This paper considers the impact of tobacco consumption on wages in the UK using data from fifteen waves of the British Household Panel Survey. Considering both overall smoker status as well as the number of cigarettes consumed, we provide estimates for the smoking wage penalty using standard regression methods, including panel estimators for fixed effects and panel instrumental variable estimators. Furthermore, we analyse the impact of stopping and starting to smoke relative to permanent smokers and non-smokers by Mahalanobis-matching. In the cross-section, we find a rather large wage penalty for smokers of about 4%. However, panel estimator and IV results show relatively few support for hypotheses linking the smoking wage penalty to either lower productivity of smokers, be it health related or not, or discrimination. Matching results suggest that starting or stopping to smoke does not affect later earnings relative to remaining either smoker or non-smoker.