The paper evaluates the effects of two labor market programs in Germany, namely the Job Creation- /Structural Adjustment Scheme and Vocational Training, on the unemployment duration of long term unemployed persons. The study uses data from the Mikrozensus Sachsen. A two step Nearest-Neighbor-Matching is employed to solve the sample selection problem. The first step is the estimation of the participation tendency to obtain potential pairs and to compute their Mahalanobis distances. For the assignment of pairs in the second step two different procedures are used: a standard technique and a new one - the iterative improvement of an initial assignment. This process is superior to the standard matching algorithms in the sense that it allows for a closer match between participants and non-participants. Including additional information about a person's employment history enables us to eliminate the bias due to unobservables. The impact of participation in a labor market program is evaluated by comparing the unemployment duration between both groups using the Cox Proportional Hazard Model. Overall we find empirical evidence that both participation in Job Creation- /Structural Adjustment Scheme and Vocational Training result in even longer unemployment.