← All podcasts
Issue 1/2024 · Európai Unió és regionális fejlődés

Convergences and the Matthew effect in European Union Member States based on the DESI index

Kovács Tünde Zita, Bittner Beáta, Huzsvai László, Nábrádi András

Published: 2024 · Forum on Economics and Business, Issue 1/2024 · Pages: 38–74

„Felzárkóznak a digitálisan lemaradt EU-országok, vagy a gazdagok egyre gazdagabbak?”

Listen to the podcast

Conversation · 7 min
HU EN Podcast language: Hungarian · EN: English version not available yet
0:00-7:26

Published: 2024 · Forum on Economics and Business, Issue 1/2024 · Pages: 38–74

The full article HU EN

OJS ↗ Full issue (PDF) ↗

This English text is a machine translation of the original Hungarian article, made with a local AI model. The original is authoritative: read the original (Hungarian)

Közgazdász Fórum · Issue 1/2024 · 38–74

Convergences and the Matthew effect in European Union Member States based on the DESI index

Abstract. Since 2014, the European Commission has been monitoring the digital progress of Member States using the Digital Economy and Society Index (DESI). By utilising the DESI annual databases, we sought to determine whether convergence can be identified among the Member States. By examining the discrepancies between the indices, we suspected the presence of the so-called Matthew effect. Furthermore, we hypothesised that the Covid-19 pandemic has an impact on changes in the DESI index. The research questions were examined using bibliometric and statistical-mathematical methods. $\sigma$-convergence analyses were used to assess the temporal reduction of differences between Member States, while $\beta$-convergence analysis was used to estimate the extent of catching up to the initial level of development. To verify the Matthew effect, we performed PCA analysis with additional $\lambda$-variances, taking into account the change in real GDP per capita. Both $\sigma$- and $\beta$-convergence were confirmed for the 2016–2021 period, and we established that the halving time of the catch-up process is approximately 20 years. In the 2016–2021 period, although the Matthew effect could not be significantly proven, its trend nonetheless suggests its existence.

Keywords: digital development, digital economy, convergence, EU, DESI index

Introduction

The competitiveness of the European economy is predicated on the support of a knowledge-based economy, as well as research and development (R&D) activities. While the former is closely linked to the development and qualitative improvement of human capital, the latter presupposes the integration and advancement of information and communication technology (ICT) (Bassanini et al. 2000). This study was prepared with the professional support of the New National Excellence Programme (ÚNKP-22-3) of the Ministry of Culture and Innovation, funded from the National Research, Development and Innovation Fund. 2 PhD, Assistant Professor, Partium Christian University, Faculty of Economics and Social Sciences, Department of Economics, e-mail: [email protected]. 3 PhD, Lecturer, University of Debrecen, Faculty of Economics, Institute of Management, Non-independent Department of Business Economics and Enterprise Development, e-mail: [email protected]. 4 PhD, Associate Professor, University of Debrecen, Faculty of Economics, Institute of Statistics and Methodology, Non-independent Department of Statistics and Methodology, e-mail: [email protected].

At the EU level, the Lisbon Strategy was the first to address the conscious development of competitiveness in 2000, followed in 2010 by the EU 2020 Strategy, within which efforts in this direction were repeatedly reinforced (European Commission 2010b). A key initiative of the latter is the Digital Agenda for Europe. To ensure the rapid and timely implementation of objectives, the proposed governance framework is based on the Digital Economy and Society Index (DESI), which has been monitoring the digital progress of Member States since 2014 (Bánhidi–Dobos 2020). In 2021, the European Commission (EC) modified the structure of the DESI, narrowing it from the previous five dimensions to four. This transformation became necessary as a result of two policy initiatives that will impact the digital transformation within the European Union (EU) in the coming years: the Recovery and Resilience Facility (RRF), adopted in February 2021, and the digital compass of March of the same year, which defines Europe's digital decade. In September 2021, the EC proposed the Path to the Digital Decade draft, through which the digital transformation of the European economy and European society can be realised (European Commission 2021d).

At present, the DESI index contains the values of four main dimensions, 10 sub-dimensions, and 33 individual indicators (Table 1), and ranks EU Member States according to a weighted system measuring digital performance, monitoring their development closely; this system was developed based on the guidelines and recommendations of the Organisation for Economic Co-operation and Development (OECD) (European Commission 2021a). The assessment published in a given year contains data from the previous year, and thus always shows the results of the already concluded year. For this reason, the DESI index values for the 2016–2021 period reflect the state of affairs for 2015–2020. The data in the index are collected by the EC from the competent authorities of the Member States, as well as from various studies prepared for the Commission.

Table or figure – see the PDF version of the issue for its exact content.

Table or figure – see the PDF version of the issue for its exact content.

Internet user skills and advanced skills, and their development (2 dimensions and 7 individual indicators)

Table or figure – see the PDF version of the issue for its exact content.

Fixed and mobile broadband internet coverage and price (4 dimensions and 10 individual indicators)

3. Digital technologies Digital intensity, corporate digitalisation and integration (desi_3) e-commerce (3 dimensions and 11 individual indicators) 4. Digital public services (desi_4)

Table or figure – see the PDF version of the issue for its exact content.

Table or figure – see the PDF version of the issue for its exact content.

For each country, the value of the DESI index is determined as the weighted arithmetic mean of the individual DESI main dimensions:

DESI = desi_1*0.25 + desi_2*0.25 + desi_3*0.25 + desi_4*0.25

The aim of the present study is to provide evidence regarding the digital development of Member States, both in aggregate and through comparison between Member States, using mathematical and statistical methods. The convergence of the economic, social, and digital development of the countries that joined the EU during the Eastern European enlargements (2004, 2007, and 2013) towards that of the older Member States is highly significant, as the results of accession must manifest precisely in this approach, otherwise known as convergence. Our null hypothesis regarding convergence is as follows:

H1: Based on the change in the DESI index, we assume that there is no convergence between the Member States.

For further analysis of the DESI index, we employed correlation analysis and principal component analysis (PCA). An advantage of this procedure is that, as a result of the PCA, the data becomes more transparent and the condensed information becomes easier to interpret (de Solla Price 1967; Merton 1988). A further advantage of the analysis is that, instead of the DESI main dimensions (weights provided by the EC), we determine the contribution of variables to the DESI index along the principal component(s) (PC). Accordingly, our second null hypothesis is as follows: H2: We assume that the contribution of the components (variables) of the first dimension obtained during the PCA analysis is identical to the weights provided by the EC.

Robert K. Merton (1968, 1988) refers to the following passage from the Gospel of Matthew as the "Matthew effect", and Derek J. de Solla Price (1967) as "cumulative advantage":

"For to everyone who has, more will be given, and he will have abundance; but from him who does not have, even what he has will be taken away" (Matthew 25:29). Both made this observation while explaining disparities experienced in the recognition of scientific work, a phenomenon which colloquial language merely describes as "the rich getting richer". Today, the concept is used to describe the self-reinforcing pattern of inequalities related to economic well-being, political power, prestige, knowledge, or any other scarce or valuable resource (Perc 2014; Rigney 2010). In this study, we examined through the change in the DESI index whether the Matthew effect can be verified alongside convergence among the EU-27 Member States. Accordingly, we formulated our third null hypothesis as follows:

H3: The mean and variance of the latent variables of the DESI index during the 2016–2021 period do not confirm the Matthew effect.

The structure of the study is as follows: in the Literature Review chapter, we reviewed those scholarly sources that deal with the analysis of DESI index values. This is followed by the Materials and Methods chapter, in which we present the statistical analytical methods used to analyse the DESI index data for the 2016–2021 period: $\sigma$- and $\beta$-convergence, as well as principal component analysis (PCA). The Analyses section contains the statistical analyses themselves, and in each case, we endeavoured to justify and explain the findings revealed. In the Results chapter, we interpreted the obtained results, which we compared with the findings of previous studies and used to test our hypotheses; finally, in the Conclusions chapter, we summarised the main lessons and identified further research directions.

Literature Review

To explore the literature analysing or discussing the DESI index, as well as to examine the specific areas investigated by individual authors and the countries involved, we deemed a bibliometric analysis to be the most appropriate method. For this purpose, we employed the freely available bibliometric software VOSviewer 1.6.17 (Van Eck–Waltman 2013), which we used to examine the keywords of articles addressing the DESI index themes, thereby uncovering the connections between them, and to analyse bibliographic couplings between countries. To perform this, we used the Web of Science Core-

Collection – Clarivate Analytics database. Our searches were limited to English-language keywords, as the aforementioned database typically contains articles and studies written in English. Our searches targeted scientific journal articles and conference proceedings containing the phrase "Digital Economy and Society" or the acronym "DESI", examining their occurrence in study titles, abstracts, and author-provided keywords during the 2015–2021 period. The examined database yielded 89 direct hits and 43 related records, which may assist in a more comprehensive interpretation of the field.

