OBJECTIVE: To explore the variation in hospital caesarean section (CS) rates for nulliparous women, to determine whether different case-mix, labour and delivery, and hospital factors can explain this variation and to examine the association between hospital CS rates and outcomes. DESIGN: Population-based cohort study. SETTING: New South Wales, 2009-2010. POPULATION: Nulliparous women with singleton cephalic live births at term. METHODS: Random effect multilevel logistic regression models using linked hospital discharge and birth data. MAIN OUTCOME MEASURES: Prelabour and intrapartum CS rates following spontaneous labour or labour induction; maternal and neonatal severe morbidity rates. RESULTS: Of 67 239 nulliparous women, 4902 (7.3%) had a prelabour CS, 39 049 (58.1%) laboured spontaneously, and 23 288 (34.6%) had labour induced. Overall, there were 18 875 (28.1%) CSs, with labour inductions twice as likely to result in an intrapartum CS compared with women with a spontaneous onset of labour (34.0% versus 15.5%). After adjusting for differences in case-mix, labour and delivery, and hospital factors, the overall variation in CS rates decreased by 78% for prelabour CSs, 52% for intrapartum CSs following spontaneous labour and 9% following labour induction. Adjusting for labour and delivery practices increased the unexplained variation in intrapartum CSs. The adjusted rates of severe maternal and neonatal morbidity were not significantly different across CS rate quintile groups, except for women in spontaneous labour, where the hospitals in the lowest CS quintile had the lowest neonatal morbidity rate. CONCLUSIONS: Differences in clinical practice were substantial contributors to variation in intrapartum CS rates. Our findings suggest that CS rates in some hospitals could be lowered without adversely affect pregnancy outcomes.