Distinct impact of education and income on habitual exercise: A cross-sectional analysis in a rural city in Japan

Education and income are important socioeconomic indicators that reflect different aspects of social hierarchy. However, only a few studies have explicitly examined how different the relationship between education and health behaviour is from that between income and health behaviour. According to the human capital theory of health investment, education would reflect knowledge assets that allow an efficient investment in health, while income would relate to the value of healthy days and/or the time cost of health investment. Since time cost and the relative price of health would differ across age strata, we examined the significance of effect modification by age strata to distinguish the effects of education on habitual exercise from the effects of income. A cross-sectional survey was conducted using a selfadministered questionnaire in a rural city in northern Japan in January 2007 (n = 3385). Logistic regression analyses were conducted to assess the association of educational attainment and household income with habitual exercise. Interaction terms of these socioeconomic indicators with age strata (<60 years versus >= 60 years) were included to test the distinctive association across age, followed by a stratified analysis. As theoretically predicted, higher income was significantly associated with habitual exercise among those aged 25-59 years, while the association was null or negative among those aged 60 and above. Education was significantly associated with habitual exercise regardless of the age groups. These results suggest that the effects of socioeconomic factors on health behaviours vary according to which socioeconomic indicators are analysed, and which age group is selected. We conclude that studies on the socioeconomic disparity of health behaviours should carefully choose socioeconomic indicators to explain specific health behaviours to reveal underlying mechanisms and provide relevant policy implications, based on explicit behavioural models. (C) 2011 Elsevier Ltd. All rights reserved.