social stratification in U.S. society, generally starting at the bottom of the social ladder. Access to valued resources—jobs, income and wealth, education, power and prestige—generally comes over decades or generations. Ethnic stratification has often given rise to racist ideologies that ascribe inherent inferiority to particular groups (Van den Berghe, 1964).
Belonging to a particular ethnic group can be linked to health in a variety of ways. Differences in genetic characteristics could produce differences in susceptibility to disease. Since racial and ethnic identity is a crucial part of personal identity, it may be associated with ways of behaving and reacting to the social environment that have implications for health. Since groups differ in their social standing—which tends to evolve over time—their access to health resources may vary. Since groups are not always treated similarly, some may experience more favorable environments for health than others. Arguably, some of these effects could be particularly notable among older people, whose racial and ethnic identification may be more deeply rooted and whose life experiences may reflect early, perhaps turbulent, history of race and ethnic relations. We consider all of these possible effects of racial and ethnic identity systematically in subsequent chapters.
Three complications in the way racial and ethnic groups are defined are important to note for their possible effects on health differences. First, selfidentification allows certain biases to enter group comparisons. For instance, consider the case of American Indians and Alaska Natives. As noted earlier, their numbers have increased rapidly, at least partly because some people are identifying with this group who had not previously done so (Sandefur et al., 2004). These individuals often do not live on designated federal reservations and have higher socioeconomic status than those who do. Though such changes in self-identification need not affect individual health, they still affect average group health, in this case reducing any health disadvantage of American Indians and Alaska Natives relative to other groups. Such selection effects are considered further below.
Second, for one important health outcome—mortality—self-identification is not the norm. Funeral home directors, sometimes with information from relatives, may make the determination of race and ethnicity of the deceased. The complication, from this and other possible sources of misclassification, amounts to published death rates that are too high for whites and blacks (by 1 percent and 5 percent respectively) and too low for American Indians (21 percent), Asian and Pacific Islanders (11 percent), and Hispanics (2 percent) (Rosenberg et al., 1999). A recent study of state hospital discharge data in California found that approximately 70 percent of hospitaliza-