This paper provides a theoretical and quantitative analysis of various types of wellknown employment subsidies. Two important questions are addressed: (i) How should employment subsidies be targeted? (ii) How large should the subsidies be? We consider measures involving targeting workers with low incomes/abilities and targeting the unemployed. To make our analysis particularly useful to policy makers, we focus on policies that are "approximately welfare efficient," i.e. policies that (a) improve employment and welfare, (b) do not raise earnings inequality and (c) are self-financing. This criterion enables us to identify policies which satisfy these favorable properties and to determine the size of the subsidies required for this purpose. We construct a simple, dynamic model of hiring and separations, derived from microfoundations, and calibrate it with German data. The calibration shows that hiring vouchers targeted at the long-term unemployed and low-income/ability workers can be approximately welfare efficient, while low-wage subsidies do not satisfy this criterion. Even in terms of inequality reduction low-wage subsidies are outperformed by targeted hiring vouchers. Furthermore, hiring vouchers targeted at the long-term unemployed are more effective than hiring vouchers targeted at low-income/ability workers. These subsidy rankings also hold if the self-financing constraint is relaxed and the government spends a given additional amount on the subsidies.