Structural Conditions of Income Inequality Convergence Within the European Union

Abstract

European integration aims to achieve spatially sustainable development across the member states. However, the success of socio-economic integration is conditioned by structural features of the economies, which, hitherto, appear highly diversified across the EU countries. The paper focuses on the structural conditions of the process of income inequality convergence. It aims to identify differences in the convergence regarding the structural conditions of the economies. To fulfil the research tasks the paper classifies the 27 European member states according to their sectional employment structures using the Ward method. It then tests the appearance of beta convergence using FE panel models for the specified clusters of economies. It also considers structural change, measured by the NAV (norm of absolute value), as a determinant of income inequality convergence. The main research period covers 2009–2021. The findings of the paper confirm that income inequality convergence occurs within the groups of economies specified by different structural conditions. Importantly, the clustering according to the similarity of the employment structure overlaps with the division along the lines of the ‘new’ and ‘old’ member states, which proves the importance of historically shaped institutions for development. However, the observed convergence does not lead to improved social cohesion. Social policy, especially in the ‘new’ member states, is not able to offset the growth in market income inequality additionally stimulated by the structural changes. It can be concluded that an urgent need to design new solutions for social policy concerning structural transformation in employment in the EU emerges.

1. Introduction

Economic and social cohesion is the main aim of the European integration process, which covers spatially sustainable development as well. This pursuit was included in the EU founding Treaties. It assumes reducing the income inequality and disparity in living standards among countries, regions, and individuals, ensures social inclusion, and involves harmonising the entire economic system and maximising the potential of its components. Therefore, the process of multidimensional convergence is an important precondition to look out for regarding European integration [1,2,3].

Simultaneously, sustainable development understood as a process aimed at improving welfare, while finding the best path of development to balance social, environmental, and economic pursuits is regarded as an important goal by the whole global community. The paper deals with economic inequality within societies and across the EU countries and structural foundations for equalising incomes. These problems are expressed either by SDG 10 by the UN (Reduce inequality within and among countries) as well as SDG 9 (Build resilient infrastructure, promote inclusive and sustainable industrialisation, and foster innovation). It fits SDG 12 (Ensure sustainable consumption and production patterns) as well, as it aims to search for such employment structures that could foster economic and social development. Therefore, the study deals with themes of sustainability in terms of enabling socio-economic development to be stable and sustainable referring to relationships between social and economic dimensions of development and focusing on convergence of income inequality and its structural foundations.

Convergence is understood as the process of eliminating disparities, and it can refer to social, technological, structural, and income convergence as well as to its real and nominal form [1,4]. Income convergence has been analysed since the 1980s, starting from Baumol [3,4,5]. It was traditionally analysed in terms of the average income per capita; however, after the study by Bénabou [6], it was also extended to the whole distribution of income, laying the foundations for research into the convergence of income inequality. Still, research into income inequality convergence is relatively rarely conducted, and the few exceptions include studies by Ravallion [7], Bleaney and Nishiyama [8], Chambers and Dhongde [9], Solarin et al. [10], and for the EU countries, Alvarez-Garcia et al. [11], Kvedaras and Cseres-Gergely [12], Arı et al. [13], Suárez-Arbesú et al. [14], and Cyrek [15].

Nevertheless, convergence, either in terms of income level or distribution, depends on prior structural convergence, described by the socio-economic relations in the labour and goods markets, among others [1]. In the classical theory of integration, structural similarity is also pointed to as a precondition for the success of the process [16]. The assumption that the initial conditions play a role in shaping the long-term outcomes leads economists to study convergence within clubs specified by some common structural features [3].

In this study, income inequality convergence is considered as being conditioned by the employment structure. Therefore, the paper aims to identify differences in the convergence regarding the structural conditions of the economies. It verifies not only whether income inequality convergence takes place, but also whether the sectoral allocation of labour influences the process. The study identifies differences in convergence between groups of economies with different structural features. It considers both market income inequality as well as inequality in income after social transfers, therefore paying attention to the role of the welfare state in mitigating inequality. Moreover, as the convergence may either follow a favourable path of limiting inequality or a path specified by growing inequality, the research also investigates the direction of changes that indicates the successfulness of European integration. Additionally, it tries to specify the influence of structural change on income inequality as well, to identify potential challenges for social policy connected with economic dynamics.

The novelty of the study lies in connecting the issues of income inequality convergence and structural change, which, albeit of fundamental character for socio-economic development, are relatively rarely analysed in the literature in relation to each other. Meanwhile, they are important determinants of European integration as they form a basis for levering well-being and social cohesion. The paper focuses on relatively neglected aspects of an analysis of spatially sustainable development as it researches the structural conditions of the process of income convergence; however, it goes beyond the traditional approach to convergence, taking into account the entire income distribution, i.e., income inequality convergence.

Generally, the research confirms that income inequality convergence appears within each group of the EU countries, as specified by a sectional structure of employment. Importantly, the clustering reveals that structural features in terms of the sectional allocation of labour closely overlap with the division in line with the ‘new’ and ‘old’ EU member states. Therefore, the historical experience of socio-economic system solutions coincides with the structural characteristics of a modern economy. Moreover, the results are ambiguous regarding the influence of structural change on income inequality, as there are differences between the two groups of countries. Structural change stimulates growth in market income inequality within the group of Central and East European countries, while within West, North, and South Europe, the influence is significant only concerning incomes after social transfers. It stresses the differences in the role of social policy across the EU countries.

The study is organised as follows. The next section reviews the literature concerning economic structures and income convergence. Therefore, it provides some theoretical basis for the research into the role of employment structures in explaining income inequality and its changes. It also presents some empirical results that concern convergence of income inequality and factors that affect the process. Then, in Materials and Methods, the methods used are described. This Section explains the conceptual framework of the study, gives methodological basis for clustering the countries and specifies the model of income inequality convergence along with the variables used. In the following section the results are presented. They cover the following: the general socio-economic situation, results of clustering the 27 EU countries concerning their employment structures, and estimations of the models for income inequality convergence and specification of the target levels of coherence. The final part concludes and discusses the main findings, refers to limitations of the study and draws some future directions of research into structural foundations of income inequality.

