Table 2 shows the relations of the three outcomes to education. In this and subsequent
analyses, the outcomes are treated as categorical variables: poor/fair physical health;
upper quintile of waist:hip ratio; lowest quintile of psychological well-being. In every
case, the group with the lowest education had strikingly worse levels than those with at
least college degrees. In all six analyses, the gradient was also clearly in evidence, as
each group had worse health than the one above it in the hierarchy. The magnitude of the
difference and the slope of the gradient suggest that educational level may be the most
important correlate of these health outcomes in this national sample.
The relations between the potential explanatory variables and the health outcomes are
reported in tables 3, 4, and 5. The tables show the effects adjusted for age and race.
Odds ratios with confidence intervals for the same relations, additionally adjusted for
education, are also shown.
 |
The results for poor/fair physical health are shown in table 3. Nonwhite women report
worse physical health than white women, but this difference by race was not seen among
men. An alternative measure of socioeconomic circumstances, household income, was related
to poor health independent of education. A third measure, poverty/unemployment in area of
residence, was classified on the basis of two variables available from the census:
proportion below the poverty line and proportion unemployed. This measure of deprivation
of the area of residence was associated with poor/fair health independent of education.
When further adjusted for household income, the odds ratios of 1.58 (women) and 1.67 (men)
for those in the worst areas, compared with the best, decreased to 1.48 (1.0-2.2) and 1.41
(0.9-2.2) for women and men, respectively.
We note with interest that mother's education, but not father's, was related to
poor/fair physical health in women, even after adjustment for their own educational level.
Conversely, in men, father's education, but not mother's, was related to poor/fair
physical health.
If poor self-reported health is indeed a measure of poor physical health, one would
expect smokers to have worse health than nonsmokers. This was the case.
Contrary to prediction, there was no relation between marital status and self-reported
ill health. There was, however, some relation between a measure of social support and
health; those who indicated that there was strain in their social relationships
experienced worse health. Perceptions of inequalities were related to ill health. Among
work characteristics, low use of skills was related to poor health in men, but not women,
after adjustment for education.
The control and self-efficacy measures were separated into mastery and constraints.
There was no relation between mastery and poor/fair health for either men or women, but
there was a strong relation with constraints: men and women with highest perceived
constraints (i.e., low control) had the worst health.
Table 4 shows the same relations for waist:hip ratio. The "middle-aged
spread" is a real phenomenon for men and women. Waist:hip ratio increases with age.
The apparent decrease in the oldest age group for men could be a survivor effect. Black
women have higher waist:hip ratio than other groups. This is not the case for black men.
As with poor/fair physical health, so with waist:hip ratio: father's education, but not
mother's, was associated with waist:hip ratio in men. However, parent's education was not
associated with waist:hip ratio in women. The poverty/unemployment index of area of
residence was associated with waist:hip ratio in women, but not in men. Strain in social
relationships was associated with higher waist:hip ratio, as was perceived inequality in
women. In both sexes, high perceived constraints were associated with higher waist:hip
ratio.
 |
The associations with psychological well-being are shown in table 5. In contrast to the
data on waist:hip ratio, neither parent's education was related in men; mother's
education, but not father's, was related to psychological well-being in women. There were
stronger relations between psychosocial factors and psychological well-being than for
other health outcomes in tables 3 and 4. This may represent overlap in the measurements.
Area of residence, which is externally assessed, was not related to psychological
well-being.
 |
Tables 6, 7, and 8 address the question of how much of the education gradient in the
outcome measures can be accounted for statistically by the potential mediators. Most of
the explanatory variables considered in univariate fashion do appear to make a
contribution. In men, the four variables that make the largest contribution to explaining
the social gradient in all three outcomes are smoking, psychosocial work characteristics,
perceived inequalities, and low control. These variables are also important in women.
 |
 |
 |
The clearest finding from each of these tables, however, is that no one factor plays a
crucial role in accounting for the gradient, but the total contribution is quite
substantial. With adjustment for all the variables simultaneously (table 9), the odds
ratio for poor/fair physical health associated with being in the lowest educational group
is reduced from 5.96 to 3.25 in men, and from 8.00 to 3.21 in women. For waist:hip ratio
and psychological well-being, the odds ratios are reduced, although the magnitude of the
effect is smaller.
