Abstract
This paper explores the changes in inequality and welfare between EU regions at the NUTS 3 level over the 2003–2011 period. Changes in absolute and relative inequalities are broken down into components explaining the effects of population change, re-ranking of regions and income growth between regional per capita incomes. Each component of inequality change is further decomposed by subgroup, revealing the contributions arising from changes within subgroups and from changes between subgroups. The decomposition of the change in absolute inequality is used to develop a decomposition of the change in welfare between EU regions.
Similar content being viewed by others
Notes
The European Regional Development Fund, the Cohesion Fund and the European Social Fund provide the necessary investments to meet EU regional policies aimed at reducing economic and social disparities and promoting inclusive growth. Together these Funds will allocate almost a third of the total EU budget over the 2014–2020 period (The European Commission 2015).
As noted by Milanovic (2006), the population-weighted approach is often used for measuring international inequality since it requires only information on per capita incomes and population sizes. An alternative approach measures international inequality without weighting per capita incomes by the respective population sizes. If this approach were used in our study, all regions would be treated as equally important irrespective of their population sizes.
A complete review of the various approaches to the calculation of the Gini index is provided by Xu (2004).
The change in the Gini index between two points in time is decomposed by Bönke et al. (2010) to explain the effects of changes in various income sources on the change in overall income inequality.
If \(s_{\mathbf{a},ij}\) is positive and \(i>j\), \(s_{\mathbf{a},ji}\) is negative and equal to \(-s_{\mathbf{a},ij}\) as matrix \({\mathbf {S}} _{\mathbf{a }}\) is skew-symmetric.
Suppose that \({\mathbf {y}} _{t}=\left( y_{1,t}=5.5,y_{2,t}=5,y_{3,t}=3.5,y_{4,t}=1\right) ^{T}\) contains the per capita incomes of four regions in t sorted in decreasing order and that \({\mathbf {p}} _{t}=\left( p_{1,t}=0.4,p_{2,t}=0.2,p_{3,t}=0.1,p_{4,t}=0.3\right) ^{T}\) includes the corresponding population shares in t. Let \({\mathbf {y}} _{t+1\left| t\right. }=\left( y_{1,t+1\left| t\right. }=6,y_{2,t+1\left| t\right. }=8,y_{3,t+1\left| t\right. }=4,y_{4,t+1\left| t\right. }=2\right) ^{T}\) be the per capita incomes in \(t+1\) sorted by the decreasing order of the respective per capita incomes in t. The vector of the \(t+1\) per capita incomes sorted in decreasing order is \({\mathbf {y}} _{t+1}=\left( y_{1,t+1}=8,y_{2,t+1}=6,y_{3,t+1}=4,y_{4,t+1}=2\right) ^{T}\) and the vector of the corresponding population shares is \({\mathbf {p}} _{t+1}=\left( p_{1,t+1}=0.25,p_{2,t+1}=0.35,p_{3,t+1}=0.15,p_{4,t+1}=0.25\right) ^{T}\). Consequently, \({\mathbf {p}} _{t\left| t+1\right. }=\left( p_{1,t\left| t+1\right. }=0.2,p_{2,t\left| t+1\right. }=0.4,p_{3,t\left| t+1\right. }=0.1,p_{4,t\left| t+1\right. }=0.3\right) ^{T}\). For example, when considering the two richest regions by per capita income in \(t+1\), we obtain \(s_{\mathbf{a},21}=0.0075\) (and obviously \(s_{\mathbf{a},12}=-0.0075\)) in the \(4\times 4\) matrix \({\mathbf {S}} _{\mathbf{a }}\). This suggests that the population change increases the weight of the absolute inequality between the two regions.
Since \({\mathbf {R}} _{\mathbf{a }}\) is skew-symmetric, \(r_{\mathbf{a},ji}\) is equal to \(-2p_{i,t\left| t+1\right. }p_{j,t\left| t+1\right. }\).
Referring to the numerical example shown in footnote 6, re-ranking occurs between the two regions with the highest per capita incomes in t when moving to \(t+1\). The \(4\times 4\) matrix \({\mathbf {R}} _{\mathbf{a }}\) detects the re-ranking by showing \(r_{\mathbf{a},21}=2p_{2,t\left| t+1\right. }p_{1,t\left| t+1\right. }\) and \(r_ {\mathbf{a},12}=-2p_{2,t\left| t+1\right. }p_{1,t\left| t+1\right. }\), whereas its remaining elements are equal to zero; that is, \(r_{\mathbf{a},21}=0.16\) and \(r_{\mathbf{a},12}=-0.16\).
