LS Working Paper 73 DEATH AND THE SALESWOMEN: AN INVESTIGATION OF MORTALITY AND OCCUPATIONAL IMMOBILITY OF WOMEN IN THE LONGITUDINAL STUDY OF ENGLAND AND WALES by Anne Harrop and Heather Joshi Social Statistics Research Unit The City University Northampton Square London EC 1V OHB ACKNOWLEDGEMENTS ‘ l l i s study is part of an ESRC hnded programme, Health Ineaualities Amona Women and Their Families (R000232797), undertaken at the Centre for Population Studies, London School of Hygiene and Tropical Medicine, 99 Gower Street, London WC1E 6AZ. The permission of OPCS to use the Longitudinal Study is also gratehlly acknowledged, as are helphl comments from Rosamund Weatherall. The coding for occuaption in 1981 was supplied by Clare Ward. May 1994 ABSTRACT Occupational differences in the mortality of women in the Longitudinal Study of England and Wales are explored. Occupations are classified into the expansion of Social Class originally proposed for the Women and Employment Survey. This survey also provides direct evidence on the likely history of occupational mobility leading up to the occupations in 1971, which are followed up to 1989 in the mortality analysis. We find occupational mortality differentials which, though modest, are greater than those reported earlier. When the analysis is confined to groups least likely to have experienced occupational mobility before 1971 or between censuses, estimates of mortality ratios become more diEerentiated though less well determined. For example, among those who were in the same occupation at both censuses, and in hll-time employment in 1971, estimated occupational mortality differentials appeared as wide as among men. Thus women's occupational mobility may serve partly to obscure (and protect from) specific occupational hazards. However, relatively poor health ofwomen in semi-skilled factory jobs is confirmed in this study, especially for long-term incumbents. INTRODUCTION "In view of the intermittent and transitory character of much female occupation, it may be doubted whether the difficulty will ever be overcome". Stevenson (1927). The analysis of mortality by type of paid occupation is important for the study of social stratification and the prevention of particular hazards, but, for women, has still not reached any clear resolution. Women's mortahty differs from mens's, and so do their occupations. This paper presents a new approach to identifling women's occupations with particularly high or low risks of death. A woman's work history is often intermittent and her current occupation transitory. It is therefore unclear whether the intrinsic hazards of a particular occupation are best revealed by looking at the relative mortality of women in given occupations at one point in time. Such differentials as may appear could reflect general differences in lifestyle. Underlying differences associated with long tenure of particular occupations may be masked by the mixed histories of those whose occupations are less fixed. The latter issue is addressed here. We use two data sets with a common occupational classification, related to mortality in one and to occupational mobility in the other, the OPCS Longitudinal Study (LS) and the Women and Employment Survey (WES) respectively. The classification adopted is an expansion of the Registrar General's Social Class. It splits the big category of women's occupations into office and shop jobs. The latter are also of interest as a common destination of the downward occupational mobility that WES documented (Dex, 1987:70). This paper is one of the first to make use of the mortality follow-up to 1989 linked in the LS to the 1971 Census. The retrospective work histories collected by the Women and Employment Survey in 1980 enable us to infer the employment histories that preceded the snapshots of occupation recorded at the 1971 Census. 1971 is taken as the point of reference as the only date before 1980 for which Census data are available. It also maximizes the period of follow-up. Routine analysis of the occupational mortality of men has been conducted in England and Wales since 1861. The first attempt to analyze that of women did not occur until 1908 (Registrar 1 General, 1908). In 193 1 the occupational mortality of single women was considered for the first time (Registrar General, 193 1) and married women's own occupation was first recorded in 197 1 (OPCS, 1978). The dramatic increase in the labour force participation of women in the post-war period increases the need to investigate their occupational mortality. The Decennial Supplement compares occupations recorded at death registration with the number of persons recorded in the respective occupation at the Census. However the under-reporting of occupation at death registration means that the numerator does not correspond to the denominator, the population at risk. This is termed the numerator-denominator bias (OPCS, 1986a). This bias is greater for women than for men because the recording of women's occupation at death registration is worse than that of men's. In the 1970-72 Decennial Supplement (OPCS, 1978), 20 per cent of all women aged 15-64 could be assigned to an occupation at death registration compared with 49 per cent in the 1971 Census. There are two related reasons why the recording of women's occupation at death registration is so poor, both arising from the instructions to registrars (OPCS, 1978; Roman, Beral and Inskip, 1985). The instructions state: 'in the case of a woman who is not employed at the date of her death the last 111time occupation should not be recorded unless she has been in paid employment for most of her life'. Thus, the Decennial Supplement only records an occupation if it was full time, thus excluding women