Malmquist-type indices in the presence of negative data: an application to bank branches

Maria C. A. S. Portela*, Emmanuel Thanassoulis

*Corresponding author for this work

Research output: Contribution to journalArticlepeer-review

73 Citations (Scopus)


In this paper we develop an index and an indicator of productivity change that can be used with negative data. For that purpose the range directional model (RDM), a particular case of the directional distance function, is used for computing efficiency in the presence of negative data. We use RDM efficiency measures to arrive at a Malmquist-type index, which can reflect productivity change, and we use RDM inefficiency measures to arrive at a Luenberger productivity indicator, and relate the two. The productivity index and indicator are developed relative to a fixed meta-technology and so they are referred to as a meta-Malmquist index and meta-Luenberger indicator. We also address the fact that VRS technologies are used for computing the productivity index and indicator (a requirement under negative data), which raises issues relating to the interpretability of the index. We illustrate how the meta-Malmquist index can be used, not only for comparing the performance of a unit in two time periods, but also for comparing the performance of two different units at the same or different time periods. The proposed approach is then applied to a sample of bank branches where negative data were involved. The paper shows how the approach yields information from a variety of perspectives on performance which management can use.

Original languageEnglish
Pages (from-to)1472-1483
Number of pages12
JournalJournal of Banking and Finance
Issue number7
Publication statusPublished - Jul 2010


  • Bank branches
  • DEA
  • Directional distance functions
  • Luenberger indicators
  • Malmquist indices
  • Meta-frontier
  • Negative data in DEA


Dive into the research topics of 'Malmquist-type indices in the presence of negative data: an application to bank branches'. Together they form a unique fingerprint.

Cite this