Public Employment Service (PES) units and caseworkers fundamentally shape the treatment of individual job seekers by applying specific strategies (mixes) of labor market policies. The fact of the joint existence of several types and levels of policies is usually ignored in policy evaluations. This paper applies a novel approach to jointly model two types and two layers of policies (ex-ante) and of treatments (ex-post). It empirically assesses the role of case worker and of PES policies for the job seekers earnings and employment outcomes in the 3.5 years after unemployment entry. We exploit the substantial variation in (the intensity of) policy use across caseworkers and PES agencies in Switzerland, relying on a large register data base covering a fourth of the full unemployment inflow from 2000 to 2005. We propose a method to estimate the (unknown) intended policies by types (carrots and sticks) using actual treatment realizations. Then, we relate these policies to the short- and longer-run outcomes of the individuals. The joint estimation reveals that the intended policies, for carrots and for sticks, both have significant impact on unemployment exit, earnings, and employment stability. Their interaction is of importance too: It rejects the hypothesis that earnings losses by an intense sanction regime could be compensated by intensifying training. The negative sign of the interaction effect on earnings suggests to keep either the intensity of supportive policies or of punitive policies low. Joint with Rafael Lalive and Gerard van den Berg.