What is Manning et al's concept of the "Nonsmoking smoker". Why is it useful? How is it developed? Why may it not be accurate? How do the authors deal with the potential inaccuracy?
hypothetical comparison group who do not smoke but who are like smokers in a number of different key variables that could influence costs.
control variables include age, sex, race, education, drinking habits, exercise habits, family size, income, self-assessed measures of physical, mental and general health and seat belt use.
estimate an upper bound for the differences in cost