Generating well-being and efficiency: Evidence from Italy

Economic Analysis and Policy 65 (2020) 262–275

Contents lists available at ScienceDirect

Economic Analysis and Policy journal homepage: www.elsevier.com/locate/eap

Full length article

Generating well-being and efficiency: Evidence from Italy Graziella Bonanno a , Giovanni D’Orio b , Rosetta Lombardo b ,

∗

a

Department of Economics, Vanvitelli University of Campania, Corso Gran Priorato di Malta, 81043 Capua (CE), Italy Department of Economics, Statistics and Finance ‘‘Giovanni Anania’’, University of Calabria, Ponte Pietro Bucci - Cubo 0C - 87036 Arcavacata di Rende (CS), Italy b

article

info

Article history: Received 6 September 2019 Received in revised form 6 February 2020 Accepted 7 February 2020 Available online 12 February 2020 JEL classification: C33 I31 P48 R11 Keywords: Multidimensional well-being Composite indicators Efficiency Stochastic frontier

a b s t r a c t The paper aims to shed light on the geography of well-being in Italian regions and explain the distance of a given region from an efficiency frontier. We build a composite regional well-being index over the period 2010–2015. Then using the index as output, we estimate a well-being generating function to rank Italian regions in terms of efficiency in attaining well-being. The rankings confirm the divide between Northern and Southern regions as regards overall well-being and efficiency. Our findings indicate that regions more dependent on external financing achieve lower efficiency scores. Current failures should not be used to reinforce selfish localism; they should rather stimulate the search for more effective policies to reduce disparities. © 2020 Economic Society of Australia, Queensland. Published by Elsevier B.V. All rights reserved.

1. Introduction The need to understand the mechanisms through which people’s lives can improve along with that of society as a whole is one of the key concerns underlying the debate on well-being (Veneri, 2019). Over the last decade, the Stiglitz–Sen–Fitoussi report (2009) has given new impulse to the research that considers well-being (WB hereafter) as a multidimensional phenomenon concerning several dimensions of people’s lives beyond the usual income-related aspects. The measures of multidimensional WB proposed in the recent literature can be grouped into two different approaches: the dashboard of indicators and the composite index. Both present strengths and weaknesses (Ciommi et al., 2013). A dashboard of indicators provides a detailed picture of WB and avoids any loss of information; however, because of the high number of indicators considered, it does not allow for a simple comparison across countries or regions in a country and over time.1 A composite index is useful for measuring the performance of a country (or region) over time even if synthesizing all the information in a single number might leave important aspects hidden2 (Bleys, 2012). Composite indicators, however, are widely used. Macroeconomic or aggregate measures of economic and non-economic dimensions of quality of life are, indeed, usually weighted and aggregated following different statistical methodologies to form synthetic indices of WB domains. These, in turn, are combined in order to obtain an overall synthetic indicator. ∗ Corresponding author. E-mail address: [email protected] (R. Lombardo). 1 One of the most important attempts is the Better Life Indicator proposed by OECD. 2 One of the best known is the Human Development Index. https://doi.org/10.1016/j.eap.2020.02.006 0313-5926/© 2020 Economic Society of Australia, Queensland. Published by Elsevier B.V. All rights reserved.

G. Bonanno, G. D’Orio and R. Lombardo / Economic Analysis and Policy 65 (2020) 262–275

263

Until recently, multidimensional WB has been mainly studied at country level and a number of countries proposed country specific WB measures. However, many of the features that influence WB are likely to be locality-specific and hence spatially variable. Within the same country, people have different access to collective provisions depending on where they live. People living in the same region share a common cultural, political and socio-economic environment, which contributes, alongside individual characteristics, to life satisfaction. The WB of individuals living in the same country might differ from one region to another (Aslam and Corrado, 2011). The measurement of efficiency in the generation of WB, furthermore, is a neglected aspect in the literature, but it is prominent when looking for policy implications of a WB geography. The aim of this paper is to try to fill these gaps in the literature. To this end, it compares the Italian regions as regards WB and measures their efficiency in reaching WB. Italy is characterized by significant socio-economic differences between Northern and Southern regions. To investigate whether there is also a dualism for WB is worthwhile, but it would be even more instructive to find out whether regions at the top of a WB ranking are the most efficient in generating it. Some regions, in fact, might appear at the top of a WB ranking, without being efficient in terms of WB generation. In the literature, a number of papers propose composite indices, a few of which are based on Italian data and present a ranking of Italian regions or provinces, yet to the best of our knowledge, no one has treated the problem of efficiency in WB generation through parametric methods. With these considerations in mind, we adopt the composite index approach and proceed in two steps. In the first we compute, through the factor analysis (FA hereafter), a composite index of WB starting from the insights that emerge from a project carried out by the Italian National Institute of Statistics (ISTAT) in conjunction with the National Council for Economy and Labour.3 In the second step, the evaluated regions (Decision-Making Units) are treated as producers of output (WB) given some inputs. We compare the ‘‘competitiveness’’ of regions through the estimated levels of technical efficiency in generating WB as the ability of the regions to transform their ‘‘capitals4 ’’ in WB for their further development. A WB generating function is estimated by using the stochastic frontier approach (SFA hereafter) in the specification proposed by Battese and Coelli (1995). The computed WB composite index is used as the dependent variable of the WB generating function. The paper is structured as follows: Section 2 reviews the literature; Section 3 illustrates data and methodologies; Section 4 discusses the results and Section 5 concludes. 2. Related literature In the literature on WB two strands of research can be distinguished: studies that look at subjective WB and those that focus on aggregate measures of objective WB. The former relies upon individuals’ self-reported perceptions of happiness (Blanchflower and Oswald, 2011). The second considers WB as a multidimensional phenomenon concerning several dimensions of quality of life. The paper deals with to the second strand of research, in which several methods of measurement have been proposed. A number of papers are based on composite indicators calculated as weighted averages of variables and sub-indices (OECD, 2011, among others). Other works are based on mixed statistical strategies with the principal component analysis (PCA hereafter) used to assess the internal coherence of the various domains and the weighted average of the partial indices to calculate the respective composite indicators (Annoni et al., 2012). The Human Development Index, first proposed by the United Nations in 1990, is the most common alternative to GDP for measuring WB at country level. It is calculated as a mean of three indicators (per capita income, health, and education). In 2010, the indicators used to measure the three dimensions of HDI changed together with the way to summarize the indicators in an overall index (from an arithmetic to a geometric weighted mean). As the index neglects a number of important dimensions of quality of life, it is often ‘‘augmented’’ to include these. Among the composite indicators, the Index of Economic Well-Being (IEWB), developed by the Centre for the Study of Living Standards (Osberg, 1985; Osberg and Sharpe, 2005), measures WB in terms of command over resources, and covers four dimensions: consumption flows, stocks of wealth, economic equality and economic security.5 Osberg and Sharpe (2005) compare trends in the index and its components to trends in GDP per capita and in HDI for seven countries from 1980 to 2001 and find that IEWB advanced at a significantly lower rate than GDP. The European Well-Being Index (EWI) uses the FA and adopts the social indicators approach (Ivaldi et al., 2016). In their quantitative exercise, the authors rank the European Union 27-Countries according to their EWI score. They are aware that an index of this type offers a description of the national reality as a whole, not focusing on the important regional differences that distinguish each country. 3 This project produced a database covering 12 dimensions of ‘‘Equitable and Sustainable Well-Being’’ (whose Italian acronym, BES, is used hereafter). 4 We mean all the sets of inputs that will be used in the paper such as economic, social, relational, environmental, human capital and so on. 5 Each dimension of economic WB is itself an aggregate of many underlying trends. Consumption flows include current market consumption, government spending, and unpaid work. The second dimension captures the accumulation of productive assets: per-capita capital stock, research and development, natural resources, and human capital. The economic equality dimension deals with both income distribution and poverty. Economic security refers to the degree of riskiness of future economic conditions. The overall index is obtained by adding the four equally weighted dimensions (See Osberg and Sharpe, 2005 p.329).

