Modelling Regional Welfare Efficiency Applying Conditional Full Frontiers

Abstract

This paper demonstrates an evaluation of welfare policies and regional allocation of public investment using the recent developments in efficiency analysis and statistical inference. Specifically, the efficiency of the welfare policies of the Greek prefectures for the census years of 1980, 1990 and 2000 are compared and analyzed. The paper, using bootstrap techniques on unconditional and conditional full frontier applications, indicates that there are major welfare inefficiencies among the prefectures over the three census years. The analysis reveals that the increase of population density over the years has a negative impact on the welfare efficiency levels of the Greek prefectures.

RÉSUMÉ Cette communication démontre une évaluation des politiques sociales et de l'affectation régionale d'investissements publics, sur la base de développements récents dans l'analyse du rendement et des conclusions statistiques. Plus spécifiquement, nous comparaisons et nous analysons l'efficacité des politiques sociales des préfectures grecques pour les années de recensement 1980, 1990 et 2000. En appliquant des techniques de rééchantillonnage à des applications à frontières intégrales inconditionnelles et conditionnelles, la communication démontre que les années de recensement ont été marquées par l'existence d'une grande inefficacité sur le plan social, parmi les préfectures. L'analyse révèle que l'augmentation de la densité de la population a, au fil des années, eu un effet négatif sur l'efficacité des préfectures de la Grèce sur le plan social.

R esumen Este estudio demuestra una evaluación de las políticas de beneficios sociales y la distribución regional de la inversión pública usando los recientes desarrollos en análisis de eficiencia e inferencia estadística. Específicamente, se compara y analiza la eficiencia de las políticas de beneficios sociales de las prefecturas griegas en los años de censo 1980, 1990 y 2000. El estudio que usa técnicas bootstrap en las aplicaciones de frontera completa incondicionales y condicionales indica que existen importantes ineficiencias en los beneficios sociales entre las prefecturas durante los tres años censados. El análisis revela que el aumento de la densidad de la población a través de los años tiene un impacto negativo en los niveles de eficiencia de los beneficios sociales en las prefecturas griegas.

Keywords:

• Welfare efficiency
• regional policies
• conditional DEA
• bootstrap techniques
• kernel density estimation

Keywords:

• C02
• C14
• R15

Acknowledgements

The authors would like to thank Professor Bernard Fingleton and three anonymous reviewers for their helpful and constructive comments. Any remaining errors are solely the authors’ responsibility.

Notes

1. The data can be retrieved from: http://www.economics.gr

2. All the population data are recorded for the year 2000.

3. For a range of procedural issues on the practical application of DEA, see Dyson et al. (2001).

4. One of the main reasons why our study uses data from the censuses of the last three decades is due to the fact that the effect of some of the inputs used (like the number of public schools per 1,000 students) are observable in the long run.

5. It must be noted, however, that the causality between infrastructure and GDP per capita in the short run and in the long run can also be a bi-directional one (Canning & Pedroni, 2008).

6. Coelli & Perelman (1999) and Coelli et al. (2005) suggest that the choice of orientation has only a minor influence upon the scores obtained. In addition, the two measures provide the same value under the CRS assumption (Coelli et al., 2005, p. 180). Furthermore, Coelli et al. (2005) suggest that the choice of orientation must be made according to which quantities (outputs or inputs) the managers have most control over.

7. Several authors have pointed out the essence of bootstrap techniques as an alternative method of conducting inference where the sample size is not large or sampling distributions are analytically intractable, due to nonlinearity or pretesting, etc. (McCullough & Vinod, 1993, 1998; Halkos & Tzeremes, 2010b).

8. For the theoretical background to the statistical properties of the conditional estimators, see Jeong et al. (2010).

9. For crucial discussion on kernel selection and bandwidth choices, see also Daraio & Simar (2005, 2007a, 2007b) and Bădin et al. (2010).

10. According to Zelenyuk & Zheka (2006, p. 149) the CRS assumption enables us to obtain a greater discriminative power, which may result in larger variation of the regressand (in our case for the second stage of the analysis). In addition, the use of the CCR model implies a common assumption made in economic theory.

11. From the second stage analysis the regions of Thessalonikis, Kerkiras and Attikis have been excluded due to the fact that their population density values are extremely high and they ‘mask’ the overall effect of population density levels on the other regions’ welfare efficiencies. For instance, the regions’ average values without considering those three regions are about 47 residents per km² for every recorded year. In contrast, for the region of Attiki the population density value for the year 1980 is 6,809 residents per km², for 1990 it is 6,781 and for 2000 it is 7,294. The same phenomenon is also reported for the regions of Thessalonikis and Kerkiras.