who had been employed part time. At the 1971 Census 39 per cent of women in paid employment were part timers (42 per cent in 1991 (Employment Gazette, 1992)). Secondly women whose employment is intermittent are not classified to an occupation at death. These discrepancies between the Census and death registration render mortality rates by woman's occupation in the standard Decennial Supplement method near meaningless (McDowall, 1983). The inception of the Office of Population Censuses and Surveys (OPCS) Longitudinal Study (LS) following the 1971 Census broke new ground for the study of occupational mortality. It is a prospective study of persons enumerated at the Census. Using the occupation information provided at the Census linked to register deaths the occupational mortality rate can be calculated matching population at risk to outcome. Women's occupational mortality has been studied for deaths in the first 10 years of follow up using this source (Moser et al, 1990). 2 This study suggests that women in six occupation orders (hrnace, forge, foundry, rolling mill workers; engineering and allied trades workers; textile workers; makers of other products; drivers of stationary engines, cranes etc; and warehousemen, storekeepers, packers and bottlers) had standardised mortality ratios above that of all women aged 15-59. However this difference was only significant for the relatively few women working in engineering and allied trades. Five occupation orders had mortality levels sigdicantly lower than all women, food drink and tobacco workers; clerical workers; sales workers; administrators and managers; professional, technical workers and artists. Moser and colleagues conclude that although significant differences in mortality between occupation groups were evident, these were more likely to be associated with a common lifestyle rather than to specific occupational hazards. Other work on the LS indicates that the social class gradient in mortality is less severe for women than it is for men (Goldblatt, 1990). The previous analysis of occupational mortality of women in the LS suffered from two limitations. The first relates to the classification of occupations, the second to the lifetime perspective on women's occupations. The classification of occupations was developed when the vast majority of employees were men working in manufacturing industries. This classification is unsuitable for classifling women's occupations for several reasons. Occupational segregation means that women's employment is concentrated in a relatively small number of occupations which differ markedly from men's. In the 1971 LS, 75 per cent of women were concentrated in just four occupational of 26 occupational orders (Moser, Pugh and Goldblatt, 1990). These large categories for women were also small categories for men. Paid work remained sexually segregated. Over 80% of the labour force in 1971 were in occupations in which their own sex was over-represented (Hakim 198 1, using 223 occupations). Though the classifications detects large feminized occupational groups they do not distinguish very finely the skills and responsibilities that women's jobs encompass, covering a wider range of tasks than is the case of male dominated job categories. The classification often fails to differentiate between women doing differing jobs in the same industry. Neither does it distinguish between full and part time employment, nor include categories for unwaged labour in the home (Martin and Roberts, 1984). Alternative occupational classifications for women have been developed. Martin and Roberts 3 made a simple modification of the RGs occupational classification by subdividing categories accounting for a large proportion of women's employment. This discriminates more finely among some women's jobs (Martin and Roberts, 1984). The resulting schema has 12 categories. S d a r l y Dale, Arber and Gilbert (1985) developed a classification scheme which differentiated women's occupations more finely by distinguishing between full time and part time employment. Neither of these classifications have been used to study mortality. A third classification of women's occupations was developed by Roberts and Barker. Using cross sectional data they identsed 'work positions' which combined women's paid employment with their fertility histories (Roberts and Barker, 1985). A similar classification has been used to consider the effect, for women's mortahty, arising fi-omthe dual burden of domestic responsibilities and paid employment (Weatherall et al, 1993). The second limitation of the earlier analyses of women's occupational mortality is partly met by such joint treatment of occupatiori and fertility. The problem is that a woman's paid occupation may not necessarily reflect her long-term way of Me in the same way that a man's paid occupation does. While employment is assumed to be of central importance to a man's life, it may not be for women, whose domestic responsibilities are also assumed to have greater, if not paramount importance. Though most women are in paid employment for most of their working lives, at any one point in time a large proportion of women are out of the labour force, for reasons largely relating to their domestic responsibilities. After such breaks women often return to occupations that do not reflect their skills or training, experiencing occupational downward mobility (Dex, 1987). Thuty-seven percent of mothers in the Women and Employment Survey with a job before and after childbearing returned to jobs at a lower level than the one they held before their first birth (Martin and Roberts, 1984). Similar incidences of downward mobility after childbearing have been found in the National 'Training Survey and the