In the studied papers, the software used identified 592 keywords; subsequently, similar keywords were merged using a so-called thesaurus file, leaving us to work with 568 keywords. With a set minimum occurrence rate of 5, 20 keywords reached the specified threshold. The network of connections between the relevant keywords is illustrated in Figure 1 and should be interpreted as follows: the size of the nodes indicates the frequency of occurrence, the thickness of the lines connecting the nodes indicates the frequency of the co-occurrence of individual keywords, and the distance represents the strength of the connection between individual keywords. The VOSviewer software distinguishes the clusters identified during the analysis using colours.

Table or figure – see the PDF version of the issue for its exact content.

Table or figure – see the PDF version of the issue for its exact content.

Table or figure – see the PDF version of the issue for its exact content.

Table or figure – see the PDF version of the issue for its exact content.

The primary keyword of the first cluster, indicated in red, is "ICT" – information and communication technology; it most frequently appears alongside the terms/phrases "European Union", "technologies", "education", and "digital skills", suggesting that the key to the proliferation of information and communication technology (ICT) lies in education, through which digital skills are acquired. Several authors contend that the use of ICT tools is indispensable for the development and growth of the EU economy (Fernández-Portillo et al. 2020; Jovanović et al. 2018; Stofkova–Hraskova 2017). Furthermore, as early as 2010, the EC itself has included ICTs as key elements within its digital agenda regarding EU objectives (European Commission 2010a). The lagging behind of the 54+ age group in the field of digital skills represents a serious challenge, as, in its absence, this aforementioned age group is effectively marginalised from those employed in sectors where the presence of digital technology is significant (Zdjelar et al. 2019).

Also related to this is the quasi-exclusion of women from the digital market economy, as the vast majority of ICT professionals are male, thereby further increasing the divide among those possessing digital skills (Hadad 2017; Kamberidou–Pascall 2019; Martínez-Cantos 2017). The closing of the gap for these two social groups through education is otherwise a cardinal point of the EC's Digital Compass programme, launched in 2021 with a target of 2030 (European Commission 2021b). Tkác

(2018) studied the extent to which measures concerning digital public services contribute to the implementation of the Digital Single Market (DSM) strategy adopted by the European Council in 2015. According to their analyses, as a result of the measures applied, public trust in administration has increased. If data is made visible—primarily among organisations providing public services—it would rationalise document processing, thereby saving time and costs (Jendricko–Mesaric 2019; Stofkova–Stofkova 2019).

The second cluster, indicated in green, links the area of the DESI index (which also appears as a keyword) with the digital economy and its fundamental characteristic, competitiveness (Pyroh et al. 2021). The concept of sustainable development and Industry 4.0 are also situated within this cluster (Table 2). Zdjelar et al. (2019), MacGregor Pelikánová (2018), and Esses et al. (2021) also argue that the EU economy must meet not only competitiveness but also sustainability criteria, as the notion of infinite and unlimited growth is merely a utopia (Pope Francis 2015). When formulating national strategies, an economy employing digital tools must be taken into account in order to achieve economic and social development (Jakovic et al. 2021; Laitsou et al. 2020; Rozite et al. 2019).

In her analysis of Austria's competitiveness, Čiefová (2017) concluded that it is a function of long-term conditions, meaning that shifts occur only very slowly. Following the impact of Industry 4.0 on logistical performance, Moldabekova et al. (2021) argue that the development of human capital, the integration of digital technologies, and dense, robust digital coverage are essential. The second group of studies, classified into the green cluster, aligns with the pillars of sustainability and development. While the DESI index clearly illustrates the digital progress of Member States, it focuses less on sustainability issues.

In the blue-coded cluster 3, the most frequent keyword is "digitalization", the impact of which on innovation (which also appears as a keyword) is unquestionable. Luchko (2020) and Jovanović et al. (2018) contend that the cultural habits of individual Member States are also decisive in the process of digitalisation. Utilising Hofstede's (1984) cultural dimensions, they concluded that in those states where individualism and levels of hierarchy are high, and where risk-taking and indulgence are also higher, the level of economic digitalisation is also higher (Ivanovic-Dukic et al. 2019). In contrast, the level of digitalisation is lower in traditional and future-oriented societies. Therefore, cultural specificities must be taken into account, as must the level of development of the various dimensions of the DESI index. Consequently, Rakićević et al. (2019) consider the formulation of intelligent policies for each DESI dimension to be expedient, taking into account country-specific characteristics and addressing deficiencies in a targeted manner, as development across countries varies by sector (Borowiecki et al. 2021; Giannone–Santaniello 2019; Polozova et al. 2021; Tsenzharik et al. 2020). In their analyses, Esses et al. (2021) and Bogoslov–Lungu (2020) established that the level of digitalisation in an economy and society depends on the general human development of a given country (HDI index).

The studies in the final cluster examine the extent of digitalisation adoption within individual societies and analyse the degree of impact exerted on society.

Monitoring the trends and development of the DESI index is primarily important for those Member States experiencing significant lag. For this reason, the number of relevant studies is also higher in these countries; furthermore, several states are not members of the EU. Thus, following the processing of the previously mentioned database, and this time by analysing bibliographic attachments, we can identify those countries that have enriched the relevant DESI index literature with a minimum of five studies (scientific articles or papers published in conference proceedings) (Figure 2). Out of 40 countries, 12 reached this threshold. The majority of studies originate from Romania (25), followed by Ukraine (12), and then Slovakia and Croatia (10 each).

Table or figure – see the PDF version of the issue for its exact content.

Table or figure – see the PDF version of the issue for its exact content.

Several studies conduct national-level assessments by comparing different countries or regions, based on the dimensions of the DESI index. Karnitis et al. (2019) compare the Baltic states with the EU and offer a critical assessment regarding the dimensions of the DESI index, as they argue that the primary drivers of digitalisation are data mining and modelling. Ershova, Obukhova, and Belyaeva (2020) assessed the digital economies of Russia, Ukraine, and developed nations, concluding that digital, technology-driven economies will be the engines of growth. Countries that digitalise their production and economic activities can gain immense economic advantages. Similarly, through the example of Russia, Lomakina et al. (2021) suggest that the role of economic resources undergoes a transformation in a digital society: the role of natural resources decreases, the development of renewable resources comes to the fore, and furthermore, human resources, knowledge, and entrepreneurial skills are appreciated. At the corporate level, Szőllősi, Béres, and Szűcs (2021) formulated the finding that significant improvements in profitability, competitiveness, and the reduction of negative environmental impacts can be achieved through the complex application of the most modern technologies—automation, robotics, and digitalisation—alongside the necessary expertise acceptable at a macro-economic level. Jurčević, Lulić, and Mostarac (2020) compared the digital development of the Croatian economy with other EU Member States, noting that the Croatian economy has not yet reached the desired level of competitiveness; one reason for this may be that the Republic of Croatia lags significantly behind the EU average in terms of broadband internet coverage (Ivanovic–Ambros 2016). According to Grinberga-Zalite and Hernik (2019), Poland appears better prepared for the future challenges of the digital economy in terms of human capital than Latvia, a claim they support with R&D expenditure as a percentage of gross domestic product (GDP). Latvia's digital economic and social performance lags behind that of the other Baltic states (Vidruska 2016).

Material and Methods

Calculation of $\sigma$- and $\beta$-convergence

Convergence calculations are now performed in numerous fields, such as income disparities, poverty, the development of social capital, environmental economics, etc. In the present study, I address two types of convergence.

First, $\sigma$-convergence, or relative dispersion, which shows the average relative—expressed as a percentage—deviation of the DESI index values from the mean. If the dispersion of development levels decreases over time, $\sigma$-convergence exists. Another well-known convergence indicator is $\beta$-convergence, introduced into the empirical growth literature by Barro and Sala-i-Martin (1992) and Barro, Sala-i-Martin, Blanchard, and Hall (1991). This growth indicator is based on the neoclassical growth theories of Ramsey (1928), Solow (1956), and Koopmans (1963). According to the Solow-based approach, and subsequently from the perspective of endogenous growth theories since the 1980s, national economic policies and country-specific characteristics decisively influence a country's catch-up process. Growth is accelerated through the inflow of foreign capital, which over time leads to a reduction in the differences between developed and less developed, yet catching-up, countries.

$\sigma$-convergence refers to the temporal reduction of differences between regions—in this case, Member States—whereas $\beta$-convergence focuses on identifying the potential catch-up process.

$\sigma$-convergence (relative dispersion or dispersion coefficient) is a variation coefficient that enables the comparison of populations or samples with different means and differing properties. In our case, we work with a population consisting of the DESI index values for the period 2016–2021. The dispersion coefficient shows the average relative deviation, expressed as a percentage, of the observations from the mean. If the dispersion of development levels decreases over time, we speak of $\sigma$-convergence.

The calculation of relative dispersion is contained in formula (1): $\sigma_t (\%) = \frac{x}{\bar{x}} \cdot 100$ (1), where $\sigma_t$ is the dispersion of the DESI index at time $t$ and $\bar{x}$ is the mean of the DESI index.

