The Blackwell Companion to Sociology (82 page)

But regardless of personal, or even collective, values, is there any evidence that the degree of income inequality matters? Using international data from a variety of sources, Wilkinson (1986, 1990, 1992) has pioneered the investigation of this question with respect to health, life expectancy, and mortality. First, it has been shown that while there is a very substantial association between per capita gross national product (GNP) and life expectancy in developing countries, this relationship is relatively modest for developed countries. Wilkinson (1992, p. 165)

reported a correlation of 0.38 based on 1986±7 data for 23 Organization of

Economic Cooperation and Development (OECD) countries and noted that

changes in per capita GNP and changes in life expectancy during the preceding

16 years were uncorrelated (r ˆ 0.07). In contrast, using various sources of data and various measures of inequality, he and others have found substantial correlations, in excess of 0.80, between inequality and both mortality (positive association) and life expectancy (negative correlation). Data from the Luxembourg

Income Study for nine OECD countries exhibited correlations of 0.80 or higher

between life expectancy and the percentage of income received by the bottom 60,

70, or 80 percent of the population (the higher the proportion, the less unequal the income distribution).

Within the Unites States, Kaplan et al. (1996) and Kennedy et al. (1996) have

demonstrated that a variety of different measures of inequality (the Gini coefficient, the Robin Hood index, and the percentage of total income received by

those at or below the median income) are all correlated 0.50 or greater with ageadjusted state mortality rates. The more unequal a state's distribution of income is, the higher that state's mortality rate tends to be. Kaplan et al. found that the correlation was highest (r ˆ 0.74) when predicting mortality rates among those

aged 25±64. Controlling for median state income (both articles) and proportion

of the population having incomes below the federal poverty level (Kennedy et al., 1996) did not substantially alter the raw correlations. This is critical because if those states with greater income inequality were poorer or had higher poverty

rates, their higher mortality rates might be attributable to the usual SES gradient.

These studies indicate that it is not simply poverty per se that increases

mortality risk, at least in economically advanced societies. The authors, including Wilkinson, suggest that relative economic well being may be more important

than absolute levels. However, it is possible that not only the relatively deprived but also those in the middle of the income distribution are at higher risk of death if they live in a less equal society. This possibility could be tested in a data set, such as the NLMS, that contains individual mortality data if appropriate aggregate measures of average income and income inequality were added.

Perhaps the most intriguing evidence supporting a causal relationship between

economic inequality and mortality comes from longitudinal analyses examining

changes in income distribution (Wilkinson, 1992). In an analysis of 12 European

Community countries, increases between 1975 and 1985 in the proportion of the

population living on less than half the national average disposable income, a

cutpoint commonly used to define relative poverty, were strongly associated

354

Joseph E. Schwartz

with smaller increases in life expectancy (r ˆ À0X73). From a graph of the results (Wilkinson, 1992, p. 166, figure 3), it appears that over this ten-year period, a change of 4 percent in the proportion living in relative poverty is associated with about a one-year change in life expectancy for the entire population. In a second analysis of six OECD countries, increases in the percentage of total disposable

income received by the bottom three quintiles (the bottom 60 percent of the

population) were strongly associated with increases in life expectancy (r ˆ 0.80).

While the proportion of income received by the bottom 60 percent increased by

more than 2 percnet during the 1970s in Japan, the life expectancy was increas-

ing by almost 3.5 years. During the same period, the proportion of income

received by the bottom 60 percent decreased by about 1 percent in Great Britain

while the life expectancy increased by about two years. The notion that redis-

tribution of as little as 3 percent of income might result in an increase of 1.5 to 2.0 years in life expectancy for the entire population is certainly provocative.

Further research is necessary to ascertain which segments of the population ± for example, which age groups and which SES groups ± would see the greatest

changes in life expectancy.

Kawachi et al. (1999) have integrated components of several sociological

literatures to propose a model hypothesizing that the amount of absolute

deprivation (poverty, unemployment), relative deprivation (inequality), and

social disorganization (lack of social cohesion) in a community jointly influence the rates of violent and property crimes, poor health, and the age-adjusted

mortality rate of the community. Using aggregate, state-level data from several

sources, they had shown previously (Kawachi et al., 1997) that differences in

income inequality among states are closely related to the percentage of residents who endorsed the following statements: ``most people would try to take advantage of you if they got a chance,'' ``you can't be too careful in dealing with

people,'' and ``people mostly look out for themselves'' (all r b 0.70) in the 1986±

90 General Social Survey. They interpret such statements of cynicism and mis-

trust as indicators of a relative lack of social cohesion, or social capital, and suggest that this may play a mediating role in the relationship between income

inequality and mortality. Such analyses might help to explicate the social pro-

cesses underlying previous macrosociological investigations of the consequences

of social inequality (Blau and Blau, 1982; Blau and Schwartz, 1984).

I conclude this section with two comments. First, if the degree of economic

inequality influences the social cohesion, crime rates, and even disease morbidity rates and life expectancy of a society, then significant social costs (what economists call `èxternalities'') accompany the presumed economic benefits of a free

market economy. In such a case, the issue of how much income inequality is

appropriate should no longer be primarily a philosophical or economics ques-

tion, but rather a political question that concerns the well-being and public

health of the entire populace.

