Numbers for Pascal: explaining differences in the estimated benefits of the Doha Development Agenda

Economists use partial and general equilibrium trade simulation models to estimate the impact of changes in domestic policies and international trade rules. During the WTO Doha Development Agenda (DDA) negotiations economists have produced many different estimates of the gains that would result from global trade liberalisation scenarios. However, these estimates differ quite widely even for apparently similar liberalisation scenarios. The result is confusion about the true magnitude of the gains from trade liberalisation, and a reduction in the perceived credibility of the theories and models that economists use. We apply meta-analysis to a dataset extracted from 110 studies that present simulated assessments of global trade liberalisation scenarios under the DDA. Initial meta-regression analysis demonstrates that covariates that capture model characteristics, the nature of the data used in the modelling exercise, and the nature of the simulated liberalisation scenarios can explain roughly one-third of the variance in the dependent variable ‘simulated global welfare change’. We test whether additional explanatory power can be obtained by adding information about the authors of the simulation studies. We find significant fixed effects for the top 20 authors in the field. We interpret this as evidence that leading authors in the field employ model specifications that reflect their individual preferences and beliefs about how economies function and the impact of liberalisation, specifications that are hidden in the complex interactions of simulations models and therefore difficult to capture in a meta-analysis. We use these results to generate a confidence interval for the gains that would result from trade liberalisation under the DDA.