Discussion
In this study of a U.S. national sample there is an inverse social gradient in the
three measures of health: self-reported physical health, waist:hip ratio, and
psychological well-being. As with studies of mortality differentials (Marmot, Shipley, and
Rose 1984), the point to emphasize is that there is a gradient in morbidity. People in the
lowest socioeconomic group, comprising, in this study, those who did not complete high
school, have worse health than others, but the social differentials do not stop there.
Each educational group has worse health than the one above it in the hierarchy.
Does It Matter Which Measures of Health?
For these analyses, we chose three measures of "health." The first,
self-reported physical health, is attractive as a global measure of health status.
Although it tells us little of specific biomedical problems and is potentially subject to
the vagaries of all subjective measures, it has been shown in many studies to be a
powerful predictor of mortality (Idler and Angel 1990; Appels et al. 1996). It is likely,
therefore, that determinants of self-reported health will prove to be determinants of more
specific biomedical disorders.
Waist:hip ratio was chosen as a more specific biomedical marker because of its
demonstrated relation with diabetes mellitus, cardiovascular disease, and mortality
(McKeigue, Shah, and Marmot 1991). The evidence suggests that central obesity may have
different metabolic effects from a more general pattern of obesity. It is the typically
male pattern of obesity, but it also characterizes menopausal women. The findings here
show clearly the increase in waist:hip ratio with increasing age. Whereas there is some
fall-off in men in the oldest age group (65-74), this is not the case for women. It is of
interest that studies of nonhuman primates show social differentials in central adiposity:
subordinate female rhesus macaque monkeys have a greater degree of central adiposity than
do dominant female monkeys further up the hierarchy (Shively et al. 1987).
Numerous prior studies have documented age, gender, class, and cultural differences in
aspects of well-being (Ryff and Singer 1998). For the present investigation, the central
question was the extent to which lower educational level was linked with lower profiles of
well-being, thereby extending the social inequalities literature to consider not just
increased likelihood of negative health, but also diminished chances for positive mental
health among those occupying lower positions in the educational hierarchy. Diminished
psychological well-being among the less educated further contributes to their
vulnerabilities.
Does It Matter Which Measures of Class?
There are at least two reasons for examining the relation of socioeconomic status to
health: description and explanation. Description is a pragmatic exercise: use what works.
The argument here is that any measure of socioeconomic status will "do" to
describe the relation between social position and health. From this perspective, education
has two distinct advantages. Superficially, at least, it has the attraction of a graded
scale that lends itself to obvious cut-off points with social meaning: did not complete
high school; high school graduate; some college; college degree or higher. Other
classifications, like those based on occupation, need further work to put them into this
simple form (Erikson and Goldthorpe 1992). An exception can be found in the Whitehall
studies, which indicated that a grading based on occupation was a better predictor of ill
health than was education, perhaps because the occupational hierarchy in the civil service
is defined precisely; hence, employment grade was a more precisely graded measure than
years of education.
The second advantage of using education as a measure is that it deals, to a large
extent, with the issue of health selection. In principle, it is possible to argue that
people in worse health end up in a lower socioeconomic position and that this accounts for
the relation between socioeconomic status (SES) and health. This position becomes less
plausible when education is used as the marker for SES. Educational status is by
definition achieved when people finish full-time education. This precedes the development
of ill health for most people. For health status to determine the relation of education to
health, ill health in childhood would have to affect both educational achievement and
health in adulthood. Although this may happen, the evidence suggests that it is a minor
part of the explanation for the relation between socioeconomic status and ill health
(Wadsworth 1986; Power, Manor, and Fox 1991; Power and Hertzman 1997). In our preliminary
analyses for this paper, we included self-reported physical and mental health at age 16.
Despite the fact that recall of past health status is likely to be influenced by present
health status, thus overstating the connection between past and present health, these
variables did not eliminate the SES gradient in health in adulthood.
Turning from description to explanation, the particular measure of socioeconomic status
may indeed matter. It is no longer simply the pragmatic question of what works but,
rather, what may convey meaning: insight into how the social structure may translate into
forces that determine a social gradient in health status. Different measures of social
status convey different meaning (Bartley et al. 1996; Krieger and Fee 1996). In this data
set we have, as yet, made only a limited exploration of alternate measures of
socioeconomic status. Household income was related to the three health measures
independent of educational level, and the relation of education to health was independent
of household income. We do not interpret education, or income, as necessarily being
causal. The existence of independent relations with health does not necessarily mean that
level of income or number of years of education is the cause, in the sense that providing
more income or increasing educational opportunities necessarily reduces inequalities in
health. They may be, but our approach to speculation on causation is to ask which factors
may mediate the relation between education and health.
