Bio-based and Applied Economics 6(2): 183-208, 2017 ISSN 2280-6180 (print) © Firenze University Press ISSN 2280-6172 (online) www.fupress.com/bae Full Research Article DOI: 10.13128/BAE-18518 Quality of life and territorial imbalances. A focus on italian inner and rural areas Paola Bertolini, Francesco Pagliacci* Department of Economics “Marco Biagi”, Università di Modena e Reggio Emilia, Italy and CAPP (Centro di Analisi delle Politiche Pubbliche), Italy Date of submission: 2016 28th, June; accepted 2017 8th, March Abstract. The Italian National Strategy for Inner Areas stresses the importance of improving socio-economic conditions of people as the only way to reverse negative demographic trends in those areas. In this respect, improving quality of life (QoL) may represent a key driver. This work provides a statistical tool to measure existing gaps in QoL levels across Italian NUTS 3 regions, by focusing on inner areas. Being QoL a multidimensional concept, a composite indicator is computed following a non- compensatory approach: the QoL Mazziotta-Pareto Index. Firstly, we consider the var- iability of this indicator across Italy, with respect to the presence of inner areas. This analysis breaks down the supposed negative relationship between QoL and presence of inner areas, which the paper proves to be mostly overlapping with rural ones, by controlling for sub-national structural divides. Secondly, spatial aspects make the pic- ture more complex. Neighbourhood affects QoL at local level and through global, and local indicators of spatial autocorrelation, groups of NUTS 3 regions sharing similar QoL levels with their neighbours, are detected. From a policy perspective, locked-in paths among neighbouring regions can influence the effectiveness of place-based poli- cies. Keywords. Inner areas, rural areas, quality of life, spatial effects. JEL codes. O18, R00, R10, R11. 1. Introduction Across European countries, geographical differences in terms of economic and social development may also affect Quality of Life (QoL). QoL is similar to the multidimension- al concept of wellbeing, being a function of people’s life circumstances (MEA, 2005). Thus, it does not comprise just economic aspects (e.g., meeting people’s basic material needs): it may also refer to social networks, people’s health, their sense of worth and the sustainabil- ity of the environment on which they depend (Cagliero et al., 2011; Costanza et al., 2008; *Corresponding author: francesco.pagliacci@unimore.it 184 P. Bertolini, F. Pagliacci Petrosillo et al., 2013). At EU level, QoL shows wide territorial imbalances, for instance among urban and rural areas (Eurofound 2014). Even the EU Rural Development Policy has traditionally stressed the importance of the quality of life in rural areas. Nevertheless, when focusing on QoL territorial imbalances, the urban-rural divide is just part of the story. Even the concept of ‘Inner Areas’, which has been introduced by the Italian government, may play a role (Barca et al., 2014). The idea behind this concept is rather simple: as stressed by Christaller (1933), cities and larger towns have always pro- vided population with essential services (e.g. education, health, mobility). According to the model of economic growth that had occurred in Italy since the end of World War II, those urban hubs have been attracting more and more people also because of the variety of services they could offer (Barca et al., 2014). Conversely, minor municipalities and oth- er remote areas have started lagging behind. Suffering from geographical (and economic) remoteness and being affected by negative demographic trends, they have experienced a steady deprivation of essential services, which, in turn, has made the population decrease faster. These trends have led to negative effects such as: population abandonment and reduction of economic activities, disaggregation of the fabric of society, increasing costs in terms of land management. Despite the surge of counter-urbanization processes since the 1980s (Dematteis, 1986; OECD, 2009), most of rural and remote areas still suffer from these drawbacks (Bertolini et al., 2008; Copus et al., 2015), whose costs are paid by the country as a whole (Barca et al., 2014). Thus, besides traditional North-South socio-eco- nomic divides, other kinds of spatially divergent dynamics still affect Italy, suggesting the existence of local core-periphery patterns. To tackle this situation, the Italian government launched a specific National Strate- gy for Inner Areas, firstly aimed at defining them. According to it, inner areas are those municipalities that, being located at some considerable distance from major urban poles, suffer from a limited provision of essential services. Thus, such a definition essential- ly refers to a spatial (hence, more conventional) theoretical framework for inner areas, although a wide part of the European literature now points out ‘aspatial’ models of periph- erality (Copus, 2001; Kuhn, 2015; Bock, 2016; Noguera and Copus, 2016). Conversely, inner areas as defined by the national strategy (namely, on a spatial basis) tend to share rural and agricultural traits (Barca et al., 2014). Since its launch, this strategy has fuelled the attention of Italian policymakers towards the need for improving social and economic conditions of people living in inner areas, as the only way to reverse negative demographic trends. Assuring a good performance of the local labour market, creating new forms of employment, and enhancing QoL levels represent the only way to cut emigration from inner areas and to attract new people and households (Barca et al., 2014). Given these important policy implications, this paper provides some simple statisti- cal tools for policy analysis, and in particular for assessing and measuring existing gaps in QoL levels across Italy, with a specific attention to rural and inner areas. As QoL is a mul- tidimensional concept, its measurement poses three major methodological issues (OECD, 2008), which this paper explicitly tackles: i) defining the most appropriate territorial level of analysis, according to available data; ii) returning a composite and comprehensive QoL indicator, whose variability across Italy can be eventually assessed; iii) properly stressing the role of spatial spillovers in influencing such a variability. 185Quality of life and territorial imbalances The rest of the paper is organised as follows. Section 2 introduces the concept of inner areas, as defined by the Italian National Strategy for Inner Areas. Section 3 tackles the main measurement issues linked with a spatial definition of inner areas, suggesting some ways to assess the importance of inner areas at NUTS 3 level. Section 4 provides a synthetic indicator of QoL, discussing main methodological approaches and returning main results. Main relationships between inner areas and QoL are also described. Section 5 focuses on spatial issues, by introducing into the analysis the role of spatial neighbour- hoods. Section 6 concludes the paper. 2. The italian national strategy for Inner Areas In 2014, the Italian government launched the National Strategy for Inner Areas as a way to promote innovative projects within remote municipalities. In this framework, remoteness is assessed in terms of lack of essential services, considered as constituents of the EU ‘citizenship’ (Barca et al., 2014). Focusing on the provision of services, the strategy defines inner areas on the basis of their geographical distance from those centres (i.e. large cities) providing services. Thus, rather than an ‘aspatial’ definition of inner peripheries, a spatial approach is mostly adopted. Referring to service provision as a key element to classify the territory is not a com- pletely new approach: it was introduced in 2008 by the EU DG Regio, with the goal of better classifying rural areas in comparison to the official OECD classification, which had been mainly based on population density (Dijkstra and Poelman, 2008). DG Regio com- bined the OECD population criterion with an indicator of distance and it considered the driving time (namely 45 minutes) to reach a city of at least 50.000 inhabitants, as a main centre of services. In contrast to the DG Regio classification, the methodology suggested by the Italian national strategy does not consider rural conditions or number of inhabitants of cities: it only focuses on the effective availability of services at municipal level, defining inner areas in terms of their