This paper compares two different ways of doing policy evaluation: on the one hand, quasi-experimental methods (or "ex-post" evaluations) which exploit the introduction of a reform and identify its effect by comparing treated and untreated individuals; on the other hand, structural models (or "ex-ante" evaluations) which are based on economic theory and predict the effect of potential reforms by using the estimates of behavioural parameters. The comparison is carried out using an empirical case. In 1998, in Norway, a major welfare reform changed the rules of the most generous benefit for lone parents: it increased the amount of the benefit and introduced working requirements. Using a quasi-experimental evaluation approach, it is found a positive effect of the reform on lone mothers' employment. In this paper, I estimate a static structural model of work and welfare participation decisions and compare the results using the two different approaches. Despite the differences in the assumptions I make for the two models, results are fairly comparable.