MRC National Survey of Health and Development (Stewart and Greenhalgh, 1984, Joshi and Newel1 1987). These jobs are often transitory. In due course, some women recover occupational status. Thus, particularly for women, a cross sectional snapshot may not record a person in their main occupation. Recording a woman in a temporary occupation will introduce error to the assessment of mortality in that occupation because she will not have been exposed for a reasonable length of 4 time to any hazards that may be associated with that occupation. Thus, occupational hazards may be masked iflong term incumbents are swamped by occupationally mobile workers doing similar jobs on a less permanent basis. We bring the classification devised by Martin and Roberts to the study of mortality, to address both the problems of classification and mobility. We use it to draw on evidence from the Women and Employment Survey about occupational and family building history to distinguish women in the LS whose occupations are less likely to be transient. It also enables us to investigate whether sales workers are more like manual workers than other non-manual women workers as far as mortality analysis is concerned. In the standard social class scheme they cannot be distinguished from other "junior" non-manual workers. The next section describes our sources of data and our classification of occupation. The following one sets out the methods to be used in our two-stage analysis. The fist stage, on occupational mobility, is reported in the fourth part of the paper, and the fiflh reports findings from the LS on the relation of occupation in 1971 to mortality and also to permanent sickness in 1981. The findings are discussed and summarized in the concluding section. SOURCES AND CLASSIFICATION OF DATA 1. Sources of Data The two data sets used here are the Women and Employment Survey (WES) and the OPCS Longitudinal Study (LS). The first is used to identifL sub-groups within occupations that are relatively stable, ie. include few women whose main paid occupation was different. As occupational mobility is associated with domestic responsibilities, women's fertility data are used in the search for stable sub-groups. The LS provides evidence on the mortality of women in such sub-categories of occupation. The Women and Employment Survey was conducted in 1980 by OPCS. It is a representative sample of 5 5 8 8 women who were aged between 16 and 59 in 1980 and resident in a private household in Britain. It achieved a response rate of 83 per cent. The Women and Employment Survey collected retrospective lnformation on marriage, child bearing and job histories. Each job 5 that a woman had ever had was allocated to one of the 12 occupational categories in the survey's own classification mentioned above. For this paper (and the published report) child care occupations are combined with semi-skilled domestic work, resulting in 1 1 occupation categories. These are shown in Table 1 . This classification has been used for all analyses of both sources presented here. The LS is a one per cent sample of the population enumerated in England and Wales at the 1971 Census. Its geographical coverage therefore differs from the WES which also includes Scotland. Though it might have been desirable to omit the Scottish women from the WES analyses to maintain strict comparability they were retained to boost sample size. There is no evidence of Scottish employment patterns differing greatly from the rest of Great Britain. Another difference between the data sets is lack of information from the 1971 Census on childbearing by women who had never married. As the proportion of mothers of dependent children who were never married women was low in 1971 (1%) this discrepancy is not serious (OPCS, 1986). From the job histories in WES, we extracted occupational careers since leaving school until April 1971, the time from which survival and mortality can be followed up in the LS. Thus for each woman in WES her occupation at April 1971 was considered as the current occupation she would have reported had she been in the LS. Details of her previous jobs, including whether they were full time or part time, and their type of occupation were recorded along with the duration. An occupational history to 1971 can only be considered for women aged less than 5 1 in 1971 as data in 1980 was only collected from women aged under 60. Marital and fertility histories up until April 1971 were also extracted. 'The 1971 Census classified 49 per cent of women aged 15-64 to a specific occupational unit. W E S suggests that a much larger fraction would have had paid occupation(s) at some time. The low proportion of women classified by the Census is due to the way occupational information was collected: on 'main employment last week, or of the most recent job if retired or out of work' with the caveat that 'housewives who did not have a job last week should not provide any information'. Thus, data on women's occupation in the 1971 Census principally relates to women in a job in the week before Census, and not many others who might have reported an occupation in WES. To 6 ensure comparability we restrict our analyses to those in employment in April 1971. Information gained from the WES is linked to groups of women in the LS, not individual LS members. In the LS information collected at death registration is linked to individuals' Census records. In this paper we use data obtained in the follow up sample of women aged 16-49 at the 1971 Census, to examine all cause mortahty in the period 1976-1989. The upper age limit was chosen to maintain consistency with the analyses from the Women and Employment Survey. Deaths in the period 1971-75 have been discarded to allow for