model specification, the confounding relationship of smoking with BMI, and possible differences in the distribution of body fat (Stern et al., 1990). Some generally negative effect of obesity on health—though not necessarily moderate overweight—is likely (Casper, 1995; Rössner, 2001), though attending to ethnicity clearly complicates the issue.
Alcohol consumption may also vary in its effects across racial and ethnic groups. As with BMI, alcohol consumption has a U-shaped or J-shaped relationship to mortality, at least beginning around age 35. The number of drinks associated with minimum mortality risk rises with age (White et al., 2002). Epidemiological studies have also found a minimum for deterioration in cognitive performance, which begins around four to eight drinks a day for men and two to four drinks a day for women (Dufouil et al., 1997; Elias et al., 1999; Kalmijn et al., 2002). For older Japanese Americans in Hawaii, however, cognitive performance declines at lower levels, beginning at about one drink a day (Bond et al., 2001; Galanis et al., 2000). A genetic factor may be involved: the fact that 50 percent of Japanese and Chinese lack the active form of aldehyde hydrogenase (ALDH2) and therefore have a lower alcohol elimination rate (Bond et al., 2001; Eckardt et al., 1998).
Health-related risk behaviors could get worse for groups with large proportions of immigrants. As noted above, acculturation is related to increases in both smoking and obesity, which are initially lower among immigrants than natives. Nor does rising socioeconomic status among minorities always reduce risky behavior. Consider the relationship between obesity and education. Educational levels are rising, from which one might infer the spread of healthier behavior. However, higher education increases physical activity more for whites than blacks, as well as reducing alcohol consumption more among whites (Gallant and Dorn, 2001). Furthermore, within each educational level, obesity is also rising—and it is rising faster among blacks than whites (Himes and Reynolds, 2002). Future generations of the older adults will be made up of individuals whose youthful habits, for whatever environmental and generational reasons, will have been less favorable to maintaining health.
We have considered major behavior risk factors that affect chronic illness. Other behaviors may also be relevant. Substance abuse and unsafe sex, for instance, may also vary across groups and produce racial and ethnic differences (Anderson, 1995). These behaviors are of greater concern among younger adults, but they are also relevant for older people. In a British sample of people over 50 years old, for instance, 7 percent reported behavior that put them at risk of contracting sexually transmitted diseases (Gott,