A Malmquist-Based Approach to Change in Local Economic Well-being

A Malmquist-Based Approach to Change in Local Economic Well-being PILAR MURIASa, SIMONE NOVELLOb and FIDEL MARTÍNEZ-ROGETa aDepartment of Applied Economics, University of Santiago de Compostela, Rúa Burgo das Nacións s/n, E-15782 Santiago de Compostela, A Coruña, Spain bCESUGA/University College of Dublin, Rúa Obradoiro, 47, E-15190 A Coruña, Spain Contact: [email protected]; fi[email protected] Published in: Regional Studies, Volume 50, 2016 - Issue 8 https://doi.org/10.1080/00343404.2015.1016414 The text that follows is the unformatted authors’ final version Abstract The focus on the multifaceted nature of wellbeing has increased substantially in recent years. This paper employs Data Envelopment Analysis (DEA) combined with the Malmquist index to examine the temporal change in the economic wellbeing of Spanish provinces over the ten–year period between 1996 and 2006. This method not only allows the aggregation of multiple indicators and the estimation of wellbeing shifts over time, but also shows the sources of variation in local economic wellbeing. Our results seem to suggest that whilst economic wellbeing has improved in the great majority of provinces, this advancement stems largely from a more favourable environment than from the specific capacity of each province to reduce its relative distance from key benchmarks. Keywords: Economic wellbeing; temporal change; Data Envelopment Analysis; Malmquist Index; Spain JEL: R11; C61 1. Introduction Scholars’ recent disaffection with the dominant economic focus of local development has led to a call for a broader range of notions including social welfare, social equality, sustainability or social cohesion (MORGAN, 2004; BRISTOW, 2005; PETRAKOS et al, 2005; DUNFORD, 2005; PIKE et al, 2007; HUDSON, 2007; STIGLITZ et al, 2009; TOMANEY et al, 2010). Regional imbalances expand far beyond GDP-per capita differences to key economic and social aspects, such as employment, educational levels, infrastructure endowment, equality, access to technology and innovation, as well as provision of social services. A new multidimensional view of wellbeing has spawned the development of composite indicators enalbing policy-makers and practitioners to measure, analyze and compare the relative wellbeing of regions and other territorial scales (MURIAS et al, 2006; MURIAS et al, 2012; MARCHANTE and ORTEGA, 2006). The main challenge confronting these studies regarded the importance of each dimension in the construction of a composite local wellbeing indicator. However, each territorial unit is characterized by specific policies, social values and historical backgrounds, prioritising the various dimensions behind the concept of wellbeing. Thus, when comparing the wellbeing of places a common weighting scheme for all the units under study may not be appropriate, since individual societies subjectively assess their definition and targets for local development (PIKE et al, 2007). Whilst advancing the wellbeing and local development research agenda, these studies fail to define patterns of local wellbeing, or the mechanisms underlying its improvement or worsening. These approaches implicitly assumethat space is the main reason why societies prioritize certain wellbeing dimensions. However, as PIKE et al, (2007) states, “different kinds of local and regional development connect to socially determined and normative principle and values that differ geographically and change over time” (page 1265, emphasis added). While supportive of regional development studies’ defining concern with the uneven spatial distribution of local development, we begin by acknowledging that places are historically contingent (PRED, 1984; PAASI, 2004) and therefore the importance they attach to various dimensions of wellbeing varies not only in space, but also time. It may be reasonable to expect societies’ emphasis on single socio-economic indicators to vary in the light of environmental or policy changes and therefore analysis of changes in wellbeing should be based on a weighting system “which allows weights to vary, across objectives, over places and through time” (LOVELL et al, 1995, page 508). Informed by the pluralist and multidimensional conception of local development in regional studies and economic geography (PIKE et al, 2007; TOMANEY et al, 2010; BRISTOW, 2005; HUDSON, 2005), we propose a method to assess the shift of wellbeing conditions of local territorial units, and to examine whether their spatial structures allow them to capitalize on opportunities from wider socio-economic conditions. The broader research agenda is the need “to grasp the form and nature of local and regional development, the basic understandings of what it is, what it is for and, in a normative sense, what should be about” (PIKE et al, 2007, p. 1254). HUDSON (2005) argues that “focusing on the determinant of change and the specific institutional form of different areas could help unveil the reasons as to why places change in varying ways – some as winners and some as losers – within the structural limits of a capitalist political economy. As such, it could help to highlight on the key question as to why some territories experience different trajectories of change, of how the fortunes of different areas relate to one another”. BRISTOW (2005) argued that “more empirical research is needed to understand and identify the conditions which enable some regions to adapt successfully, while others remains “locked-in” to fatal development path”. This paper proposes a Malmquist index based on Data Envelopment Analysis (DEA) to assess shifts over time in economic wellbeing of different provinces. DEA preserves the inherent idiosyncrasies of local units and considers changes in perceptions in economic wellbeing. DEA endogenously determines the weights of each dimension of economic wellbeing for each time period, considering the peculiarities of each unit and assessing it in the most favourable light. Drawing on DEA measures, the Malmquist index calculates the variation in societies’ economic wellbeing from year t to year t+1. Furthermore, this approach identifies two components of change in local economic wellbeing. The first is unit specific, attributable to the improvement/regression of a unit in the production of economic wellbeing regarding top performers which act as a benchmark. The second affects all units, including the top performers, the result of a more/less favourable