Article Text

Download PDFPDF

Work, permanent sickness and mortality risk: a prospective cohort study of England and Wales, 1971–2006
  1. Bola Akinwale1,
  2. Kevin Lynch2,
  3. Richard Wiggins3,
  4. Seeromanie Harding4,
  5. Mel Bartley5,
  6. David Blane1
  1. 1Department of Primary Care and Social Medicine, Imperial College London, London, UK
  2. 2Office for National Statistics, Titchfield, Fareham, UK
  3. 3Department of Quantitative Social Science, Institute of Education, University of London, London, UK
  4. 4MRC Social and Public Health Sciences Unit, University of Glasgow, Glasgow, UK
  5. 5Department of Epidemiology and Public Health, University College London, London, UK
  1. Correspondence to Bola Akinwale, Department of Primary Care and Social Medicine, Imperial College London, St Dunstan's Road, London W6 8RP, UK; b.akinwale{at}


Background In recent decades, labour market participation has fallen in men, with large amounts of this decline accounted for by increases in permanent sickness. There is speculation that the rising numbers of permanently sick incorporate more people with less severe conditions than was previously the case. This paper examines the relationship between labour market position and subsequent mortality around State Pension Age.

Methods Using linked census and death records in the ONS Longitudinal Study, samples of men aged 55–69 and women aged 50–64 were selected from each decennial census, 1971–2001 and their health followed up. Differences between the employed, unemployed and economically inactive in age-specific death rates, Standardised Mortality Ratios and odds of reporting limiting long-term illness were examined.

Results Labour market activity in late middle age has changed since 1971. For example, the proportion of men employed at ages 60–64 years has fallen by 39%, and the proportion permanently sick has more than doubled. Despite this change, there has been stability in the RR of mortality between labour market positions. Working people have the lowest risk of premature death, while, relative to working people, the permanently sick continue to have mortalities around three times higher among men and four to five times higher among women.

Conclusion The evidence does not support the notion that the permanently sick are becoming less seriously ill. The persistence of the group's raised mortality suggests that measures aimed at encouraging later life employment should ensure provision of work environments suitable for people with chronic illnesses.

  • Labour market position
  • mortality
  • early old age
  • ONS longitudinal study, mortality SI
View Full Text

Statistics from


Socio-demographic change in historic proportions is transforming early old age in Britain. In England, life expectancy at age 65 increased by more during the final 30 years of the twentieth century than during the preceding 70 years.1 This change is more marked among men, but it is also present among women. In addition, near-universal male employment up to the State Pension Age has been replaced by more diverse labour market participation in a deindustrialised labour market.2 Only a minority of men aged 60–64 years were in paid employment around the turn of the new century, although the proportion of older workers has increased in recent years.3 At the same time, women's labour market participation increasingly resembles that of men, with a shrinking proportion being described as housewives.4

While these changes have been predominately beneficial, they raise issues such as how to fund pensions,5 prolong active ageing and maximise Third Age quality of life.6 Accordingly, at a time when labour market participation at older ages has been falling sharply, the UK government has responded with a range of initiatives to prepare for population ageing including a package of measures designed to enable people to retire later. Health is an important part of these challenges and should be given due weight in debates about the appropriate policy response. This paper, therefore, focuses on understanding whether those economically inactive nowadays are healthier than their equivalents in past decades. We examine the relationship between labour market position around the State Pension Age and subsequent mortality and morbidity in England and Wales using longitudinal data for the period 1971–2006. In the UK, State Pension Age (SPA) is currently 60 and 65 for women and men respectively. For many currently in the workforce, it will soon be 65 for both sexes, and will rise progressively for both sexes to 68 by 2046. This study has international relevance because most rich countries share the same combination of decreasing labour market participation in late middle age, and debates about the appropriate policy responses.

It is known that there is a relationship between labour market position—that is whether a person is employed, self-employed, unemployed or not looking for work—and subsequent health across the years of working life.7 8 Much debate has been concerned with understanding whether observed associations between labour market position and health are causal or reflect selection bias.9 If the relationship between labour market position and health reflects selection, poorer health among the unemployed compared with workers may reflect the fact that people in poorer health are more likely to lose their jobs, while those with better health are more likely to stay in work. Selection may also be due to factors other than health such as education level and socio-economic position,10 which are known to be independently associated with health.

