indicators could be achieved based on a reliable system. Therefore, the Nomenclature of Territorial Units for Statistics (NUTS) was introduced in the 1970s by Eurostat for regional statistics purposes and sensible comparisons between members of the European community [1]. Nowadays, since it is grounded on benchmark population ranges, the NUTS system comprises three levels: NUTS 1 divides each country into macro-regions; NUTS 2 divides each country into development regions; and NUTS 3 regions correspond to existing national administrative divisions, which may be called “counties” (Hungary, Romania), “states” (Germany), “departments” (France) or “provinces” (Belgium, Bulgaria, Italy, Spain).

Empirical analyses in this study were conducted on a sample of 62 NUTS 2 regional divisions from 11 Central and Eastern Europe countries, which belong to the European Union: Bulgaria, Croatia, Czech Republic, Estonia, Hungary, Latvia, Lithuania, Poland, Romania, Slovenia and Slovakia. The time span considered for the analyses was the decade 2011–2020.

The novelty of this study is that it scrutinizes the phenomenon of economic growth in the long run for CEE regions from the standpoint of economic, social and infrastructure determinants. In this sense, without sufficient financial resources at hand, employment opportunities for young people and other age groups, equal educational opportunities across regions and proper internet facilities for daily and work activities, regional economic growth would stall or even relapse. The message of this empirical research stems from the very title. Namely, economic growth from Central and Eastern Europe regions is mainly driven by working hours, income obtained by private households, young people’s position in the labor market and education [2].

For this study, the phenomenon of economic growth was captured by the following indicators of macroeconomic performance: regional gross domestic product, regional gross value added and gross fixed capital formation. The predictors that drove changes in the evolution of economic growth were the following: educational attainment level, employment rate of young people, internet broadband access, income of private households and employment—thousand hours worked per year.

As thoroughly motivated in the manuscript, the choice of estimating results based on first-differenced generalized method of moments (GMM) was grounded on its multiple benefits that secure reliable and unbiased econometric results for panel data. In addition, multicollinearity was ruled out with the help of variance inflation factors, besides the routine pairwise correlation analysis. The lack of serial correlation in the idiosyncratic error term was supported by the Arellano–Bond AR(1) and AR(2) tests.

This research study emphasizes important factors that boost the economic performance of EU members belonging to the CEE group. Such insights could assist other CEE countries that aim to join the EU community block or are willing to enhance sustainable economic growth in the upcoming years.

The remainder of the paper is as follows. The section titled Literature Review browses the most relevant studies on economic growth, with a particular focus on regional comparisons. The section called Materials and Method provides details on the country sample, period of analysis, set of predictors and outcomes, research hypotheses and general format of the econometric models, choice of statistical estimation method. The section titled Results reports the empirical outcomes estimated with panel data modeling. The last section called Discussion and Conclusions presents relevant insights on the most notable empirical results and advances policy implications.

2. Literature Review

The next paragraphs highlight important empirical results on the topic of economic growth in the case of CEE countries and development regions. For, as the Nobel laureate in Economic Sciences Amartya Sen used to state, “economic growth without investment in human development is unsustainable—and unethical”. On the same topic, Phil Bredesen, American politician and former governor of the state of Tennessee, strongly affirmed the following: “Continuing economic growth requires both recruitment of new companies and expansion of existing businesses” [3].

The literature on economic growth captures the influence of both economic and non-economic factors driving this phenomenon [4,5,6]. If the impact of economic determinants is straightforward and generally visible in a shorter amount of time, social determinants become relevant across extended periods of time. For a case in point, as people become more educated, they amass more skills that enable them to have better jobs, open and develop businesses, which all contribute to national and regional economic growth.

Njindan [7] examined the degree to which trade openness fueled economic growth across 17 CEE countries during the period 1994–2017. The author proxied a country’s trade openness by trade share and interaction with the world, while economic growth was proxied by GDP per capita. Fixed-effects models indicated the existence of a positive link between trade openness and economic growth.

Batrancea et al. [8] examined the phenomenon of economic growth across 212 NUTS 2 regions for the EU-28 country sample during the period 2001–2020 from the perspective of wellbeing-related infrastructure. The study targeted to identify the most notable determinants of economic growth for Western regions in comparison to Central and Eastern Europe regions. Estimated empirical results showed that the economic growth of Western regions was impacted by factors such as disposable household income, housing indicator, mobility among regions and labor force aged 15–64. At the other end, the level of economic growth for Central and Eastern Europe regions was driven by factors such as air pollution, internet broadband access and housing facilities.

Using econometric models estimated with random effects and generalized method of moments, Afonso and Blanco-Arana [9] studied the degree to which economic growth (proxied by GDP per capita) was influenced by economic and financial development. The period of analysis ranged from 1990 to 2016, and the sample included developed countries belonging to the European Union and the Organization for Economic Co-operation and Development. According to empirical estimations, a solid increase in GDP per capita would be generated by more domestic credit, higher market capitalization and domestic shares turnover. In addition, economic growth would evolve under the impact of inflation, unemployment and investments in education.