spatial remoteness. In particular, inner areas deserve political attention – and a national strategy – for many reasons. Firstly, they hide wide potentials, holding an important environmental (e.g., water, forests, natural and human landscapes), cultural (historic settlements, small and rural museums, skills centres) and agricultural heritage (Barca et al., 2014). Secondly, they still represent a key part of Italy (60% of the total land area and 25% of Italian popu- lation). In order to improve their socio-economic conditions, the national strategy moves from inner areas’ main problems. In fact, most of them have been facing a steady process of marginalisation, followed by a degradation in the provision of essential services (health, education and mobility). Therefore, they are expected to increase their own marginalisa- tion, boosting national social costs in terms of hydro-geological instability, degradation of both cultural and landscape heritage, decay and soil consumption (Barca et al., 2014). As a way to reverse these trends, both enhancing QoL and improving labour market perfor- mances represent effective policy tools (Barca et al., 2014). In particular, the strategy moves from the idea that, despite common sense, some inner municipalities have been able to implement good practices, over time. Thus, if inner 186 P. Bertolini, F. Pagliacci areas’ economic marginalisation does not represent an unavoidable process, the strategy just aims to spread the knowledge about those best practices, trying to replicate them across Italy (Barca et al., 2014). In most cases, interventions involve the promotion (and preservation) of local environment and local cultural resources. To this respect, the definition of inner areas as suggested by the strategy could over- lap with the identification of rural regions, which experience a significant lack in essential services provision as well. Nevertheless, both inner and rural areas may also share some potential strengths. For instance, they both have plenty of area-specific agricultural pro- ductions, which originate from tight connections between the territory and local skills. Inner areas are home for many typical productions (PDOs and PGIs), prompting local food industry1 (Barca et al., 2014). Given the existence of such a potential, studies on rural development have often highlighted the emergence of positive tendencies (such as the increase in rural tourism and the diffusion of agriculture multifunctionality), which may prompt the development of both rural and inner areas (Hoggart et al., 1995; Pania- gua 2012), overcoming traditional urban-rural economic divides (Pagliacci, 2017). Besides a radical change in its theoretical perspective, even the implementation of this strategy is innovative, as each region is forced to select a limited number of pilot pro- gramme areas2, to promote territorial safeguarding, valorisation of natural and cultural assets (namely sustainable tourism), agricultural activities, renewable energy and energy saving, handicraft and local knowledge. As underlined above, the enhancement of QoL at local level sits at the heart of the National Strategy for Inner Areas, their socio-economic development being the main aim of the strategy. In other words, QoL emerges as an important target of this strategy. Actu- ally, the enhancement of QoL is crucial to promote local development, involving both eco- nomic growth and a greater social inclusion. As already mentioned, the ultimate objective – and guiding light – of the strategy is reversing population trends in inner and remote areas. A reversal in demographic dynamics is acknowledged as a key factor to limit social costs linked to socio-economic marginalisation, hydrogeological instability and degrada- tion of both human and environmental capital. Therefore, QoL levels cannot be ignored: actually, they represent key drivers in people’s settlement choices. It follows that assess- ing QoL divides between urban poles and inner areas represents a key issue, especially in helping policy makers in fine-tuning their own policies (Barca et al., 2012). Furthermore, QoL divides matter even within inner areas, which now show polymorphic traits, having followed differentiated trajectories of development for decades (Barca et al., 2014). Thus, assessing different needs across different areas as well as different geographic patterns may represent a great improvement to the strategy itself. 1 Foodstuffs represent cultural assets as they refer to local identities. Furthermore, new types of employment may originate, thanks to major changes in agro-food activities and in the distribution process, which may also show positive effects on the environment (Barca et al., 2014). Indeed, Common Agricultural Policy stresses cross-com- pliance as a key point (Matthews, 2013). 2 Although following nationally shared criteria, regions are in charge of identifying the neediest areas, according to a well-defined selective approach (Barca et al., 2014). 187Quality of life and territorial imbalances 3. How “inner” are NUTS 3 regions? Methodological and measurement issues 3.1 Share of inner areas at NUTS 3 level The national strategy provides a detailed and innovative methodology to classify Ital- ian municipalities. While mapping and zoning have always represented challenging tasks for policy makers, Barca et al. (2012) suggest that any place-based policy would take great advantage from more accurate indicators of existing territorial differences. Here, the iden- tification of inner areas moves from the polycentric structure of Italy, where just some main cities provide services to other municipalities that gravitate around them, each of them with its own level of spatial remoteness. Three main theoretical assumptions drive this way of mapping inner areas (Barca et al., 2014): • the network of differentiated urban centres provides the whole range of essential ser- vices, generating catchment areas according to a gravitational models (Christaller, 1933); • other minor municipalities’ degree of spatial remoteness from this network may hin- der social inclusion as well as QoL levels (inner areas in a spatial perspective); • inner areas are not homogeneous and, in fact, they are becoming more diverse with regard to their own socio-economic and territorial development (Sotte et al., 2012; Barca et al., 2014; Copus et al. 2015). For decades, they have followed different evo- lutions, according to both their natural/geographical characteristics and their relative proximity/remoteness to urban areas. In other words, different time-space evolution- ary patterns have occurred among inner areas. From a methodological perspective, identification of inner areas is a two-step proce- dure. Firstly, Italian municipalities acting as service providers are defined as those munici- palities (or groups of neighbouring municipalities) being able to provide simultaneously: i) the full range of secondary education; ii) at least one major emergency care hospital; ii) at least one medium railway station, with an average degree of uptake for regional servic- es and some long-distance journeys3. Accordingly, both urban poles and inter-municipal poles are defined as those cities (or groups of contiguous cities) that provide the whole set of these services4. Then, all remaining municipalities are classified into four different typologies (outlying areas; intermediate areas; peripheral areas and ultra-peripheral areas), according to spatial accessibility. Number of minutes taken to get from each municipal- ity to the nearest urban pole is considered to compute each band (less than 20 minutes, less than 40 minutes, less than 75 minutes, more than 75 minutes) (Barca et al., 2014). As stressed, such a classification moves from a spatial definition of inner areas, consid- ering no other socio-economic weaknesses but remoteness. Eventually, moving from this six-typology classification, a broader definition of inner areas is provided by just putting together intermediate, peripheral and ultra-peripheral areas (Barca et al., 2014). Given the purposes of this work, here we refer to this broader definition of inner areas. Nevertheless, some methodological drawbacks occur. A first issue deals with the ter- ritorial level of the analysis. Inner areas are defined at municipality level, but no reliable 3 Barca et al. (2014) provide further details on the characteristics of services under consideration. 