You are on page 1of 19

Soc Indic Res DOI 10.

1007/s11205-012-0096-7

A DEA Approach to Measure the Quality-of-Life in the Municipalities of the Canary Islands
Juan Carlos Martn Cira Mendoza

Accepted: 28 May 2012 Springer Science+Business Media B.V. 2012

Abstract The notion of the quality of life has always intrigued economists, sociologists and other researchers in the area of social science. Since the genesis of the denition of Gross Domestic Product (GDP) as a truthful measure of well-being and economic development, other sophisticated methodologies have been proposed in the literature to measure the quality-of-life (QOL) that extend in a multidimensional way this complex concept. Measuring QOL in municipalities consists in nding a set of comparable attributes that can be weighted by some metric in order to construct a synthetic index. Thus, the narrow vision obtained by a single measure as the GDP, in which differences in the QOL cannot be fully analyzed, is overcome. Based upon a renement of data envelopment analysis (DEA)the cross-efciency method, the current paper develops a synthetic QOL index that is based in 19 partial indicators which present the tradeoffs of different dimension for the 87 municipalities of the Canary Islands in Spain. Marginal rates of substitution are calculated to evaluate the tradeoffs on QOL dimensions. A method is also proposed to determine the scores chart of each municipality which can be used as a tool to policy makers in order to establish a program of improving the ranking position of the municipality identifying the critical QOL factors. Keywords Quality-of-life Well-being DEA Cross-efciency Synthetic index

1 Introduction Traditionally, since its development in the 1930s, Gross Domestic Product (GDP) has been considered as the unique and truthful measure of well-being and economic development. It is a measure geared to gauge well-being wholly based on material wealth, unable to
J. C. Martn (&) C. Mendoza Faculty of Economics, Business and Tourism, Universidad de Las Palmas de Gran Canaria, D.2.13, 35017 Las Palmas de Gran Canaria, Spain e-mail: jcmartin@daea.ulpgc.es C. Mendoza e-mail: cmendoza@becarios.ulpgc.es

123

J. C. Martn, C. Mendoza

measure how well-off people live beyond the economic activity. As Stiglitz and Sen (2009) mention, Too much emphasis on GDP as the unique benchmark can lead to misleading indications about how well-off people are and run the risk of leading to the wrong policy decisions. In this sense, these authors corroborate that market-based measures of wellbeing need to be complemented by other non-monetary indicators. The United States led the research in this eld. Works began around 1966, when President Johnson instructed the Department of Health, Education and Welfare to address the development of statistics and indicators that provide information about areas of interest. A notable study in this eld is the seminal work of Easterlin (1974), who asserted that gains in higher income did not have to be necessarily accompanied by greater happiness. At present there is a large frame of work in this area of research (Berenger and Verdier-Chouchane 2007; Gonzalez et al. 2011; Rahman et al. 2005; Royuela et al. 2003). There has been an increasingly concern in developing measures of well-being which evaluate factors that have an impact on the standard of living and go beyond purely income based estimations. In this respect, the European Commission launched a successful initiative in 2007. The conference Beyond GDP was organised with the aim of exploring how to improve the measurement of progress and well-being of nations. After its success, the Beyond GDP partners continue to work on improving these measures. Along the same lines, in 2008 the French government, concerned by the limits of GDP as an indicator of social progress and economic performance, considered creating the Commission on the Measurement of Economic Performance and Social Progress (CMEPSP), whose nal report was published in 2009 (see Stiglitz and Sen 2009). Probably the best known yardstick of well-being that complements GDP with nonmonetary indicators is the Human Development Index (HDI). It has been published by the United Nations Development Programme (UNDP) in their Human Development Reports since 1990 to date. The HDI is a composite index to benchmark countries aggregating three basic dimensions into a summary measure. These dimensions are: health, education and income. Other indices are the Genuine Progress Indicator (GPI), the Index of Economic Well-Being (IEWB) or the Quality of Life Index, for example. See Hagerty et al. (2001) and Bandura (2008) to consult a large number of indices which add non-monetary indicators to measure well-being. Quality-of-life (QOL) is a broad and complex concept studied by diverse disciplines. In the research about this notion, concretely in economics, different lines characterized by the different objectives pursued may exist, and the different methodologies employed using a great variety of factors highlight the ambiguity of the concept. But it is intensely accepted that QOL is a multidimensional measure in which the general concept of well-being is based. QOL should include economic, social and environmental dimensions of well-being. The Organization for Economic Co-operation and Development (OECD), conducted a study in 1982 to identify the common social concerns of most member countries. It established as the most important areas to measure QOL the followings: health; education and skills; employ and quality of working life; leisure; purchasing power of goods and services; physical environment; social environment; and personal security. More recently, the CMEPSP in its nal report determined which elements should belong to the list of QOL features. These elements are very similar to those of the OECD: health; education; personal activities; political voice and governance; social connections; environmental conditions; personal insecurity; and economic insecurity. To construct acceptable measures is necessary to have databases that allow the study of factors affecting QOL. Data availability constraints preclude researchers to use some variables in the QOL analysis. In the recent years the European Commission is working very hard in order to provide adequate

123

A DEA Approach to Measure the Quality-of-Life

information to analyze QOL in the European Union. Thus, at the national level, the interactive database EurLife and the European Quality of Life Survey offers data on living conditions and quality of life in most of the countries of Europe, but at the municipal level, information is still scarce. The rest of the paper is organized as follows. The next section describes the measurement of quality of life. Section 3 describes the dataset. The methodology employed to study the quality of life of Canary islands municipalities is described in Sect. 4. The results obtained are then presented in Sect. 5, and nally, Sect. 6 concludes.