One particular area that has led to some speculation is the level at which causes
operate: the individual or the group. Wilkinson points to the fact that within societies
income is related to life expectancy, whereas at the level of whole societies, countries,
or states of the United States (Kaplan et al. 1996), not income level but inequality of
income is related to life expectancy (Wilkinson 1996). This suggests that the character of
whole societies may be related to ill health; hence, the relation between measures of
social capital and mortality in the United States (Kawachi and Kennedy 1997). In the
Alameda County study, people living in poverty areas had higher mortality than people
living in other areas, independent of a range of personal characteristics, including
income (Haan, Kaplan, and Camacho 1987). In the MIDUS sample, health varied according to
area of residence, based on a classification derived from level of poverty and
unemployment in the area. This relation was independent of household income and education.
It is consistent with the proposition that characteristics of neighborhoods may relate to
differences in health status.
Explanations for Inequalities in Health
The Black Report set the debate in Britain on causes of social inequalities in
health (Black et al. 1988). The Black committee suggested four types of explanation:
artifact; selection; culture/behavior; material/structural. MacIntyre has recently
reviewed the subsequent debate, which tended to take hard positions as to which
explanation was primary (MacIntyre 1997). "Softer" positions allow for a more
balanced view of the relative contribution of each of these. We have already dealt briefly
with the issue of selection.
Although the evidence suggests that the oft-found relation between SES and health is
not the result of artifact, we should consider what role artifact may have played in the
present study. First, there is the issue of self-report bias. Both the exposures and the
outcomes were generally based on self-reports. This raises the possibility of
contamination: an unfavorable self-view of one area of life - for example, poor health -
may lead to an unfavorable view or unfavorable reporting of other areas of life.
Alternatively, there may be a common antecedent (a third variable): people in an
unfavorable socioeconomic position may tend to report unfavorably on all areas of life,
resulting in an apparent relation between them. Although this could, in all likelihood,
account for some of the relation between the constraints dimension of low control and
psychological well-being, it is less likely to account for relations like that between
parents' education and central adiposity and could not account for the relation between
area of residence, classified on the basis of census characteristics, and ill health. Even
with two self-report measures that have a high degree of subjectivity, like control and
self-reported health, contamination may not be the explanation for the association. If it
were, why would there be a strong relation between constraints and self-reported health
but not between mastery and health? It is not obvious why one measure would be more
affected by the bias of having assessed one's health as poor, or of occupying a low
status, than the other. To the extent that there is such contamination, it will overstate
the contribution to an explanation of the educational gradient and will understate the
independent effect of education.
A second issue that makes causal interpretation difficult is that the study is
cross-sectional in design. We are not, therefore, examining the determinants of incidence
of disease or, more generally, of change in health status. This leads to the possibility
of reverse causation or of survival effects. Reverse causation, akin to the health
selection issue, would arise if health determined the exposure rather than the reverse.
If, for example, central adiposity led to a feeling of lack of control, we would be wrong
in our interpretation of the association as showing the influence of psychosocial factors
on the development of a high waist:hip ratio. As we argued earlier in the discussion of
health selection, because education is the marker of socioeconomic position, reverse
causation is unlikely to be the reason for the observed socioeconomic gradient in health.
Reverse causation could, however, weaken the interpretation of whether the factors studied
are truly intermediaries between socioeconomic status and health.
The possibility of survival effects is shown through the apparent decline with age in
waist:hip ratio among the oldest men. This could result from a higher mortality among men
with the highest waist:hip ratio. These issues cannot be settled in this cross-sectional
study. Nevertheless, a causal interpretation is possible and consistent with the
hypotheses we elaborated at the start of the study, and it informed the choice of measures
that were included.
The findings are indeed compatible with a model of causation of social inequalities in
physical health and psychological well-being that suggests that no one factor is
responsible. Various factors each appear to make a small contribution to the explanation
of the social gradient, but they add up to a substantial explanation.
