The broad traits comprising the Big Five or five-factor model of personality have each been associated with longevity. Introversion (e.g., Wilson, Mendes de Leon, Bienias, Evans, & Bennett, 2004), low Conscientiousness (e.g., Friedman et al., 1993), low Agreeableness (e.g., Miller, Smith, Turner, Guijarro, & Hallet,1996), Neuroticism, (e.g., Almada et al., 1991), and low intelligence (related although not directly comparable to Intellect/Imagination or Openness; e.g., Whalley & Deary, 2001) have all been identified as mortality risk factors. The lifespan health behavior model provides one theoretical account of the mechanisms by which personality traits affect morbidity and mortality: personality influences on health status are mediated by patterns of health-enhancing and health-damaging behaviors over the life course (Smith & Spiro, 2002). Consistent with this model, there is evidence of associations between all of the Big Five traits and self-reported health practices (e.g., Booth-Kewley & Vickers, 1994; Lemos-Giraldez & Fidalgo-Aliste, 1997), although Conscientiousness has received the most attention. Low Conscientiousness has been associated with numerous behavioral risk factors, including unhealthy eating, tobacco use, and lower levels of physical activity (Bogg & Roberts, 2004). Furthermore, health behaviors such as these are associated with morbidity and mortality (e.g., Thompson, Edelsberg, Colditz, Bird, & Oster, 1999; U.S. DHHS, 2000).

There are few opportunities to test lifespan health-behavior models within the same sample, one exception being the Terman Life Cycle study. Friedman et al. (1993) found support for smoking and alcohol use as partial mediators of the effects of childhood Conscientiousness on mortality. The Hawaii cohort enables the study of life-course pathways spanning four decades from personality in childhood to health status at midlife in a multiethnic population-based sample. In a previous study of this cohort, childhood personality traits predicted adult smoking, alcohol use, body mass index (BMI), and self-rated health (Hampson, Goldberg, Vogt, & Dubanoski, in press). That study also demonstrated partial mediation of the effects of Conscientiousness on self-rated health by smoking and BMI. In the present study, we tested a more complete lifespan health-behavior model involving all of the Big Five traits and two intervening mechanisms connecting childhood personality to adult health status over the life course: educational attainment and health behaviors.

Educational attainment is associated with longevity (Adler, Marmot, McEwen, & Stewart, 1999; Elo & Preston,1996), and a healthy lifestyle may be one of the numerous benefits that education confers (Yen & Moss, 1999). Lower educational attainment is associated with less healthy eating habits (e.g., Kristal, Hedderson, Patterson, & Neuhauser, 2001), greater risk of cigarette smoking (e.g., Bergen & Caporaso, 1999), and less leisure-time physical activity (He & Baker, 2005). Educational attainment characterizes a person’s life pathway in a single, summary variable. It is typically achieved in the early adult years and remains stable, so there is no question of reverse causality in its association with later outcomes. For the Hawaii cohort, in the absence of any data collection between elementary school and midlife, educational attainment provides a proxy for numerous intervening variables during those unobserved years. Moreover, childhood personality traits are reliable predictors of educational attainment both in this cohort and other longitudinal samples (Digman, 1989; Shiner, Masten, & Roberts, 2003). The health behaviors included in the model were the three key dimensions of a healthy lifestyle: healthy eating, tobacco use, and physical activity. Adult health status was measured by three indices. Self-rated general health is a remarkably reliable predictor of mortality (Idler & Benyamini, 1997), and better self-rated health is associated with higher levels of Extraversion and Agreeableness (Goodwin & Engstrom, 2002) and lower levels of Neuroticism or Negative Affect (Costa & McCrae, 1987). Functional status, as an indicator of quality of life, is regarded as a major component of good health (Ware, & Sherbourne, 1992), and obesity (BMI) has numerous health-damaging consequences including increased risk for cardiovascular disease, cancer, and diabetes (Thompson et al., 1999).

It was hypothesized that a model depicting a life-course pathway from childhood personality traits to health status at midlife, with educational status and health behaviors serving as intervening variables, would prove a good fit to the data. In view of gender differences in the associations between personality traits and longevity (Friedman et al. 1993), the fit of the model was compared for men versus women.

This test of a lifespan model was limited by the measures available for the Hawaii cohort. The measures were all by self-report, and they were obtained concurrently in the same questionnaire, whereas the model postulates that health behaviors over a lifetime contribute to health status. However, given the difficulty of permanently changing eating and exercise habits (Katz, 2005), it is likely that reports of typical and recent health habits reflected more than a current fad. The smoking measure did not indicate amount of smoking since initiation (e.g., pack years), which would have been a more accurate measure of lifetime exposure. The measure of physical activity failed to distinguish between work-related and leisure-time activity over a 7-day period, which may account for the negative association between physical activity and educational attainment (He & Baker, 2005).