4 All NUTS 3-level capital municipalities are considered as urban poles, even when they do not provide all the aforementioned set of essential services (Barca et al., 2014). 188 P. Bertolini, F. Pagliacci QoL indicators are available at such a territorially disaggregated level. At the maximum, any analysis can refer to NUTS 3 level (i.e., 110 observations). Thus, municipal data have to be converted into NUTS 3 level data5. To return robust results, the relevance of inner areas within each NUTS 3 region is computed according to three alternative indicators. Firstly, number of municipalities is considered. Given the i-th NUTS 3 region and its n municipalities, the inner-municipality indicator (Ii) is defined as follows: I m ni jj n 1 ∑ = = (1) where j is one of the n municipalities in the NUTS 3 region i and the generic element mj can take two different values: mj = 1, when j is classified as either intermediate or periph- eral or ultra-peripheral; mj = 0, otherwise. Alternatively, both population and land area are considered. As in (1), given the i-th Italian NUTS 3 region and its n municipalities, the inner-population indicator (IPi) and the inner-area indicator (IAi) are defined as follows: IP m P P ( ) i j jj n jj n 1 1 ∑ ∑ = = = (2) IA m A A ( ) i j jj n jj n 1 1 ∑ ∑ = = = (3) where j is one of the n municipalities in the NUTS 3 region i, Pj is its population and Aj is its land area. As in (1), the generic element mj can take two values (either 0 or 1). Each indicator may range from 0 to 1: 0 stands for the absence of inner area; 1 stands for the absence of non-inner areas. Figure 1 returns the values of each of the three indicators at NUTS 3 level. While Ii and IAi show similar patterns, when focusing on population, the share of inner areas at NUTS 3 level is generally lower. Just in a few Southern NUTS 3 regions, the share of population living in inner municipalities is above 50%. A sharp North-South divide also emerges when looking at average values at regional level (Table 1). Among Italian NUTS 2 regions, Liguria, Piedmont and Lombardy share the lowest shares of population living in inner municipalities (less than 12%). On the opposite side, in three Southern regions (i.e., Basilicata, Molise and Calabria) more than 55% of their population lives in inner areas. Thus, such a North-South divide should be always taken into account in the rest of the analysis. 5 The authors are aware that such a transformation may results into concrete limitations for the analysis. In some cases, NUTS 3 regions might be internally heterogeneous. Thus, a focus on LAU 2 territorial units would be much more appropriate for this kind of analysis, if data about QoL were available. 189Quality of life and territorial imbalances Figure 1. Inner areas, share out of the total by NUTS 3 region: Number of inner municipalities (left); Inner population (centre); Inner land area (right). 1 Source: authors’ elaboration Table 1. Inner areas, share out of the total by region. Regions Inner municipalities (Ii) Inner population (IPi) Inner land area (IAi) North-West Piedmont 38.06% 11.70% 46.29% Aosta Valley 59.46% 30.50% 71.60% Lombardy 33.03% 10.69% 45.95% Liguria 43.83% 8.89% 50.52% North-East Trentino-Alto Adige 76.28% 44.93% 81.24% Veneto 33.05% 18.72% 38.06% Friuli-Venezia Giulia 39.45% 13.77% 53.79% Emilia-Romagna 41.95% 13.11% 42.84% Centre Tuscany 44.25% 13.10% 51.30% Umbria 61.96% 25.31% 48.51% The Marches 44.35% 14.77% 42.73% Latium 76.72% 28.06% 64.62% South Abruzzo 75.41% 37.05% 70.96% Molise 80.15% 61.11% 83.37% Campania 49.00% 14.70% 63.19% Apulia 54.26% 26.05% 44.92% Basilicata 96.18% 74.65% 92.32% Calabria 79.95% 55.21% 81.10% The Islands Sicily 74.62% 41.34% 73.36% Sardinia 84.35% 52.27% 84.54% Italy 51.72% 22.43% 59.77% Source: authors’ elaboration. 190 P. Bertolini, F. Pagliacci 3.2 Inner Areas and other indicators of rurality As already stressed, inner areas are expected to share important rural traits (Bar- ca et al., 2014). Having computed NUTS 3 level indicators, we can compare them with alternative indexes of rurality: Eurostat urban-rural typologies (Eurostat, 2010); the PRI (PeripheRurality Indicator) (Camaioni et al., 2013); the FRI (Fuzzy Rurality Indicator) (Pagliacci, 2017). Each indicator is built on an alternative methodology, all of them referring to the whole EU-27. Eurostat (2010) defines urban-rural typologies according to population den- sity and controlling for the presence of large cities. Such a single indicator is eventually collapsed into a discrete ordinal variable, returning three urban-rural typologies: predomi- nantly urban (PU), intermediate (IR) and predominantly rural (PR) regions. Thus, it is too rough to capture increasing rural areas’ polymorphism (Camaioni et al., 2013). Camaioni et al. (2013) compute the PRI, following a multidimensional approach. They apply a conventional principal component analysis to a 24-variable dataset (covering socio- demographic features, economic structure, land use, remoteness). Then, an ideal urban benchmark (i.e., a region being extremely urban in Europe) is identified and statistical dis- tances between any other EU region and this benchmark are computed (Camaioni et al., 2013). So, for each region, the PRI returns jointly the extent of rurality and peripherality. Eventually, the FRI stresses the concept of urban-rural continuum. It applies fuzzy logic to six input variables (covering role of agriculture, population density and landscape/ use of land) and it returns a final output (i.e., the FRI), which ranges from 0 to 1, where 0 stands for completely urban; 1 stands for completely rural (Pagliacci, 2017). The statistical relationship between indicators of inner areas and indicators of rurality can be assessed by means of Pearson correlation coefficients. Table 2 returns the corre- lation between Ii, IPi, IAi respectively and the aforementioned three indicators of rural- ity computed for Italian NUTS 3 regions6. In any specification, correlations are positive and statistically significant, also thanks to the spatial definition of inner areas adopted by the strategy. Coefficients are larger for the FRI than for the PRI, although the latter also assesses NUTS 3 regions remoteness (thus, a spatial concept, similar to the one referring to inner areas). Similar findings emerge when looking at the presence of inner municipali- ties among different Eurostat urban-rural typologies. Point-biserial correlation between each dummy variable and the presence of inner areas is consistent with expectations: cor- relation is positive for PR regions (inner areas’ share is larger in PR regions than in non- PR ones), and it is negative for both PU and IR ones. When comparing average shares of inner areas among three typologies, similar evidence is returned: One-Way ANOVA (Analysis of Variance) tests whether average values are statistically different or not.7 Tests show statistically significant differences in any specification. As a strong relationship between rural and inner areas emerges, the National Strategy for Inner Areas implicitly refers to rural areas, as well. 6 Here, just 107 observations are considered, as neither PRI nor FRI values are available for Monza and Brianza, Fermo, Barletta-Andria-Trani. Actually, those NUTS 3 regions were just instituted in 2004. 7 Preliminarily, Levene’s Test is computed. It tests the null hypothesis that groups’ variances are equal. If they are, simple F test for the equality of means in a One-Way ANOVA is performed; otherwise, Welch (1951) method is adopted. 191Quality of life and territorial imbalances Table 2. Relationships between inner indicators and indicators of rurality (PRI, FRI, Urban-rural typol- ogy) (p-values in parenthesis). Ii IPi IAi Pearson correlation coefficients: PRI (Camaioni et al., 2013) 0.522* 0.560* 0.487* (0.000) (0.000) (0.000) FRI (Pagliacci, 2017) 0.657* 0.601* 0.638*   (0.000) (0.000) (0.000) Point-biserial correlation: Urban-rural typology: PR regions 0.471* 0.538* 0.421* (0.000) (0.000) (0.000) IR regions -0.242* -0.269* -0.248* (0.012) (0.005) (0.010) PU regions -0.291* -0.341* -0.218* (0.002) (0.000) (0.024) Avg. comparison: Avg. PR regions 0.689 0.461 0.677 Avg. IR regions 0.459 0.224 0.478 Avg. PU regions 0.343 0.112 0.407 Levene’s test 0.182 6.608* 0.318 (0.834) (0.002) (0.728) One-way ANOVA  17.919* 31.324* 12.871* (0.000) (0.000) (0.000) * Statistically significant at the 5% level. Source: authors’ elaboration. 