ARE THERE ANY FRONTIERS OF RESEARCH PERFORMANCE? EFFICIENCY MEASUREMENT OF FUNDED RESEARCH PROJECTS WITH THE BAYESIAN STOCHASTIC FRONTIER ANALYSIS FOR COUNT DATA

Highlights • Measurement of efficiency of research attracts attention in scientometrics. • Data envelopment analysis (DEA) is mainly used for efficiency analyses. • The stochastic frontier analysis (SFA) will be discussed as an alternative to DEA. • A statistical simulation study shows that SFA works even in the case of count data. • Data of the Austrian Science Fund provides for data to illustrate the proposal. Abstract In recent years, scientometrics has devoted increasing attention to the question of measurement of productivity and efficiency in research. In econometrics, the question is usually examined using data envelopment analysis. Alternatively, in this paper we propose using a statistical approach, Bayesian stochastic frontier analysis (B-SFA), that explicitly considers the stochastic nature of (count) data. The Austrian Science Fund (FWF) made data available to us from their peer review process (ex-ante peer evaluation of proposals, final research product reports) and bibliometric data. The data analysis was done for a subsample of N = 1,046 FWF-funded projects (in Life Science and Medicine, Formal and Physical Sciences). For two outcome variables, a general latent research product dimension (CFACTOR) and the total number of publications (P), technical efficiency values (TE) were estimated for each project using the SFA production functions. The TE values for CFACTOR and P were on average 0.86 and 0.27, as compared with a maximum TE value of 1.0. With regard to CFACTOR, female PIs, younger PIs, and projects with longer durations have slightly higher TE than male PIs, older PIs, and projects with shorter durations. A simulation study showed the statistical behavior of the procedure under different sampling conditions.