At older ages, structural and cyclical changes in the economy are likely to have a significant bearing on the relationship between labour market position and health. During the 1980s and 1990s, there was a shift in the British economy away from manufacturing and heavy industry and towards services. Some have argued that low unemployment rates have been maintained since the early 1990s recession because non-employed people have been signed onto sickness benefits, excluding them from unemployment figures and increasing the numbers of people claiming sickness benefits11 though active labour market programmes such as the New Deals have also reduced the claimant count in Britain. Until recently, the main benefit paid to people unable to work or seek work due to long-term illness or disability was Incapacity Benefit (IB), which is awarded to those judged not fit enough to work following structured assessments. Apart from changes in who is deemed not fit for work in the benefits regime, increases in sickness benefit claimants may also be a sign of changes in work environments that have been driven by deindustrialisation, reflecting reduced scope to move workers to lighter or less demanding duties as they age than had previously been the case.2 Most people non-employed due to sickness are over 50 and more likely to have a history of manual jobs.11 These people may have real health conditions which, if not prompting their labour market withdrawal, put them at the back of the queue for re-employment when coupled with their age and skill level.2 The relationship between labour market position and health can therefore be viewed as reflecting relative advantage or disadvantage in individuals' occupational histories.9 Against this background, the present paper examines two related questions: (1) to what extent does being in work around the UK State Pension Age have a protective effect for subsequent health? (2) With increases in their relative numbers, are the permanently sick less seriously ill nowadays than was previously the case?



The present study uses data from a record linkage study, the Office for National Statistics Longitudinal Study (ONS-LS), which links census records and records of deaths, cancers and births for approximately 1% of the population of England and Wales.12 Currently, data from each decennial census 1971 to 2001 inclusive are available in the ONS-LS, along with death registrations up to the end of 2006.

We have selected people within 10 years of the UK State Pension Age (SPA) at each census; specifically, women aged 50–59 years and men aged 55–64 years. We have examined also the health of those immediately post-SPA by including men aged 65–69 and women aged 60–64 years.



Mortality is the primary health outcome examined in this study. Our analysis of mortality is based on follow-up of each census cohort for 5 years, taking into account any known exits from the population at risk due to emigration from England and Wales. The sample used is limited to study members who were traced in the NHS Central Register—more than 97% of the eligible sample at each census.

Limiting long-term illness

Information on morbidity comes from the census question on limiting long-term illness, which was collected first in the 1991 Census: ‘Does the person have any long-term illness, health problem or handicap which limits his/her daily activities or the work he/she can do? Include problems which are due to old age.’ The question was repeated in 2001 with a different wording, which referred to disability instead of handicap.

Labour market position

We use the term labour market position to refer to an individual's relationship to the labour market, where people are classified as either economically active (participating in the labour market through work or job-seeking because of unemployment) or inactive (outside the labour market due to permanent sickness, for other reasons such as family responsibilities or retirement at any age). In the 1971 and 1981 Censuses, people could classify themselves as unemployed due to temporary sickness; we have combined this small temporarily sick group with the permanently sick. Although the census measurement of labour market position has changed in other ways, it is possible to classify people into five groups that are consistent over time: in work; unemployed; retired; permanently sick; other inactive.

Social class

We adjust for socio-economic differences between the labour market positions using the Registrar General social classes, aggregated into social classes I–IIIN (non-manual) and IIIM–V (manual). In the 2001 Census, the Registrar General's classification was replaced by the National Statistics Socio-economic Classification (NS-SEC) as the official measure of social class.13 It possible to allocate study members to an approximation of NS-SEC at the 1991 Census, but this is not currently an option at earlier Censuses. Therefore, for the 2001 cohort, we have used an approximation to Registrar General non-manual and manual social classes, to allow consistency with earlier cohorts.14 Women have been assigned a social class according to the occupation of the head of their household; this is typically the woman's own occupation for never married, divorced and widowed women and the spouses' occupation for married women.

Statistical analysis

For each cohort, we measure labour market position recorded at the census and relate it to subsequent health. To examine differences in mortality between groups, we calculate age-specific mortality rates by labour market position for each census cohort. Relative differences in mortality are examined using RRs, calculated with Cox regression, and using Standardised Mortality Ratios (SMRs). For each cohort, the gender-specific death rates used for standardisation of SMRs are those of people in work. Logistic regression equations are used to model the risk of limiting long-term illness adjusted for social class. Statistically significant differences between groups are determined by observing non-overlapping CIs.

Sensitivity analysis

To investigate possible selection bias into the different labour market positions, deaths in the first 5 years after each census were ignored, and only deaths during the postcensus years 6–10 were included in sensitivity analyses. Allowing in this way for the possibility that ill people are more likely to enter certain labour market positions9 (wearing off of selection) increases confidence in the mortality differentials identified.15 As the ONS-LS's most recent mortality linkage is 2006, it was not possible to use this method for the 2001 Census.