2. Research into Structural Foundations of Income Convergence in the EU

As a structure is generally understood as a description of the relationships between parts and the whole, concerning an economic structure, the attention is focused on the proportions and relationships that characterise the economic setting in space and time. Therefore, an economic structure is usually analysed in terms of the composition of employment, output, and sectoral structure of production [17]. Moreover, as structures are transforming, the proportions and ratios associated with different sub-parts are changing, and these non-trivial changes have implications for the structural properties of the whole system [17]. Long-term changes in the composition of economic aggregates, as Krüger [18] defines structural change, are seen as the evolution from low-productivity activities to modern, higher-productivity activities [19]. Still, structural change is a complex process with its unique specificity in any single example. There are at least three models of structural change, as described by Zacher [20]: increase and displacement (when new elements displace traditional ones), transformative (when there are two-way interactions between the new and the traditional), and cumulative (which assumes complexity of the evolution, both in the form of displacement and transformation). Moreover, according to Pasinetti’s theory [21], structural change is an open-ended process, which makes alternative dynamic paths for the whole economy possible [22,23]. The complexity of structural change and its mutual relations with long-term outcomes of the whole system make it both challenging and also essential to research its role in socio-economic development.

The essential role of the structural features of the economy in the course of development can be derived even from the thoughts of classical economists, such as Smith [24] (cf. [25,26]). However, the literature focused more on structural change within the development economics, evolutionary, and structuralist approaches. The most important early contributions were made by Clark [27], Fisher [28], Fourastié [29], Kuznets [30], Leontief [31], Chenery [32], Hirschman [33], Nurkse [34], Rosenstein-Rodan [35], and Lewis [36], amongst others (cf. [17]). Since then, the structuralist approach to development economics has been renewed in the form of the ‘new structural economy’ (cf. [19,37]), and the sectoral composition was argued to be essential not only for economic growth but also for the environment and society (e.g., [38,39,40,41,42]). Nevertheless, the ‘economics of structural change’ still lacks systematisation and unification, as it covers the enormous heterogeneity of studies expressing the complexity of the structural issues [26].

Concerning the role of structures for traditional income convergence, there are numerous empirical studies into the problem. Especially, structural factors are perceived to be important in the convergence processes in the EU (see

Table 1. Summary of selected empirical papers on income/income inequality convergence in the EU.

Paper	Data	Approach	Structural Features	Results
Jena, D.; Barua, A. (2020) [43]	EU-28; 2000–2017	income convergence	considered	income convergence confirmed; convergence mainly due to convergence in income shares of the industrial and services sector.
Bolea, L.; Duarte, R.; Chóliz, J.S. (2018) [2]	EU-28; 2000–2014	income convergence	considered	convergence until 2008; the different behaviour of services, in particular, knowledge-intensive services condition income generation
Cutrini, E., Mendez, C. (2023) [44]	267 EU regions; 2003–2016	income convergence	considered	formation of convergence clubs jointly determined by structural change and spatial dependence
Cavallaro, E.; Villani, I. (2021) [45]	EU-27 (EU-28 except Malta); 1995–2018	productivity convergence	considered	the EU countries do not converge to a unique path—clustered patterns for aggregate and sector-level productivity growth; countries approaching the high-growth paths specialise in knowledge-intensive production,
Bal-Domańska, B. (2024) [3]	EU NUTS–2 regions; 2000–2021	income convergence	considered	income convergence confirmed; three regional convergence clubs—income convergence within the Industry club and the Other club, absent for the Service club
Alvarez-Garcia, S.; Prieto-Rodriguez, J.; Salas, R. (2004) [11]	EU-15; 1993–1996	income inequality convergence	not considered	income inequality convergence confirmed
Kvedaras, V.; Cseres-Gergely, Z. (2020) [12]	EU-27; 2007–2014	income inequality convergence	not considered	income inequality convergence confirmed
Arı, A.; Cergibozan, R.; Demir, C.; Yetkiner, H. (2022) [13]	EU-15; 1988–2017	income inequality convergence	not considered	income inequality convergence confirmed but at a higher level of inequality
Suárez-Arbesú, C.; Apergis, N.; Delgado, F.J. (2022) [14]	EU-27 (EU-28 except Croatia); 2007–2018	income inequality convergence	primary sector included as explanatory variable	convergence limited to some convergence clubs: four clubs and four divergent countries
Cyrek, M. (2023) [15]	EU-27; 2009–2021	income inequality convergence	considered	income inequality convergence confirmed but at a higher level of inequality; structural change determines market income inequality but not inequality after social transfers

Source: own study.

). Jena and Barua [43] claim that for income convergence to be sustainable, the lower income countries must also experience convergence in economic structures. Bolea et al. [2] stress that the structural and technological characteristics of the EU countries are highly conditioning for their inter-dependencies, economic outcomes, and convergence. Cutrini and Mendez [44] also support the thesis about the essential role of the sectoral composition for club convergence, also stressing the influence of spatial dependence. Similarly, Cavallaro and Villani [45] identified that the heterogeneity across the EU countries was due to the dissimilarity in their economies’ productive structures and concluded that the convergence of poorer countries is hindered by their lower shares of innovation-intensive and high-skilled sectors and the lower productivities of these sectors. Bal-Domańska [3] identified three different regional convergence clubs based on the knowledge and technology-based economic structure and concluded that income convergence appears only within the Industry club and the Other club (with its high share of less knowledge-intensive services (LKIS)), while it was absent for the Service club (with its high share of knowledge-intensive services (KIS)).

