In a working paper on scientific collaboration Carillo, Papagni, & Sapio (June 2012) say:
. . . what really matters for enhancing the quality of research are the flows of knowledge that arise from collaborations among researchers from different institutions and/or countries.
In this paper, we analyse the effects of research collaborations on the scientific output of academic institutions, drawing on data from the first official Italian research assessment exercise. We measure the scientific performance of a research unit as the number of publications that received an excellent grade in the evaluation process. Different aspects of scientific collaboration are taken into account, such as the degree of openness of a research team towards other institutions and/or other countries, the frequency of co-authorships, and the average size of a collaborating team. Using econometric models for count data, we find that collaborations are more effective when they imply knowledge exchange resulting from collaboration with external or foreign colleagues, are very frequent, and the collaborating teams have a small size.
The title is "Do collaborations enhance the high-quality output of scientific institutions? Evidence from the Italian Research Assessment Exercise (2001-2003)."