Class and Health

Research and Longitudinal Data

Richard G. Wilkinson, Richard G. Wilkinson

• Publisher: Routledge
• Year: 2022
• ISBN:
  9781000578676
• Edition: 1
• Topic: Scienze sociali
• Subtopic: Classi sociali e disparità economica

Originally published in 1986, this book discusses issues such as social class differences in health; the effect of unemployment on health; the relationship between income and health; how much of the class differences in death rates can be explained in terms of medically recognized factors. Presenting empirical research to resolve these issues, the book takes health to the centre of the political stage and raises fundamental issues about the direction of modern economic and social development and its impact on inequality. As relevant now as when it was first published the book reviews twenty of the most important longitudinal studies in the area of health and class that had been carried out in the UK at the time.

ONE Socio-economic differences in mortality: Interpreting the data on their size and trends

R. G. Wilkinson

Prior to the 1980s, it was widely assumed that Britain was becoming a more egalitarian society. The predominant impression was that class divisions and socio-economic inequalities were becoming less important. Although long-term changes in income distribution have been comparatively small, it seemed reasonable to assume that this was compensated for by the growth of welfare services and the increasing volume of protective and regulatory legislation.

In 1980, however, the Black Report—dealing with class differences in health—seemed to cast serious doubt on this picture. Not only did it draw attention to very large differences in death rates between occupational classes but it also suggested that these differences were not declining (DHSS 1980). The crucial figures are reproduced in Table 1.1. They show that mortality differentials, as measured by age-standardized death rates for occupational classes, have increased since the 1930s. What these inequalities amount to can be summed up in terms of differences in life-expectancy. If the 1971 age-specific death rates for classes I (professional occupations) and V (unskilled manual occupations) were applied throughout the lives of a cohort, they would produce a difference in life-expectancy at birth of just over seven years (Registrar General 1978). This amounts to a lower-class disadvantage of about 10 per cent of life.

Table 1.1 Mortality by social class 1931-81 (Men, 15-64 years, England and Wales)

Class	1931	1952	1962*	1972*	1981**
I professional	90	86	76 (75)	77(75)	66
II managerial	94	92	81	81	76
III skilled manual and non-manual	97	101	100	104	103
IV semi-skilled	102	104	103	114	116
V unskilled	111	118	143 (127)	137(121)	166

^*To facilitate comparisons, figures shown in parentheses have been adjusted to the classification of occupations used in 1951.

^**Men, 20-64 years. Great Britain.

Note: Figures are SMRs—which express age-adjusted mortality rates as a percentage of the national average at each date.

Source: DHSS (1980), Table 3.1.

The figures in Table 1.1 provided the core of the Black Report: indeed, as the best record we have of the changing social distribution of health in British society, they may be seen as its rationale. The issue of the size and trends in class differences in mortality that these figures raise is clearly crucial, not only to health but to our understanding of the direction of modern social development. The problems of measurement are complicated, however, and the different interpretations of the data are controversial and have far-reaching implications. This chapter provides a discussion of the issues and reviews what recent research, including some of that presented in subsequent chapters of this volume, can tell us about how we should interpret the key figures in Table 1.1. The issues involved can only be disentangled by a careful assessment of the evidence.

The real size and trends in class differences in mortality are important not only because health matters in itself but also because health serves as a barometer of the social and economic conditions in which people live. Though we can quantify changes in class access to housing, education, jobs, and services, and can also describe some of the wider but perhaps less tangible changes in the social and physical environment in which people live, we do not know what all these changes add up to in human terms. There is no unified summary of changing class differentials in the quality of life and human welfare. Familiar summary measures of changes in the standard of living, such as indices of real income, suffer from a number of important weaknesses. Economic indicators are largely blind to the qualitative changes in the material and social environment, which are so crucial to human welfare. Health, on the other hand, is not only sensitive to qualitative changes in material life but the accumulating research evidence on stress, boredom, inactivity, depression, and lack of close social contact shows that it is also sensitive to many psychosocial aspects of the quality of life.

To point out the value of health as a social indicator does not detract from its own importance; rather, to do so serves as a reminder that insofar as the shortening of life is associated with poor social and economic circumstances, class differences in health represent a double injustice: life is short where its quality is poor. While we may focus principally on the shortening of life as a health issue, those who live on low incomes and in poor housing are no doubt more concerned with these problems for more immediate reasons than their effects on life-expectancy.

As a record of the social distribution of health in society, there are a number of ways in which the figures in Table 1.1 may be affected by problems of measurement. An unwary acceptance of the picture they present of the size and trends in social inequalities in health may be unnecessarily alarmist; but if, after looking at the evidence, it appears that health differences really have failed to narrow, we would surely need to reassess the belief that the burden of socio-economic differentials has narrowed. Expectations of growing equality have understandably led several analysts to infer that problems of measurement have resulted in an exaggeration of the size of class differences in mortality and a masking of an assumed diminution of differentials. I shall try here to approach the evidence more even-handedly, assuming that problems in the data are as likely to have under- as overestimated the size of the differentials and their apparent resistance to narrowing during most of this century.

