Welfare inequality, regionalisation, and welfare policy : measurement and analysis for Spain

This study is focused on the changes in regional inequality in Spain over the last four decades, with emphasis on regional welfare. The two most important items of welfare in Spain are, health and education, and so these are the main focus of this study. Attention is paid to the levels and trends in inter and intra-regional disparities in the welfare components of health, education and housing. The extent to which changes in inequality with respect to welfare relate to changes in regional welfare policy is evaluated. Various methodological issues are explored in the context of measuring welfare inequality between regions. A specific procedure to measure inequality in longitudinal analyses is developed. The study is organised in three parts. The first part includes Chapter 2 and 3, and deals with regional structure and policy to provide a foundation for the analysis. The second part focuses on the methodology developed in this study and the techniques used for that purpose (Chapter 4). The third part includes all the results of the analysis (Chapter 5, 6, and 7) and the conclusion chapter (Chapter 8).Chapter 2 focuses on the development of the Spanish welfare state and its socio-economic context. A substantial part of the policy changes relating to the welfare system in Spain have resulted from the redefinition of the government's duties following the 1978 Constitution. Since the sixties the Spanish economy has been unstable and there have been important developments such as, the population explosion, and the ageing of the population. This situation has resulted in the need for significant changes in the welfare state as seen in a variety of policy changes.The devolution of power to the regions and the <em>regionalisation process</em> of the welfare state in particular are of major interest in the present study. The regional state in Spain, known as Co <em>munidades Autónomas,</em> is a decentralised policy model composed of any of the nineteen Autonomies or admisnistrative regions consisting of one or several provinces (from a total of fifty two). The development of the welfare state in Spain has involved increased autonomy for the regions in welfare issues. The nineteen regions are responsible for welfare programs relating to basic infrastructure (ports, road networks, etc). But only seven out of the nineteen Spanish regions have gained full autonomy in education and health (the largest expenditure items of the welfare state). So the regionalisation process has not been symmetric among all regions. This situation may have some implications for the inequality between regions. The impact on inequality of the regionalisation of the welfare state is therefore one of the important issue investigated in the present study. In the coming years, regions with high levels of autonomy are likely to contribute greatly to policy making since they will be responsible for modelling the structure of the welfare state.In Chapter 3, the regional policy of the European Union, and the Spanish regional policy is described in detail. Spain is today one of the leading beneficiaries of the EU's financial assistance for regional development known as the <em>Structural Funds</em> . The relevance of the European Union (EU) regional policy in mitigating existing disparities between regions is discussed. The rapid development of mechanisms for the regional support of (economically) weak regions has contributed to a reduction of inequality. The <em>Compensation Funds</em> which started in 1978 have played an important role in the regions although the Structural Funds remain more important.In Part II we discuss the selection of a measure of inequality for our study. The <em>Theil's Second measure for multidimensional inequality</em> is selected (Chapter 4). A specific procedure is developed to estimate this measure for longitudinal analyses. We use several indicators to represent each of the welfare components under consideration. This involves defining a composite index of indicators. Inequality in regional welfare is investigated focusing on the following welfare components: health facilities and health status, education facilities and education enrollment and finally, household expenditures and housing conditions. The underlying multidimensionality of the welfare components is thus taken into account. In the present study, Maasoumi's (1986) aggregator function is used to aggregate the indicators. This function enables us to reproduce the maximum amount of information contained in the original indicators. The data used relates to the following years (or periods): 1964, 1974, 1981, and 1991. There are also no studies that have done a longitudinal analysis of welfare inequality. Regional disparities in Spain over time with respect to health and/or education facilities have not been analyzed using an inequality measure. Thus it is not possible to compare inequality results from other literature sources with our results.For empirical purposes the use of Maasoumi's function requires weights associated with the indicators. Different weights are used for the different indicators. The estimation procedure for these weights developed in the present study is based on the <em>Partial Common Principal Component model</em> (PCPC) whenever appropriate or <em>Principal Component Analysis</em> (PCA) otherwise. The weights attached to the indicators