Quantification of welfare changes due to trade liberalisation play a crucial role for political decision making. However, meaningful comparisons of simulation results from different sources are difficult. Often significant differences in simulated gains from liberalisation do not serve to increase confidence in quantitative assessments based on trade models. We employ a metaanalysis of applied trade simulations under the WTO Doha Round to identify model characteristics that influence the magnitude of simulation results, and to estimate the magnitude of this influence. Findings from our simple econometric model are plausible and show that each simulation experiment represents a complex interaction of experimental settings that may not easily be understood by and communicated to non-experts. Meta-analysis proves to be a useful tool for empirically assessing this complexity.