Background and Aims The association between depression and smoking is firmly established, but how the association develops remains unclear. The aim of this study was to examine development of the smoking–depression association from early adolescence to adulthood. Design Cross‐sectional and longitudinal analyses of the smoking–depression association from adolescence to adulthood. Setting Hordaland, Norway. Participants A cohort of adolescents (initially, 924 pupils) in the Norwegian Longitudinal Health Behaviour Study (NLHB) was followed over nine data collection waves from ages 13 to 30 years. Measurements Daily smoking and depressed mood were measured in each wave. Findings In the cross‐sectional analyses, daily smoking and depression were significantly associated (P‐value range from P < 0.01 to 0.04) in eight of nine waves. In the final longitudinal model, after controlling for the time‐invariant effects of smoking and depression and of tertiary factors, the only significant paths were early adolescent smoking prediction of early adolescent depression (waves 1–2: β = 0.07, P < 0.05; waves 2–3: β = 0.12, P < 0.05) and vice versa (waves 1–2: β = 0.10, P < 0.05; waves 2–3: β = 0.08, P < 0.05). Conclusions The inter‐relationship between depression and smoking seems to be due to the reciprocal causal effects between smoking and depression that are established in early adolescence and maintained into adulthood.