environment for the production of economic wellbeing. This decomposition thus provides an alternative way of testing for patterns of territorial disparities in economic wellbeing. A Malmquist index based on DEA measures is particularly valuable in research that simultaneously considers different areas over varying periods of time, especially when the underlying purpose is to obtain an insight into the composite performance dynamics of an area in relation to others (CHERCHYE et al, 2007a). We apply this method to the calculation of change in economic wellbeing of the 50 Spanish provinces between 1996 and 2006. Our findings point to improvements in economic wellbeing, due essentially to the mainly favourable environment. Nevertheless, the scale of change reveals a North-East and South-West divide. Furthermore, leading progress in economic wellbeing was limited to a small number of innovating provinces, mainly located in Northeast Spain. Finally, we found that improvement in economic wellbeing did not bridge the gap between the top performers and other provinces. This paper contributes to the debate on local economic wellbeing and development While attempting to account for the shift of a composite indicator of economic wellbeing, our study demonstrates that places change their level of economic wellbeing through different wellbeing paths, reflecting distinctive ideological priorities, the unique changing nature of spatial structures and/or the broader social and economic processes in which units are embedded. The method proposed provides evidence regarding the extent to which an increase in wellbeing is unit specific or whether it stems from a general improvement at a country level. Furthermore, by examining whether places are bridging gaps in wellbeing in a way that reflects the priorities and idiosyncrasies of each unit, our approach offers an alternative view to the study of reduction/widening territorial disparities. Finally, from a policy perspective, it might indicate the relative performance of each province over time, providing an additional tool for assessing whether intervention policies might affect changes in economic wellbeing. Section two illustrates the theoretical framework underpinning our composite indicator of economic wellbeing. Section three explains the methodology used to estimate economic wellbeing and its change. Section four presents the data for Spanish provinces. Section five discusses the results and implications, whilst the final section addresses the conclusions. 2. The contingent and multifaceted nature of economic wellbeing The central concern of economic geographers and regional scientists is uneven socioeconomic and geographical development, which necessarily implies the transformation of the economic landscape (BOSCHMA and MARTIN, 2007). Various authors (e.g. PAASI, 2004; MACLEOD and JONES, 2001, 2007; PIKE et al 2007) have argued that conceptualizations of change view spatial structures as the outcome of historical processes, conditioned by resolutions shaped by varying social relations and relations of interests. This perspective is informed by the relational view of local and regional economic development (AMIN, 2004; ALLEN and COCHRANE, 2007; HUDSON, 2007), which posits that places are not static entities but rather the product of a complex condensation of social relations, which combine contingently to form unique and constantly changing societies. Moreover, some proponents of this perspective (e.g. PIKE et al, 2006; MACLEOD and JONES, 2006; HUDSON, 2006) have argued that each area or level of analysis presents a specific “competitiveness profile”, which unfolds against a context of priorities, historical background and socially constructed values. The key point here is that particular definitions of local economic development vary across places and their articulations change constantly (PIKE et al, 2007). Places evolve as the endogenous population of economic agents and their characteristics shift, as those agents deliberately try to influence their organizational and institutional environment (Essletzbichler and RIGBY, 2007). Some societies, following a neoliberal model may prefer institutions and organizations which incentivize short-term and selective gains, sacrificing the future for the present (PECK and THEODORE, 2007). In the last two decades Spain has fostered a model of development which led to a sharp rise in employment yet failed to focus on long-term goals such as productivity performance (VAN ARK et al, 2007). Recently, the European Commission (through for example the report by STIGLITZ et al, 2009) and scholarly research (MORGAN, 2004; OSBERG and SHARPE, 2005; DUNFORD, 2005; PIKE et al, 2007; TOMANEY et al, 2010) have recommended that an area’s economic development should be assessed on sustainability, inequality and social cohesion rather than just short-term goals. In this sense, the examination of wellbeing should include the accumulation of economic and technological capital for the benefit of future generations (OSBERG and SHARPE, 2005; STIGLITZ et al, 2009). Moreover, while levels of economic development and wellbeing are related to material conditions, a number of scholars (e.g. OSBERG, 1985; OSBERG and SHARPE, 2002; 2005; DUNFORD, 2005; PIKE et al, 2007; HUDSON, 2007) pointed out that some localities or societies may place significant emphasis income distribution. For example, TOMANEY et al (2010, page 776) argued that “local and regional development should not just be about promoting greater growth, but also about reducing levels of inequality, and that mobilising resources in lagging and/or peripheral areas may constitute a valid recipe for both greater overall growth and lower territorial disparities”i. Finally, welfare analysis has helped design varieties of local development that tackle economic insecurity (OSBERG, 1985; OSBERG and SHARPE, 2002, 2005). Income flows can vary considerably over time, so individuals’ consumption plans and access to resources are constrained by current and expected future income. In sum, for each period, local units present a variety of wellbeing definitions, theories of social change, and agents of development (PIKE et al, 2007). Although the multidimensional view highlights the varieties of economic wellbeing models, it is more important to examine how societies change their wellgeing models and this influences their position against other societies (HUDSON, 2005; BRISTOW, 2005). The concepts of path