CGE models are widely used tools for economic assessments of trade policy changes. However, overall confidence in their results tends to be low. We employ the methodological framework of meta-analysis in order to approach a quantitative comparison of CGE-based simulation results. Therefore, we compile a dataset of twelve recent Doha simulations and fit a linear regression model that explains the variance between simulation results on the regional level as a function of various modeling characteristics. The estimates are broadly in range with documented qualitative knowledge about modeling assumptions. The size of the sample limits general conclusions; however, an ongoing research project will extend the approach to a larger sample.