Computable general equilibrium (CGE) models are widely used to conduct ex-ante policy impact evaluations. However, in addition to policy design and policy stringency, structural features of the CGE models also affect the resulting estimates of policy costs. We use harmonized policy analysis results from 15 CGE models and meta-regression analysis to identify the structural variables that are significant determinants of the global and regional marginal abatement costs (MAC) for fulfilling the initial Nationally Determined Contributions (NDCs). Our results show that models with dynamic characteristics, higher regional disaggregation and with a representation of different electricity technologies estimate higher MACs. On the contrary, modelling endogenous technological change lowers the MAC estimates. Additionally, as to policy design, a statistically significant reduction in global MAC is observed with a fully linked global carbon market (45% reduction) and a climate-club of China, Japan and South Korea (4% reduction). This meta-analysis provides robust quantitative insights can help to provide a more useful and understood tool to inform policy relative to the results from a single model.