Decomposing Differences in Cotinine Distribution between Children and Adolescents from Different Socioeconomic Backgrounds

This study decomposes differences in saliva log cotinine between children/adolescents from low and high socioeconomic backgrounds using the 1997/98 cross-section of the Health Survey for England (HSE). Three decomposition methods are applied including a mean-based (Oaxaca-Blinder) decomposition method and two further methods that allow the decomposition of differences in quantiles (the quantile regression and the recentered influence function regression decomposition methods). By extending the analysis beyond differences in means, this study is able to identify the contributions of different characteristics to differences in quantiles of the log cotinine distribution. Differences in log cotinine between the two study groups are decomposed into a part explained by group differences in the distribution of characteristics (composition effect) and a part explained by group differences in the impact of these characteristics (structural effect). The composition effect accounts for a larger proportion of the total difference in log cotinine compared to the structural effect. The composition effect attributable to smoking within the home explains more of socioeconomic differences at lower quantiles indicative of passive smoking compared to higher quantiles that are indicative of active smoking while the composition effect of household income and parental smoking explains more of socioeconomic differences in active smoking compared to passive smoking. The structural effect of parental smoking and smoking within homes is indicative of underlying group differences in parents‟ compensatory behaviours that limit the impact of parents‟ risky lifestyle choices on child health.