2 The Measurement of Quality of Life Unlike GDP, QOL is often associated to the opportunities available to people, to the meaning they give to their life, and how they enjoy these opportunities available to them (Stiglitz and Sen 2009). QOL comes from both the objective dimension as well as the subjective. Although mixing both of them results promising, subjective ones are not in the scope of our measurement techniques unless appropriate surveys exist and could be available. Objective dimensions are most commonly used to study QOL at the individual level, which attempt to capture complex life-satisfaction variables by looking at different dimensions, such as socioeconomic condition, health facilities or leisure time. The measurement of quality of life may be done in different ways. Composite Indicators (CIs) are recognized as an useful tool in policy analysis and public communications. CIs are an aggregate of all variables used and should ideally measure multidimensional aspects that cannot be gauged by single indicators. A general objective of most of these indicators is to obtain the ranking of countries or regions according to some aggregated dimensions. Its use is increasingly growing due to its ability to provide comparability and illustrate complex multidimensional issues. General public often nd easier to interpret CIs than determinate common factors across many separate indicators (Saltelli 2007). But if they are poorly constructed they may lead to inappropriate interpretations and, therefore, to wrong policies. The improvement of the way the indicators are constructed and used seems to be a very important issue from both theoretical and empirical points of view. In August 2009, in the Communication GDP and beyond: Measuring progress in a changing world the Commission outlined its road map with ve key actions to improve the indicators of progress in ways that track how well-off citizens are and incorporate the best new technical and political developments. Along the same lines, the Handbook on Constructing Composite Indicators, project launched by OECD (Somarriba 2008; OECD 2008), guide builders of CIs in a set of technical guidelines about how to design and develop a composite indicator. The present study aims to create a composite indicator of the quality of life for ranking Canary Islands municipalities (NUTS IV and V) .1 The Canary Islands conform an archipelago which is one of the seventeen NUTS II of Spain. It is composed of seven islands: El Hierro, La Gomera, La Palma and Tenerife, which form the province (NUTS III) of Santa Cruz de Tenerife, and Fuerteventura, Gran Canaria and Lanzarote, which belong to the province of Las Palmas.
1

NUTS represents the initials for Nomenclature of Territorial Units for Statistics used by the European Union for statistical purposes. It is classied according to the number of population in the regions of each European country. The European division model establishes ve levels, of which NUTS I represent the highest one.

123

J. C. Martn, C. Mendoza

This research is partly based on Gonzalez et al. (2011). The development of CIs faces two main empirical challenges. The rst is the denition of a set of indicators that collect information on all aspects that inuence QOL. The second, how to weight and aggregate them in a way that the synthetic index could be used as a policy maker tool in which all the dimensions that shape the domains of QOL of each municipality could be analyzed. Based in the components of QOL proposed by the OECD and in the study of Stiglitz and Sen (2009), representative indicators of all domains described above are obtained, except for the political dimension. In this case, as the level of analysis is the municipality and all municipalities are governed by the same organism and are subject to the same rules, it does not make any sense to include that domain. In relation to the second problem, the literature contains a large quantity of methods that have been used to weight and aggregate CIs. Most synthetic indexes trust on equal weighting, which implies that all indicators have equal importance in the composite and therefore, if it is wrongly used, the composite could present an unbalanced structure. Weighting and aggregation should be done in a way according to the underlying theoretical framework and data properties. If the interest lies in analyzing different units in the same period of time, it should ideally give different weights to the indicators. The Handbook on Constructing Composite Indicators revises several weighting and aggregation techniques, such as Analytic Hierarchy Process (AHP), unobserved component models (UCM) or data envelopment analysis (DEA). The AHPis a widely used technique in multi-attribute decision making. It subdivides a complex decision-making problem into its components, and arranges them into a hierarchic order to compare to each other using a pair wise comparison scheme. The result of this process is a set of priorities between the various actions or alternatives. The weights can provide trade-offs among the indicators. The empirical use of this methodology may be consulted in Puolamaa et al. (1996), who built the Index of Environmental Friendliness using this methodology. Unobserved Component Models assume that indicators depend on an unobserved variable and an error term. In this sense, weights are set in a way that minimise the error in the composite. Using this methodology Kaufmann et al. (1999) aggregated governance indicators. DEA is a nonparametric approach that has been used by Hashimoto and Ishikawa (1993) for the QOL evaluation of the 47 prefectures of Japan. Although, DEA was initially developed to measure efciency in production, some non-standard uses of this technique have been proposed in the literature focusing on the properties of DEA as a powerful aggregating tool.

3 The Data Set In the present study a QOL synthetic indicator that ranks all the Canary Island municipalities is obtained. At this territorial level, data from the Population and Housing Census, provided by the National Statistical Institute (INE) of Spain have been obtained. Data on living conditions and quality of life of Canarian citizens are also available on the website of the Canarian Institute of Statistics (ISTAC). Our data are referred to 2001, the date on which the census was last published. Its elaboration takes place every 10 years, so that next one is about to be published. In fact, the INE is now in the process of data collection. Our sample is composed of the 87 municipalities in the seven Canary Islands. The sample size is large enough to ensure the robustness of the results. As already discussed above, the CI is usually based on a set of QOL dimensions that in our production parlance will be represented by bad dimensionsinputs or good dimensions

123

A DEA Approach to Measure the Quality-of-Life

outputs. The inputs represent a cost, since their reduction makes possible improvements in the results of the composite index of municipalities. By contrast, the outputs are a benet, since they affect positively to our measure of quality of life. Hence in this study we referred to input variables as unfavorable factors, and output ones as positive factors. Of the total 19 dimensions, 11 variables correspond to positive factors and 8 to unfavorable ones. Positive dimensions are conformed by health facilities, which include hospital and primary care centers; social care facilities, which encompass senior centers, day care centers, pensioners club, etc.; cultural and sports facilities, which incorporate theatres, museums, sport centers, etc.; and education facilities, composed by primary and secondary schools, faculties, nursery schools, etc. Education of citizens is composed of three positive factors. The average education level, is an index variable created by INE that compute the average acquisition of knowledge by citizens between 30 and 39 years old2; post-compulsory education represents the percentage of citizens that completed post-compulsory education; and university studies refers to percentage of people with an university degree. Positive economic aspects are measured by the average socioeconomic condition, an index variable that ranges from 0 to 3. It reports about the socio-economic status of citizens. It was obtained by combining information from the variables of occupation, activity and occupational status. The average area per occupant and the physical conditions of dwellings are the proxy variables that reect the positive aspects of housing. The former is the ratio between the sum of the total useful surface of the housing and the total of the residents. The second one is an index that takes values between 0 and 100, in relation to the conditions under which households are. It takes into account aspects like the state of the building, running water, sewage conditions and electricity. Finally, the last positive factor included in our analysis is the weight of the tertiary sector, as the percentage of people employed in the sector of services. As unfavorable factors of living in a municipality, different dimensions are considered that proxy the domains of personal activities, environmental conditions, personal insecurity and economic insecurity. Commuting time, lack of parks and bad communications are the variables employed to approximate the domain of personal activities; commuting time indicates the average time spent by citizens in their daily trips to their place of study or work; lack of parks captures the problems of dwellings in relation to the lack of parks, gardens, etc.; and the number of households that report having poor communications is represented by the variable bad communications. Related to environmental conditions, three unfavorable variables are included: pollution, external noises and lack of cleanliness; pollution measures the percentage of houses that notify problems of pollution or bad smells; external noises indicate the percentage of houses that report problems of external noises; and lack of cleanliness indicates the percentage of dwellings that report a poor cleanliness in the streets. Personal insecurity is approximated by the variable delinquency and vandalism. It measures the percentage of dwellings that notify problems of delinquency or vandalism in the zone. Finally, the unemployment rate is used as a proxy of the domain of economic insecurity. The guidance of both OECD and CMEPSP to select the 19 variables that are reasonably close to each domain of QOL (Fig. 1) has been taken into account. The domains represented are: health, education, personal activities (which sub-domains of commuting and housing are included in), social connections, environmental conditions, and insecurity (both personal and economic). Each domain may be approximated by more than one
2

