Much of the current discussion about the Transatlantic Trade and Investment Partnership (TTIP) is focused on the potential welfare and employment effects. Supporters of TTIP often support their argument by highlighting the optimistic results of computable general equilibrium (CGE) models. CGE-models are the methodological backbone of most ex-ante impact assessments of free-trade agreements, as for instance published by the European Commission. The objective of this paper is to assess the accurateness of ex-ante studies by scrutinizing the example of the North American Free Trade Agreement (NAFTA). The analysis suggests that a considerable gap exists between ex-ante projections and ex-post evaluations with regard to NAFTA's effects on welfare, wages and employment. Most exante models had a tendency to overestimate the benefits and underestimate the costs of free-trade. The experience of NAFTA reveals the weak credibility of ex-ante simulations. Policy makers should thus treat the formers' results with the appropriate skepticism.