As might be expected, the pace of economic growth differs across countries, regions or continents. In this context, a joint research project conducted by the NGO Friedrich-Ebert-Stiftung and the Vienna Institute for International Economic Studies [10] (p. 56) assessed the specificity of the growth model in the CEE region and emphasized six priorities for the countries in question: (a) have fewer debates between local and EU authorities concerning macroeconomic policies; (b) shift from functional specialization to profit-oriented activities; (c) capitalize the opportunities of the digital revolution; (d) fully support green transition; (e) automatize low-wage jobs; and (f) change tax systems and expand the role of public authorities.

Kvedaras [11] focused on explaining differences in economic growth between CEE countries during the period 1995–2004 by means of the Balance-of-Payments constrained growth model. According to his results, economic growth in the region was substantially higher for countries reporting a low-income elasticity for imports.

Schadler et al. [12] conducted a comprehensive analysis on the long-term growth potential and income conversion of eight Central and Eastern Europe countries that joined the EU in 2004 (i.e., Czech Republic, Estonia, Hungary, Latvia, Lithuania, Poland, Slovakia and Slovenia). According to the authors, economic growth in these CEE states was favored by aspects such as EU transfers and substantial foreign savings. Moreover, some of the key ingredients to secure constant economic growth would be institutional development and the strength of public authorities’ actions.

Marciniak, Novak and Purta [13] examined ten CEE countries for the period 1996–2017 and concluded that the socio-economic gap between CEE and Western EU members was notably reduced: while the strongest EU economies reported increases by 27%, CEE economies reported a rise of 114%. Interestingly enough, the authors noted that CEE states “are uniquely positioned to capture the digital opportunity” and that digitization could be the catalyst for boosting growth in this area.

3. Materials and Method

For the purpose of the study, the selected outcome variables that capture the phenomenon of economic growth were the following macroeconomic indicators: regional gross domestic product (RGDP), regional gross value added (RGVAD) and gross fixed capital formation (GFCF). Regional gross domestic product (RGDP) accounts for the value of all goods and services that are produced in a region across all economic branches and are addressed to final consumers. Regional gross domestic value (RGVAD) includes the regional gross domestic product without the intermediate consumption [14]. Gross fixed capital formation (GFCF) comprises the value of investments made by resident manufacturers in fixed assets, after subtracting disposals. Either tangible or intangible, fixed assets are constantly used by economic agents during the production process, for a minimum of one fiscal year. Depending on the jurisdiction, fixed assets must exceed a specific threshold amount to be included in this category of assets. As the literature posits, investments in fixed assets have a strong impact on economic growth [15,16,17] because they generally enhance manufacturing, distribution or sales facilities, therefore helping economic agents to reach more consumers and increase income revenues by trading goods and services. For that matter, capital goods investments are at the forefront of enhanced productivity [18,19].

The set of predictors included the following variables: educational attainment level (EDU), employment rate of young people (EMPYOUNG), internet broadband access (INT), income of private households (INC) and employment—thousand hours worked (HOURS). Educational attainment level (EDU) expresses the percentage of people aged 20–24 years who completed less than primary, primary or lower secondary education. Employment rate of young people (EMPYOUNG) shows the percentage of young workers who are not involved in education or training. Internet broadband access (INT) designates the percentage of households in the region that benefit from high-speed internet services. Income of private households (INC) indicates the net income obtained by household residents, which is available for spending or saving. The last predictor counts the thousands of hours worked in a region during a calendar year.

The variables were retrieved from established sources such as the Eurostat database [20] and the OECD database [21].

The sample included 62 NUTS 2 regional divisions of 11 Central and Eastern Europe members of the European Union, that is, Bulgaria, Croatia, Czech Republic, Estonia, Hungary, Latvia, Lithuania, Poland, Romania, Slovenia and Slovakia. For a detailed account of the regions, see Appendix A.

The period of analysis spanned a decade from 2011 to 2020. The importance of this period stemmed from the fact that Central and Eastern Europe (CEE) countries have developed considerably after joining the EU, with Croatia being the last CEE country to join the Union in 2013. In addition, this decade has been characterized by recessions following the 2008 global financial crisis, the 2009–2010 sovereign debt crisis, economic recovery and the recent pandemic, which affected CEE countries to a lesser extent than other EU members. In this sense, CEE countries were less vulnerable to the economic downturns than the other EU counterparts because they were less exposed to the US stock exchange, did not belong to the eurozone and their banking systems were guided by stronger Basel II regulations.