The index is calculated as the ratio of the sum of the class marks of educational level and the total population. The class marks range from 0 (illiterate) to 4.5 (Ph.D.).

123

J. C. Martn, C. Mendoza

Healthfacilities Social carefacilities

Personal
Deliquency and vandalism

Health

Education facilities Average Education Level Postcompulsory education Universitystudies

Insecurity
Unemployment rate Socio-economic condition

Education
Housing

Economic

Environmental conditions
Pollution Accoustic Pollution Lack of cleanliness Weight of tertiary sector

Quality of Life

Lack of parks Badcommunications

Personal activities

Average area per occupant Physical conditions of dwellings


Commuting time

e Social connections and governance

Commuting

Cutural and sportfacilities

Fig. 1

Domains and variables used to measure quality of life

variable. The variables health facilities and social care facilities are proxies for health. Education is accounted for by the variables education facilities, average education level, post-compulsory education and university studies. Personal activities domain is represented by the variables lack of parks and bad communications. Commuting time and housing, sub-domains of personal activities, are measured by commuting time for the former, and by average area per occupant and physical conditions of dwellings for the last. Cultural and sports facilities indirectly approximates the domain of social connections. Environmental conditions is represented by pollution, external noises, lack of cleanliness and indirectly by weight of the tertiary sector. It is assumed that the greater the weight of the service sector in economic activity of the municipalities is, the less weight of other sectors will be, especially the industry. Therefore, less ecological pressure on the environment will be exerted. Delinquency and vandalism measures personal insecurity. Economic insecurity is covered by unemployment rate and socioeconomiccondition. As mentioned above, political voice is not included in our model, as the territorial units are all subject to the same rules of government. It is out of the scope of this paper to investigate how political voice or governance could be incorporated in the analysis of QOL. However this could be a promising line of research for the future due to the special events we are observing in the EU after the nancial crisis, as political parties seem to lose part of their previous power in favor of the markets. The selection of some variables to proxy this dimension such as participation rates for previous election or other political activities could be done. Table 1 shows the descriptive statistics for the positive and unfavorable factors in our model. As it can be seen, there is a great difference between the minimum and maximum values of almost all the variables. Municipal facilities present large differences between their minimum and maximum values. The municipalities of Vallehermoso, Garafa, Arico, Alajero, Betancuria and Artenara do not present any health facilities. However, Vilaor

123

A DEA Approach to Measure the Quality-of-Life Table 1 Descriptive statistics Mean Positive factors Health facilities (HEALTH) Social care facilities (SOCIAL) Education facilities (EDUC) Cultural and sports facilities (CULT) Average education level (AEL) Post-compulsory education (POST) University studies (UNIV) Socio-economic condition (ASC) Weight of tertiary sector (TERT) Average area per occupant (AREA) Physical conditions of dwellings (LIVCOND) Unfavorable factors Lack of cleanliness (DIRT) Lack of parks (GREEN) External noises (NOISE) Pollution (POLLUT) Delinquency and vandalism (CRIME) Unemployment rate (UNEMP) 27.14 41.03 19.97 13.01 11.95 14.00 19.34 11.98 7.79 11.00 0.18 0.55 0.73 0.18 0.00 Puntagorda Puntagorda Artenara Artenara Puntagorda 65.29 80.08 48.57 37.86 51.34 Vega de San Mateo Betancuria Santa Cruz de Tenerife Santa Cruz de Tenerife Santa Cruz de Tenerife Puntagorda 12.73 8.43 16.26 9.05 0.00 0.00 Vallehermoso, Garafa, Arico, Alajero, Betancuria and Artenara Vilafor, Garafa, La Guancha, Fasnia, Brena Baja, Barlovento, Arico, Alajero, Mogan and Artenara Vallehermoso, Tijarafe, San Sebastian, San Juan de la Rambla, Fasnia, Arico and Alajero Vilaor, San Miguel de Abona, Garafa, Frontera, Fasnia, Brena Baja, Brena Alta, Barlovento, Arico, Alajero and Antigua El Tanque Yaiza 116.41 49.80 Vilaor Yaiza SD Min Municipality Max Municipality

11.70

14.86

0.00

97.61

Yaiza

10.05

14.01

0.00

81.95

Hara

2.62 33.59

0.15 8.64

2.28 13.39

2.98 54.30

Santa Brgida Santa Brgida

0.09 0.90 68.03 32.14

0.03 0.08 7.98 3.89

0.02 0.66 49.67 24.45

El Tanque Puntagorda La Victoria de Acentejo Tazacorte

0.22 1.12 85.78 47.62

Santa Brgida Pajara Puerto de La Cruz Hara

58.68

5.82

45.80

Puntagorda

76.81

Brena Alta

15.90

4.70

5.16

Pajara

36.25

123

J. C. Martn, C. Mendoza Table 1 continued Mean Bad communications (COM) Commuting time (TIME) 18.57 SD 10.87 Min 1.89 Municipality Valle Gran Rey Max 48.90 Municipality Pajara

