Using DEA and worst practice DEA in credit risk evaluation

TY - JOUR

T1 - Using DEA and worst practice DEA in credit risk evaluation

AU - Paradi, Joseph C.

AU - Asmild, Mette

AU - Simak, Paul C.

PY - 2004/3

Y1 - 2004/3

N2 - The purpose of this paper is to introduce the concept of worst practice DEA, which aims at identifying worst performers by placing them on the frontier. This is particularly relevant for our application to credit risk evaluation, but this also has general relevance since the worst performers are where the largest improvement potential can be found. The paper also proposes to use a layering technique instead of the traditional cut-off point approach, since this enables incorporation of risk attitudes and risk-based pricing. Finally, it is shown how the use of a combination of normal and worst practice DEA models enable detection of self-identifiers. The results of the empirical application on credit risk evaluation validate the method. The best combination of layered normal and worst practice DEA models yields an impressive 100% bankruptcy and 78% non-bankruptcy prediction accuracy in the calibration data set, and equally convincing 100% and 67% out-of-sample classification accuracies.

AB - The purpose of this paper is to introduce the concept of worst practice DEA, which aims at identifying worst performers by placing them on the frontier. This is particularly relevant for our application to credit risk evaluation, but this also has general relevance since the worst performers are where the largest improvement potential can be found. The paper also proposes to use a layering technique instead of the traditional cut-off point approach, since this enables incorporation of risk attitudes and risk-based pricing. Finally, it is shown how the use of a combination of normal and worst practice DEA models enable detection of self-identifiers. The results of the empirical application on credit risk evaluation validate the method. The best combination of layered normal and worst practice DEA models yields an impressive 100% bankruptcy and 78% non-bankruptcy prediction accuracy in the calibration data set, and equally convincing 100% and 67% out-of-sample classification accuracies.

KW - Credit risk

KW - Data envelopment analysis

KW - Layering or peeling technique

KW - Worst practice DEA

U2 - 10.1023/B:PROD.0000016870.47060.0b

DO - 10.1023/B:PROD.0000016870.47060.0b

M3 - Journal article

AN - SCOPUS:3843113429

SN - 0895-562X

VL - 21

SP - 153

EP - 165

JO - Journal of Productivity Analysis

JF - Journal of Productivity Analysis

IS - 2

ER -