A large number of research articles have looked into data sharing and other ways of participating in open science practices. These articles have in common that willingness or past open science practice are usually measured as self-reports in a survey. We define open science practices as either sharing self-collected data or referencing re-used data in a published paper. We use a unique dataset of social science researchers with which we can identify whether open science practices have been actually carried out, rather than self-reported willingness or past behavior. With this dataset we are also able to distinguish between impact of journal characteristics and personal characteristics on actual open science practices. Doing so, we apply multilevel regression to account for the nested structure of the data. Multilevel analysis is usually recommended when a significant part of the variance is explained by between-cluster differences. We found that this is the case for open science practices of authors in different journals. In addition, we have access to more than 500 open answer statements that we categorize and analyze for those participating and for those not participating in open science practices. Both analyses will provide evidence of incentives and barriers to actual open science practices in the social sciences.
Perry et al. (Tue,) studied this question.