24.92

7.33

12.74

Puntagorda

58.63

Tejeda

Source: Own elaboration

shows 116.41 health facilities for ten thousand inhabitants. The same happens to educational, cultural and sports and social care facilities, which present some municipalities with no such facilities, although other small municipalities have more than 49 facilities for ten thousand inhabitants. Santa Brgida is highlighted for the level of studies of its inhabitants. This municipality presents the highest values at average education level, post-compulsory education and university studies. Santa Cruz de Tenerife shows the highest values in relation to the problems of dwellings in external noises, pollution, delinquency and vandalism. On the other hand, Artenara is the most quiet and less polluted municipality, as it presents the lowest values at external noises and pollution. A municipality that requires special attention is Puntagorda. Puntagorda suffers from deciencies in two positive factors and in the unemployment rate, but presents very good conditions with some of the lowest values in some unfavorable factors. Socio-economic condition and physical conditions of dwellings present the lowest values. However, the municipality is considered to be the cleanest, safest and the one with more green areas by its inhabitants, as it can be seen for the values of lack of cleanliness, lack of parks, and delinquency and vandalism. This municipality also shows the highest unemployment rate and, in turn, it has the lowest value at commuting time. Because of these disparities, it is necessary to assign correctly the weights to the variables that nally will inuence the composite index.

4 Methodology In this paper, given the nature of the attributes used to measure the quality of life of municipalities and its multidimensionality, DEA is an intuitive method to look at. Charnes et al. (1978), dene the DEA methodology as a mathematical programming model applied to observed data that provides a new way of obtaining empirical estimates of extremal relationships such as the production functions and/or efciency production possibility surfaces that are the cornerstones of modern economics. Since then, numerous applications employing the DEA methodology have been presented involving a wide variety of contexts: education, health care, banking, armed forces, sports, transportation, agriculture, retail stores and electricity suppliers. Originally designed to evaluate decision making units (DMUs), which use multiple inputs to produce multiple outputs, without a clear identication of the relation between them, DEA has progressed throughout a variety of formulations and uses to other kind of industries.3
3

DEA can be applied to scenarios where the data cannot be strictly interpreted as inputs or outputs or there is no direct functional relationship between the variables. In such situations, a general guideline to the classication of the variables is that variables for which lower levels are better are considered inputs, while outputs are those variables for which higher amounts are better.

123

A DEA Approach to Measure the Quality-of-Life

The ability of DEA to model multidimensional relationships among multiple inputs and multiple outputs without considering a basic functional form assumption makes DEA very popular within a wide variety of areas. Some DEA applications are in the area of socioeconomic performance, such as city and nation performance (Charnes et al. 1989; Sueyoshi 1992; Golany and Thore 1997; Zhu 2001; Gonzalez et al. 2011). It can be seen that DEA is particularly appropriate when information on how to weight multiple factors is not clear or even unknown. We do not intend to cover the basic aspects of DEA models. A good introduction to DEA notation, formulation and geometric interpretation can be consulted in Charnes et al. (1994), Ali and Seiford (1993), Coelli et al. (1998) and Cooper et al. (2000). As discussed therein, a model can be described by the envelopment surface, orientation of the model, invariance of units, and efciency measurement. There are three basic DEA models: variable returns to scale (VRS), constant returns to scale (CRS) and additive model. These can be used to seek which ones of the n DMUs determine the frontier of the envelopment surface, and are deemed efcient. The units that do not lie on the frontier are inefcient and the measurement of the grade of inefciency is determined by the selection of the model. The choice of a DEA model depends on some assumptions regarding the data set to be used and in some prior knowledge about the issue under analysis. The data set has to describe the activities of the units in the better possible way. It is especially important to have some idea about the hypothetical returns to scale that can exist. This knowledge is going to determine the envelopment surface constant return to scale CRS or variable return to scale VRS4 of the model. After the selection of envelopment surface, researchers then usually face how to select an orientation of the model to determine the measurement of the efciency. There are three basic orientations: input, output and equal. An input orientation focuses on proportional decrease of the input vector, the output orientation adjusts the proportional increase of the output vector and the equal orientation do not discriminate the importance or the possible increase of output or decrease of input. DMUs, involved in the study, do determine the selection of the orientation. In DEA analysis, it is generally assumed that there are n production units to be evaluated in our case cities or municipalities, using amounts of m different inputs (unfavorable factors) to produce quantities of s different outputs (positive factors). Specically, the oth production unit consumes xio units of input i (i = 1 to m) and produces yro units of output r (r = 1 to s). The oth production unit can now be described more compactly with the vector (Xo,Yo), which denote, respectively, the vectors of favorable and positive factors for the city o. Next, we consider the dominance comparisons for this particular city using the data set as a reference. DEA consider the dominance of the linear combinations of the n cities P P considered in the analysis, i.e., k kk Xk ; k kk Yk , with the scalar restricted to be nonnegative.5 The city o is dominated, in terms of unfavorable factors, if at least one linear combination of cities shows that some of these ones can be decreased without worsening off the rest of unfavorable or positive factors. The city o is dominated in terms of positive factors if at least one linear combination of cities shows that some positive factor can be increased without worsening off the rest of unfavorable and positive factors.
4

CCR and BCC acronyms are sometimes used in reference to CRS and VRS models. The acronyms come from the initial of the authors of the papers that employed these two different envelopment surfaces (Charnes et al. 1978 and Banker et al. 1984). The different assumptions about the scalar produce distinct envelopment surfaces: VRS, CRS or extensions of these basic models.

123

J. C. Martn, C. Mendoza

Thus, the method serves to split up a set of cities into two subsets: efcient and inefcient cities. The method also serves to calculate the level of inefciency of a given inefcient city. Majors and policy makers of a municipality can affect the quality of life of citizens using different planning strategies, such as improving the environmental conditions and reducing the personal insecurity of the citizens. In the current study, multiple attributes related to the QOL are considered in forms of DEA inputs/outputs. As a result, these various attributes are integrated and balanced by DEA-based measures. We rst analyze a VRS-DEA output orientation model to measure the QOL. Formally, the multiplier-DEA VRS output efciency for the city o is calculated through the following linear programming problem: Pm v x vo i1 i io min Ps
vl