Despite the numerous research stressing the role of economic structures in the development and processes of convergence, the relations between the structural composition of an economy and income inequality appear as an issue that still needs to be recognised. It is partly connected with the later emergence of the concept of convergence in income inequality, which was only proposed relatively recently by Bénabou [6]. Empirical findings for income inequality convergence in the EU (see Table 1) are inconclusive, as it has both been confirmed (e.g., by [11,12,13]) and limited only to some convergence clubs (e.g., [14]). Still, in this vein of research, the role of economic structures as determinants of income inequality convergence is nearly absent. Nevertheless, the role of structural change in income convergence in the EU has recently been empirically investigated by Cyrek [15].

In the structuralist tradition, Kuznets [46] identified the inverted-U relationships between GDP per capita and social inequality with the moderating role of structural change (the pattern is known as the Kuznet’s curve). According to Kuznet’s findings, at the early stage of development, there are only a few branches taking advantage of the economic growth, which increases inequality, while the following periods of spillover effects contribute to reducing inequality. Similarly, Schumpeter’s theory [47] of sector polarisation explains the role of innovations in new sectors, leading to an initial increase in inequality (cf. [48]). More recently, attention is being paid to the spillover perspective, as it is argued that job creation and income increase depend on both the sectoral total factor productivity (TFP) growth and cross-sectoral (input-output IO) multipliers [49].

The theory of structural dynamics ties the transformation of production to the positions of individuals in the labour market and, therefore, to the whole configuration of the society (e.g., [50,51,52], cf. [22]). It has been pointed out that certain trajectories of structural change may lead to undesirable outcomes, such as increased social disparities resulting from the loss of productive capacity and stable jobs (cf. [21,22]). It is expected that, in the long-term, structural change would create job opportunities in more productive sectors and overall increases in the employment level, and, as a result, would raise the income level of the population and lead to a more equal distribution. However, in the short-term, the higher demand for skilled workers for the expanding high-productivity sectors would cause an increase in labour inequality and, therefore, in total inequality. Therefore, the main point of connection between the sectoral composition and income inequality is the labour market and its institutions [53]. Adding the different roles of human capital in various activities is an important ingredient in understanding some key features of structural transformation [54]. Moreover, in the modern globalised world, it is also essential how an economy is interconnected with the global value chain. Some researchers (e.g., [55]) find that the growing complexity of an economy can bring about lower inequality. There are also studies that examine the role of trade in income convergence via its impact on structural transformation [43]. However, the discussion concerning specialisation-diversification issues is still open.

Generally, the literature prompts a complex pattern of relationships between structural change and income inequality regarding numerous interfering factors. All this brings about the need for further research into the matter.

3. Materials and Methods

The study focuses on the convergence in income inequality within 27 European Union countries in the years 2009–2021; however, it assumes that common structural features of the economies present a precondition for the process. Therefore, to consider structural features as conditioning convergence, some separate models are specified for different groups of the EU countries.

The sectional structure of employment makes a base for specifying groups of the EU countries that share common features and thus may form convergence clubs. The clusters have been identified using the Ward method with Euclidean distance. The economic structure of each economy is described by employment data classified into 19 sections (A-S) of NACE Rev. 2. The classification is based on data describing the average structure of employment in each country for 2008–2021, which allows to obtain clusters forming a stable base for the whole period. However, it is necessary to be aware of dynamic patterns of structural characteristics which make possible some shifts between clusters when comparing detailed years of analysis. The clustering for 2008 and 2021 are thus presented for comparisons.

The approach assuming that structural features form essential conditions for income inequality convergence influences the choice of the research period. To base on comparable data, it was necessary to start the analysis from 2008, as the NACE Rev. 2 classification of economic activities was introduced then. As structural change incorporated into the models of income inequality convergence is a dynamic process, it was necessary to calculate the variable representing the year-to-year change and the modelling period was limited to 2009–2021. The period thus reflects a situation influenced by the global financial crisis and the COVID-19 pandemic which both have their impact on the socio-economic development of the EU countries. These external shocks strongly determined both structural features and inequality. Therefore, the paper reflects the development processes in times of unexpected changes in socio-economic conditions and adjustments to them. The shocks connected with the Russian military aggression against Ukraine create another disturbance of the processes; however, the situation seems to be of a different (less universal) character leading to more diversified reaction across the EU member states. Because of geographical, historical, and political reasons, the shocks induced by military conflict could differently influence each of the EU countries—with strong distinction of the post-Soviet bloc, which make the assumptions of fixed clustering over the whole period more questionable. Therefore, the years after 2021 are to be included in future research.

The paper focuses on the problem of income inequality convergence; however, the distribution of income is strongly influenced by institutional setting. Therefore, the choice of data describing the category of income has essential meaning in modelling the convergence. Data about income inequality, measured by the Gini coefficient, were used to verify whether income inequality convergence appears across each of the EU’s structural clusters. Two alternative income categories were taken into account, e.g., equivalent disposable income before and after social transfers. Therefore, two alternative Gini coefficients have been used: Ginimar, for market incomes, and Ginisoc, for incomes after social transfers.

The study tests convergence, adopting the idea of beta convergence. It originally assumes that convergence occurs when a country with an initially lower level of income tends to grow faster and thus catches up and is evaluated using growth-initial regressions (e.g., [56,57,58], cf. [1]). This approach has also been used since the work of Bénabou [6] to test for convergence of income inequality. In a positive way, income inequality convergence occurs when a country with an initially higher level of inequality tends to decrease it at a higher speed. A regression model has been estimated to identify income inequality convergence in a form:

$$∆{Gini}_{it}=\alpha +\beta {Gini}_{it-1}+{\epsilon }_{it}$$

(1)

where β is a convergence parameter—when β adopts a negative sign and is statistically significant, it confirms convergence.