dependence, lock-in and path creation may provide useful frameworks for examining a place’s ability to reduce (or exacerbate) its wellbeing gap compared with other territorial units. Places may be subject to quite different mechanisms of path dependence, such as increasing returns, localized learning and spillovers, and externalities of localization. A particular form or outcome of this is the notion of lock-in, which captures the trend for the geographical structure of the economy to exhibit historical ‘quasi-fixity’. MARTIN and SUNLEY (2006) indicated several ways of avoiding lock-in processes. The creation of new technological paradigms within the region may allow locations to establish new indigenous development paths. Renewal processes may emerge as results of the role played by larger firms, whereby supply networks and financial resources offset the lack of assets and resources that characterized the economies of small firms. Change can also be the result of research institutions and human capital and the territory’s ability to retain skilled labour. In addition, heterogeneity among technologies, industries and institutions may provide the redundant capabilities that, when recombined and reworked, prove useful in adapting to environmental change. Questions of change imply that places improve because they transform the economic and social structure, becoming the loci of innovative practices. However, as indicated by literature on economic growth and productivity (KUMAR and RUSSELL, 2002; LOS and TIMMER, 2005; VANDENBUSSCHE et al, 2006), we also need to focus on other types of change processes which recognize that units are likely to develop a dominant economic or industrial position that not only shapes their own development, but also impacts favourably on that of other areas. Consequently, the improvement in wellbeing of a particular local society reflects shifts in the wider wellbeing frontier, or the maximum wellbeing attainable in a specific time-period. Processes of frontier shift are the diffusion of new organizational forms, discourses, radical new technologies, from outside a territorial unit. For example, a number of authors have highlighted the use of competitiveness discourse within the political agenda on the European economic dilemma (BRISTOW, 2005, RODRIGUEZ-POSE and SANDALL, 2008). The allure of this discourse of competitiveness lies partly in its comprehensibility. Like the discourse of globalization, local competitiveness presents a relatively structured set of ideas, often in the form of implicit and sedimented assumptions, on which policymakers can draw when formulating strategy and legitimize strategy pursued for quite distinct ends. This discourse fits in with the trend towards decentralized, bottom-up approaches to economic development policy and the development of university-business relations, industrial districts, strong innovation networks and rural tourism. The case of consultant agencies, for example provides local authorities and business associations with technical expertise that allows them to implement practices developed elsewhere. However, a follower location can only benefit from this knowledge spillover if it operates at a similar level of knowledge or capital intensity. The benefits of assimilating best practices and new technology is therefore not immediate, because adopters need learning and investment capital to adopt technologies that can enhance productivity and development levels. We aim to illustrate from a multidimensional view that a Malmquist index drawing on DEA measures (FÄRE et al, 1994) can be used to assess the shift – and the mechanisms underpinning it – in the economic wellbeing of local units. DEA is used to aggregate different dimensions, endogenously assigning weights to indicators, in accordance with the profile that each unit presents for each year studied. DEA evaluates the performance of each unit with respect to the best performance attainable which itself may vary over time. The Malmquist index can be used to measure the units’ shift in levels of wellbeing. Moreover, this approach allows us to identify whether economic wellbeing variations result from specific changes in the relative performance of individual units, or from a more/less favourable environment for the generation of economic wellbeing affecting the performance of all the units. The analysis of these two components yields a frontier version of changes in local economic wellbeing. 3. A Malmquist-based Approach for the Analysis of Change in Economic Wellbeing 3.1. Cross-sectional perspective: Data Envelopment Analysis and the economic wellbeing index The present application of DEA moves away from the traditional context of productive efficiency (Charnes et al 1978), aggregating partial indicators into a single performance measure (Hashimoto and Ishikawa, 1993; Hashimoto and Kodama, 1997; Despotis, 2005; Cherchye et al, 2004, Murias et al, 2008; 2012; Zhu, 2001). Traditional outputs will be replaced by partial indicators, whilst input is based on a single resource with an equal value for all units and years: a "helmsman" (Koopmans, 1951; LOVELL et al, 1995). The helmsman represents a bureaucracy or decision-making apparatus in every unit. Size or organizational structure is irrelevant, because the focus is on apparatus performance, regardless of the amount of resources used to achieve this performance. In our case the helmsman relates to the provincial socio-political structure: political and administrative apparatus and economic and social networks (Spanish provincial councils, chambers of commerce, employers’ confederations, regional trade union offices, etc.). All provinces have a similar structure, although their organisational methods and size may vary, related directly, albeit not perfectly, to the population variable. This paper focuses on the capacity of such structures to provide services which can increase boost residents’ economic wellbeing. A single input of equal value is considered for all provinces and all years (normalized to 1 for greater simplicity). Other papers adopting this approach are Mahlberg and Obersteiner (2001), Coelli et al (2008); Reig-Martínez (2013), Zanella and Camanho (2013) or Ferrier and Trivitt (2013). DEA defines a best practice frontier for the production of economic wellbeing, as the envelope for the observed partial indicator combinations of all the units in the sample, under the constraint that all the well-performing units support the envelope (Mahlberg and Obersteiner, 2001). This frontier represents the technological limits of what a unit can achieve in terms of economic wellbeing and serves as a benchmark, with DEA endogenously selecting for every unit the aggregation weights that minimize its distance from this benchmarkii. The mathematical expression of the model can be summarized asiii: ( ) 1− = k kk θyxD max)),(( s.t rk n j krjjyy ∑ = ≥ 1 θλ r=1…q [1] ∑ = ≤ n j j 1 1 λ 0≥ j λ j=1…n Where 1≤θk≤∞, yrk denotes the value of the r-th indicator for unit k, λ j is a positive scalar, n are the number of provinces and q the number of partial indicators considered. The optimum value of D (xk,yk)= 1/θ*k represents the composite index for unit k When a province is on the best-practice frontier it achieves a score of 1 in the composite index. Values below 1 imply that the province is beneath the frontier and economic wellbeing performance can be improved. Existing literature offers numerous alternative weighting and aggregation methods (Nardo et al, 2005). Many require priori consent regarding weight values, which can be especially difficult to obtain (CHERCHYE et al, 2004, 926). DEA instead estimates weights endogenously according to the observed performance of each unit and year, respecting the fact that units may have varying priorities that may change over time Thus, the particularities of the provinces and their evolution are respected. In this paper we estimate an economic wellbeing index for Spanish provinces using DEA and correcting results for bias using bootstrapping (Efron, 1979), following the procedure from Simar and Wilson (1998). In doing so, FEAR library (Wilson, 2008) on R was used. 3.2. A Dynamic Perspective: The Malmquist index and the temporal variation in the economic wellbeing index Evolution in economic wellbeing over time was estimated using the Malmquist productivity index. The Malmquist index (MI) allows for either the change in the relative position of the unit regarding the best practice frontier or the change of the frontier itself (Zaim et al, 2001; Cherchye et al, 2007a; Coelli et al, 2008; Kortelainen, 2008; Hashimoto et al, 2009) to be taken into consideration. An MI based on DEA measures referring to the technology at periods t and t+1 can be defined asiv: ( ) ( ) ( ) ( ) 2 1 1 111 11 11 , , , , ),,,(                         = + +++ ++ ++ ttt ttt ttt ttt tttt yxD yxD yxD yxD yxyxMI [2] Improvements in the capacity to generate economic wellbeing imply a Malmquist index greater than unity. Retardation in terms of economic wellbeing over time yields a Malmquist index of less than unity. If there is no change in economic wellbeing from period t to period t+1, the Malmquist index equals one. Following Färe et al (1994), this expression can usefully be factored in the following wellknown form: ( ) ( ) ( ) ( ) ( ) 2 1 1111 11111 11 , , , , , ),( ),,,(                         = ++++ +++++ ++ ttt ttt ttt ttt ttt ttt tttt yxD yxD yxD yxD yxD yxD yxyxMI [3] The first component, outside the brackets, represents the extent to which a province can reduce the distance from its best practice benchmark (on the frontier) in period t+1 compared with period t. This improvement depends exclusively on the specific performance of the province itself (Cherchye et al, 2007a). This component is usually called the ‘catch up’ component as its improvements indicate ‘catching up’ to the frontier (Färe et al, 1994). The second component in brackets is called the “frontier shift” and represents variations in best possible performance between period t and t+1. The frontier shift reflects a change in best practices, or in the limits of what a province can achieve using a given level of resources. For Kumar and Russell (2002), the frontier shift component should be interpreted quite broadly, encompassing institutions and policies as well as purely technological phenomena. Even though the product of both components must by definition equal the Malmquist improvement index, those components may move in opposite directions: improvements in economic wellbeing may be observed in a province because of its specific better performance even if there is a less favourable environment than in the initial period. Moreover, a province can experience a regression of economic wellbeing even when better practices have become possible, because it is unable to make complete use of them (Cherchye et al, 2007a). As in section 3.1 we complemented the Malmquist estimates with bootstrapping, in this case following Simar and Wilson (1999). 4. Research Setting The methodology proposed in this paper addresses multidimensional constructs and is of particular value when analyzing areas exhibiting some degree of cultural, social and economic heterogeneity. Spain offers an ideal context as it is a nation-state that has experienced difficulties in its state building processes. The result is a fragile sense of national identity and significant territorial disparities, due to specific and often unique institutional and industrial processes (GIORDANO and ROLLER, 2004). Secondly, Spain’s ongoing devolution processes are a key reform for the modernization of the public sector. Thirdly, the institutional and political context of intergovernmental relations between local and regional authorities is well-developed. We consider the geography of economies and societies as the outcome of political and institutional endeavours as well as the result of a dense condensation of social relationships (ALLEN and COCHRANE, 2007). Research in local development (TORTOSAAUSINA et al, 2005; MURIAS et al, 2006; MAZA et al, 2009; HIERRO and MAZA, 2010) noted that major territorial differences would be overlooked if Spanish Autonomous Communities were considered. Spanish regions are of widely differing sizes displaying a highly uneven social, economic, and industrial heterogeneity. Provinces have therefore been the preferred unit of analysis. Institutionally, they are entrusted with ensuring the principles of social and economic solidarity and balance among municipalities, especially the smallest ones. In increasingly depopulated rural areas provinces have been a key tool in integrating municipal services. Moreover, studies analysing the social transformation of rural Spain (e.g. HOGGART and PANIAGUA, 2001; PANIAGUA, 2008; 2009; PINILLA et al, 2008) pointed out that it has favoured the formation of a new provincial middle class, made up of managers and professionals, generally with a high level of education, who