Econometric models were estimated as the following:

$$Z_{nt}=b_0+b_1{X}_{1nt}+b_2{X}_{2nt}+b_3{X}_{3nt}+b_4{X}_{4nt}+b_5{X}_{5nt}+b_6{X}_{6nt}+{\delta }_n+{\theta }_t+{\epsilon }_{nt}$$

with

• $Z$ indicating the outcome variable that captures the phenomenon of economic growth;
• $X$ indicating explanatory variables, which drive the phenomenon of economic growth;
• n indicating CEE regional subdivisions;
• t indicating the period of analysis;
• ${\delta }_n$ indicating time-invariant region-specific fixed effects;
• ${\theta }_t$ indicating fixed effects that account for economic downturns;
• ${\epsilon }_{nt}$ indicating the error term.

The following research hypotheses will be tested:

Hypothesis 1:

There is a significant relationship between RGDP and educational attainment level, employment rates of young people, internet broadband access, income of private households and employment—thousand hours worked.

Hypothesis 2:

There is a significant relationship between RGVAD and educational attainment level, employment rates of young people, internet broadband access, income of private households and employment—thousand hours worked.

Hypothesis 3:

There is a significant relationship between GFCF and educational attainment level, employment rates of young people, internet broadband access, income of private households and employment—thousand hours worked.

The method selected to estimate econometric results on the CEE panel data was first-differenced generalized method of moments (GMM) for a number of benefits reported in the literature [22,23,24]: (1) it provides a better control for endogeneity issues; (2) it removes biases of “unobserved and time invariant country fixed effects”; (3) instrumental variables eliminate correlations between predictors and errors; (4) it is grounded on “minimal assumptions”; and (5) it yields consistent results. The estimated models included among predictors the dependent variable with a lag of one period, thus aiming to control for the dynamic effects of the studied phenomena.

The econometric models also entailed lagged versions of the outcomes and predictors, which served as instrumental variables. Arellano–Bond tests, namely, AR(1) and AR(2), investigated the existence of serial correlation in the idiosyncratic error term. Overall, one expects a positive impact of the selected explanatory variables on the outcome variables.

A panel dynamic least squares (DOLS) method with fixed effects was also applied as a robustness check.

4. Results

4.1. Descriptives

Various descriptive statistics were determined for the chosen independent and dependent variables, based on which variable distributions were characterized (

Table 1. Descriptive statistics.

	RGDP	RGVAD	GFCF	EDU	EMP-YOUNG	INT	INC	HOURS
Mean	9.712	109.426	8.142	12.898	76.861	73.278	9.212	13.941
Median	9.723	108.500	8.180	11.600	78.500	75.200	9.244	13.828
Maximum	11.449	144.300	9.979	29.400	96.900	96.000	10.738	15.167
Minimum	7.992	95.000	6.011	2.000	45.000	17.000	7.406	13.086
Std. dev.	0.629	7.160	0.667	5.939	9.935	12.837	0.587	0.477
Skewness	−0.115	1.208	−0.229	0.532	−0.647	−1.058	−0.382	0.397
Kurtosis	3.282	5.789	3.856	2.637	3.201	4.719	3.682	2.303
Jarque–Bera	3.303***	156.014***	21.595**	13.513**	20.389***	139.325***	24.085***	21.698***
Observations	620	275	549	257	285	450	551	466

Note: ***, ** indicate statistical significance at the levels of 1% and 5%.

).

Table 1 displays statistics related to central tendency measures and deviation, distribution shape and amplitude, and distribution normality.

Among the outcome variables, the one with the largest variation was regional gross value added, while the smallest variation was reported by regional gross domestic product. In the case of predictors, internet broadband access had the largest variation, while employment—thousand hours worked registered the smallest variation. Analyzing the distribution shape of outcome variables based on skewness values, it was concluded that two of the dependent variables were skewed to the left and one to the right. As for predictors, the majority were skewed to the left (except for educational attainment level and employment—thousand hours worked). While analyzing the kurtosis values, it can be stated that all outcome variables had leptokurtic distributions. In the case of independent variables, the majority also had leptokurtic distributions.

The Jarque–Bera test investigated whether variables were distributed normally. According to the results, all variables of interest (except for regional gross domestic product) were non-normally distributed at the 1% level.

4.2. Correlation Analysis

A correlation analysis was performed to identify potential multicollinearity problems (

Table 2. Correlation analysis.

	RGDP	RGVAD	GFCF	EDU	EMP-YOUNG	INT	HOURS	INC
RGDP	1
RGVAD	0.323*	1
GFCF	0.935***	0.393*	1
EDU	0.570**	0.153	0.533**	1
EMP YOUNG	0.494*	0.042	0.521**	0.559**	1
INT	0.501**	0.505**	0.619**	0.435*	0.373*	1
HOURS	0.784**	0.151	0.607**	0.319*	0.207	0.036	1
INC	0.970***	0.328*	0.924***	0.570**	0.514**	0.513*	0.739**	1

Note: ***, **, * indicate significance at the 1%, 5% and 10% levels.