The models of convergence also make it possible to specify the level of income inequality the countries converge to, which is −α/β, or, when a logarithmic form of the variable is used, exp(−α/β) (e.g., [59]).

The basic model of convergence is then browsed to cover different determinants of income inequality. As the main focus of the study is to assess the role of structural features in the convergence process, a variable that represents a dynamic approach to structural features is incorporated into the convergence model. NAV (norm of absolute value) measure is used for structural change in the form specified by Kukuła [60] as:

$$\mathrm{N}\mathrm{A}\mathrm{V}={\displaystyle \frac{\sum _{i=1}^k\left|{\alpha }_{it}-{\alpha }_{i(t-1)}\right|}{2}}$$

(2)

where α is the share of each section of employment in the total employment, with i (i = 1, …, k) denoted for sections, and t is a specified year from the period 2009–2021.

Moreover, other variables that are identified in the literature as determinants of income inequality are used as control variables and included in the model. They cover the following: GDPpc—the general level of economic development as measured by GDP per capita, as it is assumed that wealthier societies focus their efforts on equalising income distribution; edu—the share of people with higher education, which represents the level of human capital, which is perceived as increasing incomes, and with growing human capital resources, an economy improves income distribution; gov—government spending as a percentage of GDP, as the state intervention is aimed at decreasing inequality; op—the sum of export and import values as a percentage of GDP, which measures the openness of the economy, as globalisation is perceived as one of the crucial modern determinants of inequality.

Finally, the convergence model takes the form of:

$$∆{Gini}_{it}=\alpha +\beta {Gini}_{it-1}+\delta {NAV}_{it-1}+\gamma Z_{it}+{\epsilon }_{it}$$

(3)

where NAV stands for structural change as the main determinant of convergence, while Z represents the control variables, successively: GDPpc, edu, gov, and op.

The models are estimated as panel fixed effects (FE) models. The choice is based on diagnostic panel tests that make it possible to decide between the panel OLS model, or the FE or RE panel models.

All data used in the study are derived from the Eurostat database.

4. Results

4.1. General Socio-Economic Situation—Descriptive Statistics

The general characteristics of the socio-economic situation across the 27 EU countries within the 2009–2021 period are described by data in

Table 2. Descriptive statistics of data used in the convergence models (2009–2021, 27 EU countries).

	Ginimar	Ginisoc	NAV	GDPpc	edu	gov	op
Avg.	49.13	29.79	0.020	25,568.83	26.95	45.93	1.40
CV (%)	9.31	12.96	50.69	65.74	28.33	15.25	63.71
min	37.2 (Slovakia 2018)	20.9 (Slovakia 2018, 2020)	0.005 (Germany 2018)	4970 (Bulgaria 2009)	11.2 (Romania 2009)	24.3 (Ireland 2019)	0.45 (Italy 2009)
max	61.6 (Greece 2013)	40.8 (Bulgaria 2019)	0.074 (Romania 2021)	84,750 (Luxembourg 2016)	45.2 (Ireland 2021)	64.9 (Ireland 2010)	4.80 (Lithuania 2020)

Source: own study based on Eurostat data.

. It is important to stress that social policy in the member states plays an important role as it limits income inequality from its average level of 49.13 before social transfers to only 29.79 after social transfers. However, the influence seems to be strongly diversified across countries as the coefficient of variation increases from merely about 9% for market income Gini to 13% for Gini after social transfers. Inequality, independently of the measure used, is the lowest in Slovakia, while its highest levels were noted in Greece in 2013 (income before social transfers) and in Bulgaria in 2019 (income after social transfers). Even stronger differences, however, are observed concerning dynamics of structural changes with the variation of 51%. Mature economies, such as Germany in 2018, experienced slower structural evolution, while it seems to be more dynamic in the ‘new’ EU members, such as Romania in 2021, especially in periods of overcoming crises (as 2021 expresses coverage after the COVID-19 pandemic). Essential differences between the EU countries concern their level of general socio-economic development, as the GDPpc variation is about 66% and ranges from about EUR 5000 in Bulgaria at the beginning of the period analysed to nearly EUR 85,000 in Luxembourg in 2016. The 27 EU countries are characterised by considerable resources of human capital, as the average share of the population with higher education is above one quarter, with the best performing example of Ireland in the last year of the period with a share of 45%. The results are both improved in time and with socio-economic development and the poorest results were noted for Romania in 2009. The analysed period is characterised by important changes in state policy as well, with usually growing state interventionism. Especially strong shifts were observed in Ireland that was specified both by the highest government spending (in 2010, which could be connected with the financial crisis aftermath) and the lowest (in 2019). There are also essential differences across the EU countries concerning the openness of their economies, with a general trend towards a growing share of external exchange in GDP. The most closed economy was the one of Italy in 2009, while the most open one was Lithuania in 2020, which expresses the general rule of the increasing openness of smaller countries.

4.2. Structural Features as Determinants of the EU Clustering

Structural features of an economy may determine the income inequality through intersectional differences in both the income for the human resources and the rates of return on invested capital. Supporting unique working conditions and based on unique qualifications, different sections of economic activity may simultaneously create opportunities to achieve above-average earnings while also generating barriers to inclusion into professional activity. Different levels of innovativeness of each section may also induce diversified profits for Schumpeterian entrepreneurs. As a consequence, different structural features may result in differences in income inequality convergence.

To identify common structural conditionings, the 27 EU countries were grouped with the use of the Ward method based on their sectional employment structure. The basic clustering uses the average shares of employment for the whole 2008–2021 period and forms stable frames to analyse the specificity of income inequality convergence within the groups. This average structure is representative of the entire period. However, it is also important to identify changes in a pattern of structural similarity and therefore alternative clustering were conducted separately for 2008 and 2021 (Figure 1).