264

G. Bonanno, G. D’Orio and R. Lombardo / Economic Analysis and Policy 65 (2020) 262–275

The Mazziotta–Pareto Index (MPI)6 is based on the assumption that the individual components cannot be substituted.7 It rules out the unit of measurement and the variability effect. To normalize the values around the mean it uses a nonlinear function and penalizes observations that are relatively far from the mean. The Adjusted MPI (AMPI),8 based on a re-scaling of the individual indicators by a Min–Max transformation, is the method through which ISTAT synthetizes the variables of the BES. Among composite indices based on Italian data, there is the Italian IEWB proposed by Berloffa and Modena (2012), a revised version of the IEWB that includes the share of temporary workers in the economic security dimension, and the age-wage gap in the equality dimension.9 The authors used the index to analyse the trends in WB in Italy and in the Italian region Lombardy over the 1995–2007 period. They found that WB advanced at a lower rate than GDP and employment insecurity and intergenerational inequality affected these trends. Among the measures of multidimensional WB at a regional level, the Regional WB Index (Ferrara and Nisticò, 2015) synthetizes ten dimensions of quality of life through a PCA in a two-step approach starting from fifty-seven BES elementary indices. The index is used to compare the dynamics of regional WB in Italy with those of per capita GDP. Ferrara et al. (2019) explore the correlation between the Regional WB index and its sub-indexes and happiness and find a positive and statistically significant association between happiness and multidimensional objective WB in general, and with some specific sub-dimensions of objective WB. The composite indices, however, reflect average population performance, without revealing anything about inequality. A number of these add up population-level average indicators, which fail to differentiate between groups with cumulative disadvantages concentrated in multiple dimensions of WB and groups in which disadvantages are spread over individuals sporadically (Yang, 2014). The common assumption of full substitutability of the different dimensions is often inappropriate. Furthermore, according to Mazziotta and Pareto (2019), PCA and FA — being methods that ignore the polarities, i.e. the sign of the relation between the indicator and the concept to be measured — are not suitable for constructing composite indices of WB. To construct a composite indicator, they argue, a measurement model is needed. The correct approach for measuring WB is a formative one.10 The paper is related to a second group of articles dealing with efficiency measures for the assessment of DecisionMaking Units (DMUs). The concept of efficiency is open to different interpretations (Aigner et al., 1977; Coelli et al., 2005), while there is consensus in considering efficiency as the degree of proximity of an actual production process to an optimum standard. Efficiency can be thought of as the ability of a decision unit to minimize the amount of input for the production of a certain output or to maximize the amount of output given a certain amount of input, for any level of technology. In order to estimate the WB ranking, we adopt the Stochastic Frontier Approach (SFA) that allows a region to be distant from the frontier also for randomness (Aigner et al., 1977; Meeusen and van de Broek, 1977). In this, the SFA differs from the Data Envelopment Analysis (DEA), which assumes that the distance from the frontier is entirely due to inefficiency.11 A further advantage of SFA derives from the specification of Battese and Coelli (1995), which enables a cleaner efficiency measure to be obtained by comparing it with the model where one first estimates inefficiency using a frontier and, second, uses the estimated efficiency-score as the dependent variable in subsequent regression (Greene, 1993). Indeed, it is a simultaneous estimation of the frontier and the inefficiency equation, in contrast with the standard

6 From the name of authors who proposed it in 2013. 7 That is, it does not allow for compensation among them. 8 Introduced in Mazziotta and Pareto (2013) and re-adapted in Mazziotta and Pareto (2016). 9 Economic security includes unemployment rate and temporary employment. The equality dimension includes three components: poverty intensity; income inequality and intergenerational inequality. The index is a weighted average of four dimensions, with weights subjectively determined according to the relative importance attributed to each dimension (See Berloffa and Modena, 2012, p.761). 10 Measurement models can be formative or reflective. In a formative approach, individual indicators are causes of an underlying latent variable. Therefore, causality is from the indicators to the concept and a change in the phenomenon does not necessarily imply variations in all its measures. Indicators are not interchangeable. In a reflective model, individual indicators denote effects of an underlying latent variable. Therefore, causality is from the concept to the indicators and any change in the phenomenon causes variation in all its measures. Individual indicators are interchangeable (Diamantopoulos et al., 2008). 11 Since efficiency is evaluated in relation to best practice, the key concerns in this field of research come from the methods. A common criterion of classification distinguishes between parametric and non-parametric approaches. Parametric methods assign density functions to the stochastic component of the model, while nonparametric methods define the deterministic part only. Whereas the SFA is the most used parametric method that assigns a distribution to the error term and allows inference, the Data Envelopment Analysis (DEA) is the most used non-parametric method. Inference, however, is not specific to SFA because of advances in bootstrapping in the DEA procedure (Simar and Wilson, 2007). The difference between SFA and DEA lies in the fact that the latter does not assign a distribution function to the error term. Another criterion is based on how the distance from the frontier should be understood. In this respect, we have stochastic or deterministic methods. The first group admits that a DMU may be far from the frontier due to randomness and/or inefficiency. In other words, a stochastic method, such as the SFA, enables the decomposition of the error into two parts, one attributable to inefficiency and the other to random error. When using a deterministic approach, the distance from the frontier is attributed entirely to inefficiency. The determinist approach ignores, in other words, the existence of purely random disturbance, which may be, for example, due to measurement errors or unforeseen events.

G. Bonanno, G. D’Orio and R. Lombardo / Economic Analysis and Policy 65 (2020) 262–275

265

two-step approach. In this respect, Lensink and Meesters (2014) and Wang and Schmidt (2002) argue that the twostep approach considers inefficiency independently and identically distributed in the main frontier equation, while it is determined by other variables in the inefficiency equation.12 3. Data and research methodology In this section, we illustrate the data and the ‘‘two-step’’ research methodology. Section 3.1 describes the Dataset. Section 3.2 defines the methodology on which to build the WB variable as a composite index, while Section 3.3 explains the WB generation model deriving from the use of the SFA. 3.1. Data Table 1 describes the main variables and the expected sign of the effect of the variable on WB. Variables without a star (49) are used in the FA to generate the WB indicator used as dependent variable of SFA. The independent variables (input) of SFA are 12 synthetic indicators each of them built by ISTAT using all the variable in Table 1 respecting the dimensions presented in Table 4 (e.g., the synthetic index Education has been built by ISTAT using the 5 elementary indicators Education 1 to 5) We shall, however, focus on the role of certain variables which, albeit seldom used in previous work, are quite significant. In our opinion the integration of these rather neglected variables could well represent a way forward for future research. Economic well-being The set of indicators on economic WB contains information on Average disposable income (per capita) of consumer households and on an Index of inequality of disposable income (see Table 1). While the effects of income inequality differ across various dimensions of WB, reducing economic inequality will generally help to improve the WB of society overall. We consider not just ‘‘economic’’ inequality but even indicators related to education, health and participation in the labour market to monitor the problem of exclusion, which was addressed in the definition of the so-called ‘‘Laeken indicators’’, established by the European Council in December 2001. These indicators help measure the progress made by Regions on a number of agreed objectives, such as the fight against poverty. Social exclusion is characterized as a fracture of the bond of solidarity and social participation between the society as a whole and the marginalized groups. The various programmes set up to address this grave issue will play a role in future strategies to enhance WB. Inequality concerns the distances that separate each individual from others in society. These distances are not necessarily simple gaps in aggregates of an economic nature. There are forms of non-economic inequality that may have significant implications in terms of WB. A disabled person, for example, can be excluded a priori from education or access to work; family and/or ethnic origin can mark the destinies of people by constraining their possibilities and freedom; gender often leads to differentiated and unequal life paths regardless of individual characteristics. These phenomena have not only unfair consequences for those directly involved, but also for society as a whole, because they do not allow everyone to fully deploy their skills, thereby impoverishing the quality of life in general. Inequality, therefore, is present not only in income but also in politics. Consideration of the factors that give rise to social stratification, therefore, is of the upmost importance. Social relations, subjective well-being and relational goods These indicators are also helpful. The theory of modern relational goods raises questions that are simple but of fundamental importance for the definition of specific targets in a WB indicator. The production of relational goods, the multiplication of socialization and support opportunities that may reduce the discomfort of minors, young people, the elderly and families are, in all respects, essential areas of WB. With these categories, we can consider different aspects of WB:

• WB as a quality of the overall life of a society; • WB as social capital reconnected to human resources; • WB as the point of arrival of development models. These three paths characterize everyday reality but they are scarcely (or insufficiently) represented in the various models of individual and collective WB measurement. Environment, landscape and crime Environment and landscape may influence quality of life. There are numerous examples in the literature of how people’s stress level and mental health can be affected by images or contact with nature. At the same time, individuals or groups can feel a deep sense of loss and dissatisfaction when their local environment is degraded by unregulated, 12 Although we are aware that many methodological advances have been proposed in some recent contributions (Greene, 2017; Kumbhakar et al., 2017; Wheat et al., 2019), we use the standard specification proposed by Battese and Coelli (1995), whose use is quite consolidated in the literature.