), which is an important aspect when aiming at obtaining unbiased empirical estimations. In addition to determining pairwise correlations, variance inflation factors were computed in relation to all independent variables.

According to Table 2, the dependent variables that registered the highest pairwise significant correlation were RGDP and GFCF, while the lowest significant correlation was set between RGDP and RGVAD.

Focusing on the correlation between independent variables (because of multicollinearity concerns), it resulted that the highest correlation was registered between the predictors income of private households and employment—thousand hours worked ($r=0.74$). At the same time, the predictors educational attainment level and employment—thousand hours worked reported the lowest significant correlation ($r=0.32$).

Overall, multicollinearity concerns were dismissed because no correlation coefficients of independent variables had values above 0.9 (as suggested by the literature). These insights were strengthened by the values of the variance inflation factors (VIF), which were below the threshold of 5.

4.3. Econometric Models

This section details the econometric models estimated with first-differenced generalized method of moments (GMM) that test the research hypotheses. The results concerning the evolution of the outcome variables regional gross domestic product (RGDP), regional gross value added (RGVAD) and gross fixed capital formation (GFCF) under the impact of five relevant determinants are presented in

Table 3. Econometric models for the dependent variables RGDP, RGVAD and GFCF.

	Model RGDP (M1): $\begin{array}{ll}{\mathit{R}\mathit{G}\mathit{D}\mathit{P}}_{\mathit{n}\mathit{t}}& ={\mathit{b}}_0+{\mathit{b}}_1{\mathit{R}\mathit{D}\mathit{G}\mathit{P}(-1)}_{\mathit{n}\mathit{t}}\\ & +{\mathit{b}}_2{\mathit{E}\mathit{D}\mathit{U}}_{\mathit{n}\mathit{t}}+{\mathit{b}}_3{\mathit{E}\mathit{M}\mathit{P}\mathit{Y}\mathit{O}\mathit{U}\mathit{N}\mathit{G}}_{\mathit{n}\mathit{t}}\\ & +{\mathit{b}}_4{\mathit{I}\mathit{N}\mathit{T}}_{\mathit{n}\mathit{t}}+{\mathit{b}}_5{\mathit{I}\mathit{N}\mathit{C}}_{\mathit{n}\mathit{t}}\\ & +{\mathit{b}}_6{\mathit{H}\mathit{O}\mathit{U}\mathit{R}\mathit{S}}_{\mathit{n}\mathit{t}}+{\mathit{\delta }}_{\mathit{n}}\\ & +{\mathit{\theta }}_{\mathit{t}}+{\mathit{\epsilon }}_{\mathit{n}\mathit{t}}\end{array}$	Model RGVAD (M2): $\begin{array}{ll}\mathit{R}{\mathit{G}\mathit{V}\mathit{A}\mathit{D}}_{\mathit{n}\mathit{t}}& ={\mathit{b}}_0+{\mathit{b}}_1{\mathit{R}\mathit{G}\mathit{V}\mathit{A}\mathit{D}(-1)}_{\mathit{n}\mathit{t}}\\ & +{\mathit{b}}_2{\mathit{E}\mathit{D}\mathit{U}}_{\mathit{n}\mathit{t}}+{\mathit{b}}_3{\mathit{E}\mathit{M}\mathit{P}\mathit{Y}\mathit{O}\mathit{U}\mathit{N}\mathit{G}}_{\mathit{n}\mathit{t}}\\ & +{\mathit{b}}_4{\mathit{I}\mathit{N}\mathit{T}}_{\mathit{n}\mathit{t}}+{\mathit{b}}_5{\mathit{I}\mathit{N}\mathit{C}}_{\mathit{n}\mathit{t}}\\ & +{\mathit{b}}_6{\mathit{H}\mathit{O}\mathit{U}\mathit{R}\mathit{S}}_{\mathit{n}\mathit{t}}+{\mathit{\delta }}_{\mathit{n}}\\ & +{\mathit{\theta }}_{\mathit{t}}+{\mathit{\epsilon }}_{\mathit{n}\mathit{t}}\end{array}$	Model GFCF (M3): $\begin{array}{ll}{\mathit{G}\mathit{F}\mathit{C}\mathit{F}}_{\mathit{n}\mathit{t}}& ={\mathit{b}}_0+{\mathit{b}}_1{\mathit{G}\mathit{F}\mathit{C}\mathit{F}(-1)}_{\mathit{n}\mathit{t}}\\ & +{\mathit{b}}_2{\mathit{E}\mathit{D}\mathit{U}}_{\mathit{n}\mathit{t}}+{\mathit{b}}_3{\mathit{E}\mathit{M}\mathit{P}\mathit{Y}\mathit{O}\mathit{U}\mathit{N}\mathit{G}}_{\mathit{n}\mathit{t}}\\ & +{\mathit{b}}_4{\mathit{I}\mathit{N}\mathit{T}}_{\mathit{n}\mathit{t}}+{\mathit{b}}_5{\mathit{I}\mathit{N}\mathit{C}}_{\mathit{n}\mathit{t}}\\ & +{\mathit{b}}_6{\mathit{H}\mathit{O}\mathit{U}\mathit{R}\mathit{S}}_{\mathit{n}\mathit{t}}+{\mathit{\delta }}_{\mathit{n}}\\ & +{\mathit{\theta }}_{\mathit{t}}+{\mathit{\epsilon }}_{\mathit{n}\mathit{t}}\end{array}$
Estimation Method	GMM	GMM	GMM
Constant	-	-	-
RGDP(−1)	0.602*** (6.8582)	-	-
RGVAD(−1)	-	0.3297 (1.0149)	-
GFCF(−1)	-	-	0.3817*** (5.5433)
${EDU}_{nt}$	−0.0009 (−0.6969)	0.1711 (0.5034)	−0.0131** (−2.2004)
${EMPYOUNG}_{nt}$	0.0012** (2.2238)	0.0816 (0.6495)	0.0073** (2.8945)
${INT}_{nt}$	0.0005 (0.6160)	−0.0119 (−0.0550)	0.0029 (0.6683)
${INC}_{nt}$	0.1996** (2.1954)	26.8671* (1.8121)	0.8567*** (5.8345)
${HOURS}_{nt}$	1.1101*** (10.4826)	91.4919*** (5.6548)	1.0818*** (6.6819)
Cross-section effects	Fixed	Fixed	Fixed
J-statistic	19.1076	4.68	17.61
Prob(J-statistic)	0.5148	0.4561	0.6132
AR(1)	0.0839	0.6242	0.3849
AR(2)	0.2226	0.7259	0.2705
Instrument rank	26	11	26