s:t:m P i1 Ps vi xij vo ! 1 j 1; . . .; n where vi ; lr ! 0; vo free


r1

r1

lr yro

lr yrj

The set of constraints requires that the same weights, when applied to all the cities, do not provide any city with efciency lower than one. The solution to this minimization problem is not unique. It can be shown that if there exists a solution (v, l) to the above problem, then there exist an innite number of solutions because ((/v, /l), / C 0) is also a solution to the problem (Coelli 1996). Since, there are an innite number of solutions for the dual variables (multipliers), it is necessary to formulate an equivalent linear programming program which avoids this problem. In this sense, the following problem is resolved for each city: min s:t: m P
i1 s P r1 m P vl i1

vi xio vo
s P r1

vi xij vo lr yro 1

lr yrj ! 0

j 1; . . .; n

where vi ; lr ! 0; vo free P A city is in the frontier if and only if s vi xio v 1 in optimality. The constraint i1 Ps r1 lr yro 1 is known as a normalization constraint, and the weighted input and output are called virtual input and virtual output, respectively. See Seiford and Thrall (1990) for a detailed discussion of these models. The efciency ratio ranges from 1 to innity. Thus, each city will choose weights so as to minimize self-efciency, given the constraints and DEA, instead of (subjectively) combining each single attribute ratio between unfavorable and positive factors, provides a single virtual ratio, weighting individual factor by optimal multipliers. 4.1 Cross-Efciency DEA Model Sexton et al. (1986) were the rst to develop the cross-efciency evaluation matrix, initiating the subject of ranking in DEA. Doyle and Green (1994) validated this method, saying that decision makers do not always have a reasonable prior knowledge from which to estimate assurance regions for multipliers, and thus they recommended the

123

A DEA Approach to Measure the Quality-of-Life

cross-efciency evaluation matrix for ranking units. The cross-efciency evaluation method simply calculates the efcient score for each city n times, using the virtual multipliers obtained in each of the n linear programming programs resolved before. The results of all the DEA cross-efciency scores can be summarized in a cross-efciency matrix as following: Pm i1 vik xij vk ; k 1; . . .; n; j 1. . .n 3 hkj Ps r1 lrk yrj Thus, hkj represents the score given to city j in the DEA run of city k, i.e., the performance of city j is evaluated using the weights obtained for city k. Note that all the elements in the matrix are in the range one to innity, and the elements in the diagonal, hkk , represent the standard DEA efciency score (the elements in the diagonal are equal 1 for efcient cities and greater than 1 for inefcient ones, according to a conventional DEA methodology). Sexton et al. (1986) established a set of secondary goals for either aggressive or benevolent DMUs. In this context, a DMU could be considered aggressive if it minimizes self-efciency and at a secondary level maximizes the other DMUs crossefciency scores. The benevolent secondary objective would be able to equally minimize all DMUs cross-efciency scores. The cross-efciency ranking method in this DEA context uses the results of the crossefciency matrix hkj in order to rank all Canaries municipalities. There are different synthetic indexes that can be used to rank the performance of the cities. In this paper, we P " will use the average cross-efciency score given to city j dened as: hj 1 n hkj . k1 n However, averages are not the only possibility. There are other standard univariate summaries, such as, median, variance or some other quantile point that could also be applied. These measures represent the performance of cities better than the standard DEA efciency score hjj . This is based on the fact that all the elements of the cross-efciency matrix have been considered so all virtual multipliers are important in order to obtain a synthetic measure of performance; meanwhile only includes the virtual multipliers of the city that is being evaluated. Furthermore, all the cities are evaluated with the same set of weight vectors. The minimum value of cross-efciency is 1, which occurs when city j is efcient in all the runs, i.e., all the cities evaluate unit j as efcient. In order to rank the units, we can simply assign the city with the lowest score a rank of one and the unit with the highest score a rank of n. While DEA scores hjj are non-comparable, since each element uses " different weights, hj score can be used in comparisons because it utilizes the weights of all the units. However, this feature is also one of the principal drawbacks of this method, since the evaluation subsequently loses its connection to the multiplier weights (Adler et al. 2002).

5 Results This section shows how DEA models [e.g., models (2) and (3)] can be employed and analyzed to characterize the QOL across the municipalities of the Canary Islands. The discussion is carried out via four studies: (1) a VRS-DEA output model is proposed to calculate inefciency represented by the values obtained as the QOL score and the cities that lie in the frontier and can be considered the ones where QOL is optimal; (2) a crossefciency ranking method is utilized to identify a citys QOL status in which the set of multiple efcient cities is reduced determining better QOL scores; (3) factors will be

123

J. C. Martn, C. Mendoza

analyzed for some cities in order to identify critical benchmarks that can be object of future policy planning in order to improve the performance; (4) a method is proposed to obtain adequate multipliers for each of the factors to obtain QOL for the region and each of the islands, as well as, the percentage of the most important attributescritical factors, so each city under evaluation can choose a proper strategy to improve its performance. The VRS-DEA output model was run to obtain an initial best practice frontier. This is a necessary step to know which municipalities are located on the frontier and to calculate the optimal multipliers that will be used in the subsequent step. Table 2 shows the rst attempt to measure the QOL score for all the efcient cities and the ten worst performers. It can be seen that there are 46 municipalities out of 87 that can be considered efcient. This result shows that the discrimination power of this method is quite limited and that a further renement is needed. Looking at the results of the ten worst cities it is highlighted that most of the cities are located in the province of Tenerife and that Telde is the only municipality that is located in the province of Las Palmas. However, the inefciencies are not very signicant because in most of the cases is lower than twenty percent. The city that presents the worst QOL score is La Victoria de Acentejo with a value of 1.22. This city is located in the island of Tenerife. To better perform in the future, this municipality should improve (at least) by 22 per cent. For municipalities that lie in the frontier, the QOL score attained is equal to 1, so they cannot make any (relative) improvement, given the data observed and the structure of the model employed. Some of them belong to the frontier because they are excellent places to live in many or all the dimensions considered (e.g. Puntallana and Agulo). In turn, other frontier municipalities do not excel in any dimension but have a good balance between drawbacks and advantages (e.g. Teguise and Haria). However, some other municipalities are considered efcient because they excel in some dimension although they present low marks in others and, therefore, the QOL score should be analyzed with the X-DEA model. This model reects the natural structure of municipalities in the Canary Islands, as well as two other characteristics that have been highlighted by Zhu (2001): rstly the excessive number of unfavorable and positive factors, and secondly the weight exibility in model (2). Numerous methods have been proposed to reduce the number of frontier DMUs if this is seen as necessary. For example, we may incorporate some weight restrictions, such as cone- ratios or assurance regions (Charnes et al. 1989). However this type of methods requires additional explicit information on tradeoffs among inputs and outputs elicited from expertise of policy makers. Unfortunately, the current study does not have access to this type of information. Therefore, the current study develops a model based on X-DEA as an alternative way to implicitly express the tradeoff information and further reduce the number of frontier QOL cities. To increase the information provided by our rst attempt and achieve a higher degree of congruence or consensus in the optimal multipliers employed in the evaluation of QOL, we propose X-DEA as a valid model to overcome the aforementioned limitations of VRSDEA. The results of the X-DEA (Table 3) show a dramatic reduction in the number of municipalities that can now be considered QOL efcient no municipality can be considered to lie on the frontier. It can also be seen that there are signicant changes in the region of the worst QOL performers. It is clear that this model is an adequate method to measure the QOL score for each municipality and that its ranking power is maximum. Another interesting characteristic is that now the two champion performers are Puntallana and Agulo, two municipalities located in the province of Tenerife and both of them were considered QOL efcient.