266

G. Bonanno, G. D’Orio and R. Lombardo / Economic Analysis and Policy 65 (2020) 262–275

Table 1 Variables, meaning and expected signs. Source: ISTAT Indicator Economic Economic Education Education Education Education Education

WB 1 WB 2 1 2 3 4 5

Environment 1a Environment 2 Environment 3a Environment 4 Environment 5 Environment 6a Environment 7 Environment 8 Health 1 Health 2 Health 3a Health 4a Health 5 Innovation 1 Innovation 2 Innovation 3 Labour quality 2 Labour quality 3 Labour quality 4 Labour quality 5a Labour quality 6 Labour-occupation Landscape 1a Landscape 2 Landscape 3 Landscape 4a Politics 1 Politics 2 Politics 3 Politics 4a Politics 5a Politics 6a Politics 7 Security 1 Security 2 Security 3 Security-murders Services quality 1a Services quality 2 Services quality 3 Services quality 4 Services quality 5 Services quality 6 Social relationship Social relationship Social relationship Social relationship Social relationship Social relationship

1

1 2 3 4 5 6

Social relationship 7a Social relationship 8a Subjective Well-being 1 WB & minimum conditions 1

Meaning

Exp sign

Average disposable (per-capita) income of consumer households Index of inequality of disposable income Children of 4–5 years attending kindergarten People aged 25–64 who completed at least second grade secondary school People aged 30–34 who have obtained a university degree People aged 18–24 who completed middle school and are not included in a training programme People aged 25–64 who participated in education and training activities in the 4 weeks prior to the interview Total drinking water losses Urban waste sent to waste disposal site Urban quality of air - PM10 dust Availability of urban green spaces People ≥ 14 very or fairly satisfied with local environment Extension of protected areas Electricity consumption generated by renewable sources Urban waste subject to recycling Life expectancy at birth Life expectancy in good health at birth Physical status index for people ≥14 years Psychological status index for people aged 14 years Life expectancy without limitations in activities at the age of 65 years Research intensity Employees with scientific-technological university degree Employees in creative businesses Employees on temporary contracts and employees who started their current job at least five years previously Rate of incidence of low paid employees Rate of incidence of non-regular employees Index of Satisfaction with job Share of involuntary part-time employees on total employees Employment rate of the population aged 20-64 years Current municipal public expenditure for cultural heritage Index of illegal buildings Share of agri-tourism People ≥14 considering their home area affected by degradation People ≥14 expressing confidence in the Italian Parliament People ≥14 expressing confidence in the judicial system People ≥14 expressing trust in the political parties People ≥14 expressing confidence in police and fire brigade Share of women elected to Regional Councils Average duration of trials in ordinary courts Index of overcrowding of prisons Burglary rate Pickpocketing rate Robbery rate Murder rate Beds in residential care Children 0-2 years who used child services Families that signalled difficulties to access at least 3 essential services Families that reported shortages in water supply Seats-km available in all types of public transport Percentage expressing satisfaction grade ≥8 with public transport People ≥14 expressing satisfaction with family relationships People ≥14 expressing satisfaction with their friendships People ≥14 having relatives, friends or neighbours to rely on People ≥14 participating at least one social activity in previous 12 months People ≥14 very or fairly satisfied with the their area environmental situation People ≥14 that talk or who are informed about politics at least once a week, who participated online in consultations or votes on social or political problems or have read and posted opinions on social or political problems on the web in previous 3 months People ≥14 that carried out voluntary activities for associations or groups in previous 12 months People ≥14 who financed associations in previous 12 months People ≥ 14 with a satisfaction score for life between 8 and 10 People living in families with severe material deprivation

+ − + + + − + − − + + + + + + + + + + + + +/− + − − − + − + + − +/− − + + + + + − − − − − − + + − − + + + + + + + +

+ + + −

(continued on next page)

G. Bonanno, G. D’Orio and R. Lombardo / Economic Analysis and Policy 65 (2020) 262–275

267

Table 1 (continued). Indicator

Meaning

Exp sign

WB & minimum conditions 2 WB & minimum conditions 3 WB & minimum conditions 4

People living in overcrowded housing without services and with structural problems

−

Subjective evaluation index of economic difficulty

−

People < 60 living in very low labour intensive families

−

a

Variables with star are Extra variables used by Istat, together with the other 49 unstarred, to build, by AMPI method, the 12 synthetic indicators used as input in SFA. Variables with star are not used in FA for data unavailability.

industrial pollution and others uninvited and unwelcome developments. Crime is also important since WB and health could be affected when someone is a victim of a crime and the same applies, albeit to a lesser extent, to those living in a high crime area. Politics This set of variables can be seen as a proxy of equal opportunities in politics. Two extra variables relating to the effectiveness of the judicial system and prison overcrowding are also considered. Innovation This set of variables illustrates the Research intensity of each Region, the share of Employees with scientifictechnological university degree and the share of Employees in creative businesses. Service quality The regional framework of WB should include accessibility to services which is, in fact, particularly important for assessing WB across regions (Veneri, 2019). The set of variables considered is linked to the use and perception of quality of public services such as Transportation, Childcare, Residential services for the elderly, and Water. 3.2. The synthetic well-being indicator Considering WB as a multidimensional phenomenon is a tricky business (Ivaldi et al., 2016). The construction of composite indicators is complex because of two aspects in particular: firstly, the selection of relevant domains of WB and weights given to each domain in the aggregation procedure and, secondly, the choice of an adequate method for the aggregation.13 We compute a composite WB index through the FA, a worthwhile instrument for selecting a set of variables that explain as much as possible of the phenomenon concerned.14 Analytically, if we have p variables X1 , . . . , Xp measured on a sample of n subjects, then variable xs can be written as a linear combination of m factors F1 , . . . , Fm where m < p: xs = ks1 F1 + · · · + ksm Fm + w

(1)

where ks are the factor loads for variable xs ; w is the part of variable xs not explained by the factors. FA condenses the information contained in a matrix of correlation or variance/covariance; it aims to identify statistically the latent dimensions of the observed phenomenon. The composite WB index is computed starting from 49 variables15 as reported in Table 1. 3.3. The well-being generation function We consider the 20 Italian regions and their ranking by estimating a stochastic frontier in the specification proposed by Battese and Coelli (1995). Function G(.) indicates the link between the WB and the dimensions X composing it: WBit = G(X )ev−u

(2)

13 To limit arbitrariness in choosing WB dimensions and weights, we consider the insights that emerge from the BES project and opt for equal weight for each dimension. Equal weighting is the preferred procedure when the theoretical scheme assigns to each indicator the same adequacy in defining the variable to be measured and when the empirical knowledge is not sufficient for defining specific weights (Decancq and Lugo, 2013). 14 The FA enables the study of correlations between large numbers of variables, by grouping them around factors, so that they are arranged on factors highly correlated with each other (Dillon and Goldstein, 1984) and allows us to explain the variance of the phenomenon under analysis without requiring the estimation of parameters. It can summarize a set of sub-indicators while preserving the maximum possible proportion of the total variation in the original set. 15 Table 1 contains 63 elementary indicators used by ISTAT to build, by AMPI method, the synthetic indicators representative of the 12 BES dimensions. We consider 49 elementary indicators to compute the WB index because the data do not cover the considered time interval for the remaining 14 (starred in Table 1). The 12 synthetic indicators representative of BES dimensions will be successively used in SFA.

