Using rich longitudinal matched employer-employee data for Belgium, we provide a first investigation of the impact of sickness absenteeism on firms' productivity. To do so, we estimate a production function augmented with a firm-level measure of sickness absenteeism that we constructed from worker-level information on nonworked hours due to illness or injury. We deal with the endogeneity of inputs and sickness absenteeism by applying a modified version of the semiparametric control function method developed by Ackerberg, Caves, and Fraser (2015), which explicitly takes firm fixed unobserved heterogeneity into account. Our main finding is that, in general, sickness absenteeism substantially dampens firms' productivity. However, further analyses show that the impact varies according to several workforce and firm characteristics. Sickness absenteeism is more detrimental to firm productivity when absent workers are high tenure or blue collar. Moreover, it is especially harmful to industrial, capital-intensive, and small enterprises. These findings are consistent with the idea that sickness absenteeism is more problematic when absent workers have in-depth firm-/task-specific knowledge, when the employees' work is highly interconnected (e.g., along the assembly line), and when firms face more organizational limitations in substituting absent workers.