Basing our empirical work on the British New Earnings Survey Panel Data between 1990 and 1996, we show that overtime hours of male workers contain significant individual effects. We also show that using suitable techniques to deal with the lagged overtime variable serves to alter radically the estimated speed of adjustment of overtime to its desired level. Our results are consistent with firms either guaranteeing the length of weekly overtime or following institutionalised custom and practice in their overtime arrangements. They are far less supportive of traditional demand-side analyses of overtime working.