The basic clustering results in specifying two groups of countries. The first group (11-CEE) consists of 11 countries and covers nearly all EU members that acceded in 2004 or later. These economies are located in Central and East Europe and are characterised by similar historical post-socialist conditions. The second group (16-WNSE) covers 16 countries, mostly the ‘old’ EU members (with the exception of Cyprus and Malta), that are geographically located in West, North, and South Europe. The common experience of these countries makes their employment structures similar. It appears that clustering according to the criterion of economic structure is almost identical to distinction as the period of EU membership and historical experience of institutional systems.

Nevertheless, there are some different patterns of structural development, which can be identified considering also structures in 2008 and 2021. The initial year of the analysis was specified by two clusters of 13 and 14 countries. The first group consists of mainly the CEE countries; however, Italy and Portugal were also included. Similarly, when considering the last year of the analysis—2021—two clusters of 15 and 12 members are derived. The first cluster covers not only the CEE countries but also Portugal, Austria, Italy, and Germany (forming a common sub-cluster). The comparisons suggest that dynamic year-to-year employment shifts between economic activities can transform structural patterns in reaction to external shocks and cyclical changes. Moreover, it is possible to observe that the most important changes in clustering (

Table 3. Stability of clusters—countries that have changed clusters concerning alternative classifications (2008 vs. 2008–2021 vs. 2021).

Changes in Clusters
2008 vs. 2021		2008 vs. 2008–2021		2021 vs. 2008–2021
Number (Countries)	Share	Number (Countries)	Share	Number (Countries)	Share
2 (Austria, Germany)	7.4%	2 (Italy, Portugal)	7.4%	4 (Austria, Italy, Germany, Portugal)	14.8%

Source: own study based on Eurostat data.

) were specified within the final period of the analysis and were connected with the COVID-19 shock. When comparing the average clustering to this based on 2008, there were only 2 countries that changed their group, but when clustering from 2021 is set against the average one the shifts concern 4 countries. Interestingly in the whole period—when comparing the groups formed based on 2008 vs. 2021 data, it can be observed that only Austria and Germany changed their clusters belonging. It may suggest that these mature economies experienced the important structural changes triggered by the pandemic, which disturbed their earlier patterns of development. Meanwhile, employment structures in Italy and Portugal reveal relatively low resilience to external shocks, indicating at still not well-established modern industrial structures in these economies. Nevertheless, their long-term economic ties within the European Communities make their average structural pattern closer to other ‘old’ member states than to the ‘new’ countries. Therefore, when considering longer periods the structural patterns reveal strong similarity to institutional experiences of the economies. Therefore, the clustering based on the average structures seems to better explain other patterns in socio-economic development.

It is also important to investigate the typical structure characteristic for each of the clusters. The countries forming the 11-CEE group are specified by a more traditional structure of employment (Figure 2). The share of labour engaged in agriculture and industry is greater, while, amongst service sections, only transportation features a higher employment share. Especially lower is the share of employment in such service sections as healthcare, modern professional, scientific and technical services, finance, or accommodation and food service activities. The difference results from the higher level of wealth of the societies in the 16-WNSE, which allows to use welfare services (meeting higher needs) to a greater extent and is based on the advanced and well-established economic ties that support business functions in the countries.

4.3. Income Inequality and Its Convergence Within the EU Clusters

The clusters of the EU countries also reveal significant differences concerning the scale of income inequality. The market income inequality is higher in the more affluent 16-WNSE group (with the average value for the years 2009–2021 at the level of 50.10) than in the poorer 11-CEE one (with the average of 47.72). Meanwhile, the income inequality after social transfers follows a different pattern and is lower in the 16-WNSE group (with the average value for the years 2009–2021 reaching 29.38) than in the 11-CEE group (with the average of 30.39). Therefore, the welfare state policy in the affluent mature economies of the ‘old’ EU members is more effective at levelling the adverse effects of market mechanisms than in the ‘catching-up’ countries of Central and Eastern Europe.

Considering the achievements towards social cohesion, which is an essential goal of European policy, it is important to identify whether convergence in income inequality is occurring within each of the clusters. The models of beta convergence in income inequality estimated for both groups of the EU countries (

Table 4. Parameters of models of convergence in inequality of disposable income before social transfers for the groups of 16-WNSE countries and 11-CEE countries in 2009–2021 (panel models with fixed effects, robust standard errors).

