OBJECTIVE To identify risk factors for culling of dairy cows from eight New South Wales dairy herds. METHODS A longitudinal population study of dairy cow culling in eight non-seasonally calving dairy herds in the Camden district of New South Wales. Cox's proportional hazards model was used to evaluate various risk factors for culling for a specific reason (sales, deaths, reproductive failure, disorders of the udder and low milk production). RESULTS Age at first calving was not a significant risk factor for culling. Milk production in the first lactation greater than the population mean did not influence length of productive life overall, but was associated with a greater hazard of removal for disorders of the udder. Risk of culling for reproductive failure differed significantly between farms, and was not related to events in the previous lactation such as calving-to-first service interval or calving-to-conception interval. Shorter calving intervals were associated with increased risk of removal for low milk production and disorders of the udder. CONCLUSIONS Longitudinal surveys to accurately identify reasons for removal from a wide range of herds, identification of herds with low culling rates (especially for reproductive failure and udder disorders), and the identification of practices associated with these culling rates would be worthwhile to the Australian dairy industry.