With a narrower definition of behavior risk factors, less of the health differences can be accounted for. In a separate analysis of self-reported health among individuals aged 51-61 years (Smith and Kington, 1997), a cluster of behavioral factors—the most important of which was BMI—account for only about one-sixth of the black disadvantage relative to whites, and it did not affect the Hispanic disadvantage. Among individuals aged 70 and older, the effect of behavior risk factors on self-reported health is slightly greater, with the black disadvantage reduced by one-third and the Hispanic disadvantage by close to one-fifth. These results are largely independent of socioeconomic status.
The role of behavior risk factors is even smaller in a study by Warner and Hayward (2002) of a sample of older men. They find that, once socioeconomic and demographic factors are taken into account, behavior factors do not explain the mortality gap. The opposite is true: with socioeconomic and demographic factors controlled, black risk behavior appears more favorable for health than white behavior, mainly because, with socioeconomic status controlled, blacks smoke less.
Such comparisons are mostly confined to black and whites. Given the somewhat contradictory patterns—Hispanics and Asians both having lower mortality rates than whites (but in one case exhibiting generally riskier behavior and in the other case less risky behavior)—it seems unlikely that, whatever the effects of these behaviors, they account for a substantial portion of health differences.
Could the effects of similar behaviors be different across racial and ethnic groups? Systematic study of this question has not been done, though hypotheses to this effect have been offered (e.g., Pampel, 1998). Work on obesity and alcohol consumption may be used to illustrate the possibilities and the uncertainties.
Obesity among older blacks, according to several studies summarized by Stevens (2000), is less of a risk factor for mortality than among whites (see also Sanchez et al., 2000). This differential effect appears mainly to involve black women. Although a dose-response relationship can be shown between BMI (above a minimum BMI of around 23) and mortality rate for whites of either gender, no such effect is visible for black women in several studies (e.g., Calle et al., 1999; Stevens, 2000; Stevens et al., 1998). It has also been suggested that obesity may even have a protective effect among older Hispanics (Stern et al., 1990). There are contradictory studies, however. For instance, Allison et al. (1997) find the same association of BMI with mortality for blacks and whites among men and women 70 years and older. And Grabowski and Ellis (2001) find, for a predominantly white sample of people of the same age, no association of elevated BMI with mortality. Several explanations are possible for such conflicting findings, including restricted samples, the uneven effect of selection, variations in