How Should We Measure Individual Researcher'S Performance Capacity Within And Between Universities - Social Sciences As An Example? A Multilevel Extension Of The Bibliometric Quotient (Bq)

The assessment of individual research performance has become a major attraction for bibliometric researchers in recent years, and is dominated by the classic bibliometric indicator approach (e.g., h-index). Alternatively, a psychometric measurement approach is favored, which considers measurement errors. It is assumed that the "researcher's performance capacity" as a personal trait and competency is responsible for the individual research performance, which might vary randomly due to measurement errors. Five individual-level bibliometric variables served as items (e.g., number of articles in top 5%) to measure the competency. The central question of this contribution is how much variance in the "researcher's performance capacity" is explained by differences between universities/subfields. With bibliometric data (Scopus) for a sample of 1,071 social scientists with Swiss university affiliations a one-dimensional scale ("Bibliometric Quotient", BQ) was created by means of a psychometric model, which has a high, but not perfect, reliability of r(tt) =.84. The items were most suitable for scientists scoring above average. About 33% of the variance of the BQ is due to differences between the universities/subfields, and only 7% of the variance is due to differences between universities alone. A ranking only of Swiss universities in the social sciences does not necessarily make sense.