“…Each model adds a set of covariates to the preceding one, in a sequence that reflects the chronological realisation of these channels: as shown by Balia and Jones (2011), lifestyles such as cigarette smoking are likely to be acquired before the attainment of academic qualifications, which, in turn, influence socioeconomic status later in life. For each health outcome, the specifications that account for all these factors are of the form: health outcome i,age46/42 = α + β 1,i * (type and characteristics of school) +β 2,i * (childhood health) + β 3,i * (ability) +β 4,i * (parental background) +β 5,i * (local area/other control variables) +β 6,i * (lifestyles age33/42 ) +β 7,i * (highest edu.…”