Depending on data source, estimates of hours of work give widely different results both as to level and change. In this paper three alternative measures of hours worked are used to estimate a simple labour supply function to investigate if estimated wage rate and income effects are data dependent as well. The measures used include those from time-use surveys and those from regular surveys. The latter are based on the responses to a question about normal weekly hours of market work. The results suggest that estimates of the wage rate effects become much smaller when measures of normal hours are used compared to data collected for a well-defined time period close to the date of interview, such as time-use data. The income effects appear less sensitive to the choice of data.