Note: Robust t-statistics are shown in parentheses. Statistical significance at 10%, 5% and 1% levels is indicated via the symbols *, **, ***. The null hypothesis of heteroscedasticity was rejected by the White test.

.

The first econometric model investigated the phenomenon of regional gross domestic product across countries in Central and Eastern Europe. Regional gross domestic product with a lag of one period also counted as independent variable. The set of instrumental variables included the following: regional gross domestic product with a lag of two periods; educational attainment level with a lag of one period; employment rates of young people with a lag of one period; internet broadband access with a lag of one period; income of private households with a lag of one period; and employment—thousand hours worked with a lag of one period. Multicollinearity problems were not identified since the values of all variance inflation factors did not exceed the threshold of 10.

Based on the Arellano–Bond AR(1) and AR(2) tests, the first-differenced generalized method of moments estimator revealed the lack of first-order autocorrelation and second-order autocorrelation. Since the AR(2) test was not significant, it indicated that the GMM estimator was valid, along with its instruments. Estimations showed that three independent variables elicited significant changes in regional gross domestic product. That is, when the employment rate of young people increased by a single unit, regional gross domestic product would increase by 0.001 units. Should income of private households improve by one unit, regional GDP would augment by 0.19 units. At the same time, in case employment—thousand hours worked increased by one unit only, regional GDP would considerably increase by 1.11 units. The lagged dependent variable also reached a significant level. The other predictors did not play a relevant role in the evolution of the phenomenon. Judging by the values of the J-statistic test $(J=19.11,p=0.51)$ and the Arellano–Bond test for AR(2) $(p=0.22)$, it can be stated that changes in RGDP under the impact of the independent variables were significant. Therefore, the first research hypothesis was supported by the empirical data on Central and Eastern Europe regions.

The second econometric model investigated regional gross value added and it included the lagged version of this outcome variable. Instrumental variables numbered the following: regional gross value added with a lag of two periods; educational attainment level with a lag of one period; employment rates of young people with a lag of one period; internet broadband access with a lag of one period; income of private households with a lag of one period; and employment—thousand hours worked with a lag of one period. Like the previous model, all variance inflation factors were below the threshold of 10, thus indicating a lack of multicollinearity problems (as expected).

According to the Arellano–Bond AR(1) and AR(2) tests, no first-order or second-order correlations were identified through the first-differenced generalized method of moments. The AR(2) test was not significant; hence, it can be concluded that the GMM estimator and its instrumental variables were valid. Empirical estimations indicated that two predictors were responsible for major changes in regional gross value added. In this sense, when income of private households improved by one unit, RGVAD would substantially augment by 26.87 units. The biggest positive change was due to employment—thousand hours worked: should the predictor improve by one unit, RGVAD would follow the trend with 91.49 units. The impact of the other independent variables did not reach significance. Based on the values of the J-statistic test $(J=4.68,p=0.46)$ and the Arellano–Bond test for AR(2) $(p=0.73)$, it can be noted that the studied phenomenon changed significantly under the impact of predictors. Therefore, the second research hypothesis was supported by the empirical data on Central and Eastern Europe regions.