268

G. Bonanno, G. D’Orio and R. Lombardo / Economic Analysis and Policy 65 (2020) 262–275

From Eq. (1), the rank efficiency (RE) of regions derives from the ratio between the WB observed and that of the best performing region (for which u = 0): RE it = [G(X )ev−u ]/[G(X )ev ] = e−u

(3)

The Cobb–Douglas function is chosen to model the frontier. This satisfies the assumptions of non-negativity, concavity and linear homogeneity (Kumbhakar and Lovell, 2000). The WB generating function in the log-linear form is: WBit = α0 + Σj αj ln(Xjit ) + vit − uit

(4)

where WB is the well-being indicator; Xj represents the jth input, with j = 1, . . . , 12; α is the parameter to be estimated; u is the inefficiency; v is the random error. We assume that vit is normally distributed with mean zero and uit is distributed as a truncated normal. Again, vit and uit are independently and identically distributed:

vit ∼ iid N(0, σv2 )

(5)

uit ∼ N + (z ′ η, σu2 )

(6)

where z η is the linear predictor of inefficiency. The econometric specification of the inefficiency component is: ′

uit = Σk ηk zkit + eit

(7)

where z-variables are the explicative regressors of the inefficiency component. Moreover, eit is the erratic component. Efficiency is time varying, ensuring a change in the relative ranking of regions. This accommodates the case where an initially inefficient region becomes more efficient over time. 4. Research results and discussion Since the empirical strategy consists of two steps, in Section 4.1 we present the FA and discuss the results; in Section 4.2 we present the estimation of the WB efficiency scores. 4.1. Application of factor analysis For the sake of brevity, we do not report the results of the FA employed to obtain the dependent variable of our WB generating function.16 Table 2 reports the pattern matrix of the nine retrieved factors. The factor loadings for this orthogonal solution represent how the variables are weighted for each factor and the correlation between the variables and the factor. The higher the load the more significant it becomes in defining the factor’s dimensionality. A negative value indicates an inverse impact on the factor. The column ‘Uniqueness’ in Table 2, reports the proportion of the common variance of the variable not associated with the generated factors. Note that the greater uniqueness the lower the significance of the variable in the factor model. If uniqueness is equal to 1 we have a ‘‘perfect communality’’ so that variable is not significant. We may also consider uniqueness as the variance of the specific factors for the variables. Rotation involves the ‘‘common factors’’, so the uniqueness is not affected by the rotation.17 For each variable, we obtain relatively small values of uniqueness, signalling that all the 49 variables are important in the FA. The results reported in Table 2 are quite significant. For brevity, we shall only discuss the most important aspects of the first factor accounting for 48.4% of WB variance. Factor 1 is strongly and positively influenced by Labour Occupation (0.979), Economic WB 1 (0.934) and five variables belonging to Social Relations (all of which score higher than 0.80). Italian regional WB in 2010–2015 is, i.e., strongly and positively affected by the employment rate of 20- to 64-year olds and the average (per capita) disposable income of consumer households. The most important relational goods cover relationships with family, friends and people we rely on. Life expectancy in good health (Health 2) is also quite important (0.795) as well as satisfaction with the local environment. 16 This information is available upon request. Although we have chosen not to report these results, it is worth noticing that some of the eigenvalues deriving from FA are negative because the matrix is not of full rank, that is, although there are 49 variables the dimensionality of the factor space is much less. To choose the relevant factors we use the Kaiser method, retaining factors with eigenvalue greater than 1. This leads us to consider the first nine factors with a cumulative variance explanation of WB of almost 90% (0.897). 17 The rotation is an important issue in FA stability since it causes the reduction of factor loadings already relatively small, in the first phase, and the increase of the absolute values of factor loadings that predominate in the first phase. It is noteworthy to transform the factors through a process of rotation of the axes. In fact, in an un-rotated solution every variable is explained by two or more common factors, while in a rotated solution each variable is summarized by a single common factor (Ivaldi et al., 2016).

G. Bonanno, G. D’Orio and R. Lombardo / Economic Analysis and Policy 65 (2020) 262–275

269

Table 2 Factor loadings (pattern matrix) and unique variances. Source: Own elaborations on ISTAT Data. Variable

Factor 1

Factor 2

Factor 3

Factor 4

Factor 5

Factor 6

Factor 7

Factor 8

Factor 9

Uniqueness

economic well-being 1 economic well-being 2 education 1 education 2 education 3 education 4 education 5 environment 2 environment 4 environment 5 environment 7 environment 8 health 1 health 2 health 5 innovation 1 innovation 2 innovation 3 labour quality 2 labour quality 3 labour quality 4 labour quality 6 labour-occupation landscape 2 landscape 3 politics 1 politics 2 politics 3 politics 7 security 1 security 2 security 3 security-murders services quality 2 services quality 3 services quality 4 services quality 5 services quality 6 social relationship 1 social relationship 2 social relationship 3 social relationship 4 social relationship 5 social relationship 6 subjective well-being well-being & minimum well-being & minimum well-being & minimum well-being & minimum

0.934 −0.764 0.219 0.689 0.549 −0.520 0.593 −0.504 −0.030 0.726 0.211 0.765 0.671 0.795 0.687 0.441 −0.252 0.585 −0.742 −0.932 −0.920 −0.589 0.979 −0.873 0.587 −0.318 −0.457 0.031 0.035 0.446 0.429 −0.348 −0.624 0.828 −0.878 −0.798 0.400 0.522 0.838 0.835 0.856 0.876 0.806 0.734 0.621 −0.834 −0.490 −0.826 −0.920

0.124 0.240 −0.672 0.439 0.517 −0.308 0.211 −0.247 −0.283 −0.440 −0.598 0.067 0.112 −0.018 0.090 0.596 0.778 0.452 −0.158 −0.063 0.102 0.466 0.024 0.020 −0.025 0.194 −0.297 −0.262 −0.043 0.573 0.749 0.308 −0.080 0.014 −0.022 −0.063 0.478 −0.664 −0.115 −0.170 −0.322 0.166 −0.270 −0.192 −0.570 0.011 0.194 0.043 −0.068

0.208 0.103 0.243 −0.176 −0.278 0.343 −0.425 −0.023 −0.288 −0.313 −0.257 −0.190 −0.418 0.099 −0.124 0.338 −0.090 0.293 −0.048 −0.001 −0.058 −0.441 0.069 −0.209 −0.047 0.425 0.604 0.588 0.707 −0.020 0.307 0.365 0.175 0.113 0.007 −0.066 0.358 −0.210 0.102 0.053 0.032 0.057 −0.098 0.113 0.244 −0.082 −0.232 0.075 −0.152

0.006 0.023 −0.137 0.398 0.421 −0.408 0.232 0.272 0.414 0.157 0.147 −0.349 0.092 0.184 0.037 −0.067 0.291 0.135 0.289 0.089 0.107 −0.076 0.003 0.191 0.330 0.589 0.225 0.632 −0.120 −0.223 −0.055 0.210 −0.006 0.033 0.128 0.007 −0.177 0.081 −0.081 −0.006 0.051 −0.223 −0.011 0.172 0.070 −0.039 −0.011 −0.168 −0.046

0.067 0.268 0.204 −0.059 −0.137 0.276 0.316 −0.392 −0.021 −0.201 0.133 0.292 −0.050 0.178 −0.084 0.245 0.193 −0.005 0.236 0.086 0.226 0.180 −0.081 0.032 0.416 −0.176 0.083 0.019 −0.347 −0.144 0.025 0.396 0.215 −0.148 0.073 0.085 0.152 0.084 0.185 0.188 0.187 0.009 0.393 0.329 0.081 0.078 0.236 0.162 0.124