Since 1971, there has been a substantial change in distribution of labour market position around the State Pension Age. Table 1 shows a pattern of increasingly early labour market withdrawal. For example, the proportion of men in work at ages 60–64 years has fallen markedly, from 78.4% in 1971 to 47.5% in 2001. Correspondingly, the proportion of men retired at ages 60–64 has risen, from 7.2% in 1971 to 24.7% in 2001, while the proportion permanently sick has more than doubled, from 9.0% to 19.7%.

Table 1

Percentage distribution of labour market position by age group and census

Among women at the ages examined here, the other inactive group, mostly housewives, has decreased in size relative to other groups. Below the State Pension Age (SPA), considerable reductions in the proportions recorded as other inactive have coincided with large increases in the proportions in work. By 2001, women (61.6%) were almost as likely as men (78.3%) to describe themselves as retired after the State Pension Age. Like men, the proportion of women who are permanently sick has increased steadily in the pre-SPA decade of life. For example, among women aged 55–59 years, the proportion permanently sick rose from 3.4% in 1971 to 12.4% in 2001.

Table 2 shows the relationship between labour market position and mortality using age-specific death rates. There were substantial decreases in the absolute risk of death in all groups over the period 1971–2001; for example, there was a 66.1% fall in the mortality of men in work at ages 60–64. Retired men and women had the largest absolute reductions in mortality between 1971 and 2001, particularly among people who were approaching the State Pension Age (around 73% for men aged 60–64 and 70% for women aged 55–59). Despite these sizeable reductions in absolute mortality, the pattern of relative mortality risk remained remarkably stable during 1971–2001, being persistently lowest among those in work and highest among the permanently sick for both men and women at working ages (there were a few exceptions, where CIs overlap).

Table 2

Age-specific mortality rates of men and women around State Pension Age (SPA), 1971–2001

Such stability in the RR of mortality by labour market position is illustrated in table 3, which shows for each age-specific group, relative risks of mortality using people in work as the reference category. Working people consistently had the lowest risk of mortality; in the small number of cases where the mortality of a group was lower than working people—for example, unemployed men at ages 65–69—these differences were not statistically significant. Mostly, the mortalities of permanently sick men were around three times higher than those of men in work; and significantly higher than those of the other labour market groups. Likewise, permanently sick women had mortalities typically four to five times higher than women in work depending on census and age-specific group.

Table 3

Relative risks (RR) of mortality by labour market position (reference group: in work) for men and women around State Pension Age, 1971–2001

Along with changes in the distribution of groups in the labour market, there was some change in the RR of mortality for groups falling between the lowest mortality of working people, and the highest of the permanently sick. Table 4 shows Standardised Mortality Ratios (SMRs) by labour market position for people in the 10 years approaching State Pension Age. In 1971, the mortality of retired men (SMR=175) and women (SMR=211) was significantly raised compared with working people. Over the period 1971–2001, large reductions in absolute mortalities among retired people reduced their relative mortality risk to the extent that, by 2001, only working people had lower mortalities than retired people. Among men, the relative mortality risk of those in the other inactive group fell also over the period, at a time when the mortality of unemployed men increased relative to those in work. Among women, the relative morality of those other inactive or unemployed was higher by 2001 compared with earlier decades.

Table 4

Standardised Mortality Ratios for men aged 55–64 and women aged 50–59: 1971–2001

Table 5 reports the odds of limiting long-term illness 10 years after the census for men and women in the various labour force positions at the 1981 and 1991 Censuses. The overall pattern is the same as for mortality; namely, the best health among those in work, greatly increased odds of subsequent limiting long-term illness among the permanently sick and somewhat worse health than those in work among the other labour market groups.

Table 5

ORs (95% CI) for limiting long-term illness at next census: men aged 55–64 and women aged 50–59, 1981 and 1991

Our sensitivity analysis did not alter significantly the substantive findings. During years 6–10 postcensus, among both men and women, the mortality of those in work was lowest, and the permanently sick had the highest rate, with intermediate mortality rates for the other three labour market groups. Of contemporary interest, the mortality disadvantage of the unemployed when compared with those in work was revealed by this sensitivity analysis, because it allowed health selection into the labour market to wear off16–18(See online appendix 1).


The years of the ONS Longitudinal Study cover profound changes in British society. The first data in 1971 belong to the post-World War II settlement of full employment among men, who mostly worked until the State Pension Age, with a large proportion of women describing themselves as housewives. Subsequent census linkages have recorded deindustrialisation,19 globalisation,20 feminisation of the workforce21 and growing labour market exit before the State Pension Age.22 In parallel with these changes, the study's linkage to death registration has revealed quite remarkable reductions in middle-aged mortality, mostly of around 60%, depending on age, gender and labour market position.