In terms of the third econometric model, the examined phenomenon was gross fixed capital formation. Again, the model comprised gross fixed capital formation with a lag of one period. The following predictors were accounted as instrumental variables: gross fixed capital formation with a lag of two periods; educational attainment level with a lag of one period; employment rates of young people with a lag of one period; internet broadband access with a lag of one period; income of private households with a lag of one period; and employment—thousand of hours worked with a lag of one period. Multicollinearity did not pose any problems for estimated results since variance inflation factors registered values below 10.

Empirical analyses (i.e., Arellano–Bond AR(1) and AR(2) tests) showed the lack of first-order correlations and second-order correlations. Moreover, the first-differenced GMM estimator and the set of instrumental variables were valid based on the AR(2) test $(p=0.27)$. Four independent variables had significant impact on the phenomenon, along with GFCF lagged by one period (which turned to be significant). Hence, when EDU improved by one unit, GFCF mitigated by 0.01 units. If EMPYOUNG increased by one unit, GFCF would augment by 0.007 units. Positive changes were elicited also by INC and HOURS: should INC improve by one unit, GFCF would increase by 0.86 units; should HOURS improve by one unit, the phenomenon would considerably augment by 1.08 units. The other predictor had no major role in the evolution of GFCF. The J-statistic test ($J=17.61,p=0.61)$ and the Arellano–Bond test for AR(2) $(p=0.27)$ suggested that GFCF was significantly influenced by the chosen predictors. The third research hypothesis was thus supported by the empirical regional data.

In addition, to ensure the strength of the estimates, a robustness test was applied to each of the three models. Namely, a panel dynamic least squares (DOLS) method with fixed effects was conducted (see

Table 4. Robustness check via the panel dynamic least squares (DOLS) estimator.

	Model RGDP (M4): $\begin{array}{ll}{\mathit{R}\mathit{G}\mathit{D}\mathit{P}}_{\mathit{n}\mathit{t}}& ={\mathit{b}}_0+{\mathit{b}}_1{\mathit{E}\mathit{D}\mathit{U}}_{\mathit{n}\mathit{t}}\\ & +{\mathit{b}}_2{\mathit{E}\mathit{M}\mathit{P}\mathit{Y}\mathit{O}\mathit{U}\mathit{N}\mathit{G}}_{\mathit{n}\mathit{t}}+{\mathit{b}}_3{\mathit{I}\mathit{N}\mathit{T}}_{\mathit{n}\mathit{t}}\\ & +{\mathit{b}}_4{\mathit{I}\mathit{N}\mathit{C}}_{\mathit{n}\mathit{t}}+{\mathit{b}}_5{\mathit{H}\mathit{O}\mathit{U}\mathit{R}\mathit{S}}_{\mathit{n}\mathit{t}}\\ & +{\mathit{\delta }}_{\mathit{n}}+{\mathit{\theta }}_{\mathit{t}}+{\mathit{\epsilon }}_{\mathit{n}\mathit{t}}\end{array}$	Model RGVAD (M5): $\begin{array}{ll}\mathit{R}{\mathit{G}\mathit{V}\mathit{A}\mathit{D}}_{\mathit{n}\mathit{t}}& ={\mathit{b}}_0{+\mathit{b}}_1{\mathit{E}\mathit{D}\mathit{U}}_{\mathit{n}\mathit{t}}\\ & +{\mathit{b}}_2{\mathit{E}\mathit{M}\mathit{P}\mathit{Y}\mathit{O}\mathit{U}\mathit{N}\mathit{G}}_{\mathit{n}\mathit{t}}+{\mathit{b}}_3{\mathit{I}\mathit{N}\mathit{T}}_{\mathit{n}\mathit{t}}\\ & +{\mathit{b}}_4{\mathit{I}\mathit{N}\mathit{C}}_{\mathit{n}\mathit{t}}+{\mathit{b}}_5{\mathit{H}\mathit{O}\mathit{U}\mathit{R}\mathit{S}}_{\mathit{n}\mathit{t}}\\ & +{\mathit{\delta }}_{\mathit{n}}+{\mathit{\theta }}_{\mathit{t}}+{\mathit{\epsilon }}_{\mathit{n}\mathit{t}}\end{array}$	Model GFCF (M6): $\begin{array}{ll}{\mathit{G}\mathit{F}\mathit{C}\mathit{F}}_{\mathit{n}\mathit{t}}& ={\mathit{b}}_0{+\mathit{b}}_1{\mathit{E}\mathit{D}\mathit{U}}_{\mathit{n}\mathit{t}}\\ & +{\mathit{b}}_2{\mathit{E}\mathit{M}\mathit{P}\mathit{Y}\mathit{O}\mathit{U}\mathit{N}\mathit{G}}_{\mathit{n}\mathit{t}}+{\mathit{b}}_3{\mathit{I}\mathit{N}\mathit{T}}_{\mathit{n}\mathit{t}}\\ & +{\mathit{b}}_4{\mathit{I}\mathit{N}\mathit{C}}_{\mathit{n}\mathit{t}}+{\mathit{b}}_5{\mathit{H}\mathit{O}\mathit{U}\mathit{R}\mathit{S}}_{\mathit{n}\mathit{t}}\\ & +{\mathit{\delta }}_{\mathit{n}}+{\mathit{\theta }}_{\mathit{t}}+{\mathit{\epsilon }}_{\mathit{n}\mathit{t}}\end{array}$
Estimation Method	DOLS	DOLS	DOLS
${EDU}_{nt}$	8.6011 (0.1281)	0.1199 (0.8101)	−0.0054** (−1.2391)
${EMPYOUNG}_{nt}$	0.0007** (2.0231)	0.0169 (0.2579)	−0.0006 (−0.3152)
${INT}_{nt}$	0.0006 (1.2451)	−0.1260 (−1.1623)	0.0039 (1.2112)
${INC}_{nt}$	0.4649*** (9.0175)	47.6382*** (10.6645)	1.0369*** (7.9760)
${HOURS}_{nt}$	0.5905*** (7.6100)	43.7625*** (3.2359)	0.5708*** (1.4411)
Cross-section effects	Fixed	Fixed	Fixed
R-squared	0.9997	0.8587	0.9853
Adjusted R-squared	0.9993	0.8035	0.9795
F-statistic	592.8722	9.6798	168.2956
Prob(F-statistic)	0.0000	0.0000	0.0000
Observations	191	193	191