4. QoL as a multidimensional concept 4.1 The Mazziotta-Pareto Index As a multidimensional concept (MEA, 2005), QoL includes both economic aspects and social-relational ones (Cagliero et al., 2011; Costanza et al., 2008; Petrosillo et al., 2013). Thus, measuring QoL is harder than measuring the presence of inner areas: it requires the challenging construction of a composite and multidimensional index (OECD, 2008; Mazziotta and Pareto, 2014). In the case of QoL, both ‘objective’ and ‘subjective’ aspects matter. The former dimen- sion refers to physical and health status, personal income, local standards of living (Mal- kina-Pykh and Pykh, 2008; Petrosillo et al., 2013). The latter focuses on individuals’ sub- jective experience of their lives (Land, 1996) as well as psychological responses (e.g., life and job satisfaction and personal happiness). Although the European Foundation for the Improvement of Living and Working Conditions follows a subjective approach in carry- ing out surveys on the level of quality of life across Europe (e.g., Eurofound, 2014), here 192 P. Bertolini, F. Pagliacci no subjective measures of QoL are included, as assessing them is rather difficult. Actually, no sociologic surveys or investigations (Shin and Johnson, 1978) are available at NUTS 3 level. Therefore, this analysis just focuses on objective QoL indicators. According to this perspective, a wide literature has already discussed the main drivers of QoL at sub-national level. In particular, urban-rural divides have been widely investi- gated (see for instance Cagliero et al., 2011; Florida et al., 2013; Shucksmith et al., 2009; Sørensen, 2014). In Italy, the most cited QoL indicator, available at NUTS 3 level, is pro- vided by the financial newspaper “Il Sole 24 Ore”. Every year, it returns a QoL indicator based on 36 single variables, grouped into six different thematic areas (economic wealth, business activities and employment, services and environment, population, crime, leisure). Despite its large popularity, this indicator suffers from some drawbacks. Firstly, it assumes perfect substitutability among original variables (i.e., a good performance in a thematic area may compensate a bad performance in another one). Secondly, different standard deviations among each variable may affect the outcome8 (Mazziotta and Pareto, 2010a; 2010b; 2016). Lastly, the set of original variables changes every year: this makes impos- sible to assess time comparisons. To tackle these drawbacks, an alternative indicator is suggested here: the Mazziotta- Pareto Index (MPI), a well consolidated indicator to assess QoL at local level. The MPI is a non-linear composite index, which transforms individual variables into a standardized indicator. It sums original data up, using arithmetic mean but adjusting it by a ‘penalty’ coefficient, which is related to the variability observed for each unit (Mazziotta and Pare- to, 2016). Accordingly, those observations showing unbalanced values of the initial vari- ables are penalised, according to a non-compensatory perspective (Mazziotta and Pareto, 2010a; 2016). In particular, here we adopt the following methodology to compute a QoL MPI. Firstly, original variables standardisation occurs. Let’s consider the original matrix X, whose generic element is xij. It has n rows (observations) and m columns (variables), which are grouped into p thematic areas. From X, a standardised matrix Z is computed (Mazziotta and Pareto, 2010a), whose generic element zij is alternatively defined as follows: z x M S 100 10ij ij x x j j = + − (4) z x M S 100 10ij ij x x j j = − − (5) where: M x nx ij i n 1 j ∑ = = and S x M n ( ) x ij i n x j 1 2 j ∑ = − = In particular, (4) is applied to those indicators that are concordant in sign with the QoL MPI; otherwise, (5) is applied. Accordingly, p sub-indicators of QoL are computed, 8 This distortion comes from the fact that the synthetic indicator is computed through distances from a bench- mark (i.e. the best performing NUTS 3 region). 193Quality of life and territorial imbalances each of them referring to a thematic area. Given h thematic areas, each of them compris- ing k variables, the h-th sub-indicator of QoL is given by: z z k ih i k h j j k , ( 1) 1 ∑ = − + = (6) The p sub-indicators z ih are then grouped together and a QoL MPI is returned as: MPI M S cvi z z zi i i= − (7) where: M z pz ih h p 1 i ∑ = = S z M p ( ) z ih h p z 1 2 j i∑ = − = cv S Mz z z i i i = The S cv z zi i product represents the most innovative aspect of this approach. It penalis- es those units showing unbalanced values of the p thematic sub-indicators (Mazziotta and Pareto, 2016). In addition, due to the standardisation provided by (4) or (5), each indica- tor’s mean is 100 and each standard deviation is 10 (Mazziotta and Pareto 2010; Aiello and Attanasio, 2004). Here, this methodology is applied to a set of 28 original variables, retrieved for each Italian NUTS 3 region. They refer to seven different thematic areas linked to QoL: • Wealth & economic competitiveness (3 indicators), • Services (3 indicators), • Labour market (5 indicators), • Neighbourhood safety (3 indicators), • Population (7 indicators), • Leisure (2 indicators), • Environment & Energy (5 indicators). Thematic areas partially overlap with the ones provided by “Il Sole 24 Ore”. Neverthe- less, original variables are open data published by the OpenCoesione (OC) dataset: the fact that the source of data is ISTAT in most cases assures full comparability of results across time. 9 (Table 3). 4.2 QoL and its sub-indicators: main territorial patterns Seven sub-indicators of QoL are returned. Each sub-indicator shows standardised values. Figure 2 shows the values of each sub-indicator across Italian NUTS 3 regions. Wealth and economic competitiveness show a strong North-South divide, confirming larger QoL 9 Replicability of the analysis over time is a key issue. Indeed, changing the set of variables under study may dra- matically affect final outcomes. 194 P. Bertolini, F. Pagliacci Table 3. List of input variables, by thematic area. Variable Definition Effect on QoL Year Source Economic wealth & Competitiveness Per capita GVA (€) Gross Value Added (current prices) per inhabitants, all sectors + 2013 Istat Per capita Export (€) Exports per inhabitants + 2014 Istat (OC) Per capita Patents Patents registered to the European Patent Office, per million inhabitants + 2011 Istat on Eurostat data (OC) Provision of services Diffusion of pre-school services % of municipalities out of the total adopting pre-school services (e.g. nursery schools) + 2012 Istat (OC) Children 0-3 attending day care and pre-school % of young children (aged 0-3 years) who use day care facilities and other pre-school services + 2012 Health emigration ratio Share of the out-migration in hospital in other regions out of total hospital admissions - 2013 Labour market Employment rate Employed persons (aged 15-64) over the number of people 15-64 (%) + 2014 Istat (OC) Elderly people employment rate Employed persons (aged 55-64) over the number of people 55-64 (%) + 2014 Youth unemployment rate Unemployed persons (aged 15-24) over the number of persons 15-24 in the labour force (%) - 2014 Unemployment rate Unemployed persons (aged 15+) over the number of persons (aged 15+) in the labour force (%) - 2014 Gender differences Differences in % points between male and female employment rates - 2014 Neighbourhood safety Istat on Ministero Interno, Dipartimento Pubblica Sicurezza data (OC) Rate of thefts Number of recorded thefts per a thousand inhabitants - 2013 Rate of robberies Number of recorded robberies per a thousand inhabitants - 2013 Rate of homicides Number of recorded intentional homicides per 100 thousand inhabitants - 2013 Population Population Density Inhabitants per km2 - 2014 Istat Old-Age dependency ratio Ratio of older dependents (people aged 65+) to the working-age population (15-64) - 2014 Ageing Index Number of persons aged 65+ per hundred persons under age 15 - 2014 195Quality of life and territorial imbalances in the North of the country. Throughout Southern regions and the Islands, just Ragusa and Cagliari show local values which are close to the national average. Provision of ser- vices is at a maximum across Emilia-Romagna and Tuscany, due to a long-lasting atten- tion to these political items (Bripi et al., 2011; Giordano and Tommasino, 2011). Con- versely, education and health services show poor performances across the South (e.g., Molise, Basilicata and Calabria) and in Lazio. Similarly, labour market performance is poor in Southern NUTS 3 regions, whereas the best performances occur across the so- called