The Effects of Size on Research Performance: A SPRU Review

CONTENTS Executive Summary ii Acknowledgements v A. Introduction 1 B. Methodological Issues 2 B.1 Unit of Analysis 2 B.2 Executive Summary i Opinions as to the relative merits of big or small units for carrying out research have varied over time, counterposing arguments for integration and breadth against those favouring specialisation, diversity and competition. In research, just as in production, there is a common if disputable presumption that larger units obtain 'static economies of scale' (disproportionately greater output – i.e. greater output per unit input) and 'economies of scope' (from the synergy derived from conducting several related activities in parallel). Set against this, smaller units may reap 'dynamic' economies of scale through greater agility and responsiveness to change. ii Recent analyses of 'knowledge production' have suggested a trend from 'Mode 1' research (i.e. research conducted by individuals or teams often drawn from a single discipline and working within a single organisation) to 'Mode 2' research (i.e. research involving greater interaction between a widening range of knowledge producers in different disciplines, institutions, sectors and countries). This has been presumed by some to favour consolidation, but the logic of the argument is actually far more complicated. iii Different forms of scale economies may arise according to the level of the unit of analysis on which one chooses to focus. This review has examined studies at the macro level of entire countries or regions, the meso level of institutions (e.g. universities) or their constituent departments, and the micro level of particular groups, teams or collaborations. iv Measuring the output of the relevant unit is complicated for two main reasons: the diversity of 'products' created (e.g. teaching as well as research in the case of universities); and the need for the assessment to take account of the quality as well as the quantity of output. Bibliometric and other measures aim to adjust for quality in a variety of ways, all of which have considerable limitations in practice. The most promising approach involves the construction of multiple measures, as pursued in Data Envelopment Analysis (DEA) for instance, and this should be encouraged in future studies. v The problem of the diverse range of 'products' is often tackled by adopting an econometric approach where the aim is to reduce the items to a common denominator. However, the various studies to date suffer from the usual econometric problem of oversimplifying causal relationships ('simultaneity biases') and from the …