0.102 0.083 0.128 0.046 0.157 0.191 0.072 0.405 −0.559 0.131 0.161 −0.222 −0.020 0.034 −0.006 −0.094 0.099 0.228 0.120 −0.059 0.095 0.207 0.080 0.028 0.111 −0.071 0.140 −0.042 −0.096 0.173 −0.027 −0.228 0.181 0.270 −0.180 0.371 −0.193 −0.050 0.159 0.143 −0.098 0.164 −0.159 0.104 0.086 −0.035 0.101 0.138 −0.009

−0.121 −0.120

0.012 −0.059 0.210 −0.064 0.011 −0.009 −0.029 −0.003 −0.273 −0.039 0.139 0.156 0.113 0.294 0.070 −0.040 −0.106 −0.083 −0.163 0.050 −0.007 −0.133 0.025 0.002 0.191 0.113 −0.092 0.096 0.102 0.315 −0.059 0.322 −0.154 −0.003 −0.021 −0.230 −0.381 0.038 −0.138 −0.163 −0.140 −0.112 −0.059 −0.083 −0.125 0.155 0.382 0.077 −0.071

0.011 0.017 −0.198 −0.170 −0.103 0.261 −0.020 0.203 0.275 0.027 −0.052 −0.140 0.266 0.124 0.255 0.033 −0.115 0.066 0.136 0.006 −0.160 0.181 −0.001 −0.158 0.083 0.086 0.148 −0.171 −0.028 0.197 0.063 0.177 −0.106 0.144 0.125 −0.093 −0.101 −0.027 0.184 0.164 0.006 −0.021 0.063 −0.064 −0.057 0.349 −0.250 0.106 0.126

0.040 0.251 0.239 0.098 0.119 0.107 0.258 0.249 0.201 0.096 0.258 0.074 0.165 0.167 0.425 0.256 0.138 0.277 0.222 0.070 0.036 0.107 0.022 0.130 0.242 0.277 0.229 0.145 0.278 0.166 0.133 0.261 0.333 0.165 0.141 0.087 0.236 0.217 0.084 0.107 0.097 0.112 0.079 0.192 0.177 0.142 0.393 0.217 0.083

conditions conditions conditions conditions

1 2 3 4

0.207 0.083 0.060 −0.251 0.067 −0.062 0.017 0.028 −0.431 0.013 0.320 −0.148 −0.036 −0.062 −0.172 −0.093 0.036 0.188 0.081 −0.063 −0.040 0.023 0.270 −0.025 −0.041 −0.007 0.255 0.262 −0.124 0.044 0.356 −0.147 0.133 0.247 0.018 −0.086 0.267 0.237 0.004 0.006 0.032 −0.248 0.122 −0.015 0.015 −0.045 −0.076

The most important factors that negatively affect the WB of Italian regions concern quality of labour (incidence of low pay (−0.932); irregular employment (−0.920) and temporary contracts (−0.742)) and pieces of ‘‘minimum conditions’’ of life (people younger than 60 living in very low labour intensive families (−0.92); people living in severe material deprivation (−0.834) and families signalling difficulties of access to at least three essential services (−0.878)). By reading together positive and negative aspects of labour, the importance of a stable job emerges. This is noteworthy since Italy’s labour market flexibility has increased starting from the late 1990s. The Treu Package (1997) and the Biagi Law (2003), in fact, relaxed the discipline for standard temporary contracts and introduced new forms of ‘‘atypical’’ nonpermanent contracts while maintaining existing rules on permanent contracts. Because of these reforms, employment grew strongly until the 2008 crisis and then again in 2014, but more than half of the new jobs were temporary (Pinelli et al., 2017). Average pay was also affected by these reforms while the rate of irregular employment was unchanged. Our finding, however, shows that WB decreases if the job is temporary, irregular and low paid. Therefore, labour market flexibility creates more employment, but does not seem to increase WB. WB declines as the level of inequality (Economic Well-being 2) increases (−0.764).

270

G. Bonanno, G. D’Orio and R. Lombardo / Economic Analysis and Policy 65 (2020) 262–275

Table 3 Rankings by WB index retrieved from the factor analysis. Source: Own elaborations on ISTAT data. Rank Region

2010

Region

2011

Region

2012

Region

2013

Region

2014

1.551 1.011 0.915

Trentino Veneto Friuli

1.521 0.961 0.918

Trentino Friuli Veneto

1.607 0.924 0.873

Trentino Friuli Lombardy

1.784 1.087 0.986

0.872

Lombardy

0.891

Lombardy

0.830

Trentino Friuli EmiliaRomagna Veneto

1.759 1.015 0.981

0.789

Trentino Friuli EmiliaRomagna Veneto

1.606 0.993 0.921

4

Trentino Friuli EmiliaRomagna Lombardy

0.950

0.946

5

Veneto

0.789

Lombardy

0.807

0.861

Lombardy

0.903

Toscana

0.661

0.653

0.583

Valle d’Aosta Piedmont Umbria Marche Liguria Lazio Abruzzo Sardegna Molise Basilicata Puglia Calabria Campania Sicilia

0.553

0.587

0.638

0.561

0.501 0.478 0.476 0.469 −0.015 −0.320 −0.326 −0.504 −0.877 −1.153 −1.536 −1.641 −1.672

Piedmont Umbria Liguria Marche Lazio Abruzzo Sardegna Molise Basilicata Puglia Calabria Campania Sicilia

0.573 0.567 0.446 0.407 −0.084 −0.146 −0.295 −0.674 −0.861 −1.144 −1.668 −1.709 −1.781

Toscana Umbria Liguria Marche Lazio Abruzzo Sardegna Molise Basilicata Puglia Calabria Campania Sicilia

0.521 0.438 0.400 0.384 −0.068 −0.128 −0.305 −0.467 −0.954 −1.296 −1.657 −1.720 −1.841

Valle d’Aosta Piedmont Umbria Marche Liguria Lazio Abruzzo Sardegna Molise Basilicata Puglia Campania Calabria Sicilia

Valle d’Aosta Piedmont

0.714

7

Valle d’Aosta Toscana

EmiliaRomagna Toscana

0.824

6

EmiliaRomagna Valle d’Aosta Piedmont

EmiliaRomagna Veneto

0.529 0.488 0.367 0.239 −0.051 −0.129 −0.525 −0.695 −0.889 −1.277 −1.714 −1.796 −2.033

Toscana Umbria Marche Liguria Lazio Abruzzo Sardegna Molise Basilicata Puglia Campania Calabria Sicilia

1 2 3

8 9 10 11 12 13 14 15 16 17 18 19 20

0.710

2015

0.909

0.669

Valle d’Aosta Toscana

0.737 0.682

0.658 0.510 0.481 0.342 0.107 −0.153 −0.472 −0.542 −0.871 −1.270 −1.653 −1.704 −1.818

Piedmont Umbria Marche Liguria Lazio Abruzzo Sardegna Molise Basilicata Puglia Campania Calabria Sicilia

0.653 0.573 0.441 0.430 0.075 −0.169 −0.415 −0.647 −0.793 −1.230 −1.735 −1.782 −1.878

Some variables in Factor 1 such as Security 1 and 2 do not have the expected sign. Security 1 has the expected sign in Factor 3, 4 and 5, while Security 2 has the expected sign in Factor 4, 6, 7 and 8. This means that the total effect of these two variables on overall WB may not be as expected. We will return to this point in Section 4.2. The result of FA, presented in Table 3, is the WB index for the Italian regions from 2010 until 2015.18 Trentino-Alto Adige scores the highest WB over the considered period; its level of WB increases from 1.551 to 1.784. Friuli-Venezia Giulia is in second place with the score that increases from 1.011 to 1.087. Lombardy, fourth in 2010 (0.789), is third in 2015 with 0.986. Emilia- Romagna, third in 2010 (0.915), is fourth in 2015, despite a higher score of 0.946. These regions, almost all Northern, have increased their WB and their ranking is relatively stable. The last four positions are occupied by Southern regions and all of them register a reduction of WB between 2010 and 2015. Puglia moves from −1.153 to −1.230, Calabria moves from −1.536 to −1.782 also dropping a place; Campania gains a position even if its WB moves from −1.641 to −1.735. The region with the lowest level of WB is Sicily for all the six years declining from −1.672 to −1.878. The divide between the advanced North and the South is a pre-eminent feature of Italy, as confirmed by our findings. In 2010, the distance between Trentino-Alto Adige and Sicily was 3.223 (1.551+1.672) while in 2015 this distance increases to 3.662 (1.768+1.878). Ferrara and Nisticò (2015) found that Italian regions tended to become more similar in terms of WB over the period 2004–2010, while our results show that, from 2010 to 2015, territorial cohesion decreased.