are the component coefficients of the first component obtained using PCPC (or PCA). PCA has been applied for longitudinal analysis using Theil's second measure for multidimensional inequality (Maasoumi and Jeong, 1985;, Maasoumi and Nickelsburg, 1988; Zandvakili, 1992, 1999). When the periods under consideration share the same first component, the composite index is obtained on the basis of the component coefficients computed using a partial common principal component model. So the component coefficients are not sample-specific because they are the same in all the periods. In other words, the composite indexes for these periods depend on the values of the variables rather than the weights attached to variables. If the hypothesis of one partial common component is not rejected for a number of periods (for first three, and then two periods in this study), the composite index is then constructed on the basis of the maximum likelihood estimates for these periods together with the individual component coefficients for the remaining periods. When a partial principal component model does not fit the data, the composite index is based on individually computed component coefficients. Finally, the overall inequality of the Theil's second measure is computed.The Theil's second measure is applied in the present study to achieve the following objectives. First, the magnitude and direction of <em>overall inequality</em> , <em>between-region</em><em>inequality</em> , and <em>within-region inequality</em> is computed with respect to each of the welfare components under study. The wide variations in the geographic and socio-economic structure of the Spanish regions require an in-depth analysis of inequality focusing on intra- and inter-region disparities. In addition, the estimates of the composite indexes for the geographical units (regions) have been used for a <em>statistical cluster analysis</em> which identifies the similarities between one group of regions in contrast with another group of (similar) regions. The cluster analysis identifies two groups of high similar values ( <em>most-favored regions</em> ) and low similar values ( <em>least-favored regions</em> ). A picture of the geographical distribution of welfare components is obtained, and changes over time are compared. The inequality results and the results from the cluster analysis form the main findings of our study.The empirical results with respect to the welfare components are presented in Part III. <em>Health facilities</em> and <em>health status</em> are studied separately (Chapter 5). A substantial part of Chapter 5 is focused on health facilities. The inclusion of geographical effects ( <em>spatial</em><em>spillovers</em> ) resulting from the <em>contiguity</em> (or geographical proximity) between geographical units forms the major contribution of this study. Spatial spillovers across geographical areas are inevitable since individuals can commute from their own area to contiguous areas when health facilities are not available in the home area. A procedure is developed to incorporate contiguity into the analysis. The geographical units considered for contiguity are provinces which are the smaller territorial divisions of regions.In the method developed for incorporating contiguity, the level of facilities available in a certain province is considered to consist of the facilities in the own province plus the facilities located in <em>contiguous provinces</em> weighted by <em>spatial weights</em> . <em>Spatial weights</em> used here correspond to the simple inverse distance (optimal distance by road) between the provincial capital of the Spanish provinces and the provincial capital in contiguous provinces. For health facilities and health status the notion of <em>contiguous provinces</em> refers to first-order contiguous provinces connected at <em>the first order of contiguity</em> . The first order of contiguity describes two provinces that have a common boundary, common vertex or both. The use of this order of contiguity is justified as patients seek a first contact with doctors or specialized treatment and diagnosis at the nearest place to their home province.The results show improvements in inequality with respect to health facilities are between 1981 and 1991. The sharp drop in inequality coincides with the enactment of the <em>1986 Health act</em> (LGS). In addition there is also an important decline in the components of between-region inequality between 1981 and 1991. It is possible that regional policies and the devolution of power in health issues in the mid-eighties may have caused changes in the pattern of regional inequality.The regionalisation process of the health system may also have had important implications for regions with transferred powers in health issues. In these regions the results reveal that within-region inequality decreases between 1981 and 1991. So it is possible that the regional policies have resulted in a more uniform distribution of health facilities within certain regions. The geographical distribution of facilities obtained using cluster analysis reveals a North-South pattern with facilities located mostly in the North of Spain. The group of most-favoured regions consists of regions with transferred powers, regions which are <em>central places</em> like Madrid, and regions with certain socio-economic characteristics. It is suggested, therefore, that the geographical distribution of facilities may be affected by the socio-economic conditions of regions.The comparison of the contiguity and non-contiguity cases reveals that there are important spatial effects, especially among the regions situated in the North and the Centre of the Iberian Peninsula. Geographical proximity benefits only a few number of regions resulting in a dramatic increase in inequality in the contiguity case. When contiguity is not taken into account, the results for inequality show a very different impact for health policies. Inequality with respect to health status is investigated, but the results obtained are not very satisfactory possibly because of inaccuracies in the data used.