relocate in order to begin of further their careers. In those areas the main ideology includes an anti-urban component, or, at least, a clear distinction from the metropolitan areas of Madrid and Barcelona. Some of these depopulated areas have received public funding in order to redress rural decline (ESCALONA-ORCAO and DIEZ-CORNAGO, 2007) and provincial governments have encouraged settlement by foreigners (PANIAGUA, 2008; PINILLA et al, 2008). In a context of migration, depopulation and socio-demographic change, the emergence of social groups presents spatial trends on a provincial scale. PANIAGUA (2009) pointed out that “there are no important references to other spatial scales such as the European Union or even Spain, so the definition of an area is made according to its own characteristics and in relation to the provincial environment”. This does not mean that a society is limited to a province’s administrative boundaries; they may well extend into several adjacent provinces. However, the province represents the smallest scale that captures the social and economic discontinuities of Spanish regions. The theoretical framework follows the work of OSBERG (1985) and OSBERG and SHARPE (2002, 2005) who were among the first to suggest that measures of economic wellbeing should include the heritage of future generations, economic inequality and insecurity among the current generation. It is based on a subjective view of economic wellbeing: each individual society makes a subjective evaluation of objective data in reaching conclusions as to that society’s wellbeing. OSBERG’s four-dimensional framework helps to quantify subjective views regarding the nature and character of local development. The time period selected for the study was 1996-2006. Homogeneous series for the Spanish provinces are available for this period, except for the Gini index. During this period Spain experienced dramatic expansion and growth, with GDP rising more than the EU average. Between 2000 and 2003, however, Spain's growth rate decreasedv, although it subsequently rose again from 2003-2006. The period 1996-2006 is also important in terms of public intervention policies designed to counteract unemployment, social exclusion, and income distribution. Between 1994 and 1999, Spain received almost a quarter of the total Structural Funds allocated, more than any other EU member state (BOLDRIN and CANOVA, 2001). Regarding the variables used, we prefer inclusion as opposed to exclusion, as an imprecise measure is likely to embody a smaller error than omitting a variable. However, OSBERG and SHARPE (2004, page 130) pointed out that on a national level “when there is no estimate available at all, omission is unavoidable”. Data availability problems are even more frequent at provincial level: the accumulation component should include not only human and physical stocks, as in this paper, but also other indicators such as technological capital (public and private). However, the lack of homogeneous technological data prevented us from using technological capital indicators to construct the composite index. Lleida 1.046 (1.003-1.081) 0.995 (0.922-1.076) 1.044 (0.963-1.112) Tarragona 1.034 (1.005-1.052) 1.000 (0.935-1.057) 1.025 (0.966-1.092) Comunidad Valenciana Alicante 1.012 (0.978-1.021) 0.994 (0.964-1.022) 1.014 (0.971-1.043) Castellon 1.052 (1.003-1.092) 0.998 (0.921-1.076) 1.029 (0.964-1.114) Valencia 1.002 (0.969-1.015) 0.989 (0.959-1.013) 1.011 (0.974-1.040) Extremadura Badajoz 1.005 (0.999-1.007) 0.999 (0.970-1.021) 1.010 (0.982-1.033) Caceres 1.000 (0.981-1.009) 0.991 (0.954-1.021) 1.012 (0.974-1.040) Galicia A Coruña 1.010 (0.971-1.023) 0.999 (0.957-1.035) 1.015 (0.960-1.041) Lugo 1.020 (0.980-1.032) 1.006 (0.973-1.037) 1.027 (0.965-1.038) Ourense 1.030 (0.989-1.047) 1.018 (0.978-1.063) 1.022 (0.955-1.049) Pontevedra 1.007 (0.991-1.014) 0.996 (0.962-1.021) 1.014 (0.981-1.041) Madrid Madrid 1.055 (1.018-1.077) 1.000 (0.921-1.058) 1.040 (0.987-1.119) Murcia Murcia 1.016 (0.997-1.024) 1.004 (0.970-1.030) 1.013 (0.977-1.044) Navarra Navarra 1.040 (1.002-1.064) 1.000 (0.932-1.068) 1.041 (0.962-1.091) Basque Country Alava 1.030 (1.001-1.057) 1.000 (0.917-1.068) 1.026 (0.961-1.099) Guipuzcoa 1.032 (1.001-1.051) 1.000 (0.930-1.055) 1.032 (0.973-1.088) Vizcaya 1.023 (0.995-1.042) 1.001 (0.933-1.062) 1.014 (0.955-1.076) La Rioja La Rioja 1.029 (0.983-1.056) 1.002 (0.948-1.048) 1.027 (0.969-1.073) MEAN 1.021 (0.983-1.038) 0.998 (0.950-1.040) 1.020 (0.971-1.062) Except for Madrid, our findings support the North-East and South-West divide that has long characterized Spain (Figure 1). Provinces in southern regions – Andalusia, Castilla-La Mancha and Extremadura – exhibit a significantly lower growth of economic wellbeing than the “Northeast” – Catalonia, Aragon or the Basque Country. These findings may contribute to the discussion on the impact that processes of devolution and structural funds have had on unemployment as well as social and economic inclusion for the regions located in Southern Spain (see e.g. GIORDANO and ROLLER, 2003; BOLDRIN and CANOVA, 2001)ix. We found that whilst Southern Spanish regions failed to reduce their distance from the top performers, they nonetheless improved their economic wellbeing between 1996 and 2006. Figure 1. Economic Well-being Change Map Economic Well-being Change ≥ 3 th quartile Economic Well-being Change ≥ 2 nd quartile < 3 th quartile Economic Well-being Change ≥ 1st quartile < 2nd quartile Economic Well-being Change < 1st quartile The decomposition between the catch-up and frontier shift component suggests that improvements in economic wellbeing was caused by a frontier shift, exhibiting an average annual growth rate of 2%. This suggests a favourable environment for the “production” of economic wellbeing nationwide, showing genuine progress in previously unattainable wellbeing practices. The average frontier shift change was negative between 2002 and 2003 (also with both bounds of confidence intervals less than one), whereas the average catch up rate for the same period exceeds 6%, the highest for the whole decade (both bounds of confidence intervals were higher than one). The regression of the ‘average’ frontier coupled with advances in catching up suggests that low performing provinces could reduce their distance from the economic wellbeing frontierx. The contrary happens between 2001 and 2002. The average frontier shift is also clearly positive between 1999-2000 and 2005-06. As pointed out by FÄRE et al (1994), although the average frontier shift results are suggestive, they do not allow us to identify which provinces shift frontiers over time. In order to gain an insight into ‘innovator’ provinces (i.e. shift the frontier over time), we follow the procedure suggested by FÄRE et al (1994)xi. The analysis suggests that from 2001-2002 onward there is a significant drop in the innovating provinces. This result is consistent with our previous finding indicating that 2002 is the first and only year in which average provincial economic wellbeing decreased. Table 5 indicates that, except for Madrid, the innovating provinces are located in Northeast Spain (e.g. in Catalonia, Basque Country, Aragon or Navarra). Table 5. Provinces shifting the frontier Innovating Province Number of Innovating provinces 1996- 1997 Álava, Castellón, Girona, Guipúzcoa, Huesca, Lleida, Madrid, Navarra, Tarragona, Teruel, Vizcaya 11 1997- 1998 Álava, Barcelona, Castellón, Girona, Guipúzcoa, Huesca, Lleida, Madrid, Navarra, La Rioja, Soria, Tarragona, Vizcaya 13 1998- 1999 Álava, Castellón, Girona, Guipúzcoa, Huesca, Lleida, Madrid, Navarra, Tarragona 9 1999- 2000 Álava, Castellón, Girona, Guipúzcoa, Huesca, Lleida, Madrid, Navarra, Tarragona,Vizcaya 10 2000- 2001 Álava, Barcelona, Castellón, Girona, Guipúzcoa, Huesca, Lleida, Madrid, Navarra, La Rioja, Soria Tarragona, Teruel, Zaragoza 14 2001- 2002 Girona, Huesca, Navarra, Tarragona 4 2002- 2003 Huesca,Tarragona, Vizcaya 3 2003- 2004 Álava, Guipúzcoa, Lleida, Soria, Tarragona, Vizcaya 6 2004- 2005 Girona, Guipúzcoa, Tarragona, Teruel 4 2005- 2006 Girona, Guipúzcoa, Madrid, Navarra, Tarragona, Teruel 6 Provinces with both bounds of confidence intervals higher than 1 for the frontier shift component are shown in italics. Focusing on the average annual catch up rate, only 12 out of 50 provinces exhibit a positive, abliet low, catch up component. Nevertheless, confidence intervals for the average catch up estimates in all the fifty provinces include the value of one. Therefore, whilst improving their absolute performance over time, in average terms there is no statistical confidence for a reduction in relative distances. We then conducted a further analysis to test for any convergence in the economic wellbeing index between provinces. Following ZAIM et al (2001) we used a scatter diagram and a predicted regression equation to investigate whether the initial level of economic wellbeing (year 1996) is significant in determining improvement levels (Figure 2). The use of the Malmquist index as the dependent variable can be problematic because of correlation, so a bootstrapping technique following WANG et al (2006) was used. The positive and significant coefficient of the independent variable fails to show a reduction in disparities in economic wellbeing during the period. Our findings might be contrasted with the results of HIERRO and MAZA (2010) whose study of income-based indicators, found a weak reduction in disparities among Spanish provinces. These differences may be attributable to our decision to include a wider number of determinants. Specifically, our analysis used dispersion statistics that reveal that the differences are due essentially to poverty and unemployment. Whilst the first indicator grew throughout the period, the second rose between 1996 and 2001, before registering a steady fall. All other indicators remained stable throughout the period. Figure 2. Relationship between the economic wellbeing index in 1996 and cumulative change in economic wellbeing. Cumulative change calculated following Färe et al, (1994). Change in Economic Wellbeing(1996-2006) = 60.7 + 74.9*Economic Wellbeing Index 1996 050 .0 2 = R Furthermore, in line with MURIAS et al’s (2006) study on Spanish economic wellbeing, our findings indicate that administrative regions should not be interpreted as the locus of economic wellbeing. In the case of the Comunidad Valenciana, whilst Castellón is among the best performers, with an average index of 0.991 and 8 times on the best practice frontier, Valencia and Alicante have an average of just 0.890 and never rank among the best performing provinces. The improvement in Castellon economic wellbeing during the period was due mainly to the good performance of economic stability dimension – i.e. employment rate and long-term contracts variables – and a very low poverty rate. Castellón’s economy is based on the tile sector, which dominates the Spanish market (FLOR and OLTRA, 2004) and is characterized by a large number of first- and secondtier suppliers locked into dense networks of interaction (MOLINA, 2008). Spain’s construction boom led to a sharp rise in demand for ceramic tiles (MOLINA, 2008), leading to greater employment in the province and the rest of the region (MOLINA, 2008; BUDÍ, 2008). This may also explain the province’s low percentage of the population on a minimum wage. The positive impact of the construction boom on Castellón may be confirmed by the Malmquist index decomposition. Castellón’s wellbeing improvement was due to the frontier shift (2.9%), suggesting that Castellón capitalized favourably on the opportunities offered. 6. Conclusions and Further Research The starting point of this paper was the recognition that analysis of economic wellbeing should expand beyond income-level indicators towards a broader range of variables. We selected DEA as the approach capable of aggregating into a composite indicator of economic wellbeing different measures over various years. The weighting problem is handled for each province separately, and the province’s specific weights assigned to each sub-indicator are endogenously determined. Some may question this degree of flexibility, especially considering that such a weighting scheme determines a province’s ranking. However, the proposed methodology is attractive given that our aim is to preserve the inherent ambiguity and subjectivity of economic wellbeing. Weighting information stemming from policy-makers is often characterized by strong inter-individual disagreement, which is even less likely in a dynamic context (CHERCHYE et al, 2004). By widening the concept of economic wellbeing, our composite indicator can preserve the unique characteristics of local societies, which define their wellbeing according to their specific historical background and industrial infrastructure. The proposed Malmquist based approach provides an alternative assessment for shifts in local economic wellbeing over timexii. This allows us