Second, much research remains to be done before we will understand how it is

that income inequality can affect individuals' health, mortality risk, and life

expectancy. If we consider the substantial increase in income inequality that

occurred in the United States and several other West European nations during

Social Inequality, Stress, and Health

355

the last part of the twentieth century, it is clear that this was the conscious result of political decisions to reduce barriers to individuals and companies accumulating great wealth, while simultaneously increasing the risk of unemployment

and a decline in income to the majority. In short, opportunities for greater

success were accompanied by a net loss of economic security. As a result of

becoming more competitive, there has been an increase in both the number of

``winners'' and the number of ``losers,'' accompanied by an overall decrease in

public concern for the well-being of the losers. I would suggest that collect-

ive economic security is an important determinant of social cohesion. Further-

more, an unintended consequence of increased competition is increased

psychosocial stress, not only for thè`losers,'' but for almost everyone who

plays the game.

This brings us to the major transition point of the chapter. Thus far, I have

presented some of the evidence showing that SES and income inequality are

related to health and longevity. I have also indicated some of the macro- and

micro-level factors that probably mediate this relationship. Interested readers

should know that there are substantial literatures on the relationships of SES to diet and other known health-related behaviors, and a smaller literature examining the contribution of access to medical care to the SES gradient in mortality

(for example, Kogevinas, 1991; Mackenbach et al., 1989). However, in the

remainder of the chapter I review evidence that psychosocial stress plays a role in the etiology of several diseases. While I believe that a better understanding of the effects of stress on health will eventually help us to explain the SES gradient, I will not try to make this argument here.

Stress and Illness

The belief that psychological or mental stress is a contributing factor to several diseases is widespread. Though all of us have experienced stress or felt stressed on multiple occasions, there is not a single, widely accepted definition of stress.

Stress is often brought about by life events or situations, called stressors, but it can also result from contemplating past situations or anticipating future situations. Some of the more widely used questionnaires for assessing major and

minor life events are the Schedule of Recent Experiences (Holmes and Rahe,

1967), the Life Experiences Survey (Sarason et al., 1978), and the Hassles and

Uplifts Scale (Kanner et al., 1981). The Life Events and Difficulties Schedule

(Brown and Harris, 1989) is an in-depth interview procedure for assessing the

occurrence and significance of adverse events. Chronic situations pertaining to

work (for example, heavy work demands, conflict with or lack of support from

supervisor or co-workers, job insecurity, role conflict, lack of autonomy), family (marital conflict, financial strain, poor health of a family member), caregiver

responsibilities, living in a high-crime neighborhood, and poverty are all associated with increased reports of stress/distress.

Stress is thought to affect health by altering both behavior and physiology.

Stress is associated with increased alcohol consumption, increased nicotine

356

Joseph E. Schwartz

consumption, risk of relapse among those who have quit smoking, reduced

physical exercise, and adverse dietary changes, each of which increases the risk of multiple diseases and mortality. Nonetheless, most epidemiological studies

indicate that changes in these and other health-related behaviors can only

account for part of the observed relationship of stress with morbidity and

mortality.

There is a great deal of research on the physiological effects of acute stress.

While there is substantial variation across individuals and situations in the type and magnitude of response, it is well documented that acute stress is associated with: (a) activation of the sympathetic nervous system which triggers the release of catecholamines (adrenaline and noradrenaline) into the blood stream, which,

in turn, increase heart rate and blood pressure (Cacioppo and Tassinary, 1990);

(b) activation of the hypothalamic-pituitary-adrenal axis, which regulates the

supply of corticosteroids (Kirschbaum and Helhammer, 1989); (c) changes in the

number and functioning of various types of immune cells (Kiecolt-Glaser et al.,

1992); and (d) gastric functioning (Wolf and Wolff, 1947). However, relatively

shortly after exposure to an acute stressor has terminated, the sympathetic

nervous system, neuroendocrine system, immune system, and cardiovascular

system all return to pre-stress levels of activation. Thus, with one exception, it is not clear that physiological responses to acute stress are relevant to disease processes or mortality.

The exception is the possibility that acute stress may, on rare occasions, trigger a heart attack (myocardial infarction) or stroke. As might be imagined, it is very difficult to rigorously document this phenomenon. It would be nearly impossible

to prospectively study enough individuals intensively enough to test whether

such cardiovascular (CV) events are more likely to occur during or immediately

following an acute stressor than at other times. Nonetheless, despite limitations pertaining to retrospective recall bias, there is evidence from case±control studies that CV morbid events can be triggered by episodes of anger and stress (Mittleman et al., 1995).

If stress plays a significant role in other disease processes, it is almost certainly through exposure to chronic stressors (or persistently repeated exposure to acute stressors). The physiological impact on humans of prolonged exposure to a

stressor is less well known because such research cannot ethically be conducted

under laboratory conditions. However, observational data indicate reduced

immune system functioning following marital separation, divorce, or the death

of a spouse, and during unemployment (see reviews by O'Leary, 1990; Herbert

and Cohen, 1993). It still remains to be determined whether the observed

changes in the immune system are sufficiently large to alter the risk of life-

threatening infectious or autoimmune diseases.

Although the mechanisms are not well understood, some of the most convin-

cing evidence of a relationship between stress and illness comes from epidemi-

ological studies of work stress and cardiovascular disease. Excluding studies that have focused on specific occupations (for example, air traffic controllers,

teachers, nurses), recent research has focused on two models of work stress,

thè`job strain'' model and thèèffort±reward imbalance'' model.