It is perhaps not meaningful to ask precisely how much of the gradient is explained by
these variables for three reasons: the cross-sectional nature of the study; measurement
imprecision; and the mode of statistical analysis. We have treated each of these variables
as single variables with independent effects. That is simplistic. There will be complex
interactions of variables acting at various stages of the life course. This study is not
well suited for testing out models that really require a longitudinal birth cohort study
of the sort that has been conducted in Britain (Wadsworth 1991; Power, Manor, and Fox
1991). The advantage of the present study, however, is that we were able to include a rich
array of variables from multiple domains describing people's situation in midlife.
Policy Implications
Amid these caveats we should not lose sight of the potential significance of the study.
It confirms that midlife is a time when there are substantial socioeconomic differences in
health, in markers of disease risk, and in psychological well-being. These socioeconomic
differences follow a gradient. They are not confined to people at the bottom of the social
hierarchy. Although one cross-sectional study cannot, by itself, settle the issue of
causation, conclusions about the role of the potential causal factors we have analyzed
here are consistent with other evidence.
Upstream or Downstream?
An important question for policy is the degree to which we should focus on the causes
of social inequalities, on the reasons for the link between social position and the
potential mediator, or simply on the mediator itself. This can be illustrated with smoking
as, perhaps, typical of health behaviors. The evidence here suggests that smoking makes a
contribution to social inequalities in health. A "downstream" focus might
concentrate on measures to reduce smoking prevalence in the population. These would
include (increasing cigarette prices through) taxes and the creation of smoke-free
settings, in addition to health education. To the degree to which this was successful, it
would benefit the whole population.
A more "upstream" focus, on the other hand, would point to the social
gradient in smoking demonstrated in this and other studies as evidence that smoking
policies do not benefit the whole population to the same extent. This suggests that
smoking policies should take into account the social distribution of smoking (Townsend,
Roderick, and Cooper 1994). The present study did not address reasons for the social
gradient in smoking but confirms it as an area that cannot be ignored. The extreme version
of the upstream focus would suggest that the issue is not smoking but social inequalities.
Action to reduce inequalities in health should therefore focus on the causes of social
inequalities. Our analyses, described here, suggest the possibility of an approach
somewhat short of this fundamental one. By examining factors that may mediate the link
between social status and health, it provides information on possible intervention points.
Health behaviors exemplified by smoking, material circumstances related to income and
neighborhood characteristics, psychosocial factors in the workplace, lack of control, and
perceived constraints all appear to contribute to state of health.
Deprivation through the Life Course
The relation between parents' education and health of participants is consistent with a
view that the causes of health inequalities in midlife begin in childhood and may cumulate
through the life course (Bartley et al. 1994; Power and Hertzman 1997). Some speculation
is in order here. One interpretation of life course effects on health in adults is that
material deprivation in childhood has long-lasting effects. Somewhat against this material
deprivation explanation in the present study is the differential effect of parental
education in men and women. Father's education was more strongly related to ill health in
men; mother's education, in women. Although it is possible that this connection could be
linked to differential exposure to material deprivation in men and women, a more likely
interpretation is that the effect is psychosocial. Whether because of role models, or for
some other reason, a male whose father is more highly educated experiences better health,
independent of his own achieved education, whereas, in the case of a female, the mother's
education appears to be the dominant factor.
Is Education Important?
It is tempting to believe that if part of the educational gradient can be explained by
the psychosocial factors that were examined, the unexplained part must be the result of
education itself. Such a conclusion would be premature: First, the degree to which the
gradient can be explained may relate to precision of measurement and form of statistical
analysis. Second, education may be a marker, although, of course, the question is, a
marker of what? It may be a marker of the background from which people come. The evidence
we have on the effect of education independent of parent's education does not support this
explanation, but there is evidence, for example, that the best predictor of a child's
performance in school is level of parental interest (Wadsworth 1991). Education may indeed
be part of the causal pathway, in the sense that better education leads to a greater
likelihood of favorable work and living conditions, as well as habits and behaviors, in
later life. These, in turn, may lead to better health.
The fact that educational attainment may be part of the causal pathway linking
socioeconomic status to health does not by itself suggest a policy remedy. Mortimore and
Whitty (1997) discuss whether school improvement could overcome the effects of
disadvantage; their conclusion is guarded. It would be difficult to argue that school
improvement could be anything but good. Mortimore and Whitty discuss evidence, however,
that children from advantaged backgrounds derive greater benefit from school improvement
than do children from disadvantaged backgrounds. An unintended consequence of overall
school improvement could therefore be a greater gap in educational attainment between
those with more and less social advantage. The logic of investment in preschool education
is precisely to overcome this social disadvantage before children enter the school system.
