This study investigates the effect of temporary agency work on the user firm's productivity. We hypothesise that using temporary agency work to enhance numerical flexibility and to screen job candidates may increase productivity, whereas temporary workers' lower firm-specific human capital and spillover effects on the user's permanent employees may adversely affect productivity. Other than the sparse existing literature on this issue, we exploit a large panel data set and control for time-invariant and time-varying unobserved heterogeneity by using the system GMM estimator. We find a robust hump-shaped effect of the extent of temporary agency work on the user firm's productivity.