Note: Robust t-statistics are shown in parentheses. Statistical significance at 5% and 1% levels is indicated via the symbols **, ***. The null hypothesis of heteroscedasticity was rejected by the White test.

). In this context, the White test was used to rule out any heteroscedasticity concerns.

The panel DOLS revealed that, in the case of the fourth model (M4), which accounted for 99.93% of the variance in regional GDP ($F=592.87,p>0.001$), three of the predictors had a significant impact. Namely, if employment rates of young people improved by one unit, overall RGDP would increase by 0.0006 units. At the same time, should income of private households and employment—thousand hours worked follow the same ascending trend, regional GDP would significantly improve by 0.465 units and 0.591 units, respectively.

According to the fifth model (M5), which explained 80.34% of the variance in regional gross value added ($F=9.68,p>0.001$), two factors yielded positive changes into the outcome variable. In this sense, when income of private households improved by one unit, RGVAD would substantially increase by 47.64 units. Moreover, if employment—thousand hours worked added a single unit, then RGVAD would increase by 43.76 units.

Last but not least, the sixth model (M6) explained 97.95% of the variance in gross fixed capital formation. Again, the influence of INC and HOURS remained relevant: should both increase by one unit, the overall result would improve by 1.04 units and 0.57 units, respectively.

Overall, except for one result (i.e., the nonsignificant impact of EMPYOUNG on GFCF), the panel dynamic ordinary least squares method yielded very similar results to the first-differenced generalized method of moments, considering both the direction of the predictor’s impact and the level of significance. In this context, as suggested by the robustness check, a sensible conclusion would be that the econometric results are strong and reliable.

5. Discussion and Conclusions

Central and Eastern Europe countries in general, and the CEE members of the European Union in particular, have attracted the interest of scholars, renowned companies, [25,26,27] and international organizations [28,29] when it comes to the topic of economic growth. The increasing interest is straightforward since the pace of sustainable economic growth in recent decades has been mainly driven by emerging and developing economies [30]. In addition, in-depth panel data analyses of economic growth can provide important insights for countries and/or regions aiming at intensifying their economic performance.

The present study targeted the dynamics of economic growth during the last decade (2011–2020) across 11 members of the European Union that belong to Central and Eastern Europe. The country sample was chosen because these CEE nations have registered substantial development after joining the European Union and shifted their status “from policy-talkers to agenda-settlers” [31]. At the same time, they gradually mitigated development gaps when compared to established EU members.

Empirical data from 62 regions belonging to the second level of the Nomenclature of Territorial Units for Statistics (NUTS) classification were retrieved from the public databases Eurostat and OECD.

The methodological apparatus comprised methods ranging from descriptive, pairwise correlations and multicollinearity checks with variance inflation factors, and panel data models estimated with first-differenced generalized method of moments. A robustness check was employed to test the proposed relationships via panel dynamic ordinary least squares.