The results are encouraging (Schweinhart et al. 1993). The way to improve the
"returns" to educational investment of disadvantaged children may be to invest
in high-quality preschool programs.
Taking Action on the Gradient
If the link between social inequality and ill health was confined to worse health among
those in poverty, there would be little conceptual difficulty in devising potentially
effective policies. Anything that helped to relieve poverty might be expected to improve
the health status of those living in poverty. This is not to underestimate the political
and practical difficulties of devising and implementing such policies, but to emphasize
that the poverty question is conceptually tractable.
Inequality, on the other hand, poses greater conceptual difficulties in devising
policies. These findings imply, as did the Whitehall studies (Marmot, Shipley, and Rose
1984; Marmot et al. 1991; North et al. 1993), that a social gradient runs across the whole
spectrum. In this analysis, we have used education as our indicator of social position to
illustrate the gradient. Income can also be used; this can be illustrated by the panel
study of income dynamics (McDonough 1997). People in the poorest group of households (less
than $15,000 in 1993) had mortality 3.9 times higher than that of the richest group of
households (greater than $70,000). About 7 percent of the population fell into this
poorest category. By contrast, those in households with incomes in the $30-$50,000 range
had a relative mortality of 1.6, but 30 percent of the population was in this third
highest income category. The smaller relative excess therefore applied to a much larger
population subgroup. Concentrating only on those in poverty would, one hopes, benefit
them, but it would not address the question of inequalities in health in the total
population.
The MIDUS study provides a potential menu of factors that may be playing a role in
generating socioeconomic differences in health. It is to some extent artificial to
separate features of a complex social existence in order to ask which is most important.
The evidence we have presented suggests that it may not be possible. It is nevertheless
reasonable to ask where appropriate intervention points might be. If the perspective is to
improve things by narrowing social inequalities in health, it is important to ask which of
a number of factors might be amenable to intervention and might interrupt the pathways
leading to inequalities in health. No one study can answer this question, but our study
does contribute to the body of evidence (Wilkinson 1996; Blane, Brunner, and Wilkinson
1996) suggesting that the social environment is an important determinant of the health of
populations and should be the appropriate focus for action.
References
Appels, A., H. Bosma, V. Grabauskas, A. Gostautas, and F. Sturmans. 1996. Self-Rated
Health and Mortality in a Lithuanian and a Dutch Population. Social Science and
Medicine 42:681-9.
Bartley, M., L. Carpenter, K. Dunnell, and R. Fitzpatrick. 1996. Measuring Inequalities
in Health. Sociology of Health and Illness 18:455-75.
Bartley, M., C. Power, D. Blane, G. Davey Smith, and M. Shipley. 1994. Birth Weight and
Later Socioeconomic Disadvantage: Evidence from the 1958 British Cohort Study. British
Medical Journal 309:1475-9.
Black, D., J.N. Morris, C. Smith, P. Townsend, and M. Whitehead. 1988. Inequalities
in Health: The Black Report; The Health Divide. London: Penguin.
Blane, D., E. Brunner, and R.G. Wilkinson. Eds. 1996. Health and Social
Organization. London: Routledge.
Blaxter, M. 1987. Evidence on Inequality in Health from a National Survey. Lancet
2:30-3.
Bosma, H., and A. Appels. 1997. Differences in Mortality between Lithuanian and Dutch
Middle Aged Men. In East-West Life Expectancy Gap in Europe, ed. C. Hertzman.
Dordrecht, the Netherlands: Kluwer.
Bosma, H., M.G. Marmot, H. Hemingway, A. Nicholson, E.J. Brunner, and S. Stansfeld.
1997. Low Job Control and Risk of Coronary Heart Disease in the Whitehall II (Prospective
Cohort) Study. British Medical Journal 314:558-65.
Brunner, E.J., M.G. Marmot, K. Nanchahal, et al. 1998. Social Inequality in Coronary
Risk: Central Obesity and the Metabolic Syndrome. Evidence from the Whitehall II Study.
Diabetologia 40:1341-9.
Dahlgren, G., and M. Whitehead. 1995. Tackling Inequalities: A Review of Policy
Initiatives. In Tackling Inequalities in Health: An Agenda for Action, eds. M.