4.2. Application of stochastic frontier approach

The WB indicator, obtained through the FA, is used as dependent variable of the stochastic frontiers, in which we include one input for each dimension explaining the WB indicator. In Tables 4 and 5, we report the description and summary statistics of all the variables included in the frontiers and the controlling factors in the inefficiency equation, respectively.

18 To ensure that doubts do not affect the reliability of the FA, we conduct a rotation of factors. In detail, we employ the Varimax rotation matrix as robustness check and find no significant differences with un-rotated results (Abdy, 2003). For the sake of brevity, we do not report these results, but they are available upon request.

G. Bonanno, G. D’Orio and R. Lombardo / Economic Analysis and Policy 65 (2020) 262–275

271

Table 4 Description of the variables. Variables

Exp sign

Frontier WB Education Health Labour

+ + +

Economics

+/−

Social relationship

+

Politics Security

+ −

Satisfaction

+

Landscape Environment Innovation Quality of services

+ + + +

Efficiency equation Transfers dependence degree Tax autonomy degree Satisfaction average

+ − −

Residential social assistance and social health care

−

Description Dependent Variable Synthetic indicator built by ISTAT using elementary indices Synthetic indicator built by ISTAT using elementary indices Synthetic indicator built by ISTAT using elementary indices quality 2–6 listed in Table 1 Synthetic indicator built by ISTAT using elementary indices Well-being & minimum conditions 1–4 listed in Table 1 Synthetic indicator built by ISTAT using elementary indices Table 1 Synthetic indicator built by ISTAT using elementary indices Synthetic indicator built by ISTAT using elementary indices listed in Table 1 Synthetic indicator built by ISTAT using elementary indices Table 1 Synthetic indicator built by ISTAT using elementary indices Synthetic indicator built by ISTAT using elementary indices Synthetic indicator built by ISTAT using elementary indices Synthetic indicator built by ISTAT using elementary indices Table 1

Education 1–5 listed in Table 1 Health 1–5 listed in Table 1 Labour-occupation 1, Labour Economic Well-being 1–2, Social relationship 1–8 listed in Politics 1–7 listed in Table 1 Security-murders, Security 1–3 Subjective Well-being 1 listed in Landscape 1–4 listed in Table 1 Environment 1–8 listed in Table 1 Innovation 1–3 listed in Table 1 Services quality 1–6 listed in

ETS/(ETR+ETS+EET) ETR/(ETR+ETS+EET) Average satisfaction Index with post-office, public transportation timetable and health services local offices Number of structures (for 100.000 citizens) housing people in need for different reasons: elderly alone or with health problems, disabled, minors without protection, young women in difficulty, foreigners or Italian citizens with economic problems and in conditions of social hardship.

ETR = revenue from own taxes (Tribute) or taxes devolved by the State to Regions and substitutive sums ETS = revenue from own taxes (Contributions) and current account State allocations EET = revenue from capital goods (rents, profits, etc.) and profits of regional companies

Table 6 shows the model estimated following the approach of Battese and Coelli (1995), with Transfer Dependence, Tax Autonomy,19 Satisfaction Average and Residential Assistance as control variables for the level of inefficiency.20 Beginning with the frontier and as expected (see Table 4), we find significant and positive effects of all variables included in the model, with the exception of Security and Innovation that do not affect the WB generation and for Satisfaction that has a significant negative effect. Looking at the inefficiency equation, we estimate a positive effect of transfers dependence and efficiency, which indicates that regions more dependent on external financing sources achieve lower efficiency. Moreover, satisfaction and residential assistance positively affect efficiency (RE) scores, while the coefficient of Tax Autonomy Degree is not significant. The inputs elasticities are robust and significant, with exception of security and innovation. Various components of security (theft, but not murder) have an unexpected sign even in the FA. John (2011) argues that we cannot understand the value of security in terms of ‘‘freedom from fear’’. Considering security as a determinant of WB is linked to the notion of protection from harm as well as the provision of physical safety. We do not find enough 19 Tax autonomy degree is one control variable of our analysis. The reason to choose this variable is linked to the current Italian fiscal system. After the 2001 Italian Constitutional Reform and subsequent reforms that created the fiscal federalism in Italy, sources of revenues for Italian regions fall in three categories: (i) own taxes and revenues; (ii) shares of national taxes; (iii) resources from an equalization fund. Only own taxes and revenues is under direct control of regional authorities, representing the degree of their fiscal autonomy. The most important regional taxes, in terms of contribution to total regional revenues, include the tax on productive activities (IRAP, Legislative Decree 15 December 1997, No. 446) and the surcharge on personal income tax (IRPEF, Legislative Decree No. 446 of 15 December 1997). IRAP was originally introduced as a replacement of health contributions from the central government to finance regional health expenditures. Regional revenues from own taxes represent around 30% of total regional revenues in 2010 (Italian National Institute for Statistics). Of these, approximately 26% are derived from taxes on productive activities, while 19% come from the surcharge on personal income tax. In particular, the evolution of IRAP is highly correlated with the dynamics of health expenditure and the related transfers from the national government. For this reason we included also a control variable on the number of structures, every 100,000 citizens, where are housed people who are in need for different reasons: elderly alone or with health problems, disabled, minors without protection, young women in difficulty, foreigners or Italian citizens with economic problems and in conditions of social hardship. Finally, to account for ‘‘quality’’ of public expenditure, a control variable on the ‘‘quality’’ of some public services has also been used. 20 We use SFA to estimate a ranking scale of the Italian regions, thus the inefficiency component of the frontier (WB function) is the distance from the best frontier due to all factors which we can observe (observable heterogeneity) and also some factors that we cannot observe (unobservable heterogeneity). We have information about some of the observable variables, which allows us to obtain cleaner rankings than the case in which we do not use these explicative variables.

272

G. Bonanno, G. D’Orio and R. Lombardo / Economic Analysis and Policy 65 (2020) 262–275 Table 5 Summary statistics. Source: Own elaborations on ISTAT Data Variables

Obs

Mean

Std. Dev.

Min

Max

Frontier WB Education Health Labour Economics Social relationship Politics Security Satisfaction Landscape Environment Innovation Quality of services

120 120 120 120 120 120 120 120 120 120 120 120 120

−2.4e−08 104.53 101.92 95.95 100.54 99.22 100.70 101.24 95.22 97.80 102.14 96.93 98.90

1.00 8.53 8.40 10.31 12.50 10.62 4.10 8.20 10.58 12.10 6.12 9.54 10.58

−2.033 84.80 83.00 69.80 67.60 79.50 90.30 66.20 69.30 72.30 89.40 79.20 75.00

1.78 123.30 120.10 109.80 117.40 125.70 108.40 113.80 125.60 123.00 121.20 124.10 116.20

Efficiency equation Transfers dependence degree Tax autonomy degree Satisfaction average Residential assistance

120 120 120 120

0.12 0.84 57.17 5.06

0.09 0.10 6.13 5.64

0.01 0.53 39.20 0.13

0.41 0.96 66.70 21.81

Table 6 Estimation of SFA — Battese and Coelli (1995): the role of dependence from transfers. Source: Own elaborations on ISTAT Data Model 3

Coef.

