The purpose of this paper is to test forward-looking incentives against backward-looking incentives. Design/methodology/approach ? Wage growth model to estimate forward-looking effects of unpaid overtime and a probit model of participation in unpaid overtime controlling for excessive pay to estimate backward-looking effects. The authors use data form the OSA labour supply panel (years 1994, 1996 and 1998). Findings ? The importance of backward-looking incentives is demonstrated in an empirical analysis of participation in unpaid overtime. The authors show that employees who have relatively good wages now or who have had relatively good wages in the recent past participate more often in unpaid overtime. The authors also show that participation in unpaid overtime does not lead to extra wage growth. Research limitations/implications ? These results imply that involvement in unpaid overtime is to be explained from backward-looking incentives, not from forward-looking incentives. The paper concludes that backward-looking incentives deserve more attention in the economic literature, especially as they are well-accepted as work motivation devices by employees. Limitations are the length of the panel study (four years) and the fact that the data are restricted to one country (the Netherlands). Social implications ? Personnel policies should focus more on the intrinsic motivation of personnel rather than on extrinsic motivation. Originality/value ? This is the first paper to test both forward- and backward-looking incentives simultaneously.