Oxford Economic Papers 45 (1993), 332-347 MEASURING THE RESEARCH PERFORMANCE OF UK ECONOMICS DEPARTMENTS: AN APPLICATION OF DATA ENVELOPMENT ANALYSIS By GERAINT JOHNES and JILL JOHNES 1. Introduction FINANCIAL pressures imposed upon non-profit public sector organisations have, in recent years, led to a rapid expansion of interest in the measurement of the performance and efficiency of such bodies. Particular attention has been devoted to the problem of how, in the absence of market prices, to aggregate across heterogeneous inputs and outputs. On one level, this has led to the development of performance indicators (PI), each of which attempts to measure the output (input) of a group of nearly homogeneous products (factors of production). Such Pis include patient discharges in the health service, students' examination results in schools, and publication counts in higher education. On another level, the aggregation of various measures of performance poses problems which have also been the subject of much research. The difficulty of aggregating across Pis is accentuated by the lack of data concerning the weight that should be applied to each measure of performance; in the not-for-profit sector, market prices are not present to guide us. Since overall measured performance can be very sensitive to the weight attached to each individual PI (Johnes, 1990) the choice of loss function should not be determined arbitrarily. Recent developments in the field of linear programming—in particular data envelopment analysis— enables light to be shed on this issue. In the present paper we investigate both of these issues. Our primary concern is the assessment of research performance of university departments of econo- mics in the UK over the period 1984-88. It should be stressed, however, that the methods used, and many of the qualitative conclusions drawn, are likely to be of importance in a wide variety of other contexts. In particular, we address questions concerning the stability of efficient weighting schemes across alterna- tive model specifications, and consider how the techniques of cluster analysis can be used to reduce to manageable proportions a bewilderingly large array of Pis. The period covered by the study reported here coincides exactly with that used in the second research selectivity exercise conducted by the Universities Funding Council (UFC) in 1989. Relatively little is known about the decision processes used in the UFCs peer review to assess research performance across university departments of economics. We know, however, that chapters in books were assigned relatively low weight, while articles in core journals were considered substantial contributions (Royal Economic Society, 1989). For two © Oxford Unheraity Press 1993 at Lancaster University on December 4, 2014 http://oep.oxfordjournals.org/ Downloaded from