Given the scale of these changes, the comparative stability of the relative mortality risk attached to each labour market position is equally striking, including the long-term raised risk attached to unemployment among men, and the significant advantage associated with work. One change of note, however, is that marked increases in the proportions retired, before the State Pension Age, have occurred with falls in the absolute and relative mortality of the group. This has implications for the observed mortalities of the employed, and the permanently sick because those who retire early due to ill health are likely to be selected out of the labour market, into permanent sickness. Of particular interest for the present purposes is the situation of the permanently sick. They have increased as a proportion of the population in the decade before the State Pension Age,23 at which ages their absolute mortality has fallen significantly as they appear to have benefited from general increases in life expectancy. Nevertheless, while absolute death rates of the permanently sick have fallen, their relative mortality risk compared with people in work has changed little among both men and women.

The policy implications of these findings depend on the choice of comparator group. If one compares the permanently sick with their historical predecessors, their health is better. If one compares the permanently sick with their employed contemporaries, their health continues to be markedly worse, and this health gap has not changed since 1971.


This study used a large, nationally representative sample, with minimal non-response and very little loss to follow-up. The data set covers the period of deindustrialisation in Britain, from the 1971 industrial economy and post-World War II social settlement (eg, Beveridge reforms and family wage) to the globalised social and economic conditions of 2001, which is also the period during which life expectancy at middle age and patterns of labour market participation were transformed for the age groups under study. The changes analysed are local manifestations of international phenomena.


  1. Mortality, which is the main health outcome of the study, is a poor indicator of health, because the concept of health implies more than mere existence. Also, mortality ignores diseases such as osteo-arthritis, which are major causes of serious morbidity that rarely result in death. These points are answered to some extent by the study's second health outcome measure—limiting long-term illness. Although the measure was included in only the two most recent Censuses, it is reassuring that variation in limiting long-term illness is consistent with that of mortality by labour market position.

  2. Measurement of labour market position in the data set is made at Census, only once every 10 years, so nothing is known about intercensal working life. Substantial intercensal mobility between labour market positions could dilute the strength of the study's observed relationships, although as a result, the present study's results would be conservative estimates.

  3. Health selection into the various labour market positions may mean that the associations are short-term phenomena, without long-term significance. We have tried to estimate this effect in a sensitivity analysis which excluded deaths during the first 5 years of follow-up, on the assumption that the majority of those forced out of employment by acute, serious illness will have a low 5-year survival (our analysis of mortality after the 2001 Census was unable to include this safeguard, because of the short follow-up time). The size of the relative mortality disadvantage of the permanently sick, compared with those in work, was reduced somewhat by allowing in this way for the wearing off of selection, but the pattern of mortality risk by labour market position remained unchanged. The worse health of the permanently sick is a long-term phenomenon.

  4. Our study examined only the years around the State Pension Age. Other studies are required of younger age groups.


During the period 1971–2001, health at older working ages improved greatly for those in all labour market positions: employed, unemployed, retired, permanently sick or other inactive. A large increase in the proportions of men and women who were economically inactive due to long-term sickness would be expected to result in a dilution of mortality differences between this group and the employed. However, despite a reduction in the absolute mortalities of the permanently sick, no reduction in their RR of mortality was seen compared with those in work. Any attempts to encourage older, sick people back to work will need to ensure a supply of jobs with work environments suitable for people with chronic illnesses. This should inform further development of the new sickness benefit, Employment Support Allowance.

What is already known on this topic

  • The permanently sick have increased as a proportion of the population around the State Pension Age.

What this study adds

  • Despite large increases in the relative size of the permanently sick, there has been no reduction in their mortality relative to other groups.

  • Policies aimed at encouraging later life employment will need to ensure a supply of jobs with work environments suitable for people with chronic health conditions.


We acknowledge, with gratitude, the support of our colleagues at the Office for National Statistics. This work contains statistical data from ONS which are Crown copyright and reproduced with the permission of the controller of HMSO and Queen's Printer for Scotland. The use of the ONS statistical data in this work does not imply the endorsement of the ONS in relation to the interpretation or analysis of the statistical data. This work uses research datasets which may not exactly reproduce National Statistics aggregates.


View Abstract

Supplementary materials

  • Web Only Data jech.2009.099325

    Files in this Data Supplement:


  • Funding The work is funded by UK Research Council's Programme New Dynamics of Ageing (Award Number RES-352-25-0015) and ESRC International Centre for Life Course Studies in Society and Health (Award Number RES-596-28-0001).

  • Competing interests None.

  • Provenance and peer review Not commissioned; externally peer reviewed.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.