"We investigate the relationship between employees' and managers' training and firm performance using a policy intervention that randomly assigned training support to small- and medium-sized enterprises in the UK accommodation and food service sector. Because the number of firms self-selected into training exceeded available places, training was randomly assigned to some firms, resulting in a randomized natural experimental design that allowed us to identify the average effect of training on treated firms. Our empirical results suggest that employees' training had a stronger positive impact on firms' labour productivity and profitability than that of managers'."