123

A DEA Approach to Measure the Quality-of-Life Table 2 City VRS-DEA results to measure QOL QOL score City QOL score City QOL score City QOL score

Efcient cities 38004-Arafo 38047-Tijarafe 38046-Tegueste 1.00 1.00 1.00 38038-Santa Cruz de Tenerife 38028-Puerto de la Cruz 38033-San Andres y Sauces 38008-Brena Alta 35019-San Bartolome de Tirajana 38014Fuencaliente de la Palma 35028-Tas 38053-Villa de Mazo 38030-Puntallana 1.00 1.00 1.00 35005Artenara 38001-Adeje 35018-San Bartolome 38048Valverde 38003 Alajero 35033-Vega de San Mateo 35025-Tejeda 38040Santiago del Teide 35008-Firgas 1.00 1.00 1.00 38029Puntagorda 38013-Frontera 38002-Agulo 1.00 1.00 1.00

35024-Teguise 38049-Valle Gran Rey 35010-Hara

1.00 1.00

1.00 1.00

1.00 1.00

38016-Garafa 38044-Tanque (El) 38042-Silos (Los) 35007Betancuria 38027-Paso (El)

1.00 1.00

1.00

1.00

1.00

1.00

38052-Vilaor 35029-Tinajo

1.00 1.00

1.00 1.00

1.00 1.00

1.00 1.00

38007-Barlovento

1.00

1.00

1.00

38035-San Miguel de Abona 35021-Santa Brgida 38032-Rosario (El)

1.00

35015-Pajara 38037-Santa Cruz de la Palma 38004-Arafo

1.00 1.00

38009-Brena Baja 38010Buenavista del Norte 35020-San Nicolas de Tolentino 38017-Granadilla de Abona 38022-Icod de los Vinos 35026-Telde

1.00 1.00

35032Valleseco 35034-Yaiza

1.00 1.00

1.00 1.00

1.00

1.00

10 worst cities 38018-Guancha (La) 38050Vallehermoso 38026-Orotava (La) 1.10 1.10 1.13 1.13 38019-Gua de Isora 38031Realejos (Los) 1.14 1.15 38034-San Juan de la Rambla 38051-Victoria de Acentejo (La) 1.19 1.22

1.12

1.14

Efcient and the ten worst cities Source: Own elaboration

On the other side, the two worst performers are Artenara and Betancuria. They present very high gures that are consequence of dividing virtual input by very small virtual outputmultipliers that are equal to zero and factors near zero for those factors different from zero.

123

J. C. Martn, C. Mendoza Table 3 City X-DEA results to measure QOL QOL score City QOL score City QOL score City QOL score

Best QOL cities 38030Puntallana 38002-Agulo 38014Fuencaliente de la Palma 35034-Yaiza 35024-Teguise 1.10 1.10 1.14 38015Garachico 38027-Paso (El) 38029Puntagorda 35029-Tinajo 38053-Villa de Mazo 38004-Arafo 35032Valleseco 38049-Valle Gran Rey 38037-Santa Cruz de la Palma 38047Tijarafe 1.43 1.44 1.48

Worst QOL Cities 35030Tuineje 35018-San Bartolome 35008-Firgas 2.51 2.62 2.70 38009-Brena Baja 38012-Fasnia 38008-Brena Alta 35003-Antigua 38050Vallehermoso 38003-Alajero 38005-Arico 38016-Garafa 7,690.12 7,928.40 8,791.29

1.15 1.24

1.48 1.48

35026-Telde 38032Rosario (El) 35012 Mogan 38007Barlovento 38013Frontera 38035-San Miguel de Abona 38052Vilaor

2.84 3.01

9,889.91 10,126.91

38048Valverde 35010-Hara 38033-San Andres y Sauces 38001-Adeje

1.27 1.27 1.28

1.50 1.51 1.52

4.57 4,027.44 4,277.35

18,050.96 20,041.95 21,849.50

1.36

1.54

4,843.81

35005-Artenara

22,761.28

35013-Moya

1.38

1.56

5,821.92

35007Betancuria

23,537.18

Best and Worst QOL cities Source: Own elaboration

Table 4 shows the marks6 for a small sample of municipalities that can be used as a tool for benchmarking and whose value resides in the opportunity that is given to policy makers in order to design a strategy for improving the performance of their respective municipality. We show the results for three different municipalities: Santa Mara de Gua de Gran Canaria, Puntallana and Betancuria. The last two municipalities have the peculiarity of being the best and worst QOL municipalities in the Canary Islands. Santa Mara de Gua belongs to the third quartileposition 36, and it can be seen that presents a very bad performance in unemployment and commuting time. However, it presents a good performance in social and living conditions. On the other hand, Puntallana, the best municipality in Canary Islands show a list of marks with no Ds, 3 Csliving conditions, area and tertiary; 3 Bshealth, average socioeconomic conditions and tertiary; and 13 As. It can be highlighted that Puntallana is not the rst municipality in any factor but it can be seen in the list that presents a highly balanced sheet of marks. Finally, Betancuria

Marks are obtained for each of the factors as follows: A, B, C and D are given if the value belongs to the rst, second, third and fourth quartile, respectively.

