The Blackwell Companion to Sociology (81 page)

their social environment and their lifestyle of prayer, meditation, and silent

work (p. 461).

Thus, it is a mistake to conclude from the observed age gradient of BP in

Western societies that BP ``naturally'' increases with age. Similarly, it would be a mistake to conclude from the observed BP differences between Caucasian Americans and African Americans that racial differences are primarily genetic. While genes probably play a role, behavioral and other psychosocial factors also

contribute.

348

Joseph E. Schwartz

Social Class

Unlike age, sex, and race, the constructs of social class and socioeconomic status (SES) are inherently socially defined. Whether the social stratification within a society is more accurately represented as a set of hierarchically ordered discrete classes or as relative positions along a more continuous gradient, those in higher status positions generally have a higher standard of living, greater access to

scarce resources, and increased opportunities compared to those with lower

status. While rarely conceptualized as a scarce resource, good health and a

long life are two objectives to which almost everyone aspires. If those in the

lower classes tend to have poorer health and a shorter life expectancy, this would augment the social and economic inequalities that are more traditionally the

focus of social stratification research.

Governments and social reformers have gathered mortality data and analyzed

them according to a variety of indicators of social class for more than a century (for example, Dublin, 1917; Britten, 1934; Guralnick, 1962). Antonovsky

(1967) published one of the earliest systematic sociological reviews of the

relationship of social class to longevity and mortality. He summarized data

from more than 30 studies and concluded that the evidence was overwhelming

that the lower class, often defined as unskilled manual laborers, had a substan-

tially shorter life expectancy and a higher mortality rate than other social classes.

What was less clear was whether longevity and mortality rates were similar

across the remaining social classes, or whether those in intermediate classes

(typically lower-level non-manual and skilled and semi-skilled manual workers)

had mortality rates that were higher than those in the higher classes (for exam-

ple, professionals, administrators, businessmen, upper and intermediate level

managers and supervisors, and shopkeepers). Thus, an early question that

emerged from this research was: is there a relatively continuous SES or social

class gradient to mortality rates, or is there simply a marked difference between those at or near the bottom and everyone else?

Shortly after Antonovsky's review, Kitagawa and Hauser (1973) published an

important monograph on the subject. The novel feature of their study was that

they did not use the information on death certificates to assess people's social class or even their age. Instead, they matched the death certificates of individuals who died between May and August 1960 to US census data so that (a) information about the deceased and the full population would come from the same

source (census) and (b) differential mortality with respect to education and

income, in addition to occupation, could be examined. At the time of publica-

tion, their results were the most definitive available for the United States. In white and non-white males and females aged 25±64, there was a clear pattern of

lower age-adjusted mortality rates with each increment in education. The same

inverse relationship was observed between age-adjusted mortality rates and

family income in white males and females aged 25±64. Even when adjusted for

education, at least half of the income differential in mortality rates remained, and vice versa when the mortality rates were adjusted for income. This study

Social Inequality, Stress, and Health

349

also found a general, though not perfectly monotonic, inverse association in

white males aged 25±64 between age-adjusted mortality and major occupational

groups ranked roughly according to social status. Overall, this study supported

the conclusion that there is an SES gradient in mortality across the entire SES

distribution.

Some of the best evidence that mortality risk and the risk of cardiovascular

disease increase steadily as one moves down the SES ladder comes from the

Whitehall I Study of British civil servants (Marmot et al., 1984). This study

followed 18,000 male civil servants, aged 40±64, all of whom had secure office-

based jobs in and around London. Participants were classified into one of four

categories, based on the civil servicè`grade'' (ranking) of their position: (a) top administrative, (b) professional or executive, (c) clerical, and (d) other (for

example, messengers, doormen). As shown in figure 24.1, a smooth gradient in

mortality rates emerged within just a few years (Marmot et al., 1984) and the

differences from one grade to the next have persisted as the cohort has continued to be followed (Marmot et al., 1995). Thus, even when looking at only a single

industry (i.e. government), one in which employees are not exposed to absolute

poverty, industrial accidents, or toxic substances in the workplace, there are

clear differences in age-adjusted mortality risk among those in different non-

manual occupational positions. We might expect the SES gradient to be even

larger among private sector employees than among public sector employees.

This study also shows that these differences are not due to differences in just

one or two major causes of death, but exist for almost every major cause

(Marmot et al., 1984). While smoking, obesity, and elevated blood pressure

were all more common in the lower social grades, statistically controlling for

these and other risk factors reduced the estimated differences in coronary heart disease mortality among the four grades by less than 25 percent (Marmot et al.,

Figure 24.1 Mortality from all causes by year of follow-up and grade of employment, male civil servants, initially aged 40±64.

Source: M. G. Marmot and M. G. Shipley, Whitehall I Study, unpublished.

350

Joseph E. Schwartz

1984, 1995). Similar results have been reported in several other large epidemi-

ological studies (see review by Kaplan and Keil, 1993).