Both \({\mathbf {A}}_{t}\) and \({\mathbf {A}}_{t+1\left| t\right. }\) are obtained by keeping regions sorted in decreasing order of their per capita incomes in t. The difference in the \(\left( i,j\right)\)-th entry of \({\mathbf {A}}_{t}\) and the difference in the \(\left( i,j\right)\)-th entry of \({\mathbf {A}}_{t+1\left| t\right. }\) are calculated between the per capita incomes of the same two regions at times t and \(t+1\), respectively.
In the numerical example cited in footnote 6, the \(4\times 4\) matrix \({\mathbf {D}} _{\mathbf{a }}\) has \(d_{\mathbf{a},41}=0.5\) (and obviously \(d_{\mathbf{a},14}=-0.5\)). This indicates that the absolute inequality between the regions with the highest and lowest per capita incomes in t decreases from t to \(t+1\).
Duro isolated the population change component of \({\varDelta }G\) by using a general decomposition approach suggested by Esteban (1994) to measure the effect of the change in population on an inequality index.
R coincides with the Atkinson-Plotnick re-ranking coefficient used to measure re-ranking between income receivers in income distribution (Jenkins and Van Kerm 2006). The Atkinson-Plotnick re-ranking coefficient equals 0 if the ranking of income receivers is unchanged from t to \(t+1\), whereas it is equal to two times the Gini index in \(t+1\) if the ranking in \(t+1\) is the inverse of the ranking in t.
The Gini index, expressed as the half of the Gini relative mean difference in Eq. 1, was broken down by subgroup in Dagum (1997) and in Costa (2008). More recently, Mornet et al. (2013) suggested a subgroup decomposition of the \(\alpha\)-Gini, an extended version of the Gini index in Eq. 1 including a parameter of inequality aversion.
“Appendix 1” shows that \({\mathbf {A}}_{h,t+1}\) and \({\mathbf {D}}_{\mathbf{a},h}\) are obtained from \({\mathbf {A}}_{t+1}\) and \({\mathbf {D}} _{\mathbf{a }}\), respectively.
In “Appendix 1”, \({\mathbf {A}}_{gh,t+1}\) and \({\mathbf {D}}_{\mathbf{a},gh}\) are obtained respectively from \({\mathbf {A}}_{t+1}\) and \({\mathbf {D}} _{\mathbf{a }}\).
Since \(S_{a}\) enters negatively in the decomposition of \({\varDelta }W_{a}\), a negative value of \(S_{a}\) increases welfare.
Esteban et al. (2007) applied the Davies and Shorrocks procedure to divide the income distribution into subgroups in their study of polarization in five OECD countries. They found the optimal partitions for \(k=2,3,4\).
Ten new countries joined the EU in May 2004 : Czech Republic, Estonia, Cyprus, Latvia, Lithuania, Hungary, Malta, Poland, Slovakia and Slovenia. Two more countries from Eastern Europe, Bulgaria and Romania, joined the EU in January 2007. Croatia joined the EU in July 2013.
UK, Ireland and Sweden implemented a free movement policy for citizens of new EU member countries since 2004, while other EU-15 member countries retained some temporary restrictions (Kahanec and Zimmermann 2009).
In particular, subgroup 1 includes several regions of the Four Motors for Europe: Baden-Württenberg (Germany), Catalonia (Spain), Lombardy (Italy), and Rhône-Alpes (France).
A value of 0.000000 shown in tables means that the observed value is <0.000001.
Let \(\mathbf X\) and \({\mathbf {Y}}\) be \(n\times n\) matrices. The Hadamard product \(\mathbf X \odot {\mathbf {Y}}\) is defined as the \(n\times n\) matrix with the \(\left( i,j\right)\)-th element equal to \(x_{ij}y_{ij}\). The Hadamard product is the element-by-element matrix product (Abadir and Magnus 2005).
References
Abadir, K. M., & Magnus, J. R. (2005). Matrix algebra. Cambridge: Cambridge University Press.
Abel, G., & Sander, N. (2014). Quantifying global international migration flows. Science, 343, 1520–1522.
Atkinson, A. B. (1970). On the measurement of inequality. Journal of Economic Theory, 2, 244–263.
Bönke, T., Schröder, C., & Schulte, K. (2010). Incomes and Inequality in the long run: The case of German Elderly. German Economic Review, 11, 487–510.
Bosmans, K., Decancq, K., & Decoster, A. (2014). The relativity of decreasing inequality between countries. Economica, 81, 276–292.