123

Table 4 38030-Puntallana Value 5.01 4.49 6.99 6.60 8.58 0.26 14.60 17.89 12.84 17.12 12.84 38.51 57.62 30.81 0.95 2.85 39.74 0.12 64.52 1.10 1.00 1 22 57 15 20 A A C A B 7 A 24 B 51 C 54 C 58.76 43.23 0.92 2.52 24.47 0.09 72.79 23537.18 1.00 2 A 0.00 29 B 0.00 83 78 41 2 29 64 74 42 27 87 42 12 A 14.60 17 20 A 0.00 77 13 A 30.83 75 D D A D D B A B C D B B D B 36 B 14.93 41 B 3 A 3.59 23 B 8 A 80.08 87 D 0 5.16 12.74 81.95 97.61 116.41 49.8 76.81 47.62 1.12 2.98 54.3 0.22 85.78 1.10 1.00 12 A 15.94 40 B 1.89 0.55 9 A 13.15 18 A 0.18 13 A 9.16 26 B 0.18 11 A 7.57 17 A 0.73 Artenara Artenara Puntagorda Valle Gran Rey Puntagorda Puntagorda Pajara Puntagorda Hara Yaiza Vilaor Position Mark Value Position Mark Value Municipality 35007-Betancuria Best values

Individual analysis of factors for benchmarking

Factors Mark B B B C B B D D C B C A A B C B B C C C C

35023-Santa Mara de Gua de Gran Canaria

Unfavorable

Value

Position

NOISE

17.14

37

POLLUT

10.11

31

DIRT

19.02

27

COM

22.82

59

GREEN

39.87

31

CRIME

7.40

37

UNENP

18.62

69

TIME

33.31

80

A DEA Approach to Measure the Quality-of-Life

Positive

CULT

5.76

47

EDUC

12.24

25

HEALTH

6.48

48

Social

15.84

12

LIVCOND

63.02

16

AREA

32.70

29

Yaiza Brena Ala Hara

ASC

0.85

62

AEL

2.69

26

POST

38.23

28

Pajara Santa Brgida Santa Brgida Santa Brgida Puerto de La Cruz Puntallana Arafo

UNIV

0.08

45

TERT

64.13

64

Summary

Cross

1.848

36

123

VRS

1.030

59

123
Fuerteventura 0.90 3.89 2.58 2.45 8.52 3.46 5.03 7.34 0.47 0.08 0.42 0.25 9.94 3.83 3.84 8.26 3.62 0.82 6.71 3.63 8.06 4.55 0.88 6.69 4.05 10.00 0.44 0.92 0.60 0.22 9.40 3.50 3.25 8.01 5.78 1.00 6.13 0.20 0.08 1.05 0.44 7.06 8.72 5.16 6.46 4.52 4.56 1.54 9.15 6.80 0.62 0.11 0.60 0.31 10.45 3.73 3.37 9.12 6.93 1.01 6.26 7.42 5.73 4.02 1.76 1.43 0.81 1.65 2.05 1.40 4.04 5.45 3.81 3.70 1.95 1.70 4.94 0.92 8.78 7.76 0.52 0.06 0.36 0.29 10.12 3.77 3.58 8.48 6.09 0.69 6.62 0.80 1.00 0.81 0.55 Lanzarote Tenerife La Palma La Gomera El Hierro 0.19 0.91 0.62 0.68 3.47 0.40 9.10 7.25 0.33 0.11 0.81 0.42 11.75 4.71 3.57 9.47 7.07 1.03 6.90

Table 5

Multipliers/shadow prices for X-DEA QOL

Factor

Multiplier

Percentage of participation

Archipelago

Gran Canaria

NOISE

0.062

0.99

1.03

POLLUT

0.498

5.45

5.96

DIRT

0.115

2.08

2.21

COM

0.154

1.52

1.57

GREEN

0.243

6.09

6.31

CRIME

0.316

4.46

4.82

UNENP

0.846

6.71

6.97

TIME

0.732

8.54

8.84

CULT

0.125

0.43

0.33

EDUC

0.017

0.08

0.07

HEALTH

0.093

0.55

0.47

SOCIAL

0.070

0.24

0.22

LIVCOND

0.337

9.69

9.80

AREA

0.231

3.42

3.18

ASC

7.279

3.20

3.03

AEL

6.100

7.91

7.63

POST

0.324

5.62

5.61

UNIV

17.669

0.96

0.95

J. C. Martn, C. Mendoza

TERT

0.173

6.13

5.98

A DEA Approach to Measure the Quality-of-Life

is the other side of the coin. It has to seriously improve the performance on the six factors where it failsgreen areas, commuting time, cultural and sport facilities, health facilities, social facilities and post compulsory education. The last two columns of the table show the value and the municipality which presents the best performance for each of the factors under analysis. Table 5 shows the values of multipliers for X-DEA QOL model and it can be seen that the highest values are observed in the following factors: average socioeconomic conditions, average education level and university population. It is remarkable that all the factors belong to the category of positive ones. It is evident that these unitary values need to be multiplied by each of the factors that are part of the analysis. In the table, the percentage of these scores is accrued to each of the factors per island and for the whole archipelago. It can be seen that the most important factors for the overall performance of the archipelago are by order the following ones: living conditions, commuting time, average educational level and unemployment. Thus, policy makers should concentrate their policies in improving these dimensions in order to provide a better QOL to their citizens.

6 Concluding Remarks In measuring the QOL of cities, it is well known that the problem possesses a multidimensional nature and that there is always a huge controversy when a group of experts come up with a list of a small number of cities. It is obvious that there is large panoply of methods to develop measures designed to balance numerous factors that contribute to the QOL, as consequence of the multidimensional nature, as well as the often unknown relationships among various QOL factors. The current paper showed that by using DEA, one is able to develop a multidimensional QOL measure without a priori knowledge of factor relationships. A model based on X-DEA was used to better approximate the multidimensionality of QOL determining a synthetic index to measure indistinctly practical comfort for living in the municipalities of the Canary Islands. It was shown that average multipliers obtained in each run of VRS-DEA model can be used to identify critical QOL factors for a given city. Such new information is important in designing a good strategy to improve the performance on QOL. Some previous papers have ranked the performance on QOL using different approaches like benchmarking DEA and Value Efciency Analysis (VEA), which is a particular case of benchmarking DEA. To our knowledge this is the rst time that X-DEA model is used to analyze the QOL of cities. Our intention here is to provide indepth information on how to improve the QOL while offering an alternative perspective on how to measure QOL. From our experience in other related areas, one could use other alternative approaches such as the super-efciency DEA models or the virtual efciency model to obtain a rank for a set of cities. However, the meaning of such rankings needs to be carefully examined and compared with our proposal but it is beyond the scope of the present study.
Acknowledgments The authors thank Eduardo Gonzalez, professor of the Department of Business Administration at the University of Oviedo for providing most of the data used in this research. Additional gratitude extends to Professor Alex Michalos and two anonymous referees for their valuable comments. The usual disclaimer applies.