In the United States, the National Longitudinal Mortality Study (NLMS;

Sorlie et al., 1992) is a large ongoing study of mortality. A sample of nearly

1.3 million individuals of all ages was identified between 1978 and 1985 and

basic physical and demographic data, including education, occupation, and

income, were obtained. Using the National Death Index, the complete sample

is being followed prospectively for deaths. The nine-year follow-up data became

available in 1995 (Release 2, October 1, 1995) and can be used to examine the

SES gradient and update the earlier analyses of Kitagawa and Hauser (1973) and

others. Figures 24.2 and 24.3 show my estimates of the age-adjusted mortality

ratios for different education and income groups, based on Cox proportional

hazards regression analyses of all employed, 18±64 year old, men (N ˆ 162,216)

and women (Nˆ 128,865) in the NLMS. The SES gradient is clear, and similar in

magnitude to that reported by Kitagawa and Hauser (1973).

There are many potential explanations for why lower SES individuals, espe-

cially the poor, might be at increased risk for a variety of diseases and have a shorter life expectancy: more crowded living arrangements, poorer sanitary

conditions, poorer diet, poorer access to medical care, access to poorer quality medical care, and differential rates of various health-related behaviors (for

example, cigarette smoking, excessive alcohol consumption, and lack of regular

physical exercise). It is less clear why those with average levels of education or income should be at higher risk than those with above average levels.

Figure 24.2 Age-adjusted relative mortality risk, by education category (reference is 12

years of education).

Social Inequality, Stress, and Health

351

Figure 24.3 Age-adjusted relative mortality risk, by income category (reference is $10,000±15,000).

Areal Measures of Social Class

The preceding discussion has focused on individual level measures of SES:

education, occupation, grade of employment, and income. However, there is a

long tradition of treating geographic areas (nations, states, counties, neighborhoods, census tracts) as the unit of analysis and investigating whether SES

differences ± for example, in average education, average income, or average

price of housing ± predict differences in mortality rates. Historically, areal

measures of SES were used as a proxy measure for the SES of individuals living

in that area when individual level data on SES, mortality, and morbidity were

unavailable. In most analyses, poorer SES areas have been found to have higher

mortality rates. For example, Kitagawa and Hauser (1973) assigned each census

tract in Chicago to one of five socioeconomic groups based on median family

income in the US 1950 and 1960 censuses, and found that both infant mortality

rates and age-adjusted all-cause mortality rates were highest in the lowest SES

group of tracts, and next highest in the second lowest SES tracts. This pattern

held for males and females, whites and non-whites, and, for all-cause mortality, in those under age 65 and those 65 and over. There were only slight differences

among the three highest SES groupings of census tracts.

A small number of studies have simultaneously estimated the effects of indi-

vidual-level and community-level measures of SES and found that the commun-

ity-level measures have an independent effect over and above that of the

individual-level measures. Using data from the Alameda County Study, Haan

et al. (1987) found that after controlling for age, sex, race, baseline health in 352

Joseph E. Schwartz

1965, and any of four measures of individual's SES (education, income, employ-

ment status, or access to medical care), Oakland, California, residents who lived in federally designated ``poverty areas'' had an approximately 50 percent greater risk of dying during the subsequent nine years, 1965±74.

On a national scale, Anderson et al. (1997) merged census tract median

income into the NLMS data set and found, for 25±64 year old black and

white males and females, that while individuals' family incomes were more

strongly related to mortality than median census tract income, the latter also

had an independent and sizable effect. The increase in mortality risk associated with living in a low-income census tract was about twice as great for blacks (49

and 30 percent for male and female blacks) as it was for whites (26 and 16

percent for male and female whites). These results, like those of Haan et al.

(1987), strongly suggest that there are contextual or neighborhood factors

that increase the mortality risk of even high SES individuals who live in low-

income areas. However, it also shows that, within any given area, those with

higher income are at lower risk than those with low income. Together, these

results suggest that both absolute income and relative income affect mortality

risk.

There are several aspects of the physical and social environment that might

contribute to an association between neighborhood SES and poor health. Air

and water quality may be poorer in lower SES areas. Poor neighborhoods are

often located in or near industrial areas, landfills, and toxic dumps for two

reasons: (a) real estate in such areas is often less expensive, making it more

affordable to lower SES families; and (b) poor neighborhoods usually have fewer

political resources with which to resist the nearby location of polluting industries or dumping. The quantity and quality of available health care services also tend to be poorer in low SES areas.

Income Inequality and Mortality

At one level the question of whether there is too much or too little income

inequality is largely a question of values and personal philosophy. Those who

argue that there is an inherent conflict between those who have and those who do not have power, status, wealth, and control over the means of production

usually view the existing income distribution as inequitable and unjust, imposed by the powerful on the powerless. In contrast, others view income inequality as

reflecting differential rewards in a competition that is fundamentally fair ± with better qualified or more productive individuals receiving higher incomes ± and

just. To my knowledge, no sociological or economic theory can adequately

explain cross-national differences in income differentials. For example, why is

it that salary differentials, and therefore income inequality, are substantially smaller in Scandinavian countries and Japan than in the United Kingdom and

United States? These differences in inequality are even greater when one exam-

ines after-tax income. Even if there is a substantial consensus that some inequality is legitimate, there may be very little consensus on how much inequality is

appropriate. The question of how much inequality is desirable and how much is

Social Inequality, Stress, and Health

353

too much is largely a matter of opinion, and individuals' opinions are likely to vary according to their relative position in the distribution.