Coleman, D. (2008). The demographic effects of international migration in Europe. Oxford Review of Economic Policy, 24, 452–476.
Costa, M. (2008). Gini index decomposition for the case of two subgroups. Communications in Statistics - Simulation and Computation, 37, 631–644.
Dagum, C. (1997). A new approach to the decomposition of Gini income inequality ratio. Empirical Economics, 22, 515–531.
Davies, J. B., & Shorrocks, A. F. (1989). Optimal grouping of income and wealth data. Journal of Econometrics, 42, 97–108.
Duro, J. A. (2013). Weighting vectors and international inequality changes in environmental indicators: An analysis of CO\(_{2}\) per capita emissions and Kaya factors. Energy Economics, 39, 122–127.
Esteban, J., Gradín, C., & Ray, D. (2007). An extension of a measure of polarization, with an application to the income distribution of five OECD countries. The Journal of Economic Inequality, 5, 1–19.
Esteban, J. M. (1994). Un análisis de la convergencia regional en Europa. In J. M. Esteban & X. Vives (Eds.), Crecimiento y Convergencia en España y Europa (Vol. 3). Instituto de Análisis Económico y Fundación de Economía Analítica.
Fredriksen, K. B. (2012). Income inequality in the European Union. Technical Report 952, OECD Economics Department
Jenkins, S. P., & Van Kerm, P. (2006). Trends in income inequality, pro-poor income growth, and income mobility. Oxford Economic Papers, 58, 531–548.
Kahanec, M., & Zimmermann, K. (2009). Migration in an enlarged EU: A challenging solution? Technical Report 363, European Commission/Economic Papers.
Kanbur, R., & Venables, A. (2005). Spatial inequality and development. In R. Kanbur & A. Venables (Eds.), Spatial inequality and development (pp. 3–11). Oxford: oxford University Press.
Lessmann, C. (2014). Spatial inequality and development—Is there an inverted-U relationship? Journal of Development Economics, 106, 35–51.
Milanovic, B. (2006). Global income inequality: What it is and why it matters? Technical Report 26, United Nations/Department of Economic and Social Affairs
Mornet, P., Zoli, C., Mussard, S., Sadefo-Kamdem, J., Seyte, F., & Terraza, M. (2013). The (\(\alpha,\beta\))-multi-level \(\alpha\)-Gini decomposition with an illustration to income inequality in France in 2005. Economic Modelling, 35, 944–963.
Mussini, M. (2013a). On decomposing inequality and poverty changes over time: A multi-dimensional decomposition. Economic Modelling, 33, 8–18.
Mussini, M. (2013b). A subgroup decomposition of the inequality change over time. Applied Economics Letters, 20, 386–390.
Mussini, M. (2014). Decomposing inequality change from the perspective of reranking and income growth between income groups. Empirical Economics, 47, 619–637.
Mussini, M., & Grossi, L. (2015). Decomposing changes in CO\(_{2}\) emission inequality over time: The roles of re-ranking and changes in per capita CO\(_{2}\) emission disparities. Energy Economics, 49, 274–281.
Piacentini, M. (2014). Measuring income inequality and poverty at the regional level in OECD countries. Technicl report, OECD Publishing.
Silber, J. (1989). Factor components, population subgroups and the computation of the Gini index of inequality. The Review of Economics and Statistics, 71, 107–115.
Silber, J., & Weber, M. (2005). Gini’s mean difference and the measurement of absolute mobility. Metron, 63, 471–492.
The European Commission. (2015). The EU’s main investment policy. http://ec.europa.eu/regional_policy/en/policy/what/investment-policy/. Accessed June 30, 2015.
Xu, K. (2004). How has the literature on Gini index evolved in the past 80 years? Technical Report, Dalhousie University, Department of Economics Research Paper.
Zaiceva, A., & Zimmermann, K. (2008). Scale, diversity, and determinants of labour migration in Europe. Oxford Review of Economic Policy, 24, 427–451.
Acknowledgments
I thank the editor and an anonymous reviewer for insightful comments, which helped to improve the manuscript.