to identify two sources of change in economic wellbeing for each province: one reflecting a frontier shift and another showing movement towards (catching up) or away from the frontier. Drawing on data for Spanish provinces, this decomposition suggests that an environment conducive to improvement in economic wellbeing was created in Spain between 1996 and 2006, and almost all provinces exploited this opportunity, albeit to different degrees. Nevertheless, this improvement did not allow for a reduction in the relative distance between top performers and the remaining provinces. We also find that the majority of provinces defined as innovators are located in Northeast Spain, thereby confirming the North-East-South West divide. The map showing the improvement in economic wellbeing in Spanish provinces until 2006 and the comparison with the current situation is also worthy of reflection. Varying situations can be found amongst those provinces that generated the greatest economic wellbeing between 1996 and 2006. Provinces like Navarra, Álava, Guipúzcoa, Soria, Huesca or Teruel currently have the lowest unemployment rates in Spain, whilst others, namely Girona, Tarragona or Castellón, are suffering high unemployment rates. It appears that provinces that focused less on boosting education levels channelled fewer resources into R&D, opting to centre activity on the construction industry in order to generate economic wellbeing, are finding it particularly hard to tackle the economic crisis. This calls for serious reflection, as it similarities could be drawn for the whole of Spain within the framework of the EU KLEMS Project, pointing to the poor performance of a country whose growth is considered to be based on transpiration rather than inspiration (VAN ARK et al, 2007). Our findings indicate that the highest catch-up component, which allowed non-border provinces to reduce their distance from the top performers, corresponds to a period (2002- 2003) of slower economic growth. In the future, it would be interesting to apply the methodology developed here to another phase of slower economic growth, such as that subsequent to the ten-year period considered in this paper. This will allow us to test whether these periods affect the economic wellbeing of open and more competitive areas rather than territories with an economy largely based on non-marketed services and public transfers (PETRAKOS et al, 2005). References AFONSO A., SCHUKNECHT L. and TANZI V. (2006) Public sector efficiency: an international comparison, Public Choice 123, 321-347. Aghion P., Williamson J.G. (Eds.) (1998) Growth, Inequality, and Globalization: Theory, History, and Policy. Cambridge University Press, UK. ALDÁS J., GOERLICH F. and MÁS M. (2006) Gasto de las familias en las Comunidades Autónomas españolas. Pautas de consumo, desigualdad y convergencia. CIEF, Centro de Investigación Económica y Financiera, Fundación Caixa Galicia. ALLEN J. and COCHRANE A. (2007) Beyond the territorial fix: regional assemblages, politics and power, Regional Studies 41(9), 1161-1175. AMIN, A (2004) Regions unbound: towards a new politics of place, Geografiska Annaler: Series B, Human Geography 86(1), 33-44. BOLDRIN M. and CANOVA F. (2001) Inequality and convergence in Europe’s regions: reconsidering European regional policies, Economic Policy 16 (32), 205–253. BOSCHMA R. A. and MARTIN R. (2007) Constructing an evolutionary economic geography, Journal of Economic Geography 7(5), 537-548. BOUSSOFIANE A., DYSON R.G. and THANASSOULIS E. (1991) Applied Data Envelopment Analysis, European Journal of Operational Research 52, 1–15. BRISTOW G. (2005) Everyone’s a ‘winner’: problematising the discourse of regional competitiveness, Journal of Economic Geography 5, 285–304. BUDÍ, V. (2008) El distrito de la cerámica de Castellón, Mediterráneo Económico 13, 383- 407. CHARNES A., COOPER W.W. and RHODES E. (1978) Measuring the efficiency on Decision Making Units, European Journal of Operational Research 2, 429–444. CHERCHYE L., MOESEN W. and VAN PUYENBROECK T. (2004) Legitimately diverse, yet comparable: on synthesizing social inclusion performance in the EU, Journal of Common Market Studies 42, 919–955. CHERCHYE L., LOVELL C.A.K., MOESEN W. and VAN PUYENBROECK T. (2007a) One market, one number: a composite indicator for the assessment of EU internal market dynamics, European Economic Review 51, 749–779. CHERCHYE L., MOESEN W., ROGGE N. and VAN PUYENBROECK T (2007b) An introduction to ‘Benefit of the doubt’ composite indicators, Social Indicators Research 82, 111–145. COELLI T., LEFEVRE M. and PESTIEAU P. (2008) Social Protection Performance in the European Union: Comparison and Convergence. ECORE Discussion Paper 2008/20, Université Catolique de Louvain. DESPOTIS D.K. (2005) A reassessment of the human development index via data envelopment analysis, Journal of the Operational Research Society 56, 969–980. DÍAZ R. (2004) La Inversión Extranjera Directa en Madrid, Revista de Estudios Regionales 71, 59-84. DIRECCIÓN GENERAL DE UNIVERSIDADES E INVESTIGACIÓN DE LA COMUNIDAD DE MADRID (2006) Estudiar en Madrid (Curso 2005-2006). (Avaliable at: http://www.emes.es/Actualidad/Publicaciones/Curso20052006/tabid/454/Default.aspx). DUNFORD M. (2005) Growth, inequality and cohesion: a comment on the Sapir report, Regional Studies 39(7), 972–978. EFRON, B. (1979) Bootstrap methods: Another look at the jackknife, The Annals of Statistics 7, 1-26. ESCALONA-ORCAO A.I. and DÍEZ-CORNAGO C. (2007) Despoblación territorial y oferta de servicios: Diagnóstico y propuestas, in SERRANO-LACARRA C. (Coord.) Despoblación y Territorio, pp. 61-118. Centro de Estudios sobre la Despoblación y el Desarrollo de las Áreas Rurales, Zaragoza. Essletzbichler J. and Rigby D. (2007) Exploring evolutionary economic geographies, Journal of Economic Geography 7(5), 549-571. ESTATUTO DE AUTONOMÍA DEL PAÍS VASCO (1979) Ley Orgánica 3/1979. (Avaliable at: https://www.boe.es/buscar/doc.php?id=BOE-A-1979-30177). FÄRE R., GROSSKOPF S., NORRIS M. and ZHANG Z. (1994) Productivity Growth, Technical Progress, and Efficiency Change in Industrialized Countries, American Economic Review 84, 66–83. Ferrier G.D., and Trivitt J.S. (2013) Incorporating quality into the measurement of hospital efficiency: a double DEA approach, Journal of Productivity Analysis 40, 337-355. FLOR M. and OLTRA M. (2004) Identification of innovating firms through technological innovation indicators: an application to the Spanish ceramic tile industry, Research