This paper demonstrates how economic theory can be combined with state-of-the-art empirics to make quantitative statements about optimal policy. Using a large administrative data set for West Germany as well as nonparametric and semiparametric estimation techniques, I parameterize the model proposed by Pavoni and Violante (2005, 2007) for the optimal choice of different financial and activation measures for unemployed workers. The parameterized model is used to study the role of job search assistance in optimal welfare-to-work programs and assess the optimality of the West German policy in the period 2000 to 2002.