Author information
Authors and Affiliations
Corresponding author
Appendices
Appendix 1: Calculation of \(\mathbf {A}_{h,t+1}\), \(\mathbf {D}_{\mathbf{a},h}\), \(\mathbf {A}_{gh,t+1}\), \(\mathbf {D} _{\mathbf{a},gh}\)
Let \({\mathbf {w}} _{h,t}\) be the \(k\times 1\) vector with nonzero elements equal to 1 in the corresponding places of the t per capita incomes of the regions belonging to subgroup h (with \(h=1,\ldots ,r\)) in \({\mathbf {y}} _{t}\). The \(k\times k\) matrix \({\mathbf {W}} _{h,t}={\mathbf {w}} _{h,t}{\mathbf {w}} ^{T}_{h,t}\) has its \(\left( i, j\right)\)-th element equal to 1 if and only if the \(\left( i, j\right)\)-th element of \({\mathbf {A}}_{t}\) is the difference between the per capita incomes of two regions belonging to subgroup h, otherwise the \(\left( i,j\right)\)-th element of \({\mathbf {W}} _{h,t}\) is equal to 0. By using the Hadamard product,Footnote 22 the pairwise differences between the per capita incomes of the regions in subgroup h are drawn from \({\mathbf {A}}_{t}\):
The pairwise differences between the \(t+1\) per capita incomes of subgroup h in \({\mathbf {A}}_{t+1\left| t\right. }\) occupy the same entries in which the pairwise differences between the t per capita incomes of subgroup h are arranged in \({\mathbf {A}}_{t}\). Thus, \({\mathbf {W}} _{h,t}\) can also be used to select the pairwise differences between the \(t+1\) per capita incomes of subgroup h from \({\mathbf {A}}_{t+1\left| t\right. }\):
\({\mathbf {w}} _{h,t+1}\) being the \(k\times 1\) vector with nonzero elements equal to 1 in the corresponding places of the \(t+1\) per capita incomes belonging to regions of subgroup h in \({\mathbf {y}} _{t+1}\), the \(k\times k\) matrix \({\mathbf {W}} _{h,t+1}={\mathbf {w}} _{h,t+1}{\mathbf {w}} ^{T}_{h,t+1}\) selects the pairwise differences between the \(t+1\) per capita incomes of regions within subgroup h from \({\mathbf {A}}_{t+1}\):
Since \({\mathbf {D}} _{\mathbf{a }}={\mathbf {A}}_{t}-{\mathbf {A}}_{t+1\left| t\right. }\), the Hadamard product between \({\mathbf {W}} _{h,t}\) and \({\mathbf {D}} _{\mathbf{a }}\) is the matrix whose nonzero elements are the elements of \({\mathbf {D}} _{\mathbf{a }}\) measuring the change in the absolute disparities between the per capita incomes of regions within subgroup h:
The \(k\times k\) matrix \({\mathbf {W}} _{gh,t}={\mathbf {w}} _{g,t}{\mathbf {w}} ^{T}_{h,t}+{\mathbf {w}} _{h,t}{\mathbf {w}} ^{T}_{g,t}\) shows nonzero elements equal to 1 in the entries corresponding to the pairwise differences between the per capita incomes of subgroup g and those of subgroup h in both \({\mathbf {A}}_{t}\) and \({\mathbf {A}}_{t+1\left| t\right. }\); hence, \({\mathbf {W}} _{gh,t}\) selects the between-group pairwise differences from both the matrices:
and
The Hadamard product between \({\mathbf {W}} _{gh,t}\) and \({\mathbf {D}} _{\mathbf{a }}\) selects the elements of \({\mathbf {D}} _{\mathbf{a }}\) measuring the change in the disparities between the per capita incomes of regions in subgroup g and the per capita incomes of regions in subgroup h:
The \(k\times k\) matrix \({\mathbf {W}} _{gh,t+1}={\mathbf {w}} _{g,t+1}{\mathbf {w}} ^{T}_{h,t+1}+{\mathbf {w}} _{h,t+1}{\mathbf {w}} ^{T}_{g,t+1}\) has nonzero elements equal to 1 in the entries corresponding to the pairwise differences between the per capita incomes of subgroup g and the per capita incomes of subgroup h in \({\mathbf {A}}_{t+1}\). Thus, the matrix
comprises the pairwise differences between the \(t+1\) per capita incomes of subgroup g and those of subgroup h.
Appendix 2: Maps of EU Regions Divided in Two and Three Subgroups in 2007
Figure 3 shows the partition of regions in 2007 for \(k=2\). Figure 4 shows the partition of regions in 2007 for \(k=3\).
Rights and permissions
About this article
Cite this article
Mussini, M. Decomposing Changes in Inequality and Welfare Between EU Regions: The Roles of Population Change, Re-Ranking and Income Growth. Soc Indic Res 130, 455–478 (2017). https://doi.org/10.1007/s11205-015-1184-2
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s11205-015-1184-2