Recent economic events have amplified the importance of considering the appropriateness of different modelling traditions as the basis of policy advice. In the context of innovation policy, the appropriateness of agent-based modelling is explored relative to alternative modelling approaches. Validation, which is the process of evaluating the empirical adequacy of models, is raised as an important but complex factor in evaluating policy models. Analysis of validation generally supports claims that descriptive models are subject to the strongest validity tests. However, the validity-centric approach is criticized as creating models of insufficient scope to be useful for policy. Mid level policy modelling is suggested as an underexploited niche which allows sufficient scope to be useful in policy contexts while still maintaining links to empirical methods. This analysis confronts the idea that an increased focus on the validation of models is all that is required to better support policy decisions.