Benzeval, K. Judge, and M. Whitehead. London: Kings Fund Institute
Drever, F., M. Whitehead, and M. Roden. 1996. Current Patterns and Trends in Male
Mortality by Social Class (Based on Occupation). Population Trends 86:15-20.
Erikson, R., and J.H. Goldthorpe. 1992. The Constant Flux. Oxford: Clarendon.
Goldblatt, P. 1990. 1971-1981 Longitudinal Study. Mortality and Social Organisation.
London: Her Majesty's Stationery Office.
Grant, M.D., Z.H. Piotrowski, R. Chappell. 1995. Self-Reported Health and Survival in
the Longitudinal Study of Aging 1984-1986. Journal of Clinical Epidemiology 48:3:375-87.
Haan, M., G.A. Kaplan, and T. Camacho. 1987. Poverty and Health: Prospective Evidence
from the Alameda County Study. American Journal of Epidemiology 125:989-98.
Idler, E.L., and R.J. Angel. 1990. Self-Rated Health and Mortality in the NHANES-I
Epidemiologic Follow-up Study. American Journal of Public Health 80:446-52.
Kaplan, G.A., V. Barell, and A. Lusky. 1988. Subjective State of Health and Survival in
Elderly Adults. Journal of Gerontology 43:S114-S120.
Kaplan, G.A., E.R. Pamuk, J.W. Lynch, R.D. Cohen, and J.L. Balfour. 1996. Inequality in
Income and Mortality in the United States: Analysis of Mortality and Potential Pathways.
British Medical Journal 312:999-1003.
Karasek, R., and T. Theorell. 1990. Healthy Work: Stress, Productivity, and the
Reconstruction of Working Life. New York: Basic Books.
Kawachi, I., and B.P. Kennedy. 1997. Health and Social Cohesion: Why Care about Income
Inequality? British Medical Journal 314:1037-40.
Krieger, N., and E. Fee. 1996. Measuring Social Inequalities in Health in the United
States: A Historical Review 1900-1950. International Journal of Health Services
26:391-418.
Lachman, M.E., and S.L. Weaver. 1998. The Sense of Control as a Moderator of Social
Class Differences in Health and Well-Being. Journal of Personality and Social
Psychology 74:763-73.
MacIntyre, S. 1997. The Black Report and Beyond: What Are the Issues? Social Science
and Medicine 44:723-45.
Mackenbach, J.P. 1995. Tackling Inequalities in Health. British Medical Journal
310:1152-3.
Marmot, M.G., and J.F. Mustard. 1994. Coronary Heart Disease from a Population
Perspective. In Why Are Some People Healthy and Others Not?, eds. R.G. Evans, M.L.
Barer, and T.R. Marmor, 189-214. Hawthorne, N.Y.: Aldine de Gruyter.
Marmot, M.G., and M.J. Shipley. 1996. Do Socioeconomic Differences in Mortality Persist
after Retirement? 25-Year Follow-Up of Civil Servants from the First Whitehall Study. British
Medical Journal 313:1177-80.
Marmot, M.G., M. Bobak, and G. Davey Smith. 1995. Explanations for Social Inequalities
in Health. In Society and Health, eds. B.C. Amick et al. New York: Oxford
University Press
Marmot, M., H. Bosma, H. Hemingway, E.J. Brunner, and S.A. Stansfeld. 1997.
Contribution of Job Control and Other Risk Factors to Social Variation in Coronary Heart
Disease Incidence. Lancet 350:235-9.
Marmot, M.G., G. Davey Smith, S. Stansfeld, et al. 1991. Health Inequalities among
British Civil Servants: The Whitehall II Study. Lancet 337:1387-93.
Marmot, M.G., C. Ryff, L. Bumpass, M.J. Shipley, and N.F. Marks. 1997. Social
Inequalities in Health: Next Questions and Converging Evidence. Social Science and
Medicine 44:901-10.
Marmot, M.G., M.G. Shipley, and G. Rose. 1984. Inequalities in Death: Specific
Explanations of a General Pattern. Lancet 1:1003-6.
McDonough, P., G.J. Duncan, D. Williams, and J.S. House. 1997. Income Dynamics and
Adult Mortality in the United States, 1972 through 1989. American Journal of Public
Health 87:1476-83.