Based on the absolute $\beta$-convergence hypothesis, relationship (2) expresses the empirical link between the growth rate per member state and the initial level of development:

$\frac{1}{T-t_0} \sum_{t=t_0}^{T-1} \ln\left(\frac{X_{iT}}{X_{it}}\right) = \alpha + \beta \cdot \log(x_{it}) + \varepsilon_i$ (2), where $\ln\left(\frac{X_{iT}}{X_{it}}\right)$ is the average annual growth rate of the DESI index values

for member state $i$; $t_0$ denotes the starting year; $T$ the closing year; $x_{it}$ the initial development level of the DESI index

at time $t$; $\varepsilon_i$ is the error term, which is assumed to be identically and independently distributed; $\alpha$ and $\beta$ are the parameters to be estimated.

In this expression, only the initial level of development explains the rate of convergence. The strength of the effect is indicated by the sign and value of the $\beta$-coefficient. I applied ordinary least squares (OLS) univariate linear regression to estimate the $\alpha$ and $\beta$ coefficients.

However, the estimation would only be error-free in the case of a perfect relationship ($r=1$ or $r=-1$); therefore, $\varepsilon_i$ represents the estimation error. We perform the estimation of the $\alpha$ and $\beta$ coefficients using the least squares method, which is based on minimising the squared error.

The $\beta$-coefficient was considered significant only if the empirical significance level ($p$-value) was less than 5%. However, according to the concept of conditional convergence, too much information is hidden in $\alpha$ and $\varepsilon_i$, which may occasionally be justified to unpack, but in this case, we only tested for absolute convergence.

Following the estimation of the $\beta$-coefficient, we can calculate the annual speed of convergence, i.e., the constant rate of convergence (3).

$\theta = -\frac{\ln(1 + \beta \cdot (T-t_0-1))}{(T-t_0-1)}$ (3), where $\beta$ is the value of the regression coefficient; $t_0$ is the starting year, and $T$ denotes the closing year.

The value of the annual speed of convergence is used to determine the convergence half-life, i.e., the time required to cover half the distance towards full convergence within the examined area, provided the speed of convergence remains constant (Obláth & Szörfi 2008). The probable time required for a 50% reduction in the inequalities (4), assuming absolute convergence, is: $\tau = \frac{\ln 2}{\theta}$ (4), where $\theta$ is the annual speed of convergence. In the case of the long-term convergence time requirement, one must consider—assuming the recently experienced rate of catch-up persists—how much time would be needed to cover half the distance towards reaching the level of the more developed countries. This is the so-called half-life.

Determination of the correlation coefficient

The formula for the sample correlation coefficient (5) is as follows:

$r_{x,y} = \frac{\sum(x_i - \bar{x})(y_i - \bar{y})}{s_x s_y} = \frac{\sum(x_i - \bar{x})(y_i - \bar{y})}{\sqrt{\sum(x_i - \bar{x})^2 \sum(y_i - \bar{y})^2}}$

Table or figure – see the PDF version of the issue for its exact content.

where $s_{xy}$ is the sample covariance; $s_x$ and $s_y$ are the sample standard deviations. Following Karl Pearson, it is referred to as the Pearson $r$ or Pearson correlation.

The methodology of Principal Component Analysis (PCA)

PCA analysis can be traced back to Cauchy; however, in statistics, it was first

formulated by Karl Pearson (1901), who described the analysis as "the lines and planes that best fit sets of points in space". Subsequently, Hotelling (1933) also dealt with the procedure, but it only became truly well-known after the proliferation of computers, due to the complexity of the method.

The essence of the procedure is to derive a smaller number of artificial (latent), uncorrelated variables, known as principal components, from a large number of correlating variables, such that their variance is maximised. Since the principal components are uncorrelated variables, this does not necessarily mean that they are independent of one another. An advantage of the procedure is that, as a result of the PCA, the data become more transparent and the condensed information becomes easier to interpret. Principal component analysis can most closely be compared to a shadow play, where one attempts to project the shadow of a complex spatial shape onto a flat surface while losing as little as possible of the shape's characteristic properties. We first determined the correlation matrix and illustrated the more characteristic values.

PCA analysis is a multivariate system based on eigenvalue calculation, which enables an overview of the correlation system of all variables simultaneously.

The correlation matrix (R) of the variables serves as the initial data. A spectral decomposition of this matrix is performed, through which we obtain the eigenvectors and eigenvalues. Using these, the principal component variables and principal component loadings are determined. The observed variables can be grouped based on their mutual correlation and combined into an equal number of principal component variables. The principal component variables can be calculated from the standardised values of the original variables, according to relation (6).

Table or figure – see the PDF version of the issue for its exact content.

where $C_j$ = the $j$-th standardised principal component variable; $a_{ij}$ = elements of the eigenvector matrix; $Z_i$ = the column vectors of the standardised values. The eigenvector matrix is orthonormal in both rows and columns; that is, it is pairwise independent in both rows and columns, and the sum of their squares equals 1. The standardised variables are dimensionless data, which are formed by dividing the deviation of the variables from their mean by the standard deviation. The expected value of the principal component variables is zero, and their variance corresponds to $\lambda$. The principal component variables are independent of one another.

Subsequently, as many principal components as necessary to sufficiently describe the original variables are utilised. The first principal component (PC1) contains the largest proportion of the variance, which is indicated by the eigenvalue ($\lambda_1$). The first principal component is thus constructed in our model as follows:

PC1 = $a_{1desi\_1} + a_{2desi\_2} + a_{3desi\_3} + a_{4desi\_4}$

We only considered those principal components whose eigenvalue is greater than or equal to one, or which explain at least 80% of the total variance. Background variables can be identified using the principal components.

The principal component loadings express how strong the correlation is between a given principal component variable (e.g., $C_1$) and the variables under investigation. These values range from -1 to +1 and should be interpreted in the same manner as Pearson correlation coefficients. The closer their absolute value is to 1, the stronger the correlation. Their practical determination is carried out using formula (7): $pc_{ij} = a_{ij}\sqrt{\lambda_j}$ (7), where $pc_{ij}$ are the weights of the $j$-th principal component; $a_{ij}$ are the elements of the eigenvector matrix; $\lambda_j$ = eigenvalues.

The sum of the squares of the principal component loadings per row equals 1. The columns are orthogonal, and the sum of squares per column is equal to its corresponding eigenvalue.

Subsequently, we represented the positioning of the EU Member States on a two-dimensional chart within the coordinate system of the two most important principal components. The groupings were performed according to the time of accession to the Union and real GDP per capita.

For our analyses, we used version 4.0.2 of the R statistical software package.

Analyses

The $\sigma$-convergence of the DESI index

Using $\sigma$-convergence, we sought to determine whether the cohesion of the EU-27 Member States can be detected from the relative dispersion of the DESI composite indicator data during the 2016–2021 period. The data required to perform the $\sigma$-convergence calculations are contained in the DESI database (European Commission 2021). We analysed $\sigma$-convergence in three units: first, the EU-27 as a whole; then the EU-14 countries; and finally, the group of states that joined in 2004 and thereafter (Figure 3). During the 2016–2021 period, cohesion can be observed in all three units of analysis, although the rate of cohesion differs. Between 2016 and 2019, the degree of convergence was higher, ranging between 1% and 2% per annum per group. Subsequently, it falls below 1%, meaning that, for example, in the EU-14 group, the cohesion—the digital convergence of the Member States—is barely perceptible.

Table or figure – see the PDF version of the issue for its exact content.

Table or figure – see the PDF version of the issue for its exact content.

The average rate of development varies across EU Member States, ranging between 5.16% and 9.65%. Greece experienced the highest rate of development, while Latvia stands at the opposite end of the spectrum. Among the states that joined during the Eastern European enlargements, Poland shows the highest average development (9.34%), whereas among the EU-14 Member States, Luxembourg (5.98%) lags behind in terms of development. However, hasty conclusions should not be drawn from this, as the levels of development of the individual Member States also differed at $t_0$ (2016).

The $\beta$-convergence of the DESI index

During the 2016–2021 period, the EU Member States were characterised by $\beta$-convergence, namely a negative correlation between initial development levels and the average volume growth rate of the DESI index (Table 3 and Figure 4). The correlation can be considered quite tight ($R^2 = 0.5$), and from the regression equation shown in Figure 4—based on the coefficient for initial development of approximately -0.0317—it can be inferred that, regarding the EU as a whole, the average annual rate of shrinkage in the development gap was 3.45%. The regression coefficient can be considered significant, as the p-value ($3.69 \times 10^{-5}$) is smaller than $\alpha$ (0.05). Table 3. OLS regression calculation of the DESI index Observations 27

Table or figure – see the PDF version of the issue for its exact content.

Table or figure – see the PDF version of the issue for its exact content.

Table or figure – see the PDF version of the issue for its exact content.

Table or figure – see the PDF version of the issue for its exact content.

Based on this, the halving time of convergence can be determined; that is, the number of years required to complete half the journey towards full intra-regional convergence, assuming the rate of convergence remains unchanged. A constant convergence rate of 3.45% corresponds to a halving time of approximately 20 years. The data required to perform the $\beta$-convergence calculations are contained in the DESI database (European Commission 2021).

