A non-parametric efficiency and productivity analysis of transition banking

This thesis examines banking efficiency and the productivity of thirteentransition Central and Eastern European banking systems during 1998-2003 usingData Envelopment Analysis (DEA). It proposes a non-parametric methodology fornon-radial Russell output efficiency measure of banking firms, incorporating risk asan undesirable output. In addition, the proposed efficiency measure handlesunrestricted data, i. e. both positive and negative. The Luenberger productivity index issuggested, which is applicable to technology where the desirable and undesirableoutputs are jointly produced, and are possibly negative. Furthermore, the thesisaddresses the main issue in the literature on banking performance measurement,which concerns the lack of consistency in the conceptual and theoreticalconsiderations in describing the banking production process. Consequently, a metaanalysistool, to examine the choice of inputs and outputs definitions in the bankingefficiency literature, is suggested. In addition, the performance measures areestimated using three alternative definitions of the banking production processfocusing on the risk and environmental dimensions of bank efficiency andproductivity, with further comparative analysis using bootstrapping and kernel densitytechniques. Overall, the empirical results suggest that in Central and Eastern EuropeCzech, Hungarian and Polish banks were the most technical efficient banks and thebanking risk was mainly affected by external environmental factors during theanalyzed period. Productivity analysis implies that the main driver of productivitychange in the Central and Eastern European banks is the technological improvement.As meta-analysis revealed, the choice of particular approach of describing the bankingproduction process is determined not by the availability of particular input or outputvariable information but the concepts of researcher's theoretical considerations.Statistical tests and density analysis indicate that efficiency scores, returns parametersand productivity indexes are sensitive to the choice of particular approaches.