McKeigue, P.M., B. Shah, and M.G. Marmot. 1991. Relation of Central Obesity and Insulin
Resistance with High Diabetes Prevalence and Cardiovascular Risk in South Asians. Lancet
337:382-86.
Mortimore, P., and G. Whitty. 1997. Can School Improvement Overcome the Effects of
Disadvantage? London: Institute of Education.
North, F., S.L. Syme, A. Feeney, J. Head, M.J. Shipley, and M.G. Marmot. 1993.
Explaining Socioeconomic Differences in Sickness Absence: The Whitehall II Study. British
Medical Journal 306:361-6.
Pappas, G., S. Queen, W. Hadden, and G. Fisher. 1993. The Increasing Disparity in
Mortality between Socio-economic Groups in the United States 1960 and 1986. New England
Journal of Medicine 329:103-9.
Power, C., and C. Hertzman. 1997. Social and Biological Pathways Linking Early Life and
Adult Disease. British Medical Bulletin 53:210-21.
Power, C., O. Manor, and J. Fox. 1991. Health and Class: The Early Years.
London: Chapman & Hall.
Rodin, J. 1986. Aging and Health: Effects of the Sense of Control. Science
233:1271-6.
Rothman, K.J. 1986. Modern Epidemiology. Boston: Little, Brown.
Royal College of General Practitioners/OPCS/Department of Health and Social Security.
1986. Morbidity Statistics from General Practice: Third National Study 1981-82.
Series MB5. London: Her Majesty's Stationery Office.
Ryff, C.D. 1989. Happiness Is Everything, or Is It? Explorations of the Meaning of
Psychological Well-Being. Journal of Personal and Social Psychology 6:1069-81.
Ryff, C.D., and B.H. Singer. 1998. Middle Age and Well-Being. In Encyclopedia of
Mental Health, ed. H.S. Friedman, 707-19. San Diego, Calif.: Academic Press.
SAS Institute. 1990. SAS/STAT User's Guide. Cary, North Carolina.
Schweinhart, L.J., H.V. Barnes, D.P. Weikart, W.S. Barnett, and A.S. Epstein. 1993. Significant
Benefits: The High/Scope Perry Preschool Study through Age 27. Monographs of the
High-Scope Educational Research Foundation, no.10. Ypsilanti, Mich.: High/Scope Press.
Shively, C.A., T.B. Clarkson, L.C. Miller, and K.W. Weingand. 1987. Body Fat
Distribution as a Risk Factor for Coronary Artery Atherosclerosis in Female Cynomolgus
Monkeys. Arteriosclerosis 7:226-31.
Townsend, J., P. Roderick, and J. Cooper. 1994. Cigarette Smoking by Socio-economic
Group, Sex and Age: Effect of Price, Income and Health Publicity. British Medical
Journal 309:923-7.
U.S. Bureau of the Census. 1993. Poverty in the United States: 1992. Current
Population Reports, series P60-185. Washington, D.C.
Wadsworth, M.E.J. 1986. Serious Illness in Childhood and Its Association with
Later-Life Achievement. In Class and Health, ed. R.G. Wilkinson, 50-74. London:
Tavistock.
Wadsworth, M.E.J. 1991. The Imprint of Time: Childhood, History and Adult Life.
Oxford: Clarendon.
Whitehead, M., and G. Dahlgren. 1991. What Can Be Done about Inequalities in Health?
Lancet 338:1059-63.
Wilkinson, R.G. 1996. Unhealthy Societies: The Afflictions of Inequality. London:
Routledge.
Working Group on Inequalities in Health. 1980. Inequalities in Health (Black
Report). London: Department of Health and Social Security.
Acknowledgments: This study was funded by the John D. and Catherine T. MacArthur
Foundation Research Network on Successful Midlife Development. We are grateful to the
members and associates of this research network (Orville Gilbert Brim, PhD, Director) for
their contributions to designing this survey and to Maneesh Juneja for assistance with
statistical programming. MGM is supported by a Medical Research Council research
professorship.
Address correspondence: Michael G. Marmot, Department of Epidemiology and Public
Health, University College London Medical School, 1-19 Torrington Place, London WC1E 6BT,
England.
© 1998 Milbank Memorial Fund. This file may be redistributed electronically as long as
it remains wholly intact, including this notice and copyright. This file must not be
redistributed in hard-copy form. The Fund will freely distribute this document in its
original published form on request.