The fundamental assumption of $\beta$-convergence is that the growth rate of an indicator is higher in countries that were initially characterised by lower values (e.g. Greece, Poland, Romania, Italy, Croatia) and lower in those where the initial value was higher (e.g. Denmark, Estonia, Finland, Sweden). In the present case, this resulted in countries that are less developed from the perspective of the DESI index converging towards the higher performers. Regarding the countries that joined the EU during the Eastern European enlargements, the vast majority lie below the regression line during the period under investigation. Only Cyprus, Croatia, Malta, and Poland are located above the regression line, which suggests that only these four Member States succeeded in outperforming the convergence opportunities arising from their initial lag. Notable is Ireland's outperformance of the average DESI index volume growth rate (the distance from the regression line, i.e., the residual term of the equation for Ireland), which is 1.3 percentage points. Bulgaria and Latvia, however, significantly underperformed relative to the convergence opportunities during the period under investigation. Thus, the interpreted lag in the average DESI index volume growth rate was nearly 1.1 percentage points for Bulgaria and more than 1.8 percentage points for Latvia.

The year 2020 was challenging in several respects. As a result of Covid-19, a series of restrictive measures were introduced, which primarily prescribed the reduction of personal encounters. Consequently, the integration of digitalisation processes at social and economic levels, as well as the acceleration of these processes, became even more topical. Comparing the average DESI index values measured in 2019 and 2020, the greatest progress, expressed in percentage points, was achieved by Denmark (+8.28), Austria (+6.65), the Netherlands (+6.26), and Ireland (+6.2). On the opposite end, we find Latvia, Lithuania, Bulgaria, and Poland, where the absolute magnitude of growth was less than 2.5 percentage points.

Examining the average growth rate of the general DESI indices in the 2016–2021 period, we established that its magnitude was 6.98%. In relation to 2019 and 2020—that is, under the impact of the first two waves of the Covid pandemic—the magnitude of growth was nearly 9%. Without fetishising the impact of the pandemic alone, however, it is striking that compared to the average of previous years, a 25-percentage-point growth was experienced in a single year in the case of the average DESI index.

The change is significant; nevertheless, further investigations are required to statistically confirm the impact of Covid.

Correlation matrix

Using a correlation analysis, we examined the extent to which the four main indicators of the DESI index are interrelated. The correlation analysis was conducted for $t_0$ – 2016 (Figure 5), and was also repeated using the data published in 2021 (Figure 6). The upper triangle of the correlation matrix displays the values of the correlation coefficients and indicates the significance levels (*p<0.05; **p<0.01; ***p<0.001). The diagonal shows the distribution of the given variable, while the lower triangle partially illustrates the nature of the relationship between the pairs of variables, which we assume to be linear, as the correlation coefficient is only descriptive in such cases.

Table or figure – see the PDF version of the issue for its exact content.

Table or figure – see the PDF version of the issue for its exact content.

Table or figure – see the PDF version of the issue for its exact content.

Table or figure – see the PDF version of the issue for its exact content.

The correlation coefficients are generally significant, with positive, medium, and strong relationships existing between pairs of variables; thus, the four main indicators of the DESI index are closely interrelated. Internet access (desi_2) and the integration of digital technologies (desi_3) do not correlate with each other in t0. This may be explained by the fact that the integration of digital technology at the corporate level depends more on the mindset of corporate management than on the degree of network connectivity. The SME sector is significant within the EU: more than 22.8 million registered enterprises, representing 99% of the businesses operating within the EU territory, employ 84 million Europeans and contribute 50% of the EU's GDP (Clark 2021). However, the SME sector does not fully exploit the opportunities provided by digital technologies, partly due to its mindset and partly due to limited financing options.

Examining the correlations between variable pairs individually, the magnitude of the correlation coefficients between desi_1 and desi_4, as well as between desi_2 and desi_4, decreased slightly in 2021. In the latter case, the significance level also decreased.

This allows for the conclusion that digital infrastructure is developing on a larger scale than digital public services. The strength of the correlation increased considerably for variable pairs associated with the integration of digital technology (desi_3), specifically in its relationship with human capital (desi_1) (0.79*** -> 0.84***) and with digital public services (desi_4) (0.65*** -> 0.70***). Similarly, the magnitude of the correlation coefficient between desi_1 and desi_2 increased (0.56** -> 0.63***), as did the significance level.

Results of the Principal Component Analysis (PCA)

Determination of the principal components

The prerequisite for conducting a principal component analysis is that the variables must be correlated with one another as a group, and the Pearson correlation coefficients must be significantly higher than the measures of partial correlation. To verify this, we performed the Kaiser–Mayer–Olkin (KMO) test (Table 4).

The test compares the sum of squares of the Pearson correlation coefficients to the sum of squares of the partial correlation coefficients. The resulting indicator is the MSA (Measure of Sampling Adequacy), where the critical value of the test is 0.5. The tests performed confirmed that the data are suitable for conducting PCA, as both the global and individual values are above 0.5. Table 4. The Kaiser–Mayer–Olkin test on the main indicators of the DESI index in 2016 and 2021 KMO - 2016 Kaiser–Mayer–Olkin measure of sampling adequacy Call: KMO(r = data) Overall MSA = 0.73 MSA for each item = desi_1 desi_2 desi_3 desi_4 0.68 0.74 0.71 0.83

Table or figure – see the PDF version of the issue for its exact content.

Table or figure – see the PDF version of the issue for its exact content.

The essence of principal component analysis is to identify those variables or components that carry information relevant to our purposes.

In our case, the relevant, valuable information is represented by the variance between the data points, i.e., the deviations from the mean. This allows us to reduce the available data in such a way that it retains as large a percentage of the total variance as possible.

Instead of the DESI main indicators, we examined the data along principal components (PC). These principal components are directions within the current four-dimensional space that determine in which direction the data exhibits greater variance.

The first principal component (PC1) specifies the direction in which the variance of the examined data is greatest. The second principal component (PC2) is a direction perpendicular to the first principal component (PC1), along which the data variance is the second largest. The results of the analysis for 2016 and 2021 are presented in Tables 5 and 6. By 2021, the system of correlations had become even more one-dimensional. The explanatory power of the first principal component increased from 70.9% to 74.4%, whereas the second principal component decreased. This is natural, as the sum of the four components equals 100%. In both years, the two principal components explain practically the same fraction of the system of correlations (88.5%). There is merely a slight redistribution between the two dimensions. Table 5. Own-value and variance ratios 2016

Table or figure – see the PDF version of the issue for its exact content.

Table or figure – see the PDF version of the issue for its exact content.

Table or figure – see the PDF version of the issue for its exact content.

Table or figure – see the PDF version of the issue for its exact content.

PCA analysis of the main DESI index indicators (2016). The principal components are associated with indicator weightings, which provide information regarding which main indicators carry greater or lesser weight within a given principal component. In the case of PC1 (Dim1), the weights of the main indicator variables are high (Table 6) and all are positive; thus, they vary in the same direction in parallel. The indicator *desi_1* (0.9419) appears with the highest weight, meaning that in those countries where the value of *desi_1* is high, performance in the integration of digital technology at the economic and societal levels is generally better, and consequently, the DESI index is higher than in countries where the value of this main indicator is low. In the case of PC2 (Dim2), the main indicators *desi_2* (0.7132) and *desi_3* (-0.4301) appear with greater weight, and a negative relationship exists between the two indicators. In countries where the value of *desi_2*—namely, connectivity—is high, the value of *desi_3*, which represents the integration of digital technology at the corporate level, is typically low. This may be explained by the fact that the integration of digital technology at the corporate level depends more on the mindset of corporate management than on the extent of connectivity. In 2016, the principal component belonging to the first dimension describes 70.9% of the correlations, while the second dimension describes 17.6%. The contribution of the variables to a given component undergoes some transformation, yet remains quite close to the equal weighting used by the European Commission:

PC1 (1st dim) = 31.28*desi_1 + 16.3*desi_2 + 24.66*desi_3 + 27.76*desi_4 EU Member States were grouped according to the established principal components based on their accession to the Union (Figure 7), as well as the real GDP per capita realised by the Member State (Figure 8). Through this, we sought to answer: what impact do different grouping criteria have on the application of digital technology at the economic and societal levels? Which countries perform better based on the specified criteria, is there a discernible difference between them, and does a shift occur over a six-year period (2016–2021)?

The application of digital technology at the economic and societal levels is illustrated in Figure 7, with the ordering principle being the date of accession to the Union.

Based on this, we formed three groups:

– the group of those who joined between 1957–1980, consisting of the six founding Member States, supplemented by the first wave of accession in 1973: Belgium, Germany, France, Italy, Luxembourg, the Netherlands, Denmark, and Ireland;

– the Mediterranean enlargement between 1981–2000, as well as the countries that joined following the Cold War: Greece, Spain, Portugal, Austria, Finland, and Sweden; – the group of Member States that joined during the East European enlargements between 2001–2013: Cyprus, the Czech Republic, Estonia, Hungary, Lithuania, Latvia, Malta, Poland, Slovenia, Slovakia, Bulgaria, Romania, and Croatia.

Table or figure – see the PDF version of the issue for its exact content.

Table or figure – see the PDF version of the issue for its exact content.

