Speaker
Description
Publication bias is the prioritized and selective reporting of scientifically significant results. In contrast to the widespread assumption that this bias arises primarily from editorial desk-rejections, recent research indicates that authors themselves decide much more frequently not to publish or submit their insignificant results. This can mean that a) researchers only submit significant results without mentioning non-significant results (= selection bias at the hypothesis level), or b) researchers refrain from writing up and submitting studies yielding non-significant results altogether (= file drawer bias at the output level). This can be problematic because time and effort are invested in repeating research that has already been carried out, but never reported. In addition, the knowledge gained from non-significant results (by assuming the null hypothesis) is lost. In psychology, the effects of this selective pressure became evident in the "replication crisis".
The prevalence of file drawer bias in studies conducted within a representative US panel infrastructure has shrunk from roughly 65 per cent between 2002 – 2012 (Franco et al., 2014) to 28 percent between 2012 – 2018 (Moniz et al., 2023). While this bodes well for the open science movement, no comparable meta-analytic approach has investigated the prevalence of publication bias in German panel infrastructures. In our project, we investigate different aspects of publication bias using two probabilistic representative German panels (GESIS Panel and SOEP-IS, 2012 -2021). This talk will cover first insights derived from the comparative analysis of successful study proposals and their subsequent publications.