Std. Err.

z

p-value

Frontier Education Health Labour Economics Social relationship Politics Security Satisfaction Landscape Environment Innovation Quality of services Intercept

1.562 0.814 1.758 0.765 2.506 0.640 −0.047 −0.229 0.547 1.006 0.034 1.003 −47.482

0.167 0.259 0.269 0.208 0.250 0.287 0.141 0.115 0.208 0.207 0.145 0.290 1.392

9.370 3.140 6.530 3.670 10.040 2.230 −0.330 −1.990 2.630 4.860 0.240 3.460 −34.110

0 0 0 0 0 0.026 0.738 0.046 0 0 0.814 0 0

(In)efficiency equation Transfers dependence degree Tax autonomy degree Satisfaction average Residential assistance Intercept

0.452 −0.037 −0.012 −0.017 0.941

0.154 0.138 0.003 0.005 0.202

2.940

−0.270 −4.230 −3.710 4.670

0 0.788 0 0 0

sigma_u sigma_v lambda

0.042 0.063 0.658

0.029 0.017 0.046

1.420 3.790 14.440

0.157 0 0

Log-likelihood: 142.28; AIC: −244.55.

evidence to reject the null hypothesis of the coefficient of security being equal to zero. We cannot reject, therefore, the hypothesis that, over the considered period, security is not an important determinant of WB. By using the Akaike Information Criterion (AIC), we find that, from a statistical point of view, the estimated specifications are better than the basic model without efficiency determinants.21 Moreover, the correlation between our WB indicator and the efficiency scores of SFA model is very high (about 0.80), meaning that the rank calculated through the WB index is close to the rankings estimated following equation (2). Table 7 shows the RE of the SFA model. The main finding of the SFA model is in line with the empirical evidence on the role of governmental quality in determining the relationship between fiscal decentralization and regional disparities (Giannola et al., 2016). Decentralization of capital expenditure through transfers, in fact, has the potential to reduce WB disparities; however, it might not produce effective results because of problems associated with the ineffectiveness of local authorities. As Kyriacou et al. (2013) 21 We estimate a simple model without considering any efficiency determinants. The AIC statistic of model with the efficiency determinants is smaller than the AIC of the basic model. Results are available from the authors upon request.

G. Bonanno, G. D’Orio and R. Lombardo / Economic Analysis and Policy 65 (2020) 262–275

273

Table 7 Rankings by efficiency scores retrieved from SFA model: the role of dependence from transfers. Source: Own elaborations on ISTAT Data Rank Region

2010

Region

2011

Region

2012

Region

2013

Region

2014

Region

2015

0.980

Veneto

0.987

Veneto

0.984

Veneto

0.982

Veneto

0.977

Veneto

0.980

2

EmiliaRomagna Veneto

0.979

0.985

Lombardy

0.976

Lombardy

0.975

Lombardy

0.967

Lombardy

0.975

3

Lombardy

0.975

EmiliaRomagna Piedmont

0.973

0.975

Piemonte

0.966

Piedmont

0.974

Piedmont

0.970

Lombardy

0.973

0.957

0.963

0.945

0.948 0.883 0.881 0.880 0.871

Toscana Friuli Trentino Umbria Lazio

0.956 0.916 0.911 0.898 0.850

Toscana Friuli Umbria Trentino Lazio

0.916 0.869 0.869 0.863 0.818

0.930 0.881 0.865 0.860 0.840

Toscana Trentino Umbria Marche Lazio

0.906 0.879 0.878 0.863 0.858

0.936 0.931 0.920 0.902 0.891

10

Toscana Trentino Friuli Umbria Valle d’Aosta Lazio

EmiliaRomagna Toscana Lazio Trentino Friuli Umbria

0.974

5 6 7 8 9

EmiliaRomagna Toscana Trentino Umbria Friuli Lazio

EmiliaRomagna Piedmont

0.956

4

EmiliaRomagna Piedmont

0.867

0.821

Marche

0.809

Marche

0.819

Friuli

0.857

Marche

0.871

11

Marche

0.823

Valle d’Aosta Marche

0.820

Sardegna

0.802

0.773

Liguria

0.806

Puglia

0.791

Puglia

0.801

0.765

0.760

0.774

Puglia

0.782

13

Liguria

0.781

Liguria

0.798

Valle d’Aosta Molise

Valle d’Aosta Molise

0.776

12

Valle d’Aosta Puglia

0.765

Liguria

0.753

Abruzzo

0.756

0.772

14 15 16 17 18 19 20

Molise Calabria Abruzzo Sardegna Campania Basilicata Sicilia

0.772 0.768 0.762 0.761 0.749 0.730 0.727

Sardegna Molise Abruzzo Calabria Campania Basilicata Sicilia

0.766 0.758 0.745 0.736 0.726 0.723 0.702

Puglia Liguria Abruzzo Campania Basilicata Calabria Sicilia

0.748 0.746 0.728 0.719 0.706 0.691 0.683

Molise Sardegna Abruzzo Calabria Basilicata Campania Sicilia

0.753 0.748 0.746 0.714 0.693 0.692 0.664

Puglia Sardegna Liguria Basilicata Sicilia Calabria Campania

0.756 0.747 0.746 0.722 0.704 0.699 0.667

Valle d’Aosta Molise Sardegna Abruzzo Basilicata Calabria Sicilia Campania

1

0.762 0.759 0.731 0.730 0.717 0.705 0.689

point out, fiscal decentralization promotes regional convergence only in the context of good quality of government while it increases regional disparities where governance is weak. According to Mauro and Pigliaru (2013), the effectiveness of public investments and local social capital are related: If in the South social capital is lower than the national average, a project managed by central government is more effective than one managed by local authorities. The rank of Table 7 shows Veneto, Lombardy, Piedmont and Emilia-Romagna as ‘‘frontier’’ regions. The differences in the magnitude of efficiency scores between these four are slight. Trentino-Alto Adige, the region with the highest WB score, is not the region with the highest efficiency score. Valle D’Aosta loses many positions in the rank and 10 points of efficiency between 2010 and 2015. In the South, Puglia has the highest rank, even if over the 6 years considered the efficiency drops by almost 1 point. The 6 points lost in efficiency brings Campania from 18th to the bottom spot. Calabria loses more than 5 points over the 6 years, owing to ineffectiveness in the use of external finance. No Southern region improved its efficiency score over the considered period. This is the most worrying aspect emerging from our analysis. 5. Concluding remarks During the last few years, economic policy making and research have witnessed increasing interest in the evaluation of regional competitiveness. The increasing significance of this aspect deserves more attention especially because of the economic efficiency of regions representing the basis of economic success for the micro-economic level and, in general, the WB of the country. WB is a complex socio-economic phenomenon, with many definitions and quantification methods upon which the specialists have not yet reached full consensus, but the need to raise levels of WB is frequently discussed both in the economic literature and in everyday practice. This paper proposes an overall objective indicator of WB for Italian regions and estimates the level of efficiency of each region in generating WB. We are aware that the imposition of homogeneity across different groups is a limitation of the method used. Clearly, a poor individual puts a different value on one euro than a wealthy one. However, it can be misleading to consider only economic indicators to measure WB and the addition of other dimensions such as social and environmental, health etc., allows us to draw a fuller picture although caveats remain. The database has been built on a set of variables synthetized by ISTAT through the AMPI method with the assumption of non-substitutability of dimensions. We are confident that the BES starting domains used are a reliable description of WB in Italian regions.