	Models for the 16-WNSE Countries Group:
	Model 1	Model 2	Model 3	Model 4	Model 5	Model 6
	dlnGinimar
const	1.28 ***	1.24 ***	1.45 ***	1.24 ***	1.13 ***	1.22 ***
lnGinimar_1	−0.33 ***	−0.32 ***	−0.35 ***	−0.32 ***	−0.32 ***	−0.31 ***
NAV_1		0.64	0.57	0.64	0.18	0.62
GDPpc			0.00 ***
edu				0.00
gov					0.00 ***
op						−0.01
LSDV R2	0.26	0.28	0.32	0.28	0.35	0.28
Within R2	0.22	0.24	0.29	0.24	0.32	0.24
AIC	−809.20	−811.49	−822.48	−809.53	−830.54	−809.81
BIC	−753.82	−752.86	−760.59	−747.64	−768.65	−747.91
HQI	−786.77	−787.74	−797.42	−784.46	−805.47	−784.74
DW	1.60	1.65	1.66	1.65	1.77	1.66
p for F test	0.00	0.00	0.00	0.00	0.00	0.00
Diagnostic panel tests
panel OLS vs. FE	F = 3.13, p = 0.00	F = 3.29, p = 0.00	F = 4.10, p = 0.00	F = 3.14, p = 0.00	F = 3.97, p = 0.00	F = 2.67, p = 0.00
RE vs. FE	H = 36.40, p = 0.00	H = 39.71, p = 0.00	H = 53.95, p = 0.00	H = 37.95, p = 0.00	H = 49.60, p = 0.00	H = 32.01, p = 0.00
final choice	FE	FE	FE	FE	FE	FE
	Models for the 11-CEE countries group:
	Model 1	Model 2	Model 3	Model 4	Model 5	Model 6
	dlnGinimar
const	1.16 ***	1.10 ***	1.61 **	1.37 **	1.04 ***	1.11 ***
lnGinimar_1	−0.30 ***	−0.29 ***	−040 **	−0.34 **	−0.28 ***	−0.29 ***
NAV_1		1.01**	0.45	0.65 *	0.97 **	0.98 **
GDPpc			−0.00 **
edu				−0.00 **
gov					0.00
op						−0.01
LSDV R2	0.19	0.24	0.33	0.30	0.25	0.25
Within R2	0.16	0.22	0.30	0.27	0.22	0.22
AIC	−571.95	−579.30	−592.47	−586.75	−578.59	−577.56
BIC	−537.36	−541.82	−552.11	−546.39	−538.24	−537.20
HQI	−557.90	−564.07	−576.07	−570.35	−562.19	−561.16
DW	1.77	1.88	1.97	1.91	1.95	1.89
p for F test	0.00	0.01	0.05	0.06	0.00	0.01
Diagnostic panel tests
panel OLS vs. FE	F = 2.53, p = 0.01	F = 2.85, p = 0.00	F = 3.45, p = 0.00	F = 3.83, p = 0.00	F = 2.97, p = 0.00	F = 2.85, p = 0.00
RE vs. FE	H = 21.21, p = 0.00	H = 24.52, p = 0.00	H = 30.37, p = 0.00	H = 34.98, p = 0.00	H = 26.01, p = 0.00	H = 24.81, p = 0.00
final choice	FE	FE	FE	FE	FE	FE

Statistical significance at a level: *** 0.01; ** 0.05; * 0.1. Source: own study based on Eurostat data.

and

Table 5. Parameters of models of convergence in inequality of disposable income after social transfers for the groups of 16-WNSE countries and 11-CEE countries in 2009–2021 (panel models with fixed effects, robust standard errors).

	Models for the 16-WNSE Countries Group:
	Model 1	Model 2	Model 3	Model 4	Model 5	Model 6
	dlnGinisoc
const	1.27 ***	1.26 ***	1.39 ***	1.26 ***	1.19 ***	1.26 ***
lnGinisoc_1	−0.38 ***	−0.38 ***	−0.40 ***	−0.38 ***	−0.37 ***	−0.37 ***
NAV_1		0.72**	0.68 **	0.71 **	0.44	0.70 **
GDPpc			−0.00 ***
edu				−0.00
gov					0.00 ***
op						−0.01
LSDV R2	0.18	0.21	0.25	0.21	0.24	0.21
Within R2	0.16	0.19	0.23	0.19	0.22	0.19
AIC	−825.02	−829.01	−837.86	−827.01	−835.17	−827.35
BIC	−769.64	−770.38	−775.97	−765.12	−773.27	−765.46
HQI	−802.59	−805.27	−812.80	−801.95	−810.10	−802.29
DW	1.75	1.76	1.79	1.76	1.82	1.77
p for F test	0.00	0.00	0.00	0.00	0.00	0.00
Diagnostic panel tests
panel OLS vs. FE	F = 2.38, p = 0.00	F = 2.55, p = 0.00	F = 3.09, p = 0.00	F = 2.24, p = 0.01	F = 2.97, p = 0.00	F = 2.42, p = 0.00
RE vs. FE	H = 32.75, p = 0.00	H = 35.59, p = 0.00	H = 44.27, p = 0.00	H = 32.30, p = 0.00	H = 42.60, p = 0.00	H = 34.26, p = 0.00
final choice	FE	FE	FE	FE	FE	FE
	Models for the 11-CEE countries group:
	Model 1	Model 2	Model 3	Model 4	Model 5	Model 6
	dlnGinisoc
const	1.00 ***	0.98 ***	1.14 ***	1.09 ***	1.02 ***	1.01 ***
lnGinisoc_1	−0.29 ***	−0.29 ***	−0.31 ***	−0.30 ***	−0.29 ***	−0.29 ***
NAV_1		−0.53	−0.93 *	−0.82 *	−0.49	−0.55
GDPpc			−0.00 **
edu				−0.00 *
gov					−0.00
op						−0.01
LSDV R2	0.19	0.20	0.24	0.23	0.20	0.20
Within R2	0.15	0.16	0.20	0.19	0.17	0.16
AIC	−508.74	−508.25	−512.98	−511.10	−506.77	−506.48
BIC	−474.14	−470.77	−472.62	−470.74	−466.41	−466.12
HQI	−494.68	−493.02	−496.58	−494.70	−490.37	−490.08
DW	2.07	2.11	2.22	2.17	2.09	2.12
p for F test	0.00	0.00	0.00	0.00	0.00	0.00
Diagnostic panel tests
panel OLS vs. FE	F = 2.79, p = 0.00	F = 2.73, p = 0.00	F = 2.89, p = 0.00	F = 3.24, p = 0.00	F = 2.68, p = 0.01	F = 2.72, p = 0.00
RE vs. FE	H = 22.69, p = 0.00	H = 22.56, p = 0.00	H = 24.30, p = 0.00	H = 27.99, p = 0.00	H = 22.60, p = 0.00	H = 22.77, p = 0.00
final choice	FE	FE	FE	FE	FE	FE

Statistical significance at a level: *** 0.01; ** 0.05; * 0.1. Source: own study based on Eurostat data.

) unambiguously indicate that such convergence is taking place concerning both market income and income after social transfers. In all these models, the β parameter is statistically significant with a negative sign, and the sign is not influenced by introducing any of the control variables.