123

J. C. Martn, C. Mendoza

References
Adler, N., Friedman, L., & Sinuany-Stern, Z. (2002). Review of ranking methods in the data envelopment analysis context. European Journal of Operational Research, 140, 249265. Ali, A., & Seiford, L. M. (1993). The mathematical programming approach to efciency analysis. In H. O. Fried, C. A. K. Lovell, & S. S. Schmidt (Eds.), The measurement of productive efciency: Techniques and applications. New York: Oxford University Press. Bandura, R. (2008). A survey of composite indices measuring country performance: 2008 update. Working paper. United Nations Development Programme (UNDP). Resource document http://web.undp.org/ developmentstudies/docs/indices_2008_bandura.pdf. Accessed September, 2011. Banker, R. D., Charnes, A., & Cooper, W. W. (1984). Some models for estimating technical and scale inefciencies in data envelopment analysis. Management Science, 30, 10781092. Berenger, V., & Verdier-Chouchane, A. (2007). Multidimensional measures of well-being: standard of living and quality of life across countries. World Development, 35(7), 12591276. Charnes, A., Cooper, W. W., & Rhodes, E. (1978). Measuring the efciency of decision making units. European Journal of Operational Research, 2(6), 429444. Charnes, A., Cooper, W. W., Huang, Z. M., & Wei, Q. L. (1989a). Cone ratio data envelopment analysis and multi-objective programming. International Journal of Systems Science, 20, 10991118. Charnes, A., Cooper, W. W., & Li, S. (1989b). Using DEA to evaluate relative efciencies in the economic performance of Chinese cities. Socio-Economic Planning Sciences, 23, 325344. Charnes, A., Cooper, W., Lewin, A. Y., & Seiford, L. M. (1994). Data envelopment analysis. Theory, methodology and applications. Boston: Kluwer Academic. Coelli, T. (1996). A guide to DEAP version 2.1: A data envelopment analysis (computer) program. CEPA working paper 96/08. Centre for efciency and productivity analysis. University of New England: Armidale. Coelli, T., Rao, D. S. P., & Battese, G. E. (1998). An introduction to efciency and productivity analysis. Boston: Kluwer Academic. Cooper, W., Sieford, L., & Tone, K. (2000). Data envelopment analysis. A comprehensive text with models, applications, reference and DEASolver software. Norwell: Kluwer Academic Publishers. Doyle, J. R., & Green, R. (1994). Efciency and cross-efciency in data envelopment analysis: Derivatives, meanings and uses. Journal of the Operational Research Society, 45(5), 567578. Easterlin, R. A. (1974). Does economic growth improve the human lot? In P. A. David & M. W. Reder (Eds.), Nations and households in economic growth: Essays in honour of Moses Abramovitz. New York: Academic Press, Inc. Golany, B., & Thore, S. (1997). The economic and social performance of nations: efciency and returns to scale. Socio- Economic Planning Sciences, 31, 191204. Gonzalez, E., Carcaba, A., & Ventura, J. (2011). Quality of life ranking of Spanish municipalities. Revista de Economa Aplicada, 19, 123148. Hagerty, M. R., Cummnis, R. A., Ferriss, A. L., Land, K., Michalos, A. C., & Peterson, M. (2001). Quality of life indexes for national policy: Review and agenda for research. Social Indicators Research, 55(1), 196. Hashimoto, A., & Ishikawa, H. (1993). Using DEA to evaluate the state of society as measured by multiple social indicators. Socio-Economic Planning Sciences, 27, 257268. Kaufmann D., Kraay A., & Zoido-Lobaton P. (1999). Aggregating governance indicators. Policy research working papers. World Bank Institute. Resource document http://info.worldbank.org/governance/wgi/ pdf/govind.pdf. Accessed November, 2011. OECD. (2008). Handbook on constructing composite indicators: Methodology and user guide. Paris: OECD Publishing. Puolamaa, M., Kaplas, M., & Reinikainen, T. (1996). Index of economic friendliness. A methodological study. Helsinki: Eurostat. Rahman, T., Mittelhammer, R., & Wandschneider, P. (2005). Measuring the quality of life across countries. Research paper no. 2005/06. World Institute for Development Economics Research. Royuela, V., Surinach, J., & Reyes, M. (2003). Measuring quality of life in small areas over different periods of time: Analysis of the province of Barcelona. Social Indicators Research, 64(1), 5174. Saltelli, A. (2007). Composite indicators between analysis and advocacy. Social Indicators Research, 81, 6577. Seiford, L. M., & Thrall, R. M. (1990). Recent developments in data envelopment analysis: The mathematical programming approach to frontier analysis. Journal of Econometrics, 46, 738.

123

A DEA Approach to Measure the Quality-of-Life Sexton, T. R., Silkman, R. H., & Hogan, A. J. (1986). Data envelopment analysis: Critique and extensions. In R. H. Silkman (Ed.), Measuring efciency: An assessment of data envelopment analysis (pp. 73105). San Francisco: Jossey-Bass. Somarriba, N. (2008). Aproximacion a la calidad de vida social e individual en la Europa comunitaria. Doctoral thesis, Universidad de Valladolid. Resource document http://www.eumed.net/tesis/2010/ mnsa/index.htm. Accessed December, 2011. Stiglitz, J. E., Sen, A., & Fitoussi J. P. (2009). Report by the commission on the measurement of economic performance and social progress. Resource document http://www.stiglitz-sen-toussi.fr/documents/ rapport_anglais.pdf. Accessed September, 2011. Sueyoshi, T. C. (1992). Measuring the industrial performance of Chinese cities by data envelopment analysis. Socio- Economic Planning Sciences, 26, 7588. Zhu, J. (2001). Multidimensional quality-of-life measure with an application to Fortunes best cities. SocioEconomic Planning Sciences, 35(4), 263284.

123

You might also like