Poisson regression models were used to assess trends in smoking behaviour before and after the policy changes. From 2005 to 2011, the prevalence of smoking declined among high-school students by 6.8% per year compared with 3.6% decline per year in the control group.