274

G. Bonanno, G. D’Orio and R. Lombardo / Economic Analysis and Policy 65 (2020) 262–275

The regional ranking confirms the divide between Northern and Southern regions in overall WB. In the five years considered, we find consistent evidence that WB increased in the first four positions occupied by Northern regions and decreased in the last four positions, occupied by Southern regions. The most important aspects of differences in regional WB in Italy are the quality of employment, income, and social relations. Employment rate and the quality of employment, average per capita disposable income and satisfaction with family, friends and persons to rely on, in fact, positively affect the overall WB between 2010 and 2015. Labour quality matters significantly: the differences in regional WB increase if jobs are temporary, irregular or low paid. Although flexibility increases employment, it does not seem to increase WB; Italians still prefer a stable job. Inequality also matters and increased over the period analysed. Policies that affect job quality, affect WB in a positive way. Furthermore, we use the computed WB index as dependent variable of a generating function to measure the ranking of Italian regions, in terms of overall WB efficiency scores in which the inefficiency equation is simultaneously estimated, and ‘‘influenced’’ by transfers, tax autonomy, quality of public services and residential care. A North–South divide also emerges in terms of efficiency in determining WB. In fact, we find a positive effect of transfers’ dependence on efficiency, which indicates that those regions more dependent on external finance achieve lower efficiency scores. These are Southern regions: as fiscal decentralization seems to play an important role in disparities in the efficiency with which WB is generated, local and national policies on taxation need to be carefully planned. Opinions about the effectiveness of central government transfers cannot ignore the link that a progressive taxation system establishes between differences in regional GDP and differences in regional fiscal revenues. Given that the poorest regions have lower tax revenues, without transfers, this means less expenditure on public goods, and, hence, a widening divide. Since the Italian Constitution, which is based on the principle of equality, obliges governments to guarantee a uniform level of public goods across regions, even in the presence of significant differences in the ability of the different regions to finance such goods, transfers are necessary. Evidence of inefficiency of redistribution mechanisms between North and South is not a valid reason for a drastic cut in transfers. Current failures should not be used to boost selfish localism; they should rather stimulate the search for more effective policies to reduce disparities. Declaration of competing interest The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper. Acknowledgements We would like to thank the Editor, Clevo Wilson, and two anonymous referees for their useful comments. We are also grateful to seminar participants at the A.I.S.Re (Italian Association of Regional Sciences) 2018 Conference (Bolzano) for their suggestions. The usual disclaimers apply. References Abdy, H., 2003. Factor rotations in factor analyses. In: Lewis-Beck, M., Bryman, A., Futing, T. (Eds.), Encyclopedia of Social Sciences Research Methods. Sage, Thousand Oaks (CA), (2003). Aigner, D., Lovell, C.K., Schmidt, P., 1977. Formulation and estimation of stochastic frontier production function models. J. Econometrics 6, 21–37. Annoni, P., Weziak-Bialowolska, D., Dijkstra, L., 2012. Quality of Life at the Sub-National Level: An Operational Example for the EU. JRC Scientific and Policy Reports, EUR, 25630. Aslam, A., Corrado, L., 2011. The geography of well-being. J. Econ. Geogr. 12 (3), 627–649. Battese, G.E., Coelli, T.J., 1995. A model for technical inefficiency effects in a stochastic frontier production function for panel data. Empir. Econom. 20 (2), 325–332. Berloffa, G., Modena, F., 2012. Economic well-being in Italy: the role of income insecurity and intergenerational inequality. J. Econ. Behav. Organ. 81 (3), 751–765. Blanchflower, D.G., Oswald, A.J., 2011. International happiness: A new view on the measure of performance. Acad. Manag. Perspect. 25 (1), 6–22. Bleys, B., 2012. Beyond GDP: Classifying alternative measures for progress. Soc. Indic. Res. 109 (3), 355–376. Ciommi, M., Gigliarano, C., Chelli, F.M., Gallegati, M., 2013. Behind, beside and beyond GDP: Alternative to GDP and to macro-indicators. In: E-Frame EU-Project. http://www.eframeproject.eu/fileadmin/Deliverables/Deliverable3.1.pdf. Coelli, T.J., Rao, D.S.P., O’Donnell, C.J., Battese, G.E., 2005. An Introduction to Efficiency and Productivity Analysis. Springer, New York. Decancq, K., Lugo, M.A., 2013. Weights in multidimensional indices of wellbeing: An overview. Econometric Rev. 32 (1), 7–34. Diamantopoulos, A., Riefler, P., Roth, K.P., 2008. Advancing formative measurement models. J. Bus. Res. 61 (12), 1203–1218. Dillon, W.R., Goldstein, M., 1984. Multivariate Analysis: Methods and Applications (Vol. 45). Wiley, New York. Ferrara, A.R., Nisticò, R., 2015. Regional well-being indicators and dispersion from a multidimensional perspective: Evidence from Italy. Ann. Reg. Sci. 55 (2-3), 373–420. Ferrara, A.R., Nisticò, R., Lombardo, R., 2019. Subjective and objective well-being: Bridging the gap. Scienze Regionali 18 (Speciale), 575–610. Giannola, A., Petraglia, C., Scalera, D., 2016. Net fiscal flows and interregional redistribution in Italy: A long-run perspective (1951–2010). Struct. Change Econ. Dyn. 39, 1–16. Greene, W.H., 1993. Econometric Analysis, second ed. Pretence Hall, Englewood Cliffs, NJ. Greene, W.H., 2017. Econometric Analysis. Prentice Hall, Upper Saddle River, New Jersey. Ivaldi, E., Bonatti, G., Soliani, R., 2016. The construction of a synthetic index comparing multidimensional well-being in the European union. Soc. Indic. Res. 125 (2), 397–430. John, S., 2011. Security, knowledge and well-being. J. Moral Philos. (1), 68–91. Kumbhakar, S.C., Lovell, C.A.K., 2000. Stochastic Frontier Analysis. Cambridge University Press, Cambridge.

G. Bonanno, G. D’Orio and R. Lombardo / Economic Analysis and Policy 65 (2020) 262–275

275

Kumbhakar, S.C., Parmeter, C.F., Zelenyuk, V., 2017. Stochastic frontier analysis: Foundations and advances. In: Handbook of Production Economics. Springer, New York, NY. Kyriacou, A.P., Muinelo-Gallo, L., Roca-Sagalés, O., 2013. Fiscal decentralization and regional disparities: The importance of good governance. Pap. Reg. Sci. 94 (1), 89–107. Lensink, R., Meesters, A., 2014. Institutions and bank performance: A stochastic frontier analysis. Oxf. Bull. Econ. Stat. 76 (1), 67–92. Mauro, L., Pigliaru, F., 2013. Decentralization, Social Capital and Regional Convergence, Nota Di Lavoro N. 57. Fondazione ENI Enrico Mattei, Milano. Mazziotta, M., Pareto, A., 2013. A non-compensatory composite index for measuring well-being over time. Cogito Multidiscip. Res. J. 5 (4), 93–104. Mazziotta, M., Pareto, A., 2016. On a generalized non-compensatory composite index for measuring socio-economic phenomena. Soc. Indic. Res. 127 (3), 983–1003. Mazziotta, M., Pareto, A., 2019. Use and misuse of PCA for measuring well-being. Soc. Indic. Res. 142 (2), 451–476. Meeusen, D., van de Broek, J., 1977. Efficiency estimation from Cobb–Douglas production functions with composed error. Internat. Econom. Rev. 18, 435–444. OECD, 2011. Better Life Initiative. Compendium of OECD Well-Being Indicators, Organization for Economic Cooperation and Development (OECD). Paris, http://www.oecd.org/sdd/47917288.pdf. Osberg, L., 1985. The measurement of economic well-being. In: Laidler, D.E. (Ed.), Approaches to Economic Well-Being, vol. 26. University of Toronto Press. Osberg, L., Sharpe, A., 2005. How should we measure the economic aspect of well-being? Rev. Income Wealth 51 (2), 291–316. Pinelli, D., Torre, R., Pace, L., Cassio, L., Arpaia, A., 2017. The Recent Reform of the Labour Market in Italy: A Review (No. 072). DG ECFIN. European Commission. Simar, L., Wilson, P.W., 2007. Estimation and inference in two-stage, semi-parametric models of production processes. J. Econometrics 136 (1), 31–64. Stiglitz, J.E., Sen, A., Fitoussi, J.P., 2009. Report By the Commission on the Measurement of Economic Performance and Social Progress, Paris. http://www.stiglitz-sen-fitoussi.fr/documents/rapport_anglais.pdf. Veneri, P., 2019. The OECD framework and database to measure regional well-being. Scienze Regionali 18 (Speciale), 657–666. Wang, H.J., Schmidt, P., 2002. One-step and two-step estimation of the effect of exogenous variables on technical efficiency levels. J. Product. Anal. 18 (2), 129–144. Wheat, P., Stead, A.D., Greene, W.H., 2019. Robust stochastic frontier analysis: a student’s t-half normal model with application to highway maintenance costs in England. J. Product. Anal. 51 (1), 21–38. Yang, L., 2014. An Inventory of Composite Measures of Human Progress. Occasional paper on methodology. UNDP Human Development Report Office.