Third Italy (Bagnasco, 1977 and 1988), namely in the North-East and alongside the Adriatic. Neighbourhood safety shows a less sharp North-South divide. Best performances Variable Definition Effect on QoL Year Source Internal net migration rate Difference of immigrants and emigrants within the country in a year, divided per 1000 inhabitants + 2014 External net migration rate Difference of immigrants and emigrants (from/to abroad) in a year, divided per 1000 inhabitants + 2014 Istat Life expectancy at birth, males Number of years a new-born male infant would live (assuming no changes in patterns of mortality throughout its life) + 2014 Life expectancy at birth, females Number of years a new-born female infant would (assuming no changes in patterns of mortality throughout its life) + 2014 Leisure Live theatre and live music performances Tickets sold to live theatre and live music performances, per 100 inhabitants + 2007 Istat on SIAE data (OC) Tourists Number of overnight stays spent by national and foreign tourists in tourist accommodations, per inhabitant + 2013 Istat (OC) Environment and energy Water use efficiency % of water distributed to customers out of the total volume introduced into the municipality water network + 2008 Istat (OC) Waste recycling Share of municipal waste recycled out of total solid waste (%) + 2014 Istat on ISPRA data (OC) Renewable energy % of GWh renewable energy to total energy production in GWh + 2010 Air quality monitoring network Number of control stations of the air quality monitoring network, per 100 thousands inhabitants + 2012 Istat - Open Coesione Discontinuity of electricity supply Number of long-lasting interruptions in electricity supply (average number per single customer) - 2014 Istat on Autorità Energia elettrica, Gas, Sistema idrico data (OC) Source: author’s elaboration. 196 P. Bertolini, F. Pagliacci are observed across mountain areas (the Alps and the Apennines), while metropolitan and urban NUTS 3 regions show poorer performances. Population sub-indicator shows a good performance across Emilia-Romagna and Trentino-Alto Adige. Nevertheless, Southern regions do not lag behind Northern ones, despite a lower presence of foreign people. Lei- sure activities show a scattered pattern across Italy, with urban areas and many Northern Figure 2. Sub-indicators of QoL and QoL MPI, by NUTS 3 region. Economic Wealth & Competitiveness Provision of services Labour market Neighbourhood safety Population Leisure Environment & Energy QoL MPI Source: authors’ elaboration. 197Quality of life and territorial imbalances and Central Italian regions performing above the average. Lastly, when considering envi- ronment and energy, local performance is good across North-East NUTS 3 regions as well as in the Aosta Valley. In the South, Sicily and Calabria show bad performances, whereas other inner NUTS 3 perform generally better. Moving from these sub-indicators, a comprehensive QoL MPI is computed, by penal- ising those NUTS 3 regions that show more unbalanced performances. Figure 2 also returns main results for QoL MPI: most of Northern NUTS 3 regions share above-the-average levels of QoL MPI, while Southern ones generally lag behind. Rather than returning a ranking of NUTS 3 regions (which may change over time), the following sections aim to analyse existing correlations between inner areas and QoL levels. Furthermore, it is possible to notice that results would not have changed much, if we had not considered the penalty coefficient S cv z zi i Indeed, QoL MPI and the average mean of the seven indicators for each sub-thematic area are actually highly correlated. Nevertheless, the adopted procedure, although being more complex, seems to be more robust from a theoretical perspective. 4.3 QoL and inner areas: main relationships The analysis of Pearson correlation coefficients makes possible the preliminary assess- ment of the main relationship between QoL levels and the presence of inner areas at NUTS 3 level (Table 4). At national level, QoL dimensions are negatively correlated to the presence of inner areas, with the only exception of neighbourhood safety, which shows a positive relation with the presence of inner areas. A first – hence, preliminary – overlook of these results would suggest that Italian inner areas generally suffer from low levels of QoL: thus, the launch of a national strategy targeted to them is definitely good news. Furthermore, as shown in Section 3, given the aforementioned relationship between the presence of both inner and rural areas, same results are expected to hold even with respect to the rural part of the country. Nevertheless, same data may hide some more complex patterns, which could contrast this general and first overlook. Firstly, different patterns may arise at sub- national level. On average, Italian Southern regions tend to show a larger presence of inner areas than Northern ones (section 3). This could affect overall results in terms of QoL MPI, as well. Thus, it is useful to disentangle previous results by macro-groups of regions. For sake of simplicity, here we refer to the classification provided in Table 1: Table 5 shows Pearson correlation coefficients per sub-indicator and per group of regions. When disentangling by group of regions, differences between urban poles and inner areas seem disappearing. In particular, the negative relationship between inner areas and QoL no longer hold. In fact, just a few sub-indicators appear to be statistically related to QoL at a sub-national level: • North-West: a positive relation between the sub-indicator Neighbourhood safety and the presence of inner areas occurs. Actually, the presence of large and unsafe metro- politan areas plays a role. • North-East: service provision is negatively tied to the presence of inner areas at NUTS 3 level, when considering total population. Nevertheless, both ‘population’ and ‘envi- 198 P. Bertolini, F. Pagliacci ronment and energy’ are positively related to the presence of inner areas, as well as the QoL MPI. • Centre: a negative relation between QoL and the presence of inner areas affects many sub-indicators of QoL (e.g. economic wealth, service provision, labour market, envi- ronment and energy). The only sub-indicator that is positively related to the presence of inner areas is neighbourhood safety. • South: a negative relationship emerges when considering service provision and inner areas; on the contrary, safety is positively associated with a larger presence of inner areas. • The islands: relationships between QoL and presence of inner areas are never signifi- cant. Here, data confirm inner areas’ polymorphism: when controlling per single mac- ro-region, strikingly different results emerge. In the North-East, inner areas do not lag behind urban poles when referring to QoL MPI, whereas opposite findings occurs when focusing on Central NUTS 3 regions. Thus, these findings seem supporting the choice made by the national strategy about the implementation of a place-based policy, in accordance with regional governments: such a strategy seems to be more appropriate when dealing with specific problems, which may occur locally. Table 4. Pearson correlation coefficients between inner areas indicators and indicators of QoL (p-val- ues in parenthesis).   Ii IPi IAi Economic Wealth & Competitiveness -0.504* -0.534* -0.443* (0.000) (0.000) (0.000) Provision of services -0.523* -0.518* -0.478* (0.000) (0.000) (0.000) Labour Market -0.405* -0.465* -0.352* (0.000) (0.000) (0.000) Neighbourhood safety 0.310* 0.350* 0.314* (0.001) (0.000) (0.001) Population -0.112 -0.200* -0.071 (0.245) (0.036) (0.459) Leisure -0.213* -0.298* -0.193* (0.025) (0.002) (0.043) Environment & Energy -0.370* -0.388* -0.294* (0.000) (0.000) (0.002) QoL MPI -0.420* -0.470* -0.357*   (0.000) (0.000) (0.000) * Statistically significant at the 5% level. Source: authors’ elaboration. 