We investigate the link between death rates and smoking prevalence in ten developed countries with the aim of using smoking prevalence data to explain differences in country-specific death rates. A particular problem in building a stochastic mortality model based on smoking prevalence is that there are in general no separate mortality data for smokers and non-smokers available. We show how we can estimate mortality rates for smokers and non-smokers using information about the smoking prevalence in a number of developed countries, and making an additional assumption about the death rates of smokers which is motivated by empirical findings for British Doctors obtained by Doll et al. (2004). We consider this empirical investigation to be the first step towards a consistent mortality model for multiple populations, which will require modelling of country specific differences in mortality, as well as non-smokers' and smokers' mortality rates.