A model for non-homogeneous DMUs in Data Envelopment Analysis with educational applications

  1. Cecilio Mar-Molinero 2
  2. Fabiola Portillo Pérez de Viñaspre 1
  1. 1 Universidad de La Rioja

    Universidad de La Rioja

    Logroño, España

    ROR https://ror.org/0553yr311

  2. 2 Universitat Politècnica de Catalunya (UPC)
LEER Workshop ‘Efficiency in Education and the use of big data. Book of Abstracts

Publisher: Leuven Economics of Education Research

Year of publication: 2015

Congress: LEER Workshop ‘Efficiency in Education and the use of big data

Type: Conference paper

Institutional repository: lock_openOpen access editor


Data Envelopment Analysis requires all the units that are being compared to behomogeneous. One of the requirements is that all DMUs share the same inputs andoutputs. In practice this is often not the case, as some units may carry out moreactivities than others and may use more inputs to create more outputs than the rest. Inan extreme case, some Decision Making Units may specialize in a single activity whileothers may aim for diversification and perform all possible activities. We present a newmodel that can estimate efficiencies for the various activities, even when some DecisionMaking Units do not engage in some of the activities and do not share the same set ofinputs/outputs with the remaining DMUs. Here we present two versions of the model.The first one, that we call “horizontal integration” deals with activities that take place inparallel. We illustrate it with the analysis of an educational system in which some DMUsperform both teaching and research activities, some DMUs are teaching onlyinstitutions, and some DMUS only carry out research. The second type of model dealswith activities that take place sequentially, and we call this model “vertical integration”.We apply it to the study of primary schools. We address the question of whether it isbetter for children to go to a school from 4 to 11 years of age, or to a school from 4 to7, and change to another school from 7 to 11.