Estimation of the efficiency of structural funds: a parametric and nonparametric approach

Abstract

This article seeks to perform an analysis of the efficiency of the application of the structural funds in the regions classified as Objective 1 over the period 2000 to 2006, applying the techniques known as Data Envelopment Analysis (DEA) and stochastic frontier. In the first place, we are going to identify which are the most efficient regions. Finally, we will analyse the extent to which certain factors have repercussions on the efficiency such as country, geographical location and contribution of agriculture of Gross Domestic Product (GDP) and the result indicates that some of these variables have significantly influences of technical efficiency.

Keywords:

• regional policy
• structural funds
• efficiency
• DEA
• stochastic frontier

JEL Classification::

• C14
• H21
• O18
• H83

Notes

¹ A detailed literature based on DEA can be consulted at www.banxia.com/frontier/bbliography.html Analysis of the efficiency of a set of entities with DEA involves having software capable of solving the problem addressed. Free software that includes was written by Coelli (1996): DEAP and EMS, Scheel (2000): EMS. These can be downloaded from: www.uq.edu.au/economics/cepa/soltware.htm and www.wiso.unidortmund.de/lsfg/or/scheel/ems Another, commercially available program is professional Frontier Analyst (www.banxia.com); DEA Solver Pro (www.saitech-inc.com), DEA Frontier (Zhu, 2002) and Warwic-DEA.

² The main developments of the estimation methods of productive efficiency were put forward by Farrell (1957) in the Discussion section of his paper. Forsund (1999) addresses the original ideas therein and establishes connections with the determinist parametric approach, the stochastic approach and DEA.

³ See Coelli et al. (1998)

⁴ Alternatively in the studies Kopp (1981), Russell (1985) and Zieschang (1984) indexes are proposed which do not require efficiency to be measured radically.

⁵ This can be expressed by a production function of the type, for example, Cobb–Douglas or Translog.

⁶ The model can also be estimated by Corrected Least Squares (COLS) (Richmond, 1974). For a detailed review of the various methods of estimation of the stochastic frontier function, see Pastor (1995).

⁷ See Greene (1993) for the expressions for the exponential and gamma distributions.

⁸ According to Coelli (1996), this test should be applied when maximum likelihood estimation is used because, in this case, the test is of the right size.