199Quality of life and territorial imbalances 5. The role of the neighbouring space Regional patterns are just part of the story: actually, spatial effects can be modelled in a more accurate way. Italian NUTS 3 regions show a narrow extension: on average, their sur- Table 5. Pearson correlation coefficients between inner areas indicators and indicators of QoL by mac- ro-regions (p-values in parenthesis). Wealth & Competi- tiveness Services Labour Market Neigh- bourhood Safety Population Leisure Environ- ment & Energy QoL MPI North-West Ii -0.388 -0.237 0.088 0.493* -0.186 -0.016 0.062 -0.018 (0.056) (0.254) (0.676) (0.012) (0.373) (0.938) (0.768) (0.930) IPi -0.218 -0.114 0.177 0.551* -0.082 -0.098 0.274 0.170 (0.295) (0.588) (0.398) (0.004) (0.697) (0.640) (0.185) (0.418) IAi -0.362 -0.145 0.04 0.464* -0.100 0.021 0.074 0.048 (0.075) (0.488) (0.850) (0.019) (0.633) (0.922) (0.726) (0.820) North-East Ii 0.212 -0.363 0.267 0.083 0.491* 0.181 0.587* 0.430* (0.344) (0.097) (0.230) (0.713) (0.020) (0.421) (0.004) (0.046) IPi 0.017 -0.487* 0.115 0.407 0.317 0.148 0.428* 0.334 (0.941) (0.022) (0.609) (0.060) (0.151) (0.510) (0.047) (0.129) IAi 0.175 -0.378 0.239 0.177 0.452* 0.156 0.597* 0.423* (0.437) (0.083) (0.284) (0.432) (0.035) (0.489) (0.003) (0.050) Centre Ii -0.663* -0.630* -0.504* 0.295 -0.451* -0.193 -0.428* -0.623* (0.001) (0.002) (0.017) (0.182) (0.035) (0.390) (0.047) (0.002) IPi -0.697* -0.703* -0.538 0.496* -0.421 -0.376 -0.454* -0.672* (0.000) (0.000) (0.010) (0.019) (0.051) (0.084) (0.034) (0.001) IAi -0.549* -0.533* -0.421 0.279 -0.324 -0.294 -0.238 -0.531* (0.008) (0.011) (0.051) (0.208) (0.142) (0.184) (0.287) (0.011) South Ii 0.082 -0.529* 0.306 0.642* 0.194 -0.251 0.083 0.177 (0.704) (0.008) (0.146) (0.001) (0.363) (0.236) (0.700) (0.408) IPi -0.021 -0.584* 0.179 0.670* 0.071 -0.380 0.047 0.045 (0.923) (0.003) (0.402) (0.000) (0.740) (0.067) (0.826) (0.836) IAi 0.070 -0.573* 0.358 0.682* 0.207 -0.331 0.158 0.199 (0.745) (0.003) (0.086) (0.000) (0.333) (0.114) (0.462) (0.352) The Islands Ii 0.091 0.229 0.476 0.264 -0.068 0.12 0.034 0.333 (0.729) (0.376) (0.053) (0.306) (0.794) (0.646) (0.896) (0.192) IPi -0.107 0.138 0.310 0.219 -0.120 -0.07 0.171 0.180 (0.684) (0.597) (0.225) (0.398) (0.646) (0.792) (0.513) (0.488) IAi 0.222 0.221 0.421 0.080 -0.090 0.327 -0.059 0.279 (0.391) (0.394) (0.093) (0.759) (0.731) (0.200) (0.821) (0.279) * Statistically significant at the 5% level. Source: authors’ elaboration. 200 P. Bertolini, F. Pagliacci face is 2,745 km2, i.e. a square whose side is just 52 kilometre. Thus, people are used to live, work and spend part of their own leisure time across neighbouring NUTS 3 regions, and it could be misleading to consider QoL at NUTS 3 level by just focusing on the relationships between it and socio-economic features in the same NUTS 3 region. In fact, space matters (Tobler, 1970), at least in two ways. Firstly, QoL may show spatially clustered patterns, given the fact that neighbouring NUTS 3 regions tend to share similar QoL levels. Secondly, struc- tural characteristics of neighbouring NUTS 3 regions (e.g., the presence of either urban poles or inner areas among them) may also affect QoL levels, having an impact on people’s every- day life10. To this respect, these characteristics matter and should be considered separately. 5.1 Spatial autocorrelation: QoL across neighbouring NUTS 3 regions The simplest way to assess QoL differentials across neighbouring observations is rep- resented by the analysis of global and local indicators of spatial autocorrelation. Accord- ing to the first law of geography (Tobler, 1970), patterns of spatial association are formally assessed by means of the degree of dependency among observations within a given geo- graphic space (Anselin, 1988 and 1995). Global Moran’s I statistics tests for the presence of spatial dependence. It is a synthetic measure of global spatial autocorrelation, computed as follows (Moran, 1950; Cliff and Ord, 1981): I n w w y y y y y y i j N, , ijj n i n ij i jj n i n ii n 11 11 2 1 ∑∑ ∑∑ ∑ ( )( ) ( ) = − − − ∀ ∈ == == = (8) where yi and yj are observations of a given variable in locations i and j, and wij is the generic element of a (n x n) row-standardized spatial weights matrix (W) defined as fol- lows: w w w ij ij ijj n * * 1 ∑ = = (9) The generic element wij * in (9) can take two alternative values: w 1ij * = if i ≠ j and j ∈ N(i) w 0ij * = if i = j or i ≠ j and j ∉ N(i) where N(i) is the set of neighbours of the i-th region. N(i), thus W, can be identified in several alternative ways. Literature has empha- sized the fact there is no univocal preferable specification of W (Anselin, 1988). Despite alternative suitable weight matrices (e.g. those based on the nearest neighbours), here W is a first-order queen contiguity matrix. Thus, two regions are considered as neighbours only if they share a common boundary or vertex (Anselin, 1988). On average, each obser- vation shows 4.45 neighbouring regions11. 10 Pagliacci (2014) suggested this idea in a preliminary analysis on QoL patterns across urban and rural Italian NUTS 3 regions. Nonetheless, that work simply considered the rough indicator returned by “Il Sole 24 Ore”. 11 Most of Italian NUTS3 regions show either 4 or 5 neighbours. Nevertheless, the least connected NUTS 3 region has just 1 neighbour, whereas the most connected one has 9 neighbours. 201Quality of life and territorial imbalances This row-standardized spatial weights matrix (W) allows computing global Moran’s I statistic (thus their degree of spatial dependency) on both the QoL MPI and other sub- indicators of QoL. Global approaches do not allow the detection of specific regional struc- tures of spatial autocorrelation (i.e., either spatial cluster or spatial outliers): to do that, local approaches are also considered. A Local Indicator of Spatial Association – LISA (Anselin 1995; Anselin et al., 1996) is similar to the global Moran’s I statistic, but it is region-specific. It tests the hypothesis of random distribution by comparing values in spe- cific locations and values in their neighbourhood (as defined by W). Local Moran’s sta- tistics returns the distribution of local spatial clusters, which are groups of neighbouring locations showing significant LISA values. At a given significance level, such as 1%, it is possible to detect five alternative cases (Anselin, 1995): i) Hot spots (locations with high values and similar neighbours); ii) Cold spots (locations with low values and similar neigh- bours); iii) Spatial outliers (locations with high values but with low-value neighbours); iv) Spatial outliers (locations with low values but with high-value neighbours); v) Locations with no significant local autocorrelation. Table 6 returns the values for both the global and the local Moran’s I statistics, com- puted for both sub-indicators of QoL and the QoL MPI itself. A positive spatial autocor- relation occurs for all indicators but neighbourhood safety. The question thus becomes whether this general tendency to clustering yields to some given spatial clusters or not. Table 6. Global Moran’s I statistics (p-value in parenthesis) and Local Moran’s I statistics (number of NUTS 3 regions within each typology).   Global Moran’s I Local Moran’s I (LISA) Moran’s I Hot spots (i) Cold spots (ii) Spatial outliers (iii & iv) No local autocorrelation (v) Wealth 0.678* 14 11 0 85 (0.000) Services 0.763* 16 13 0 81 (0.000) Labour Market 0.809* 6 23 0 81 (0.000) Neighbourhood Safety 0.057 0 3 0 107 (0.152) Population 0.289* 7 6 0 97 (0.000) Leisure 0.215* 5 0 0 105 (0.000) Environment 0.630* 7 14 0 89 (0.000) MPI 0.802* 10 20 0 80 (0.000) * Statistically significant at the 5% level. Source: authors’ elaboration. 202 P. Bertolini, F. Pagliacci The analysis on the LISA values returns straightforward results (Table 6). In no cases, spa- tial outliers occur (confirming the sharp tendency to a positive spatial autocorrelation of observed values). In particular, neighbourhood safety and leisure are characterised by a fewer numbers of both hot and cold spots, whereas economic wealth, service provision and environment are much more clustered in space. Referring to the QoL MPI, 10 NUTS 3 regions are classified as hot spots, thus they benefit from large QoL levels even across their neighbourhood. Conversely, in 20 cases, low QoL levels are reinforced by bad per- formances even across neighbouring NUTS 3 regions. For the sake of simplicity, Figure 3 maps the spatial clusters occurring when consider- ing the comprehensive QoL MPI. Hot spots are mostly located across North-Eastern and Central Italy. Conversely, cold spots cover most of Southern regions, from Campania and Apulia to Calabria and Sicily. In particular, the presence of neighbouring NUTS 3 regions sharing similar QoL MPI low values may reinforce their lags compared to Northern Italy. 5.2 Neighbouring inner areas and neighbouring urban poles: an opposite effect In analysing spatial effects among neighbouring NUTS 3 regions, also the presence of either neighbouring inner areas or neighbouring larger urban poles may play an addi- Figure 3. Hot and cold spots – QoL MPI. Source: authors’ elaboration. 