We develop a matching model on the marriage market, where individuals have preferences over the smoking status of potential mates, and over their socioeconomic quality. Spousal smoking is bad for non-smokers, but it is neutral for smokers, while individuals always prefer high socioeconomic quality. Furthermore, there is a gender difference in smoking prevalence, there being more smoking men than smoking women for all education levels, so that smoking women and non-smoking men are in short supply. The model generates clear cut conditions regarding matching patterns. Using CPS data and its Tobacco Use Supplements for the years 1996 to 2007 and proxing socioeconomic status by educational attainment, we find that these conditions are satisfied. There are fewer "mixed" couples where the wife smokes than vice-versa, and matching is assortative on education among couples with identical smoking habits. Among non-smoking wives those with smoking husbands have on average 0.14 fewer years of completed education than those with non-smoking husbands. Finally, and somewhat counterintuitively, we find that among smoking husbands those who marry smoking wives have on average 0.16 more years of completed education than those with non-smoking wives.