immediate effects of health selection. Persons in employment tend to be healthier than their counterparts because a certain level of health is required to remain in employment (Fox and Collier, 1976). Within our sample, who were all in employment, we will have missed those who had already dropped out of employment altogether because of poor health. By discarding evidence on deaths in the immediate follow-up period we are making some allowance for any health selection of the less fit into certain occupations, eg those that are relatively undemanding and can be done part-time. 2. Classification of Occupations The classification of occupations offered in the LS in 1971 was the Registrar General's Classification of Occupations 1971. We recoded these as far as possible, to the WES framework set out in Table 1, which had been derived fiom the 1980 RG classification of occupations. This presented some problems. The RGs classification had 223 occupation codes in 1971 and 3 5 1 codes in 1981. Thus, occupations that had formed one occupation code in 1971 often formed several in 1981 destined for different WES categories. Arbitrary allocations had to be made in a few cases. The occupations of women in 1971 according to the LS and the Women and Employment Survey are compared in Table 2. Distinct differences can be observed particularly in the apparent underrepresentation of skilled workers (and corresponding over-representation of semi-skilled domestic occupations) in the WES. These differences could arise in part from the difference in classification discussed above or in recall or sampling biases in WES. Up to a point, they are also consistent with WES using more inclusive definitions of paid employment, covering jobs, such as mail order agent, or child minder, often not recorded in the Census (Joshi and Owen, 1987). The two LS 7 occupational distributions shown in Table 2 compare all employed women alive in 1971 with those who survived to 1980. As they are broadly similar, this suggests that occupational differences in survival are not a major source of the discrepancy. Indeed, it also suggests that mortality differentials among women's occupations thus defined are not very startling. METHOD The analysis is in two parts: occupational mobility in W S and the LS; and occupational differentials in mortality and morbidity, based on the LS. 1. Occupational Mobility Evidence on occupational mobility or stability up to April 1971 was taken from the Women and Employment Survey. The aim is to identifjl occupations in which the probability of a cross sectional survey recording long term incumbents is high. These would be occupations without large numbers of occupationally mobile workers. The proportion of time in spent paid employment up to 1971 in the occupation current at 1971 is the main measure used. We do not mclude time spent out of the labour force, in the denominator as we are particularly interested in occupational change within paid work. For occupations with high proportion of other work behind them, the highest and lowest occupations ever held were also considered to see whether the mobility was upward or downward. The analysis is carried out for women who were under 5 1 and in employment in 1971. The upper age limit is imposed by the design of the survey, the restriction to women in employment results from the incomplete coverage of non-employed women's occupation in the Census. We also have a direct measure of occupational stability in the LS for those who recorded occupations at both the 1971 and 1981 censuses. Definition changes between the two Censuses pose some difficulties. Moreover, this measure only indicates what a woman was doing at two points in time, it does not necessarily follow that the woman has only worked in this occupation for ten years. 2. Occupational Mortality Mortality for different groups was compared using odds ratios estimated on LS data. These were estimated using logistic regression, of a binary outcome variable, standardising for age at entry 8 and period of follow up, a method developed for the LS by Weatherall (1992). Effects of explanatory variables are estimated relative to a base category. The logistic model with main effects only was as follows: log(odds of death) = a + bl (period of follow up) + b2 (age in 1971) + b3 (occupation in 1971) + b4 (indicator of living standards) 'The basic model (Model 1) includes bl to b3, with extensions to b4 (Model 2), fitted over the sample who were in employment in 1971. Interactions were investigated by estimating parameters a to b4 in the sub-sample thought to be more stable according to the several measures of fifth type of variable indicating occupational stability. The model is fitted using binomial errors. The dependent variable is the log odds of death during a period of follow-up among survivors at the beginning of each period. Each of the explanatory variables was treated as a categorical variable. Occupation had 11 categories of which clerical workers were the base group. To standardize for age, terms were included for five year age ranges. To allow for the length of exposure and ageing after 1971, the period of follow up enters the model, grouped into three categories; 1976-1980; 1981-1985; 1986-1989. The last one ends at the latest date of deaths available at the time of writing. Car access was included in the model as a proxy for the household living standard, to determine whether any observed occupational differences might be due to particular hazards associated with that occupation could be accounted fkom a common lifestyle shared by women in that occupation. The household standard of living was proxied by a binary indicator of no cars or one or more cars. Car access was found by Moser et a1 to be a good discriminator of women's mortality (Moser, Goldblatt and Pugh, 1990) and is positively correlated with household income in the 1983 FES (Davies, Joshi and Clarke, 1993). 