203Quality of life and territorial imbalances tional role in explaining differences in QoL levels across the country. To assess it, we refer to the same spatial weights matrix (W) shown in section 5.1, in order to return the spatial lags of the aforementioned indicators of inner areas (Ii, IPi, IAi): wI w I i j N,i ij ij n i n 11 ∑∑= ∀ ∈ == (10) wIP w IP i j N,i ij ij n i n 11 ∑∑= ∀ ∈ == (11) wIA w IA i j N,i ij ij n i n 11 ∑∑= ∀ ∈ == (12) where wij is always defined as in (9). Table 7 returns Pearson’s correlation coefficients between QoL indicators and wIi, wIPi, wIAi, at NUTS 3 level. Overall national data may hide same North-South divides already pointed out, while data disentangled by group of regions provide more insightful findings. In the North-West, no indicators of QoL are correlated with the spatially-lagged share of inner areas. In the North-East, both the population sub-indicator and the envi- ronment-energy one are positively linked to the presence of inner areas in neighbouring NUTS 3 regions. Also, the QoL MPI as a whole shows a positive correlation with inner areas across the neighbourhood. On the contrary, across Central regions, most relation- ships are negative. As observed in advance, even the share of inner areas across the neigh- bourhood shows a negative correlation with economic wealth, service provision, envi- ronment and energy. Thus, in this group of regions, the presence of neighbouring inner areas plays a detrimental effect on QoL. Therefore, this divide seems increasing QoL dif- ferentials as well. In Southern regions, the presence of urban poles in the neighbourhood seems to have a positive effect just on the provision of services. Same relationship is per- fectly reversed in the Islands, where the share of inner areas in the neighbourhood plays a positive effect also on labour market performances, environment and energy and the QoL MPI as a whole. Moreover, with the only exception of NUTS 3 regions in the Centre, the share of inner areas in the neighbourhood is positively related to QoL. Thus, if inner areas do not show high levels QoL, their presence in the neighbouring space surely plays a more posi- tive role, suggesting the existence of positive spatial spillovers. 6. Conclusions Through the improvement of both social and economic conditions of people living in inner areas, the Italian National Strategy for Inner Areas ambitiously aims to reverse nega- tive demographic trends, which still affect most of them. To this respect, improving QoL represents a key issue (Barca et al., 2014) for both inner and rural areas. Indeed, the paper has singled out that in Italy they largely overlap. Nevertheless, this analysis has partially broken up the negative relationship between presence of inner/rural areas and local QoL levels. Such a result is suggested by the analysis of both the QoL Mazziotta-Pareto Index, a 204 P. Bertolini, F. Pagliacci Table 7. Pearson correlation coefficients between spatially-lagged indicators of inner areas and indica- tors of QoL, by macro-region (p-values in parenthesis) Wealth & Competi- tiveness Services Labour Market Neigh- bourhood Safety Population Leisure Environ- ment & Energy QoL MPI Italy wIi -0.629* -0.569* -0.583* 0.136 -0.155 -0.222* -0.524* -0.575* (0.000) (0.000) (0.000) (0.157) (0.105) (0.020) (0.000) (0.000) wIPi -0.615* -0.546* -0.637* 0.155 -0.150 -0.234* -0.552* -0.583* (0.000) (0.000) (0.000) (0.106) (0.117) (0.014) (0.000) (0.000) wIAi -0.604* -0.537* -0.562* 0.167 -0.216* -0.195* -0.494* -0.546* (0.000) (0.000) (0.000) (0.079) (0.024) (0.041) (0.000) (0.000) North-West wIi -0.403* -0.122 -0.356 0.221 -0.379 0.024 -0.800 -0.246 (0.046) (0.560) (0.091) (0.288) (0.062) (0.910) (0.704) (0.235) wIPi -0.070 0.252 -0.274 0.192 0.121 -0.101 0.027 0.140 (0.739) (0.224) (0.184) (0.357) (0.565) (0.629) (0.898) (0.503) wIAi -0.367 -0.057 -0.262 0.296 -0.378 -0.008 -0.014 -0.157 (0.071) (0.787) (0.207) (0.150) (0.063) (0.968) (0.946) (0.453) North-East wIi 0.192 -0.403 0.191 0.028 0.599* 0.358 0.570* 0.496* (0.391) (0.063) (0.395) (0.902) (0.003) (0.102) (0.006) (0.019) wIPi 0.161 -0.510* 0.190 0.353 0.385 0.387 0.514* 0.534* (0.475) (0.015) (0.397) (0.107) (0.077) (0.075) (0.014) (0.010) wIAi 0.176 -0.406 0.270 0.109 0.479* 0.510* 0.526* 0.580* (0.433) (0.061) (0.224) (0.630) (0.024) (0.015) (0.012) (0.005) Centre wIi -0.448* -0.581* -0.411 0.293 -0.367 -0.138 -0.524* -0.526* (0.036) (0.005) (0.057) (0.186) (0.092) (0.539) (0.012) (0.012) wIPi -0.436* -0.491* -0.373 0.155 -0.347 -0.024 -0.523* -0.480* (0.043) (0.020) (0.087) (0.491) (0.114) (0.917) (0.012) (0.024) wIAi -0.439* -0.553* -0.460* 0.249 -0.446* -0.127 -0.605* -0.561* (0.041) (0.008) (0.031) (0.265) (0.037) (0.574) (0.003) (0.007) South wIi 0.059 -0.482* 0.268 0.210 0.132 0.028 -0.097 -0.008 (0.786) (0.017) (0.205) (0.325) (0.538) (0.897) (0.651) (0.972) wIPi -0.173 -0.533* -0.001 0.243 0.213 -0.174 -0.152 -0.177 (0.418) (0.007) (0.997) (0.253) (0.317) (0.415) (0.477) (0.408) wIAi 0.075 -0.436* 0.252 0.120 0.002 0.030 -0.032 -0.041 (0.726) (0.033) (0.234) (0.578) (0.993) (0.889) (0.883) (0.850) The Islands wIi 0.068 0.500* 0.694* 0.331 0.125 0.448 0.707* 0.703* (0.795) (0.041) (0.002) (0.195) (0.632) (0.071) (0.002) (0.002) wIPi 0.394 0.570* 0.587* 0.274 0.194 0.451 0.465 0.688* (0.118) (0.017) (0.013) (0.287) (0.455) (0.069) (0.060) (0.002) wIAi 0.126 0.500* 0.644* 0.633* -0.017 0.311 0.654* 0.770* (0.631) (0.041) (0.005) (0.006) (0.948) (0.224) (0.004) (0.000) * Statistically significant at the 5% level Source: authors’ elaboration 205Quality of life and territorial imbalances composite and comprehensive indicator of QoL, computed at NUTS 3 level, and different sub-indicators of QoL (e.g. neighbourhood safety, labour market, leisure). Results suggest that, when controlling for sub-national structural divides, the expected negative relation- ships between inner/rural areas and QoL is softened. For instance, when just focusing on North-Eastern regions, a larger share of inner areas at NUTS 3 level is associated to high- er level of QoL. Furthermore, even neighbourhood safety (a key driver of QoL) is gener- ally larger in more inner/rural NUTS 3 regions than in urban ones. Accordingly, it is hard to find conclusive results about the relationship between inner areas and QoL because of at least two major issues: the way inner areas are measured and the existence of spatial aspects, which make the picture even more complex. Referring to the former issue, the computation of three indicators that aim to assess the importance of inner areas according to three different perspectives (i.e. number of munici- palities, total population, and land area) represents an important advancement in this field of study. Indeed, each of them might be suitable for analysing specific dimensions of QoL, For instance, IAi seems to be particularly suitable for considering “environment and energy” aspects, whereas IPi can be linked to the provision of services to population. Accordingly, also policy implications are expected to differ, as opposite political domains might be inter- ested in assessing the importance of inner areas at NUTS 3 level in different ways. The latter issue refers to spatial aspects. People may spend significant parts of their lives out of their own NUTS 3 region. Therefore, even the neighbouring space is expected to matter in QoL. Here, main results support this idea. Both QoL sub-indicators and QoL MPI show a positive spatial autocorrelation and it is possible to detect groups of regions whose neighbours share similar QoL levels. It follows that even the local development may be influenced by neighbouring regions’ development, as both positive and negative spatial spillovers can affect place-based policies and their effectiveness. The same holds true when considering the presence of inner areas among neighbouring regions: for instance, this work proves that being located close to a NUTS 3 region with a higher share of inner areas could have positive effects on QoL, especially in the North-East and in the South. Thus, inner areas’ diversity clearly emerges. Indeed, some of them show more socio- economic potential, even with respect to QoL drivers. Such a finding has important policy implications, even with respect to the National Strategy for Inner Areas. The top-down approach, carried out by the Italian central government, is crucial when setting policy targets. Nevertheless, it is even more important to maintain the decision-making process partially decentralised, in