Regional welfare disparities: the case of the European Union

Abstract

The regional welfare distribution in the European Union between 1993 and 1998 is examined, using several complementary methodologies. The results obtained show a reduction in regional welfare disparities over the period analysed. It is worth noting, in this respect, however, that regional productivity differences prove to be the main determinant behind observed welfare inequality in the European context. Moreover, there has also been a decline in regional bipolarization over the six-year contemplated, while the degree of observed intradistributional mobility is relatively low. The empirical evidence presented, nevertheless, reveals the importance of variables such as the national component, the spatial location, the regional productive structure or the percentage of GDP devoted to investment or to R&D expenditure, in accounting for the dynamics of the distribution under analysis.

Acknowledgements

The authors wish to acknowledge the financial support from MCYT (Project BEC2002-03941) and the Fundación BBVA.

Notes

An overview of this literature can be found in Armstron (2002) or Terrasi (2002).

Various articles of the Treaty of the European Union make specific reference to this issue. Article 2, for example, states specifically that ‘the Community shall have as its task (…) to promote (…) a harmonious and balanced development of economic activities, (…), economic and social cohesion and solidarity among Member States'. Article 158 of the same text indicates that ‘the Community shall aim at reducing the disparities between the levels of development of the various regions’.

This approach has in fact been adopted in an international setting by Ram (1992), Duro (2001), Grün and Klasen (2003) and Ezcurra and Pascual (2005).

Interested readers will find more details of this database in Eurostat (1996, 2003) or Álvarez-García et al. (2004).

A complete list of the regions considered, together with additional information in this respect, is included in the appendix.

This index has been used by Ram (1992), Duro (2001) or Ezcurra and Pascual (2005). For other alternative welfare measures that appear in the literature, see, for example, Grün and Klasen (2003).

Regional Gini indices have been obtained from the ECHP, taking the household as the unit of reference. To overcome problems arising from existing differences in household composition, the equivalence scale proposed by Coulter et al. (1992) has been used. This involves dividing the income of each household by N ^θ, where N is the number of household members and θ ∈ [0, 1] is a parameter that captures the impact of scale economies. In the present case, θ = 0.5. Likewise, all the observations have been weighted according to the number of household members.

Dalgaard and Vastrup (2001) have shown that the combined use of these two statistics is not redundant, since each can lead to a different conclusion.

It is well known that the variance of the logs can be expressed as the square of the standard deviation of the logs. Therefore, these two inequality measures are ordinally equivalent.

This is what Shorrocks (1982) calls the natural decomposition of variance.

The degree of statistical association between the Sen index and its various components has also been estimated. The results of this analysis are shown in Table A2 and allow the conclusions reached earlier to be confirmed. In particular, the Spearman coefficients of correlation between the welfare index, productivity, employment rate, activity rate and inequality in the income distribution take the following mean values over the period considered: 0.7910, 0.5651, 0.4092 and 0.5070.

For further details, see Silverman (1986, p. 47).

In order to complete this analysis, the corresponding boxplots have been estimated (Fig. A1), which confirm the information given by Fig. 1.

The first-order condition of this optimization problem is given by:

for j = 1, … , n − 1. For a more detailed analysis of this issue, however, see Esteban et al. (1999).

The optimal partition for the two-group case is characterized by the fact that the welfare level that separates the two groups coincides with the mean. By adopting this regional classification criterion, it is possible on mean to account for 72% of the total inequality measured by the Gini index. The amount of internal inequality left unexplained by the grouping is therefore the 28%.

This is a direct consequence of the dynamics of regional inequality within each of the two groups into which the original distribution has been divided. Thus, according to Table A3, the regions with a welfare level below the mean have experienced a reduction in regional disparities. In fact, the corresponding Gini index has decreased by 5%. On the other hand, in the group formed by the regions with welfare levels above the mean, regional inequality has risen over the same period. Indeed, the Gini index for this group has increased by 7%.

The importance of this issue has been highlighted, among others, by Mora (2004).

Fingleton et al. (1996), López-Bazo et al. (1999) or Cuadrado et al. (2002) estimate various transition matrices to analyse regional mobility in terms of per capita income or productivity. For the Finnish case, see Pekkala (2000).

See Kremer et al. (2001).

See Stokey and Lucas (1989).

Gaussian kernel functions have been used in all cases, while the smoothing parameter has been selected following Silverman (1986, p.86). Finally, all estimations have been obtained using the code proposed by Shuetrim (1999) to obtain the bivariate density function.

See also Quah (1996b), Rodríguez-Pose (1999) or Ezcurra et al. (2005), among others.

In this respect, see Table A1.

See, for example, Fingleton and McCombie (1998), López-Bazo et al. (1999), Fingleton (1999) or Maza and Villaverde (2004).

There are various works that have studied this issue in the European context. Interested readers might consult Paci (1997), Paci and Pigliaru (1999), Gil et al. (2002) or Rapún et al. (2004).

A complete list of the regions that form part of the different groups is included in the appendix.

Lack of relevant statistical data for all of the regions of interest has prevented a distinction in this analysis to be made between public and private investment.

As is widely known, the importance of these two factors has been repeatedly stressed in studies investigating the causes of economic growth. For a summary of this literature, see Barro and Sala-i-Martin (1995).

See the appendix for further information on this point.

This type of methodology has been used by Overman and Puga (2002) to explore the causes of regional differences in unemployment rates in the European Union.

In the discrete case, the corresponding transition matrix ought to coincide with the identity matrix.