Date created: | Last Updated:

Category: Project

Description: We propose a simple yet comprehensive conceptual framework for the identification of different sources of error in research with digital behavioral data. We use our framework to map potential sources of error in 25 years of research on reputation effects in peer-to-peer online market platforms. Using a meta-dataset comprising 346 effect sizes extracted from 109 articles, we apply meta-dominance analysis to quantify the relative importance of different error components. Our results indicate that 85% of explained effect size heterogeneity can be attributed to the measurement process, which comprises the choice of platform, data collection mode, construct operationalization and variable transformation. Error components attributable to the sampling process or publication bias capture relatively small parts of the explained effect size heterogeneity. This approach reveals at which stages of the research process researcher decisions may affect data quality most and can be used to identify potential sources of error in established strands of research beyond the literature of behavioral data from online platforms.