Around 40% of the male workforce regularly works 8 to 9 hours a week of paid overtime. This paper investigates the determinants of overtime hours in Britain over the period 1975-1999. For this purpose a panel data Tobit model is estimated using the very large panel of employees from the National Earnings Survey Panel Dataset. The empirical results show that changes in the job-mix across the economy, from high to low overtime jobs rather than within-job changes in the use of overtime, account for most of the apparent decline in the extent of overtime working over the 1990s. Within jobs, the GDP cycle has a significant impact on overtime work, while labour market conditions, represented by the unemployment rate, do not. The elasticity of total working hours with respect to wages is found to be close to zero and with respect to contractual hours close to unity. Furthermore the results show that the decline of unionisation has not altered the use of overtime.