Table or figure – see the PDF version of the issue for its exact content.

Table or figure – see the PDF version of the issue for its exact content.

The majority of the Member States belonging to the first group (indicated in red) perform very well in the application of digital technology, as they are positioned on the right side of the primary component (X-axis). In the cases of Belgium and Ireland, some lag is observable in terms of the degree of connectivity, whereas they perform exceptionally well in the application of digital technology at the corporate level (Y-axis). This cannot be said for Italy and France, as they are found on the left side of the X-axis, thus among the less developed, and they performed quite poorly in terms of both the desi_2 and desi_3 indicators.

The digital performance of the second group, indicated in green, is situated at or above the average. Greece is an exception in this regard, as it struggles with deficiencies in both the desi_1 and desi_4 indicators.

The third group, indicated in blue and enclosed in a frame, which predominantly comprises the Central and Eastern European states, generally performs poorly in the application of digital technology, despite the infrastructure being provided. The digital network infrastructure is outstanding in Latvia, and above average in the cases of Romania, Poland, and Bulgaria. However, regarding the desi_4 indicator, it can only exhibit low values, with the exception of Malta and Estonia. Estonia possesses good network connectivity and also applies digital technologies very well at the corporate level.

Figure 8 illustrates the application of digital technology at economic and social levels, with the ordering principle being the level of achieved real GDP per capita (European Commission, 2021c). Based on this, we formed three groups:

– the group of Member States with a real GDP per capita between 5,000 and 15,000 euros (red indication): Bulgaria, Estonia, Croatia, Latvia, Lithuania, Hungary, Poland, Romania, and Slovakia; – the group of countries with a real GDP per capita between 15,001 and 32,000 euros (green indication): Cyprus, Czechia, Greece, France, Malta, Italy, Portugal, Spain, and Slovenia; – the group of Member States with a real GDP per capita between 32,001 and 85,000 euros (blue indication): Austria, Belgium, Denmark, Finland, Netherlands, Ireland, Luxembourg, Germany, and Sweden.

The group achieving the lowest real GDP per capita (indicated in red and enclosed in a frame) is located on the left side of the X-axis, meaning they lag behind in terms of the desi_1 main indicator, possessing low values. Whether they fall below or above the X-axis is determined by their values related to the desi_2 and desi_3 indicators. The upper-left quadrant signifies good network connectivity, but a low degree of digital technology integration, while the lower-right quadrant represents weak performance in the two aforementioned indicators. Estonia is an exception, having been placed in the upper-right quadrant; thus, it is characterised by above-average network connectivity and also possesses above-average values in digital technology integration.

The digital performance of the Member States in the middle field, i.e., those indicated in green, can largely be described as average. Sweden performs exceptionally well in terms of the desi_2 and desi_3 indicators, and in the areas of desi_1 and desi_4, it possesses values slightly above the average. Greece and Cyprus are found slightly detached, below the X-axis in the lower-left quadrant, which means they are characterised by weak network connectivity and the application of digital technology at the corporate level.

The group of Member States with the highest real GDP per capita is located on the right side of the X-axis, although discrepancies can be found here regarding the main indicators of the DESI index. Denmark, the Netherlands, Finland, and Sweden produce outstanding values in terms of human capital digital preparedness and the application of internet services; however, enterprises in Austria, Belgium, and Ireland utilise digital technology very effectively. The positioning of Luxembourg and Germany in the upper right quadrant suggests that they are simply the best across all indicators of the DESI index, which is accompanied by an exceptionally high real GDP per capita.

Based on the obtained results, the countries are positioned at a wider distance from the mean than would be expected in a normal distribution, whether the grouping criterion is the timing of their accession to the Union (Figure 7) or the level of realised real GDP per capita (Figure 8). Taking into account the properties of normal distributions, 95% of the data should lie within an interval of ±1.96 standard deviations from the mean. Our artificially constructed indicators describe a state that deviates from the normal—one might say an unnatural state—which, when translated from the language of statistics, can be interpreted as follows: in terms of the DESI index, the low-performing Member States (Romania, Greece, Bulgaria, and Poland) and the highly advanced countries (Finland, Denmark, Sweden, and the Netherlands) are much further apart from each other than would be expected under normal circumstances. Those countries that have lagged behind are at such a disadvantage relative to the advanced ones that it appears almost insurmountable. In the study (*The $\beta$-convergence of the DESI average index*), we previously determined the halving time of convergence; that is, how many years would be required to complete half the journey towards full convergence within the region, assuming the rate of convergence remains unchanged. As a result, we obtained a halving time of 20 years. It was already at this stage that the intuition arose that although convergence exists in the case of the DESI average index, the Matthew effect is simultaneously present. The distances between the later-accessing, or poorer Member States in terms of real GDP per capita, and the earlier-accessing, or wealthier countries, are not decreasing but, on the contrary, are increasing. It is for this reason, among others, that the 2016 and 2021 DESI data were compared.

PCA analysis of the main DESI index indicators (2021) Similar to the values calculated for 2016, in the case of PC1 (Dim1), the weights of the main indicator variables are high and each is positive; thus, they vary in parallel in the same direction. The indicator desi_1 (0.9371) appears with the highest weight, meaning that in those countries where the value of desi_1 is high, performance is generally better in the integration of digital technology at the economic and social levels, and therefore the DESI index is also higher than in places where the value of this aforementioned main indicator is low. In the case of PC2, the main indicators desi_2 (0.6539) and desi_4 (-0.3289) appear with greater weight, and a negative relationship exists between the two variables. In those countries where the value of desi_2, i.e., connectivity, is high, the value of desi_4, which characterises digital public services, is generally lower. In 2021, the principal component belonging to the first dimension describes the correlations by 74.4%, while the second dimension describes them by 14.2%. The contribution of the variables to the given component is as follows: 2021 PC1 (Dim1) = desi_1 * 29.52 + desi_2 * 18.72 + desi_3 * 27.59 + desi_4 * 24.17.

The application of digital technology at the economic and social levels is shown in Figure 9, and the principle of ordering is the date of accession to the Union. The established groups were already presented in detail in the previous sub-section.

Table or figure – see the PDF version of the issue for its exact content.

Table or figure – see the PDF version of the issue for its exact content.

The majority of the member states belonging to the first group (marked in red) perform very well in the application of digital technology, as they are situated on the right side of the first main component (X-axis). Among the best-performing countries in the application of digital technology, Denmark and the Netherlands have further increased their advantage stemming from digital readiness. Belgium lags somewhat behind the other members of the group in the development of digital infrastructure; however, it performs exceptionally well in the application of digital public services (Y-axis). This cannot be said for Italy and France, as they are found on the left side of the X-axis, thus among the less developed, and they performed quite poorly in terms of both the desi_2 and desi_4 indicators.

The digital performance of the second group, marked in green, is mostly situated at or above the average. The exceptions are Greece and Finland, where the former struggles with lags in both the desi_1 and desi_4 indicators, while the latter lags behind the performance of its group in the area of network connectivity. Spain has succeeded in increasing its digital performance, and Greece has also begun to catch up, albeit thus far only in the field of digital public services, where the rate of development was nearly 10.5% between 2016–2021. The third group, marked in blue and framed, which comprises the Eastern European states, typically performs poorly in the application of digital technology, although the infrastructure is provided. The digital network infrastructure is outstanding in the cases of Romania and Hungary. However, regarding the desi_4 indicator, with the exception of Malta and Estonia, they can only demonstrate low values.

Estonia possesses good network connectivity and is also at the forefront of public service digitalisation. Comparing Figures 7 and 9, it can be said that no significant shifts are experienced in relation to 2016 and 2021. The later accession countries are situated in almost the same "region", maintaining their role as low performers.

Figure 10 illustrates the application of digital technology at economic and social levels, with the principle of arrangement being the achieved level of real GDP per capita (European Commission 2021c). Compared to the 2016 grouping, the deviation is such that Estonia and Slovakia have moved one level higher, with their real GDP per capita increasing above 15,000 euros.

Table or figure – see the PDF version of the issue for its exact content.

Table or figure – see the PDF version of the issue for its exact content.

The group with the lowest real GDP per capita (indicated in red) is located on the left side of the X-axis; thus, in terms of the main indicators of the DESI index, they perform in the very bottom tier.