9 To allow for transient holders of Census occupations two other variables were also considered: the age of the youngest child and whether the woman worked full time or part time. A second dependent variable was also given some consideration, the log odds of being reported permanently sick at the 1981 census. This was to see if this alternative, and by no means ideal, indicator of poor health picked up or amplified any of the differentials discernible in the mortality rates. RESULTS 1. Occupational Mobility The penultimate column of Table 3 shows the proportion of time in paid employment spent in the occupation recorded in April 1971. For the whole sample of women under 5 1, this is 67%. It is under 50 per cent for just three occupational categories - other intermediate non-manual, the small sample of semi-skilled factory workers and other semiskilled workers other than those in domestic type service occupations. For women in teaching and clerical occupations over 80 per cent of their time in employment was in the current occupation. Shop assistants, the saleswomen in whom we are particularly interested, appear to be the fourth most transient category, having spent 55 per cent of their time in the same occupation. The last column of Table 3 shows the amount of time spent in any paid work. This averages 146 months, just over 12 years, for the whole sample, but tends to be above this for those who were doing the least skilled occupations in 1971, and less for those in occupations requiring longer education. In the analysis of this denominator by age shown in the bottom row of Table 3, the number of months of employment rises with age, though not by as much as age itself, as few of the older women had been employed continuously. The distribution of occupations by age reported in the body of Table 3 reveals occupational segregation exists between women of different ages. Older women tend to predominate in sales and services. Younger women, who tend to be better educated, are better represented in the (broadly defined) professions p o s e r , Goldblatt and Pugh, 1990). 10 Table 3 shows occupational stability varies between the age groups. In gerieral, the older women have spent proportionally less time in their 1971 occupation than younger women. They have had a greater chance of experiencing different occupations. Nevertheless, in several individual occupations ( nursing, shop assistants, skilled and domestic type work and unskilled) the histories are more homogeneous among the oldest age group than among those aged 35-44.This could possibly reflect patterns of re-establishment of original occupations after a more mixed occupational experience in mid-life. Controlling for age does not increase stability notably for any individual age group. Both of the larger groups which appear particularly upstable among all ages (other intermediate and other semi-skilled) also appear relatively unstable ib each constituent age group. This is despite their both being "rag-bag" categories, allowing job mobility within them to go undetected. I A woman's occupational history is also associated with her domestic responbibilities. Most women who had a child before 1971 had an interrupted employment histork. Table 4 shows the proportion of time in paid work spent in the occupation recorded in 1971, for women who had never or ever had a child, by age of the youngest child aged less than 15 in 1971, and for all mothers with no child under 15 in 197 1. Sample size was insufficient to bonsider the age of the woman and the age of the youngest child jointly. Table 4 clearly shows an effect of child bearing on occupational stability.) Childless women had spent more than three quarters of their employment in their 1971 occupa/tion compared to 56% for all mothers, around 60% among mothers with children under 15. With the exception of professional women, an extremely small sample in "top jobs", women who 1 have never had a child show a higher proportion of time in the labour force spent in the 1911 occupation than the average mother in each occupational group. Teaching appears as occupation which is relatively easy to combine with motherhood. In the three occupational grqups which showed the most instability in Table 3, (other intermediate, semi skilled factory wQrkers and other semiskilled) women without children showed relatively low stability, oust unfer half all time in their 'own' occupations), but the occupational histories were particularly unstbble among mothers of dependent children. Mothers of children under 15 working in other semi-skilled occupations had spent only about one third of their paid employment in such jobs. Tablt 4 also shows stability 11 rates for women who had no child under 15 in 1971. For most occupations the rates well above those for women whose children are under 15, and close below those of the sub-set who had never had a child. Many mothers return to part time employment. Table 5 shows occupational stability by whether the woman was working full or part time in 1971. Part time work is self defined. Full-timers had spent 71% of their careers in their current occupation and part-timers 60%. This direction of difference applies generally across all occupations with a sample size in each group over 15. In the unstable group other semi-skilled (also the small sample of semi-skilled factory workers), selecting full timers increases stability but does not raise the proportion of time spent in these occupations to more than 45 per cent. Thus, up to a point, the practice of ignoring part-timers