<em>Education facilities</em> and <em>education enrollment</em> are studied in Chapter 6. With respect to education facilities, spatial spillovers are also incorporated since education is one of the most common causes of individuals commuting. But contiguity is not often taken into account in the literature on education. For computing available facilities in secondary education, the first order of contiguity is considered. The available facilities for university education consist of facilities located in contiguous provinces at the first order of contiguity plus the <em>second order of contiguity,</em> (plus the facilities in Madrid and Barcelona for 1964 and 1974). Here a second order of contiguity is defined as between two contiguous provinces, one of them being first-order contiguous (facilities in provinces adjacent to the neighbouring province).The results for inequality with respect to education facilities show that inequality has declined between 1974 and 1991. This may be due to the promotion of non-compulsory education. Over the last few decades the Spanish government has pursued a policy intended to distribute university and vocational training facilities more evenly. The increase in the contribution to inequality of between-region inequality between 1981 and 1991 may be due to the impact of regional policies.Intra-regional disparities are more important in Castilla León, Castilla la Mancha, and Andalucía. All these regions are bound by similar regional characteristics such as limited industry, abundant potential in natural resources, predominance of agriculture and their geographical situation in the Centre and South of the Iberian Peninsula. In addition, these three regions cover 53% of the Iberian Peninsula and 52% of the total land size. Regional authorities in Andalucía are responsible for education powers while there has been no devolution of power in Castilla León and Castilla la Mancha. The socio-economic characteristics of these regions appear to be more influential with respect to inequality than autonomy. The geographical distribution of facilities with respect to education has changed dramatically between the 60s and the 90s. This result from cluster analysis shows that changes in education policies have affected inequality.Spillover effects have improved the education facilities in the Spanish regions resulting in smaller values of overall inequality in the contiguity case compared to the non-contiguity. Spatial spillovers are observed in the Centre of Spain in 1991 between the region of Aragón and its first or second order neighbours (Madrid, Cataluña and Navarra). Further, the trends in inequality in the contiguity case are more in line with the policy measures than results in the non-contiguity case. So the inclusion of spillovers in the contiguity case seems to be a good approach for the study of inequality.Education enrollment in the non-compulsory education is also investigated in Chapter 6. The results for overall inequality and the inequality decomposition with respect to enrollment are very similar to those for education facilities. The results suggest that education facilities and education enrollment have been influenced by policy measures and the regionalisation process.Trends in inequality with respect to household consumption and housing conditions are analysed in Chapter 7. The results suggest that the magnitudes, and the trends for inequality that might be expected with respect to household incomes are similar to those obtained with respect to household expenditures. Inequality in household consumption and housing conditions has narrowed significantly over the last four decades. This is consistent with the changes in the economic situation which occurred during this period.Finally, Chapter 8 summarises and discusses the main conclusions based on the findings in this study. One of the main conclusions is that the procedure that we develop for the longitudinal analysis of multidimensional inequality in the welfare components is successful and performs satisfactory. In addition, spatial spillovers must be taken into account by using the procedure developed for incorporating contiguity. When contiguity is considered a more accurate picture of inequality is obtained. With respect to the empirical findings in this study, we conclude that firstly, the setting up of the welfare system and the social policies undertaken over the last few decades in education and health have had important consequences for inequality. Secondly, the impact of the regionalisation process on inequality with respect to education and health also appears to be important. New insights with respect to the relationship between welfare policies and the actual changes in welfare inequality may be provided by extending the present analysis.