In the middle tier, the digital performance of the member states indicated in green can generally be described as average. In Spain, the value of the desi_4 indicator increased compared to 2016, and a positive shift towards the mean is also observable in the case of Cyprus. In contrast, Greece is found here to have fallen away completely, positioned below the X-axis in the lower-left quadrant, which means it performs poorly across all indicators of the DESI index. The group of member states with the highest real GDP per capita is located on the right side of the X-axis, although discrepancies can still be found regarding the main indicators of the DESI index. Denmark, the Netherlands, Finland, and Sweden produce outstanding values in every respect, outperforming not only the other European member states but also the group of member states with a real GDP per capita of between 32,001 and 85,000‛®. Similar to 2016, the Union's member states are currently positioned at a greater distance than ±1.96 standard deviations from the mean. Therefore, in this case as well, we cannot speak of a normal distribution. The distances between developed and developing countries have increased slightly further by 2021, whether the grouping criterion is the timing of their accession to the Union (Figure 9) or the level of realised real GDP per capita (Figure 10). The same countries are found at the top of the rankings and at the bottom; thus, they have not only failed to overcome the disadvantage, but it has increased slightly further. To quantify this, I determined the means and variances of the DESI indicators for the 27 EU member states for the base year (2016) and the closing year (2021). In the case of the means (2016: 0.3658, and 2021: 0.5178), the extent of growth was nearly 42%, meaning that the average values increased significantly. However, the value of the variance (2016: 0.0064, and 2021: 0.0097) is larger, showing an increase of approximately 52%. The 10-percentage-point greater increase in variance also indicates that the distances from the mean value increased in absolute terms to a greater extent than what could be observed based on the grouping criteria (Figures 9 and 10). Nevertheless, the F-test did not confirm the significance of this divergence. Although the suspicion of a Matthew effect during the 2016–2021 period cannot be significantly proven, it suggests its presence as a trend. Since our model examines the change in the DESI index using symmetric assumptions, the development of more advanced countries means, by definition, that less developed countries will lag behind by the same amount in the future: the rich will become richer because the poor are losing.

Results

Through the literature review, using bibliometric analysis, we established that tracking the evolution and development of the DESI index is primarily more important for those member states where the lag is significant. Consequently, the number of relevant studies is also higher in these countries; furthermore, several states are not even members of the European Union community.

According to the first hypothesis (H10), based on the changes in the DESI indices, we assumed that there is no convergence, i.e., no closing of the gap between the Member States. We analysed $\sigma$-convergence across three units: first, examining the EU-27 as a whole, then the EU-14 countries, and finally the group of states that joined in 2004 and subsequently. During the 2016–2021 period, cohesion among the Member States is observable in all three units of analysis; however, the rate of cohesion varies (Figure 3) among the EU Member States, ranging between 5.16% and 9.65%. Greece developed at the highest rate, while Latvia stands at the opposite end. Among the states that joined during the Eastern European enlargements, Poland shows the highest average development (9.34%). Based on the DESI index, $\beta$-convergence can be identified regarding the EU as a whole, with the annual average rate of shrinkage in the development gap being 3.45%. This change can be proven to be significant. Based on this, the half-life of the catch-up process could be determined. We established that for a constant catch-up rate of 3.45%, a half-life of approximately 20 years is required.

To the best of our knowledge, other scholars have not addressed this problem, and no forecasts or data related to this have been published.

The $\sigma$- and $\beta$-convergence analyses supported the finding that convergence does indeed exist between the EU Member States in the case of the DESI index; thus, we reject the null hypothesis H1.

According to the second hypothesis (H20), we assumed that the contribution of the first dimension components (variables) obtained through PCA analysis is equal to the weights provided by the European Commission. Our calculations showed that the contribution of the variables to the given component differs somewhat from the procedure currently applied in the calculation of the DESI index, yet it remains sufficiently close to the equal weighting used by the European Commission. A similar correlation was revealed during the analysis of the 2021 data. Both investigations (2016, 2021) support the finding that the first component, namely human capital, carries a greater weight than the originally applied 25%. Based on all of this, we also rejected our second null hypothesis. Consequently, we suggest considering a modification of the equal weighting during the calculation of the DESI index. We recommend that human capital be assigned a higher weight. Previous studies related to this matter also emphasise the importance of skills, competencies, and education. Bánhidi, Dobos, and Nemeslaki (2020) used cluster analysis and MDS (multidimensional scaling) to group and rank EU member states, comparing their results with the values determined by the DESI index weights. They found that the two rankings did not differ significantly, but they raised the possibility of simplifying and rethinking the DESI dimensional structure. According to Folea (2018), Fleaca (2017), and Condruz-Bacescu (2019), within the EU's digital agenda, measures supporting the teaching of digital skills play a key role in developing the digital skills of the population, and every member state has an obligation to increase the number of people possessing basic and advanced digital skills (Jaculjakova–Stofkova 2020; Rodrigues et al. 2021; Tumbas et al. 2019). However, education alone is not sufficient, as technology must also be readily accessible to ensure the applicability and further development of acquired knowledge (Matkovic et al. 2019; Štofková et al. 2019). López Peláez, Erro-Garcés, and Gómez-Ciriano (2020) contrasted the digital skills of British youth with labour market requirements, and in their view, the digital divide between the two can only be bridged through education and the development of digital skills. In their research, Kwilinski, Vyshnevskyi, and Dzwigol (2020) found that the higher the level of digitalisation in a society, the lower the risk of poverty and social exclusion. Vyshnevskyi, Stashkevych, Shubna, and Barkova (2020), evaluating the development of the EU economy through the dynamics of GDP change, point out that the decisive role of the economy's digitalisation in the achieved economic growth rate cannot be demonstrated, given the current level of development of applied technologies and institutions. According to our studies, member states with the highest real GDP per capita produce outstanding values in terms of human capital digital readiness and the application of internet services. Companies in Austria, Belgium, and Ireland apply digital technology well, while Luxembourg and Germany are simply the best in every indicator of the DESI index, which is also accompanied by an exceptionally high real GDP per capita. An increase in the DESI index also signifies an increase in the employment rate and personal incomes, reducing the long-term unemployment rate and labour market uncertainty (Basol–Yalcin 2021; Stavytskyy et al. 2019). According to Neamtu et al. (2019) and Bejinaru (2017), Industry 4.0, together with information and communication technology, has completely rewritten the global economic system, and further transformations are expected. Taking digital aspects into account, Northern European states perform better, while Eastern Europe has fallen quite far behind (Stankovic et al. 2021). Romania lags behind the European average in several respects regarding digitalisation, both in the implementation of e-government measures (Murariu–Bedrule-Grigoruta 2020; Todorut–Tselentis 2018), in terms of the digital skills of the population (Gaftea 2016), and in the application of digital technology at the corporate level (Baltateanu 2020; Iliescu 2020). Esses et al. (2021) and Kuncová–Doucek (2018) evaluate the performance and progress of the Visegrád countries; Herman (2020) emphasises that in the V4 countries and Romania, the proportion of those employed in the ICT sector and its contribution to GDP also lag behind the EU average.

Continuing this line of reasoning, we proceed to test the third null hypothesis (H30), which posits that the mean and variance of the DESI index latent variables do not demonstrate the Matthew effect during the 2016–2021 period. By analysing the extent of the growth of the DESI indices as well as their variance, we established that the variance increased by 10% more than the DESI index itself.

After six years, the same countries are found at the top of the rankings, and the same countries at the bottom; thus, they have not only failed to overcome their disadvantage, but the gap has actually widened.

However, the F-test did not confirm the significance of this divergence. Although the suspicion of a Matthew effect during the 2016–2021 period cannot be significantly proven, its tendency suggests its existence, meaning we confirm the third hypothesis. Nevertheless, the suspicion remains, which warrants verification through more in-depth investigations.

It is recommended that the main indicators of the DESI index be subjected to individual statistical testing and analysis.

Finally, upon examining the average growth rate of the DESI indices between 2016 and 2021, we determined it to be 6.98%. In only two years, 2019 and 2020—namely during the first waves of the Covid pandemic—the rate of growth was nearly 9%. Without fetishising the impact of the pandemic alone, it is nonetheless striking that, compared to the average of previous years, a 25% increase could be observed in the DESI index within a single year. This change is substantial; however, further detailed investigations are required to statistically confirm the impact of Covid.

Conclusions

In the present study, we sought evidence of digital development in EU Member States based on DESI index values. To the best of our knowledge, no study has yet been conducted that applied $\sigma$- and $\beta$-convergence methods to the examination of the DESI index. We specifically examined the developmental lag of the Member States, but also applied $\sigma$- and $\beta$-convergence to determine the change in this developmental lag. Our results show that during the period under review, the average growth rate of EU Member States fluctuated between 5.16% and 9.65%. Statistical calculations demonstrated that differences between EU Member States decreased ($\sigma$-convergence), and a negative correlation exists between initial development levels and the average growth rate ($\beta$-convergence). We calculated that the average annual rate of decrease in developmental differences was 3.45%, implying that the halving time for convergence is nearly 20 years. The values presented here support the conclusion that the integration of digital technology is increasing at both economic and social levels within EU Member States. We validated the weighting method used by the EU and formulated a proposal for modifying the weighting of the DESI index. With the aid of PCA analysis, the contribution of each main indicator within the principal component can be precisely determined. Our calculations confirmed that the significance of human capital is greater than currently determined in the DESI index methodology, whereas the significance of connectivity (internet access) as a main indicator is slightly lower. Our investigations confirmed the paramount importance of human capital regarding digitalisation processes.

Our analyses allow us to conclude that, based on the DESI index and regardless of whether the clustering criterion is the date of EU accession or real GDP per capita, the gap between high-performing and low-performing countries is widening. However, this phenomenon, known as the Matthew effect, could not be supported by statistical methods, although its tendency suggests its existence.