order to identify the most appropriate policy tools to target the neediest areas to be targeted. Besides these considerations, this paper points out the effectiveness of the innova- tive approach chosen by the National Strategy for Inner Areas, which highlights territorial imbalances in terms of people’s needs rather than territorial features. Indeed, just the pro- vision of essential services to the population is seen as the main engine for local develop- ment, now and in the future. Such an approach would allow both scholars and policymak- ers to go beyond traditional urban-rural divides, which in fact are mostly considered by EU policies (such as the Rural Development Policy). Although providing partially over- lapping results, a focus on inner areas seems stressing inter-sectoral policies as the best answer to overcome territorial divides and to cope with population changes, both in Italy and in the EU. 206 P. Bertolini, F. Pagliacci Acknowledgements Authorship may be attributed as follows: Sections 1, 2 and 6 to Paola Bertolini; Sec- tion 3, 4 and 5 to Francesco Pagliacci. References Aiello, P. and Attanasio, M. (2004). How to Transform a Batch of Single Indicators to Make Up a Unique One? Atti della XLII riunione scientifica della Società Italiana di Statistica (Sessioni plenarie e specializzate): 327-338. Anselin, L. (1988). Spatial Econometrics: Methods and Models. Dordrecht: Kluwer Aca- demic Publishers. Anselin, L. (1995). Local Indicators of Spatial Association – LISA. Geographical Analysis 27: 93-115. Anselin, L., Bera, A.K., Florax, R.J.G.M. and Yoon, M.J. (1996). Simple Diagnostic Tests for Spatial Dependence. Regional Science and Urban Economics 26: 77-104. Bagnasco, A. (1977). Tre Italie: La Problematica Territoriale dello Sviluppo Economico Italiano. Bologna: Il Mulino. Bagnasco, A. (1988). La Costruzione Sociale del Mercato. Bologna: Il Mulino. Barca, F., McCann, P. and Rodríguez-Pose, A. (2012). The Case for Regional Development Intervention: Place-Based versus Place-Neutral Approaches. Journal of Regional Sci- ence 52(1): 134-152. Barca, F., Casavola, P. and Lucatelli, S. (eds.) (2014). A strategy for inner areas in Italy: definition, objectives, tools and governance. Materiali Uval Series 31.www.dps.teso- ro.it/Aree_interne/doc/Nota%20 Territorializzazione%20AI_03%20marzo_2013.pdf Bertolini, P., Montanari, M. and Peragine, V. (2008). Poverty and Social Exclusion in Rural Areas. Bruxelles: European Commission. Bock B. (2016), Rural Marginalisation and the Role of Social Innovation: A Turn Towards Nexogenous Development and Rural Reconnection. Sociologica Ruralis doi: 10.1111/ soru.12119 Bripi, F., Carmignano, A. and Giordano, R. (2011). La qualità dei Servizi Pubblici in Italia. Occasional Paper 84, Roma: Banca d’Italia. Cagliero, R., Cristiano, S., Pierangeli, F. and Tarangioli, S. (2011). Evaluating the Improve- ment of Quality of Life in Rural Areas. Paper presented at 122nd EAAE Seminar, Ancona (Italy), 17-18 February. Camaioni, B., Esposti, R., Lobianco, A., Pagliacci, F. and Sotte, F. (2013). How Rural the EU RDP is? An Analysis through Spatial Funds Allocation. Bio-based and Applied Economics 2(3): 277–300. Christaller, W. (1933). Die Zentralen Orten in Süddeutsch-Land. Jena: Gustav Fischer. Cliff, A. and Ord, J.K. (1981). Spatial processes: Models and applications. London: Pion. Copus A. (2001),.From Core-Periphery to Polycentric Development; Concepts of Spatial and Aspatial Peripherality European Planning Studies 9(4): 539-552. Copus, A., Melo, P.C., Kaup, S., Tagai, G. and Artelaris, P. (2015). Regional Poverty Map- ping in Europe – Challenges, Advances, Benefits and Limitations. Local Economy. DOI: 10.1177/0269094215601958. 207Quality of life and territorial imbalances Costanza, R., Fisher, B., Ali, S., Beer, C., Bond, L., Boumans, R., Danigelis, N.L., Dickin- son, J., Elliott, C., Farley, J., Elliott Gayer, D., MacDonald Glenn, L., Hudspeth, T.R., Mahoney, D.F., McCahill, L., McIntosh, B., Reed, B., Turab Rizvi, A., Rizzo, D.M., Simpatico, T. and Snapp, R. (2008), An Integrative Approach to Quality of Life Measurement, Research, and Policy. S.A.P.I.EN.S. 1(1). Online since 19.12.08. http:// sapiens.revues.org/169. Dematteis, G.(1986). Urbanization and Counter - Urbanization in Italy. Ekistics 53(316/317): 26-33. Dijkstra, L. and Poelman, H. (2008). Remote Rural Regions. How Proximity to a City Influences the Performance of Rural Regions. EU - Regional Focus n.1. Eurofound (2014). Quality of Life in Urban and Rural Europe. Luxembourg: Publications Office of the European Union. Eurostat (2010). A Revised Urban-Rural Typology. In Eurostat, Eurostat regional yearbook 2010. Luxembourg: Publications Office of the European Union: 240-253. Florida, R., Mellander, C. and Rentfrow, P.J. (2013). The Happiness of Cities. Regional Studies 47(4): 613-627. Giordano, R. and Tomasino, P. (2011). Public Sector Efficiency and Political Culture. Working papers Bankitalia, 786. Roma: Banca d’Italia. Hoggart, K, Buller, H. and Black, R. (1995). Rural Europe; Identity and Change. London: Edward Arnold. Kühn M. (2015). Peripheralization: Theoretical Concepts Explaining Socio-Spatial Ine- qualities. European Planning Studies 23:2, 367-378. Land, K.C. (1996). Social Indicators and the Quality of Life: Where Do we Stand in the Mid-1990s? SINET 45: 5-8. Lucatelli, S., Carlucci, C. and Guerrizio M.A. (2013). A Strategy for the ‘Inner Areas’ of Italy. Proceedings of the 2nd EURUFU (Education, Local Economy and Job Oppor- tunities in Rural Areas) Scientific Conference. Asti (Italy), 8 October. Malkina-Pykh, I.G. and Pykh, Y.A. (2008). Quality-of-Life Indicators at Different Scales: Theoretical Background. Ecological Indicators 8: 854-862. Matthews, A. (2013). Greening Agricultural Payments in the EU’s Common Agricultural Policy. Bio-based and Applied Economics 2(1): 1-27. Mazziotta, M. and Pareto, A. (2010a). La Sintesi degli Indicatori di Qualità della Vita: Un Approccio non Compensativo. Paper presented at “Congresso Qualità della vita. Rif- lessioni, studi e ricerche in Italia”, Firenze, Italia. Mazziotta, M. and Pareto, A. (2010b). Measuring Quality of Life: An Approach Based on the Non-Substitutability Of Indicators. Statistica e Applicazioni 8(2): 169-180. Mazziotta, M. and Pareto, A. (2016). On a Generalized Non-compensatory Composite Index for Measuring Socio-economic Phenomena. Social Indicators Research 127: 983–1003. Mazziotta, M. and Pareto, A. (2016). Methods for Constructing Non-compensatory Com- posite Indices: A Comparative Study. Forum for Social Economics 45(2-3): 213-229. Millennium Ecosystem Assessment (MEA) (2005). Ecosystems and Human Wellbeing. Washington, DC: Island Press. Moran, P.A.P. (1950). Notes on continuous Stochastic Phenomena. Biometrika 37: 17–23. Noguera, J. and Coups, A. (2016). Inner Peripheries: What Are They? What Policies Do They Need? Agriregionieuropa 45: 10-14. 208 P. Bertolini, F. Pagliacci OECD (2008). Handbook on Constructing Composite Indicators. Methodology and User Guide. Paris: OECD Publications. OECD (2009). OECD Rural Policy Reviews: Italy. Paris: OECD Publications. Pagliacci, F. (2014). L’Erba del Vicino è davvero più Verde? Un’Analisi sulla Qualità della Vita nelle Province Urbane e Rurali. AgriRegioniEuropa 39: 54-57. Pagliacci, F. (2017). Measuring EU Urban-Rural Continuum Through Fuzzy Logic. Tijd- schrift voor economische en sociale geografie 108(2): 157-174. Paniagua, A. (2012). The Rural as a Site of Recreation: Evidence and Contradictions in Spain from a Geographical Perspective. Journal of Tourism and Cultural Change 10(3): 264-275. Petrosillo, I., Costanza, R., Aretano, R., Zaccarelli, N. and Zurlini, G. (2013). The Use of Subjective Indicators to Assess How Natural and social Capital Support Residents’ Quality of Life in a small Volcanic Island. Ecological Indicators 24: 609-620. Shin, D.C. and Johnson, D.M. (1978). Avowed Happiness as an Overall Assessment of the Quality of Life. Social Indicators Res. 5: 475-492. Shucksmith, M., Cameron, S., Pichler, F. and Merridew, T. (2009). Urban-Rural Differenc- es in Quality of Life across the EU. Regional Studies 43(10): 1275-1289. Sørensen, J.F. (2014). Rural–Urban Differences in Life Satisfaction: Evidence from the European Union. Regional Studies 48(9): 1451-1466. Tobler, W. R. (1970). A Computer Movie Simulating Urban Growth in the Detroit Region. Economic Geography 46: 234-240. Welch, B.L. (1951). On the Comparison of Several Mean Values: An Alternative Approach. Biometrika 38: 330-336.