Policy 33, 323-336. FRUTOS L.M., HERNÁNDEZ M.L. and CASTELLÓ A. (2009) Desarrollo y turismo rural: una perspectiva sobre Aragón, Serie Geográfica 15, 93-115. GIORDANO B. and ROLLER E. (2003) A comparison of city region dynamics in the UK, Spain and Italy: more similarities than differences?, Regional Studies 37(9), 911–927. GIORDANO B. and ROLLER E. (2004) Té para todos? A comparison of the processes of the devolution in Spain and the UK, Environment and Planning A 36, 2163–2181. GOMEZ M. and ETXEBARRIA G. (2000) Panorama of the Basque Country and its competence for self-government, European Planning Studies 8(4), 521-535. GONZÁLEZ S. (2005) The politics of the economic crisis and restructuring in the Basque Country and Spain during the 1980s, Space and Polity 9(2), 93-112. GOERLICH F. and MAS M. (1999) Medición de la desigualdad: contribución a una base de datos regional. Instituto Valenciano de Investigaciones Económicas, Valencia. HASHIMOTO A. and ISHIKAWA M. (1993) Using DEA to evaluate the state of society as measured by multiple social indicators, Socio-Economic Planning Sciences 27, 257–268. HASHIMOTO A. and KODAMA M. (1997) Has livability of Japan gotten better for 1956–1990?: A DEA approach, Social Indicators Research 40, 359–373. HASHIMOTO A., SUGITA T. and HANEDA S. (2009) Evaluating shifts in Japan’s quality of life, Socio-Economic Planning Science 43(4), 263–273. HIERRO M. and MAZA A. (2010) Per capita income convergence and internal migration in Spain: Are foreign-born migrants playing an important role?, Papers in Regional Science 89(1), 89–107. HOGGART K. and PANIAGUA A. (2001) Rural Restructuring in Spain?, Journal of Rural Studies 17, 63-80. HUDSON R. (2005) Re-thinking change in old industrial regions: reflecting on the experiences of North East England, Environment and Planning A 27, 581–596. Hudson R. (2006) Regional devolution and regional economic success: enabling myths and illusions about power, Geografiska Annaler 88B, 159–171. HUDSON R. (2007) Regions and regional uneven development forever? Some reflective comments upon theory and practice, Regional Studies 41(9), 1149–1160. INE (2008) Social Indicators. 2008 Edition. Spanish Statistical Office. http://www.ine.es IVIE, BBVA Foundation (several years) El stock de capital de capital en España y su distribución territorial (1964-2002). http://www.ivie.es/banco/stock.php KOOPMANS, T.C. (1951) An analysis of production as an efficient combination of activities, in KOOPMANS T. C. (Ed.) Activity analysis of production and allocation, pp. 33-97. John Wiley & Sons, New York. KORTELAINEN M. (2008) Dynamic environmental performance analysis: A Malmquist index approach, Ecological Economics 64, 701–715. KUMAR S. and RUSSELL R. (2002) Technological change, technological catch-up, and capital deepening: Relative contributions to growth and convergence, The American Economic Review 92(3), 527–548. KUZNETS S. (1955) Economic growth and income inequality. American Economic Review 45, 1–28. LOS B. and TIMMER M. (2005) The ‘appropriate technology’ explanation of productivity growth differentials: An empirical approach, Journal of Development Economics 77(2), 517- 531. Lovell C.A.K. and Pastor J.T. (1999) Radial DEA models without inputs or without outputs, European Journal of Operational Research 118, 46–51. LOVELL C.A.K., PASTOR J.T. and TURNER J.A. (1995) Measuring macroeconomic performance in the OECD: a comparison of European and non-European countries, European Journal of Operational Research 87, 507–518. Lundberg, M. and Squire, L. (2003) The simultaneous evolution of growth and inequality. The Economic Journal 113, 326-344. MACLEOD G. and JONES M. (2001) Renewing the geography of regions, Environment and planning. D: Society and Space 19(6), 669-695. MacLeod G. and Jones M. (2006) Mapping the geographies of UK devolution: institutional legacies, territorial fixes and network topologies, in Tewdwr-Jones M. and Allmendinger, P. (Eds.) Territory, Identity and Spatial Planning: Spatial Governance in a Fragmented Nation, pp. 335–352. Routledge, London. MAHLBERG B. and OBERSTEINER M. (2001) Remeasuring the HDI by Data Envelopment Analysis, Working Paper No. IR-01-069. International Institute for Applied Systems Analysis (IIASA), Laxenburg, Austria. MARCHANTE A.J. and ORTEGA B. (2006) Quality of life and economic convergence across Spanish Regions 1980–2001, Regional Studies 40, 471–483. MARTIN R. and SUNLEY P. (2006) Path dependence and regional economic evolution, Journal of Economic Geography 6(4), 395-437. MAZA A., VILLAVERDE J. and HIERRO M. (2009) Regional Productivity Distribution in the EU: Which are the Influencing Factors?, European Planning Studies 17(1), 149-159. MOLINA F. X. (2008) Los distritos industriales en la Europa Mediterránea. Las diferencias entre Italia y España, Mediterráneo Económico 13, 183-201. MORGAN K. (2004) Sustainable regions: Governance, innovation and scale, European Planning Studies 12(6), 871–889. MURIAS P., MARTINEZ F. and DE MIGUEL J.C. (2006) An economic wellbeing index for the Spanish provinces: a data envelopment analysis approach, Social Indicators Research 77, 395–417. MURIAS P., DE MIGUEL J.C. and RODRÍGUEZ D. (2008) A composite indicator for university quality assesment: The case of Spanish Higher Education System, Social Indicators Research 89, 129-146. MURIAS P., NOVELLO S. and MARTINEZ F. (2012) The regions of economic wellbeing in Italy and Spain, Regional Studies 46(6), 793-816. Nardo M., Saisana M., Saltelli A., Tarantola S. (2005): Tools for Composite Indicators Building. Work paper EUR 21682 EN. Joint Research Centre, European Commission, Ispra. OSBERG L. (1985) The measurement of economic wellbeing, in LAIDLER D. (Ed.) Approaches to Economic Well-being 26. Royal Commission on the Economic Union and Development Prospects for Canada. University of Toronto Press, Toronto. OSBERG L. and SHARPE A. (2002) International comparisons of trends in economic wellbeing, Social Indicators Research 58, 349–382. OSBERG L. and SHARPE A. (2004) Has economic well-being improved in Canada and the United States?, in WOLFF E.N. (Ed.) What has Happened to the Quality of Life in the Advanced Industrialized Nations?. Levy Economics Institute/Edward Elgar, Cheltenham. OSBERG L. and SHARPE A. (2005) How should we measure the ‘economic’ aspects of wellbeing?, Review of Income and Wealth Series 51(2), 311–336. PAASI, A. (2004) Place and region: looking through the prism of scale, Progress in Human Geography 28(4), 536-546.