Using a German employer-employee matched panel data set this paper examines the effects of High Performance Workplace Systems (HPWSs) on labor productivity (defined as sales per worker) and labor efficiency (defined as the inverse of unit labor costs). The estimation results indicate that simple cross-sectional estimates of the effects of implementing HPWSs on labor productivity are biased downward due to unobserved time-invariant establishment effects and the endogeneity of the used measure for innovative workplace practices. The latter bias appears to be quantitatively more important. Results from estimating a correlated random coefficient model further suggest that a potential bias in the 2SLS-estimates due to self-selection seems to be negligible. The estimated effects of HPWSs on labor productivity are economically important and rising over time. However, corresponding positive effects of HPWSs on labor efficiency occur only in the long run. Finally, due to rising wages associated with the adoption of HPWSs, the effects of these systems on labor efficiency are smaller than the corresponding effects on labor productivity.