In spring 2006, the Danish government introduced a policy that required married long-term social assistance recipients to work 300 hours in non-subsidised employment during a 2-year period in order to remain eligible for benefits. The intention was to ‘make work pay’ for unemployed immigrant women. This study evaluates how this policy influenced their transitions to employment and other benefit schemes by applying a competing risk duration model on Danish administrative data. The results show that the new rules not only had the anticipated impact on the women concerned, but the policy also had an unforeseen effect on the work effort of social welfare workers. Many moved their clients to other benefit schemes instead of applying the sanctions. This was particularly the case for the weakest among the unemployed and in municipalities, which were headed by left-wing mayors or that received extra funds from the central government. Although the new policy did increase transitions to employment, this effect was stronger in municipalities that provided extra support.