Supplementing the descriptive statistics currently used for the analysis of the DESI index with the methods presented in this study may assist in linking macroeconomics with solutions.

Bibliography

Baltateanu, A. M. 2020. Digitalization. Global Challenges and Economic Prosperity Example: Romania. In: Bratianu, C.–Zbuchea, A.–Anghel, F.–Hrib, B. (eds.) Strategica - Preparing for Tomorrow, Today. Bucharest: Tritonic Publ House, 520–530. Bánhidi, Z.–Dobos, I. 2020. Az Európai Unió digitális gazdaság és társadalom indexének statisztikai elemzése. Statisztikai Szemle 98(2), 149–168. Bánhidi, Z.–Dobos, I.–Nemeslaki, A. 2020. What the overall Digital Economy and Society Index reveals: A statistical analysis of the DESI EU28 dimensions. Regional Statistics 10(2), 42–62.

Barro, R. J.–Sala-I-Martin, X. 1992. Convergence. Journal of Political Economy 100(2), 223–251.

Barro, R. J.–Sala-I-Martin, X.–Blanchard, O. J.–Hall, R. E. 1991. Convergence Across States and Regions. Brookings Papers on Economic Activity 22(1), 107–182.

Basol, O.–Yalcin, E. C. 2021. How does the digital economy and society index (DESI) affect labor market indicators in EU countries? Human Systems Management 40(4), 503–512.

Bassanini, A.–Scarpetta, S.–Visco, I. 2000. Knowledge, technology and economic growth: recent evidence from OECD countries. OECD Economics Department Working Papers 259. Paris: OECD Publishing.

Bejinaru, R. 2017. Knowledge strategies aiming to improve the intellectual capital of universities. Management & Marketing-Challenges for the Knowledge Society 12(3), 500–523.

Bogoslov, I. A.–Lungu, A. E. 2020. The Relationship between Entrepreneurship and Digitalization - Spotlight on the EU Countries. Studies in Business and Economics 15(3), 5–15.

Borowiecki, R.–Siuta-Tokarska, B.–Maroń, J.–Suder, M.–Thier, A.–Żmija, K. 2021.

Developing Digital Economy and Society in the Light of the Issue of Digital Convergence of the Markets in the European Union Countries. Energies 14(9), 2717.

Čiefová, M. 2017. Austria’s Competitiveness: Assessing Changes Compared to the Past Year. In: Cibuľa, A.– Klenka, M.–Vlková, E. (eds.) International Relations 2017:

Current issues of world economy and politics. Smolenice: Publishing Ekonóm, 155–161.

Clark, D. 2021. Number of SMEs in the European Union 2008-2021, by size. https:// www.statista.com/statistics/878412/number-of-smes-in-europe-by-size/, accessed: 07. 10. 2021.

Condruz-Bacescu, M. 2019. E-learning Market in Romania. In: Roceanu, I. (ed.) New Technologies and Redesigning Learning Spaces. Bucharest: Carol I National Defence University Publishing House, 49–56.

De Solla Price, D. 1967. Network of Scientific Papers. Science 149(3683), 510–515.

Ershova, I.–Obukhova, A.–Belyaeva, O. 2020. Implementation of innovative digital technologies in the world. Economic Annals-XXI 186(11-12), 28–35.

Esses, D.–Csete, M. S.–Németh, B. 2021. Sustainability and Digital Transformation in the Visegrad Group of Central European Countries. Sustainability 13(11), 5833.

European Commission 2010a. The European Digital Agenda. Brussels: European Commission.

European Commission 2010b. Europe 2020 A Strategy for Smart, Sustainable and Inclusive Growth. Brussels: Communication from the Commission.

European Commission. 2021a. Digital Economy and Society Index 2021 Methodological Note. https://www.juntadeandalucia.es/institutodeestadisticaycartografia/digitalizacionsociedad-digital/metodologia/Notas-Metodologicas-DESI.pdf, accessed: 2021. 11. 19. European Commission 2021b. Digital Compass 2030: the European way to implement the digital decade. Brussels: European Commission.

European Commission 2021c. Real GDP per capita. https://ec.europa.eu/eurostat/ databrowser/view/sdg_08_10/default/table?lang=en, accessed: 2021.10.21. European Commission 2021d. State of the Union: Commission proposes a Path to the Digital Decade to deliver the EU’s digital transformation by 2030. Brussels: European Commission.

European Commission 2021. Digital Economy and Society Index. Digital Agenda for Europe. Brussels: European Commission.

Pope Francis 2015. Laudato Si: On Care for Our Common Home. Budapest: Szent István Társulat.

Fernández-Portillo, A.–Almodóvar-González, M.–Hernández-Mogollón, R. 2020. Impact of ICT development on economic growth. A study of OECD European Union countries. Technology in Society 63(C), 101420.

Fleaca, E. 2017. Entrepreneurial Curriculum through Digital-Age Learning in Higher Education - A Process-based Model. TEM Journal–Technology Education Management Informatics 6(3), 591–598.

Folea, V. 2018. European public policies in the area of the digital economy and society: country performance analysis. In: Hájek, P.–Vít, O. (eds.) CBU International Conference Proceedings. Prague: CBU Research Institute, 120–128.

Gaftea, V. 2016. Awareness in eEducation Given by Strategies and Indicators. In: Roceanu, I. (ed.) eLearning Vision 2020! Volume II. Bucharest: Carol I National Defence University Publishing House, 593–600.

Giannone, D.–Santaniello, M. 2019. Governance by indicators: the case of the Digital Agenda for Europe. Information Communication & Society 13, 1889–1902. Grinberga-Zalite, G.–Hernik, J. 2019. Digital Performance Indicators in the EU. In: Markevica, A. (ed.) Research for Rural Development. Jelgava: Latvia University of Life Sciences and Technologies, 183–188.

Hadad, S. 2017. Business Digitization in the Romanian Economy. In: Pinzaru, F.– Zbuchea, A.–Bratianu, C.–Vatamanescu, E. M.–Mitan, A. (ed.). Shift! Major Challenges of Today’s Economy. Bucharest: Tritonic Publishing House, 453–466.

Herman, E. 2019. The Influence of ICT Sector on the Romanian Labour Market in the European Context. In: Moldovan, L.–Gligor, A. (eds.) Procedia Manufacturing. Amsterdam: Elsevier Science BV, 344–351.

Hofstede, G. 1984. Cultural dimensions in management and planning. Asia Pacific Journal of Management 1, 81–99.

Hotelling, H. 1933. Analysis of a complex of statistical variables into principal components. Journal of Educational Psychology 24, 417–441.

Iliescu, M. E. 2020. Barriers to Digital Transformation in SMEs: A Qualitative Exploration of Factors Affecting ERP Adoption in Romania. In: Bratianu, C.–Zbucea, A.– Anghel, F.–Brib, B. (eds.) Preparing for Tomorrow, Today. Bucharest: Tritonic Publishing House, 453–461.

Ivanovic-Dukic, M.–Stevanovic, T.–Radenovic, T. 2019. Does digitalization affect the contribution of entrepreneurship to economic growth? Zbornik Radova Ekonomskog Fakulteta U Rijeci-Proceedings of Rijeka Faculty of Economics 37(2), 653–679. Ivanovic, M.–Ambros, F. 2016. The Economic Impacts of Broadband Access Investment in Croatia. In: Križanović, K. (ed). Vision and Development. Osijek: Ekonomski Fakultet Osijeku-Fac Economics Osijek, 911–921.

Jaculjakova, S.–Stofkova, K. R. 2020. Analysis of Basic Digital Skills of Citizens in Selected Region. In: Chova, L. G.–Martínez, A. L.–Torres, I. C. (eds.) INTED2020 Proceedings. Valencia: IATED Academy, 6141–6149.

Jakovic, B.–Curlin, T.–Miloloza, I. 2021. Enterprise Digital Divide: Website e-Commerce Functionalities among European Union Enterprises. Business Systems Research Journal 12(1), 197–215.

Jendricko, K.–Mesaric, J. 2019. The Significance of European Structural and Investment Funds for the Republic of Croatia in the Area of Information and Communications Technologies. In: Turkalj, Z. (ed.) Interdisciplinary Management Research XV (IMR) Osijek: Josip JurajStrossmayer University of Osijek, 1632–1655. Jovanović, M.–Dlačić, J.–Okanović, M. 2018. Digitalization and society’s sustainable development–Measures and implications. Zbornik radova Ekonomskog fakulteta u Rijeci: časopis za ekonomsku teoriju i praksu 36(2), 905–928.

Jurčević, M.–Lulić, L.–Mostarac, V. 2020. The Digital Transformation of Croatian Economy compared with EU Member Countries. Ekonomski vjesnik/Econviews–Review of Contemporary Business, Entrepreneurship and Economic Issue 33(1), 151–164. Kamberidou, I.–Pascall, N. 2019. The digital skills crisis: engendering technology– empowering women in cyberspace. European Journal of Social Sciences Studies 4(6) 1–33.

