To measure education, the Eurobarometer survey asks respondents at what age they finished school. The responses are coded as (0) still studying, (1) no education and age 13 and under, (2) age 14, (3) age 15, (4) age 16, (5) age 17, (6) age 18, (7) age 19, (8) ages 20-21, (9) ages 22-24, and (10) ages 25 and older. Another variable that equals 0 for those still studying and 1 for all others complements the education variable. Controlling for both variables shows the effect of completed education independent of those still studying.⁹ This measure of age of finishing school aims to avoid problems of comparability across diverse education systems that affect measures of formal degrees. However, the age measure may overstate the attainment of those finishing at a later age because of slowness and problems in school rather than advanced degrees.

The measure of occupation uses the EB classification of current or last job in the following categories: (0) no job; (1) farmer or fisherman; (2) unskilled manual, servant; (3) skilled manual; (4) service—hospital, restaurant, police; (5) supervisor; (6) shop owner, craftsman, self-employed; (7) traveling—salesman, driver; (8) work mainly at a desk; (9) middle management—department head, junior manager, teacher, technician; (10) business proprietor, partner or full owner of a company; (11) general management, director, top management; and (12) professional—lawyer, doctor, accountant, architect. The measure treats the categories as a continuous scale, and, given the diverse mix of occupations in some of the categories, the ranking has some arbitrariness. However, the measure relates closely to smoking, and rearranging categories (4 and 7, for example) does little to change the results. Much as for education, a second occupational variable that equals 1 for those with a current or former job and 0 for those never having done any paid work complements the occupation measure.

To measure economic standing, the surveys ask about ownership of goods rather than income. A scale based on the proportion of the following goods owned by the respondent has an alpha reliability of .764: household phone, mobile phone, television, DVD player, music CD player, computer, Internet connection, car, and paying for an apartment or house. Given reporting errors common in usual income measures, goods-based measures do better to predict smoking (Schaap et al., 2008).

Other control variables include age or years since birth treated as a quadratic term to reflect the increase and decrease in smoking prevalence