Finding Doppelgängers in Scopus: How to Build Scientists Control Groups Using Sosia

Abstract

The construction of control groups of scientists is often a daunting effort. This paper presents sosia, an open-source Python-based software designed to query efficiently the Scopus database via RESTful API. sosia searches for researchers with publication profiles similar to a given researcher up to a given year based on all main standard bibliometric indicators. The user can choose flexibly a set of parameters to restrict the search to more or less narrow boundaries upfront and obtain additional similarity indicators to select a subset of authors after the search. Advanced settings also allow to narrow the search to a list of affiliations and to minimize the possible errors arising from ambiguous author profiles. One basic search can be set up in a few command lines and the average time of computation goes between 60 and 300 minutes. We discuss the functioning, characteristics, limitations and possible extension of the software.

Keywords: Statistical Software, Control Group, Diff-in-Diff, Scopus

JEL Classification: C00, A14