Karnitis, G.–Virtmanis, A.–Karnitis, E. 2019. Key drivers of digitalization; EU context and Baltic case. Baltic Journal of Modern Computing 7(1), 70–85. Koopmans, T. C. 1963. On the concept of optimal economic growth. New Haven: Cowles Foundation Discussion Papers, 392.

Kuncová, M.–Doucek, P. 2018. Comparison of ICT Development in V4 Countries. In: Jedlička, P.–Marešová, P.–Soukal, I. (eds.) Hradec Economic Days. Kralove: University of Hradec Králové, 513–522.

Kwilinski, A.–Vyshnevskyi, O.–Dzwigol, H. 2020. Digitalization of the EU Economies and People at Risk of Poverty or Social Exclusion. Journal of Risk and Financial Management 13(7), 142.

Laitsou, E.–Kargas, A.–Varoutas, D. 2020. Digital competitiveness in the European Union era: The Greek case. Economies 8(4), 85.

Lomakina, O.–Kookueva, V.–Makarenko, A. 2021. Redistribution of economic resources in the digital society. Business and Society Review 126, 25–35. López Peláez, A.–Erro-Garcés, A.–Gómez-Ciriano, E. J. 2020. Young people, social workers and social work education: the role of digital skills. Social Work Education 39(6), 825–842.

Luchko, M. 2020. Positions of Nordic Countries through the Prism of International Ratings. Contemporary Europe-Sovremennaya Evropa 3, 83–95.

Macgregor Pelikánová R. 2018. Fostering innovation–a myth or reality of the EU in 2018. In: Staníčková, M.–Melecký, L.–Kovářová, E.–Dvoroková, K. (eds.) Proceedings of the 4th International Conference on European Integration 2018. Ostrava: VŠB - Technical University of Ostrava, 965–973.

Martínez-Cantos, J. L. 2017. Digital skills gaps: A pending subject for gender digital inclusion in the European Union. European Journal of Communication 32(5), 419–438. Máté 25,29 1993. New Testament, Eleventh, revised edition. Pannonhalma: Bencés Kiadó.

Matkovic, P.–Maric, M.–Rakovic, L.–Sakal, M. 2019. Enhancing Digital

Competences in Higher Education. In: Chova, L. G.–Martínez, A. L.–Torres, I. C. (eds.) INTED2020 Proceedings. Valencia: IATED Academy, 7939–7944.

Merton, R. K. 1968. The Matthew effect in science: The reward and communication systems of science are considered. Science 159(3810), 56–63.

Merton, R. K. 1988. The Matthew effect in science, II: Cumulative advantage and the symbolism of intellectual property. Isis 79, 606–623.

Moldabekova, A.–Philipp, R.–Reimers, H. E.–Alikozhayev, B. 2021. Digital technologies for improving logistics performance of countries. Transport and Telecommunication Journal 22(2), 207–216.

Murariu, M.–Bedrule-Grigoruta, M. V. 2020. Digitalisation – A Sine-qua-non Condition for the Romanian Public Management. In: Tofan, M.–Bilan, I.–Cigu, E. (eds.) European Finance, Business and Regulation International Conference (EUFIRE). Iași: Editura Universității „Alexandru Ioan Cuza” din Iași, 621–631.

Neamtu, D. M.–Hapenciuc, C. V.–Bejinaru, R. 2019. The Impact of Digitalization on Business Sector Development in the Knowledge Economy. In: Dima, A. M. (ed.) Proceedings of the International Conference on Business Excellence. Bucharest: Sciendo, 479–491.

Obláth, G.–Szörfi, B. 2008. Macroeconomic convergence in the new EU member states. In: Kolosi, T.–Tóth, I. G. (eds.) Social Report. Budapest: Tárki.

Pearson, K. 1901. LIII On Lines and Planes of Closest Fit to Systems of Points in Space. Philosophical Magazine and Journal of Science 2, 559–572.

Perc, M. 2014. The Matthew effect in empirical data. Journal of the Royal Society, Interface 11(98), 20140378.

Polozova, T.–Kolupaieva, I.–Sheiko, I. 2021. Digital Gap in EU Countries and its Impact on Labour Productivity and Global Competitiveness. In: Mačí, J.–Marešová, P.– Firlej, K.–Soukal, I. (eds.) Hradec Economic Days. Hradec Králové: University of Hradec Králové, 659–670.

Pyroh, O.–Kalachenkova, K.–Kuybida, V.–Chmil, H.–Kiptenko, V.–Razumova, O. 2021. The Influence of Factors on the Level of Digitalization of World Economies.

International Journal of Computer Science and Network Security 21(5), 183–191.

Rakićević, J.–Rakićević, A.–Poledica, A. 2019. Logical clustering approach for analysing digital economy and society performance of countries. In: Štěpnička, M. (ed.) 11th Conference of the European Society for Fuzzy Logic and Technology (EUSFLAT 2019). Prague: Atlantis Press, 550–557.

Ramsey, F. P. 1928. A mathematical theory of saving. The Economic Journal 38(152), 543–559.

Rigney, D. 2010. The Matthew effect: How advantage begets further advantage.

Columbia: University Press.

Rodrigues, A. L.–Cerdeira, L.–Machado-Taylor, M. D–Alves, H. 2021. Technological Skills in Higher Education-Different Needs and Different Uses. Education Sciences 11(7), 326.

Rozite, K.–Balina, S.–Freimane, R. 2019. Digital Competence Rating and Economic Development in the EU. In: Romaniva, I. (ed.) New Challenges of Economic and Business Development - Incentives for Sustainable Economic Growth. Riga: University of Latvia, 701–708.

Solow, R. M. 1956. A contribution to the theory of economic growth. The Quarterly Journal of Economics 70(1), 65–94.

Stankovic, J. J.–Marjanovic, I.–Drezgic, S.–Popovic, Z. 2021. The Digital Competitiveness of European Countries: A Multiple-Criteria Approach. Journal of Competitiveness 13(2), 117–134.

Stavytskyy, A.–Kharlamova, G.–Stoica, E. A. 2019. The Analysis of the Digital Economy and Society Index in the EU. Baltic Journal of European Studies 9(3), 245–261.

Stofkova, K. R.–Stofkova, J. 2020. Use of Open Data in the Development of the Digital Economy in the Knowledge Society in the Era of Globalization. In: Kliestik, T (ed.) Globalization and Its Socio-Economic Consequences - Sustainability in the Global Knowledge Economy. Rajecke Teplice: EDP Sciences. Stofkova, Z.–Hraskova, D. 2017. Digital skills in period of digital economy. In:

Petranova, D.–Rybansky, R.–Mendelova, D. (eds.). Marketing Identity. Trnava: Univ Ss Cyril & Methodius Trnava, 407–425.

Štofková, Z.–Soltes, V.–Stofkova, J. 2019. Survey of the Citizens Digital Skills in the Region. In: Chova, L. G.–Torres, A. M. (eds.) International Conference of Education, Research and Innovation. Proceedings. IATED, 7127–7132.

Szőllősi, L.–Béres, E.–Szűcs, I. 2021. Effects of modern technology on broiler chicken performance and economic indicators – a Hungarian case study. Italian Journal of Animal Science 20(1), 188–194.

Tkác, M. 2018. Digital Single Market Strategy and its Impact on Trust in Public Administration. In: Doucek, P.–Chroust, G.–Oškrdal, V. (eds.) IDIMT-2020 Digitalized

Economy, Society and Information Management. Kutna Hora: Trauner Verlag, 277–286. Todorut, A. V.–Tselentis, V. 2018. Digital Technologies and the Modernization of Public Administration. Quality-Access to Success 19(165), 73–78.

Tsenzharik, M. K.–Krylova, Y. V.–Steshenko, V. I. 2020. Digital transformation in companies: Strategic analysis, drivers and models. St. Petersburg University Journal of Economic Studies 36, 390–420.

Tumbas, P.–Sakai, M.–Pavlicevic, V.–Rakovic, L. 2019. Digital Competencies in Business Informatics Curriculum Innovation. In: Chova, L. G.–Martínez, A. L.–Torres, I. C. (eds.) International Technology, Education and Development Conference (INTED 2019), Valencia: IATED Academy, 9655–9664.

Van Eck, N. J.–Waltman, L. 2013. VOSviewer manual 1., retrieved: 2021. 07. 29. Vidruska, R. 2016. The Digital Economy & Society Index and Network Readiness Index: Performance of Latvia on European Union Arena. In: Bruna, I. (ed.) New Challenges of Economic and Business Development - Society, Innovations and Collaborative Economy. Riga: University of Latvia, 901–916.

Vyshnevskyi, O.–Stashkevych, I.–Shubna, O.–Barkova, S. 2020. Economic growth in the conditions of digitalization in the EU countries. Studies of Applied Economics 38(4), 1–9.

Zdjelar, R.–Hrustek, N. Z.–Sumpor, M. J. E. 2019. Sustainable development and active ageing in EU countries–bridges and gaps. In: Hammes, K.–Machrafi, M.–Samodol,

Authors

  • Kovács Tünde Zita
  • Bittner Beáta
  • Huzsvai László
  • Nábrádi András

Text extracted automatically from the PDF; for tables, figures and exact layout see the PDF of the issue.