Aims: We examined the effect of retirement transition on changes in smoking, identified trajectories of smoking around the retirement transition, and investigated factors predicting the membership in the trajectories. Methods: This longitudinal cohort study included 1,432 current or former smokers who entered into statutory retirement in 2000–2011 and who filled out two to four questionnaires sent at four-year intervals. Effect of retirement on smoking was analysed as a non-randomized pseudo-trial in which we compared the likelihood of quitting and relapsing smoking between two subsequent survey waves among those who retired and did not retire. We used latent class analysis to identify trajectories of smoking status and smoking intensity (low: <10 cigarettes/day or high: ⩾10 cigarettes/day), and multinomial logistic regression models to assess pre-retirement factors associated with smoking trajectories. Results: Retirement transition was associated with 1.7-fold odds of quitting smoking (95% confidence intervals 1.3−2.2) compared with no retirement transition. We identified three smoking status trajectories: ‘sustained non-smoking’ (61% of the participants), ‘sustained smoking’ (23%) and ‘decreasing smoking’ (16%). For 489 baseline smokers, we identified three smoking intensity trajectories: ‘sustained high intensity smoking’ (32% of the participants), ‘sustained low intensity smoking’ (32%) and ‘decreasing high intensity smoking’ (35%). Living outside an inner urban area predicted membership in the ‘decreasing smoking’ versus ‘sustained smoking’ trajectory. Conclusions: Smokers are more likely to quit smoking during transition to retirement than before or after it. Characteristics of the smoking environment may affect smoking behaviour around retirement.