Moreover, some robustness checking allowed to test whether the COVID-19 pandemic essentially influenced the convergence (

Table 6. Robustness checking—parameters of models of convergence in inequality in 2009–2019 (panel models with fixed effects, robust standard errors).

	Models for the 16-WNSE Countries Group:		Models for the 11-CEE Countries Group:
inequality of disposable income before social transfers
	Model 1	Model 2	Model 1	Model 2
	dlnGinimar
const	1.35 ***	1.27 ***	1.29 ***	1.18 **
lnGinimar_1	−0.34 ***	−0.33 ***	−0.33 ***	−0.31 **
NAV_1		0.96		1.00 **
LSDV R2	0.33	0.37	0.21	0.27
Within R2	0.27	0.31	0.17	0.23
AIC	−711.62	−718.99	−481.13	−488.10
BIC	−659.34	−663.63	−448.72	−452.99
HQI	−690.39	−696.51	−467.98	−473.86
DW	1.44	1.52	1.54	1.68
p for F test	0.00	0.00	0.01	0.03
inequality of disposable income after social transfers
	Model 1	Model 2	Model 1	Model 2
	dlnGinisoc
const	1.49 ***	1.47 ***	1.17 ***	1.15 ***
lnGinisoc_1	−0.44 ***	−0.44 ***	−0.34 ***	−0.33 ***
NAV_1		1.03 **		−0.56
LSDV R2	0.21	0.26	0.22	0.23
Within R2	0.19	0.24	0.17	0.18
AIC	−701.94	−710.24	−416.67	−416.18
BIC	−649.67	−654.89	−384.26	−381.08
HQI	−680.72	−687.76	−403.52	−401.94
DW	1.58	1.58	1.78	1.83
p for F test	0.00	0.00	0.00	0.00

Statistical significance at a level: *** 0.01; ** 0.05. Source: own study based on Eurostat data.

). It revealed that the process remains similar in the case of excluding the last years of analysis (2020 and 2021) as well as in the basic case reflecting the external shock. It indicates that the pandemic did not change the general pattern of income inequality convergence for the whole period. Nevertheless, the β parameters are a little bit lower in models with the pandemic years included, therefore suggesting some disturbing effect of the COVID-19 situation. The difference is stronger concerning the income after social transfer, indicating that social cohesion policy appeared less efficient in times of the pandemic and illustrating the effect of an autarkic turn across the EU member states.

The estimated models, however, make it possible to identify some specificity of the income inequality convergence within each of the clusters. The speed of convergence, concerning both the market income and the income after social transfers, appears higher in the 16-WNSE cluster than in the 11-CEE group. The member states characterised by a shorter period of market economy often shape their institutions in a different way and this fact can reduce the dynamics of levelling the income inequality differences. In the ‘old’ member states, multidimensional integration is more advanced, and results in faster convergence of income inequality.

Moreover, there are essential differences concerning the role that changes in the sectional structure of employment play for the convergence in each cluster. Within the 11-CEE cluster, structural change appears as a significant factor stimulating changes in market income inequality, while this is not the rule for the 16-WNSE cluster. Dynamic market change, in the form of labour reallocation, strengthens the increase in market income inequality within the group of ‘new’ member states. Nevertheless, the effect is absent, or even opposite (with a negative sign), concerning income after social transfers in the 11-CEE. Therefore, the essential influence of the welfare state policy within the group of countries can be concluded. It reduces the economic disparities that are created by market mechanisms of labour reallocation.

On the other hand, within the 16-WNSE group, structural change is not a significant determinant of change in market income inequality, while it appears as a stimulating factor concerning inequality in income after social transfers. The relationship reveals that income inequality within the cluster is more long-term in nature. Inequality is not purely stimulated by recent structural change regarding market distribution, rather the inequality is transferred to social policy by sectional relations. Structural features must have created some specific institutions that took a decisive role in modelling social transfers. Thus, structural change influences the distribution of income, but the effect is moderated by the modifying mechanisms of redistribution. Moreover, it is possible that structural changes (e.g., the rise of the gig economy, automation of mid-skill jobs) are creating new forms of income precarity that mature welfare systems are not designed to handle, thus making transfers less effective at reducing inequality. It reveals a necessity to reform the established institutional solutions to make them capable of mitigating problems of poverty and exclusion in new conditions of a knowledge economy.

The robustness checking (Table 6) allows to conclude about stable relations concerning the influence of structural change on the convergence independently of the pandemic shock. The signs and statistical significance of the parameters reflecting structural change remained unchanged. Nevertheless, the stimulating role of structural change appeared to be stronger in the period before COVID-19 in the case of the ‘old’ member states concerning income after social transfers (parameter equal to 1.03 vs. 0.72). It proves that social policy in the mature economies responded to the pandemic in a different way than in the ‘new’ countries, reducing the channel of structural change effect.

4.4. Target Levels of Income Inequality in the Convergence Models

The estimated models of income inequality convergence make it possible to specify a target level of coherence that is expected to be a common prospect of converging countries (

Table 7. Target levels of income inequality as compared to 2021.

	Income Inequality Before Social Transfers		Income Inequality After Social Transfers
	16-WNSE	11-CEE	16-WNSE	11-CEE
The average level in 2021	50.74	46.98	29.10	29.90
The target level	50.54	47.65	29.24	29.90

Source: own study based on Eurostat data.

).

Once again, there are different prospects derived from the models concerning income inequality levels within the two clusters of the EU countries. Therefore, structural features differentiate the evolution of income inequality.

In the mature economies of the 16-WNSE, market income inequality is expected to slightly decline, while in the 11-CEE, conversely, it appears that market income inequality increases, and the effect is stronger than that for the 16-WNSE. The initial level of income inequality before social transfers, as specified in 2021, is much higher in the 16-WNSE countries than in the 11-CEE. Therefore, the convergence translates into an increase in inequality within the countries with a lower initial level and decreases in the inequality within the countries with a higher initial level. However, the former process prevails. It suggests also convergence of income inequality across the whole EU, which was in fact identified by Cyrek [15].