Three research hypotheses were tested in relation to the phenomenon of regional economic growth, and empirical results supported these hypotheses. Economic growth was proxied by regional gross domestic product, regional gross value added and gross fixed capital formation. The set of predictors dealt with education level, the involvement of young people in the labor market, available net income per household, high-speed internet facilities and overall hours spent at work during a year.

In relation to the first research hypothesis, the strongest driver of regional economic growth measured via regional gross domestic product was income of private households, followed by the employment rate of young people. As people earn more money per household, they have more financial resources available for spending on goods and services, thus boosting the regional production of goods and services. As might be expected, over time, such GDP increases cause inflation, which can have negative effects if it is not monitored by public authorities. At the same time, regional economic growth can be boosted if more young people get involved in the labor market [32,33]. They have numerous features that benefit both startups and established companies in today’s dynamic and competitive economy: (a) they have more flexible schedules; (b) they are eager to put the necessary time, effort, creativity and enthusiasm in order to climb the corporate ladder; (c) they possess multiple digital skills; (d) they are highly knowledgeable on social media platforms, which is a crucial requirement for the digitalization process; and (e) they have lower salary requirements because of no or little work experience. Furthermore, especially in Central and Eastern Europe, it is customary for undergraduate and graduate students to work while studying with the purpose of covering tuition and living expenses (often in large cities, hosting thousands of students), gaining experience to advance their careers or even starting their own businesses.

Concerning the second research hypothesis, the phenomenon proxied by regional gross value added was considerably impacted by the number of hours worked during one year, followed by the income of private households. In this context, as the annual number of working hours increases due to a spike in the number of employees, the gross value added of economic activities created across various sectors follows the same positive trend. For a case in point, in 2020, the main economic activities that fueled the EU gross value added revolved around the following: public administration, defense, education, human health and social work; industry (without construction); wholesale and retail trade, transport, accommodation and food; real estate; information and communication; and finance and insurance [34]. Similarly, more money obtained by private households incentivize providers of goods and services in CEE regions to increase the value of their wares and, consequently, to benefit from these financial resources. It goes without saying that market trends single out the economic activities able to generate higher value added. In this context, due to the input from fintech companies, the banking sector has been witnessing the emergence of the so-called value-added services, and CEE countries are at the forefront of this trend, alongside Scandinavian states and Singapore. Hence, regional gross value added is also generated by services such as: phone payment facilities; online taxpaying; digital signature on submitted documents; real estate counselling; travel apps; taxi orders; fuel and toll gate payments; and subscription management for numerous digital accounts [35].

As for the third research hypothesis, economic growth was proxied by gross fixed capital formation. Among the predictors with the largest influence, one could mention employment—thousand hours worked, private household income and educational attainment level. As expected, a higher number of hours worked by employees would be capitalized if companies intensified investments in fixed assets (e.g., accounting and business management software, production facilities, buildings, equipment, sustainable green assembly lines, delivery vehicles, computer networks and agricultural land) that could accelerate production processes and increase their long-term efficiency. This would ultimately translate into improved and more competitive products and services at the regional level. Like the previous case of gross value added, provided that individuals gain a higher net income, companies could acquire more fixed assets to expand production lines, open new facilities in other regions, increase sales and attract more customers willing to spend their higher income. An interesting result was obtained in relation to educational attainment level: the higher the share of young people with a lower level of education (potential employees), the lower the investments of companies in fixed assets. A possible explanation for this outcome is that less skilled employees require lower capital investments as compared to high skilled workers. The latter are usually well versed in using nontangible assets, high-end technology, fintech and digital equipment, which may raise the fixed asset costs of companies. At the same time, investments in fixed assets could mitigate on the account that, especially in CEE countries, less developed regions number higher percentages of less skilled workers, lack education or training opportunities and are not attractive to potential employers [36] (p. 40). Last but not least, less skilled people could choose to migrate abroad in search of a higher income, since their countries of origin offer low payments for unskilled jobs.

The originality of this study resides in that it examines economic growth in Central and Eastern Europe regions with the help of determinants related to working time, citizens’ income levels, rate of young employees, and education. The empirical results are in line with other studies from the literature, which acknowledge the positive influence of household income [8], employment of young people [37] and working hours [38]. At the same time, OECD countries registered a different result: as the number of working hours increases, the rate of economic growth decreases [39].

The limitations of the study could be summed up in the following. The period of analysis spanned one decade starting with 2011. Upcoming studies could focus on additional decades and conduct comparative analyses between EU pre-accession and post-accession of CEE countries. Moreover, the country sample could be expanded with more states in Central and Eastern Europe, which are not members of the European Union. At the same time, other variables such as migration, population aging, natural resource depletion [40] or inflation could be included among predictors.

In conclusion, the regional economic growth of Central and Eastern Europe countries [41,42,43,44,45] is important because it provides these EU members with a solution to close the development gap when compared to Western EU members. Furthermore, insights on the particularities of CEE economies could be used to capitalize their advantages in the EU single market.