Workers who use computers earn more than those who do not. Is this a productivity effect or merely selection? Using the Canadian Workplace and Employee Survey, we control for selection and find a wage premium of 3.8% for the average worker upon adopting a computer. This premium, however, obscures important differences in returns to computer adoption across education and occupation groups. We find that long-run returns to computer use are over 5% for most workers. Differences between short-run and long-run returns may suggest that workers share training costs through sacrificed wages.