The situation is different with regard to income inequality after social transfers. It is initially lower in the more affluent economies of the 16-WNSE, and the inequality is expected to slightly grow within the group, suggesting decreasing efficiency of traditional welfare policy in conditions of new forms of social division (especially connected with digital technology development and new job creation). Meanwhile, within the ‘new’ member states, an initial high level of inequality is expected to stabilise in value.

Therefore, the identified patterns of convergence do not lead to growing social cohesion within the European society. Even though the disparities across countries are to be reduced, generally, income inequality is expected to increase. A social policy that strongly reduces inequality through transfers, especially within the 16-WNSE, seems to not be able to smooth the disparities enough in the poorer countries of the 11-CEE. Simultaneously, its influence is weakening in the ‘old’ 16-WNSE. It is therefore expected that, along with dynamic structural change, social inequality will increase across the whole EU.

5. Conclusions

The conducted research under income inequality convergence concerning the structural conditions of the EU countries makes it reasonable to draw some conclusions about some important phenomena.

First, considering the structural features of the EU economies, their classification leads to distinguishing groups that are similar in view of their experiences concerning the date of accession to the EU and their history of socio-economic systems. The observation points to historically shaped institutions as important determinants of economic structures in terms of the sectional allocation of labour and, as a result, the market distribution of income.

Second, income inequality convergence occurs within both clusters of the EU countries specified based on their sectional structures of employment. This tendency has been confirmed concerning both inequality in income before and after social transfers. Simultaneously, the convergence in income inequality is faster within the group of ‘old’ EU member states; however, these economies experience a higher level of market inequality. It suggests that the observed pattern of structural development in the EU initially increases income inequality in the market and then the effect weakens, and this is an expected path for the ‘new’ member states to follow.

Third, the pattern of convergence is specified by an increase in income inequality in the economies with a lower initial level, while income inequality decreases in the countries that are starting with more serious problems of distribution. However, the increase in income inequality appears to be stronger than the decline, which does not lead to better social cohesion. The influence of social policy on smoothening income distribution through transfers is much stronger in the ‘old’ EU countries but is weakening, while it is not strong enough in the group of the ‘new’ EU countries to offset the growing market income inequality and ensure lower inequality in income after social transfers. The problem may increase in time along with new structural patterns of development into a knowledge-based economy that bring new social tensions which are not addressed by traditional institutional settings.

Finally, the influence of structural change on the change in inequality differs between the group of the ‘new’ 11-CEE countries and the ‘old’ 16-WNSE countries (e.g., between groups of countries with different structural features). Within the 11-CEE, structural change stimulates an increase in market income inequality, while the effect is insignificant concerning inequality in income after social transfers. Meanwhile, within the 16-WNSE group of the ‘old’ member states, structural change influences inequality only concerning income after transfers. Therefore, the essential differences in the mechanisms of the policy on social transfers are identified between the two groups of countries. They could be strengthened by differences in structural advancement into a knowledge-based economy that brings about new lines of social divisions and economic exclusion and therefore new challenges for social policy.

The main findings of the study are generally consistent with previous research that concludes about income inequality convergence within the EU countries but with growing level of inequality (e.g., [11,12,13,15]). Nevertheless, the study pays attention to a rather neglected factor of structural features in explaining the process. It stresses that the sectional distribution of employment cannot be omitted when researching the income inequality and the patterns of convergence can be influenced by the structural specificity of a country.

However, the paper is not free from limitations. First, the period of analysis is relatively short and includes unusual shocks (e.g., the financial crisis aftermath, COVID-19) that could influence the results as they both disturbed the structural patterns of development and brought new problems of socio-economic exclusion. Moreover, it does not consider the shock connected with the Russia-Ukraine war that seems to noticeably affect the structural conditions of income inequality convergence in Europe, especially in the east EU countries. Future research could address the problems connected with the exogenous shocks (e.g., with dummy variables) and consider a longer period. Second, the endogeneity problem may influence the results. It could be connected with a potentially bi-directional relationship between structural change and inequality. High levels of inequality might, for example, hinder human capital development and thus slow down structural transformation towards a knowledge-based economy. Addressing these problems is a task for future research that could use new methods of statistical estimation of the processes (e.g., instrumental variables). Furthermore, the paper is based on simple models with relatively low explanatory power (low R square levels resulting from limited number of variables in each model) and give only some basis investigation into the determinants of income inequality convergence. They could omit important factors (e.g., the political orientation of governments that influence both social spending and industrial policy) and extending the range of determinants is also an open space for future research. Another potential direction of research is to browse the extent of metrics specifying income inequality. It would be possible to use alternative income inequality measures such as quantile ratio, the Atkinson or Theil index as well as focus on poverty rates or material deprivation indicators to provide a more in-depth picture of social cohesion. Therefore, the paper is just an attempt to pay attention that structural features and change are not indifferent to socio-economic cohesion and opens a space for future scientific discussion.

To conclude, an urgent need emerges to design new solutions for social policy concerning the role of the structural features of the economy and their changes. It is important to support employability in dynamically changing market conditions influenced by accelerating technological progress and structural shifts. It requires the inclusion of all social groups in lifelong learning to leverage the level of human capital and increase the flexibility to adapt to new technologies. An important part of the efforts should be specified by new institutional solutions enabling the creation of ‘good’ workplaces that ensure stable, just income, and social security. Flexicurity models of employment fit the recommendations. The dual education solutions may also help to adapt better to the new market conditions.

The study pays attention to the institutional basis of relationships in the labour market as well as solutions concerning the redistribution and transfer policy as important determinants of social cohesion across the EU. Structural patterns of human resource engagement reveal themselves as an indicator of institutional experiences and require further in-depth research.