We report empirical evidence from the first field experiments to be conducted in Germany with program and control groups between 1999 and 2002. The evaluated program called "Targeted Negative Income Tax (TNIT)" is a time-restricted employee subsidy for means-tested welfare recipients. We focus on a unique data set on welfare recipients in Mannheim and estimate the treatment effect of TNIT on participation probability. The average treatment effect is significant and lies between 6.6 and 6.8 percentage points. Since January 1st, 2005, TNIT can be offered to all means-tested long-term unemployed people in Germany by public case managers.