What, then, does the evidence tell us about the interpretation of the figures in Table 1.1? A potential weakness of these figures that has always been recognized comes from the fact that they depend on dividing the number of deaths in occupations stated on death certificates by the number of people in each occupation as recorded at census. Inaccurate occupational descriptions at either point will give rise to a considerable but usually random mismatch. More systematic biases may, however, creep in from sources such as a respectful desire among the next of kin (who are the informants) to ‘promote the dead’.

Most of these problems were dealt with in the first publication (Fox and Goldblatt 1982) from the OPCS Longitudinal Study (LS). By following a 1 per cent sample of people identified at the 1971 census, the LS enabled deaths to be classified according to the occupations given at census. Using truly comparable numerators and denominators in the calculation of death rates, the results show 1971 mortality differentials much like those in Table 1.1. Part of the 1981 differential may, however, be spurious.

In chapter 2 of this volume, Marmot provides completely independent evidence of class differences in death rates among a cohort of civil servants classified by their employment grades. Using a different classification, he found the mortality gradient among civil servants (shown in his Table 2.1) is considerably steeper than that between the Registrar General’s social classes, recorded in Table 1.1. As Marmot points out, there is a more than threefold difference in mortality between the lowest and highest grades in the civil service. As with the LS, the data on civil servants do not suffer from the problems of relating occupations on death certificates to separately determined population numbers in those occupations. The employment grade of everyone in the study was established at one point in time and individual deaths were related to the employment grade originally stated for that individual. As the study is of a well-defined

group of employees and was carried out with the co-operation of the employer, it is likely that the occupational information is very accurate. It will, therefore, be free both of biases and of the fudging of differences that results simply from random occupational misclassification.

A more difficult question about the way we should interpret the observed class differences in health concerns the possible contribution of social mobility. As early as 1955, Illsley published evidence suggesting that social mobility discriminated in favour of the healthy and against the unhealthy (Illsley 1955). More recently, Stern has expressed this possibility in terms of a theoretical model, showing how, under various assumptions, social class differences in health could occur simply as a result of the healthy moving up and the unhealthy moving down the social scale (Stern 1983). This is an important issue because it suggests that to some extent inequalities in health are not attributable to structural socioeconomic inequalities. Instead of saying that people are less healthy because they are in lower classes, it suggests that they are in lower classes because they are less healthy.

To analyse this possibility, we must start off by distinguishing between two different ways in which social mobility could be selective in relation to health. A person’s chances of upward or downward mobility could be influenced either directly by their current, or manifest, health status, or indirectly through selection according to factors associated with health, such as height or education. In the first case, long-term illness or disability during any period in a person’s working life may affect their job prospects. Among young people, it may restrict initial job choices, while later in life it may force people to move to less demanding and perhaps lower-status jobs. The mobility chances of those who suffer from chronic illnesses or disabilities throughout their lives would be most affected. An important factor that reduces the size of this effect on the recorded class differences in death rates is that the social class classification is based on economic activity. People who are not economically active are excluded from the classification (except in special tabulations which may classify wives by husbands’, or children by fathers’, occupations). What may be more important to the class differences in health, as they are recorded, is the impact of the much greater burden of illness later in life. Social class at death—based on a person’s last full-time occupation as stated on the death certificate—may be substantially affected by the downward mobility of those who are forced by illness to take up less demanding jobs later in life.

A paper by Fox, Goldblatt, and Jones (published here as chapter 3) examines this possibility using more data from the LS. By classifying deaths during the period 1976-81 according to the person’s social class at the 1971 census, the authors have been able to cut out the effect of mobility during the last 5-10 years of life, when the impact of ill health would be at its maximum. After comparing the size of class differentials in death rates produced by this method with the 1971 figures reproduced in Table 1.1, the study concludes that social mobility later in life contributes very little to the disparity in death rates between classes. This does not necessarily mean that there is no deterioration in the job prospects of those who do contract chronic illness towards the end of their working lives; it means that the proportion of people affected is too small to have a major influence on the overall figures.

This result is supported by data from other studies. Marmot’s civil servants in chapter 2 show that the mortality gradient is maintained even ten years after the original classification by employment grade. American data from the Veterans’ Administration show the same pattern: during a 23-year follow-up of some 85,000 veterans discharged from the US army in 1946, the mortality differentials by military rank showed ‘little if any change’ (Seltzer and Jablon 1977: 563).

Wadsworth’s paper (chapter 4), based on the cohort of British births born in one week in 1946, tackles the problem of social mobility in relation to...