at death registration has some rationale. They tended, at least in 1971, to have more of a mixed record behind them than most full-timers, but full-timers did not have totally consistent histories either. Along with age, it is clear from Tables 4 and 5 that occupational stability is related, for most occupations, to the absence of dependent children and to full time employment. In terms of identifLing an efficient global discriminator of occupational stability it is worth considering whether working full time or not having a dependent child is a better measure than working full time and not having a dependent child. These results are shown in Table 6. A dependent child is defined as aged under 15 in 1971, then the minimum school leaving age. Overall, for women in full time jobs without dependent children, 75 per cent of their time in paid employment was in the occupation recorded in 1971 compared with 70 per cent of the broader group who either have full-time work or no dependent child, and 67% for the sample as a whole. Neither of these criteria is spectacularly successful at picking out the unmixed occupational history. Table 6 shows that working full time and not having a dependent child is a better discriminator of occupational stability, as far as it goes, within most, but not all occupational categories. For other intermediate occupations never having a child is marginally better (see Table 4). For teachers, the envelope of either full time employment or no dependent children is better at isolating unmixed histories than their intersection. Full time employment alone is 12 marginally better still. Where did the mobile workers come from? Many of the other intermediate non-manual jobs had been upwardly mobile, only 3% of the time before 1971 had been in the top 3 categories. Many of these other intermediate jobs are more senior office positions which recruit from clerical workers. Of the time up to 1971 44%had been spent in the same occupation, 34% in clerical jobs, 7% in shop jobs and 12 % in various manual jobs. Little of the mobility into this occupation was ' downwards' because the occupations above it, for example, teaching or nursing, tend to need specific qualifications and training, which help protect against downgrading (Joshi and Newel1 1987). The other large mobile group, other semi-skilled had spent the majority of their time semi-or unskilled manual jobs (35% plus the 41% in their own category), but downward mobility was also evident. 25% had been in 'higher' occupations, 10% in clerical and jobs. The small group of factory workers had very mixed previous experience, both higher and lower, with clerical jobs accounting for almost as much time (27%) as their current occupation (3 1%). The time behind shop workers not spent in shop work (45% altogether) was predominantly in lower manual occupations (27%). Higher occupations, almost all clerical, accounted for 14% of the balance, or 19% if factory and skilled manual work are ranked above shop work (see Joshi 1984). Thus, although women's employment histories have a reputation for downward mobility, this was not a majority experience among those employed in 1971. In a cross-sectional snapshot, most women were in occupations in which they had spent the majority of their working lives. Supplementary evidence on occupational stability is drawn form the LS itself in Table 7. This takes women under 5 1 in 1971 who were in paid employment at both Censuses and shows those recorded in the same occupation group at both Censuses as a percentage of women in that occupation in 1971. Overall 58 per cent of women in paid employment at both Censuses were recorded in the same occupation at both points in time, a lower degree of homogeneity here, where occupation is recorded for only two points, than the 67% recorded in Tables 3-6, where every month of the occupation history is counted. Probably for this reason, Table 7 has generally smaller estimates throughout than Table 3, broadly retaining the same pattern. The rate of 13 repeating occupation at two Censuses ranges from 90 per cent of teachers to 11 per cent of intermediate workers, of whom an exceptionally small number satisfjr the criterion of having occupations at two Censuses. Apart from this surprisingly small group, those with relatively low rates of repetition are all of the manual categories, not just other semi-skilled workers and factory workers (better represented here), who appear unstable in both analyses. Of those who were shop assistants in 1971 with employment at 1981, half were shop workers at both dates. In summary, many occupational categories appear to be reasonably stable. Teachers, clerical workers, professionals, nursing, and semi-skilled domestic work all had incumbents in 1971 who had worked more than seven tenths of their paid time in the same category. In all but the last of these, more than half those with two census jobs had repeated the 1971 occupation at the 1981 census. The three most unstable occupational categories in the WES analysis were other intermediate, semi-skilled factory work and other semi-skilled. Each accounted for well under half their incumbents' paid time. Shop assistants, along with the unskilled, had spent about half their working lives in their 1971 occupation. As indicators of occupational stability, both working full time and not having a child aged under 15 in 1971 appear to discriminate roughly, though not perfectly. Both these two categories tend to over-represent younger women. 2. Health Outcomes and Occupation in the LS Occupational Mortality 1. All Employed in 1971 The sample of 49,088 women who were classified in the LS as being in paid employment in 1971 have been followed up to record any mortality from 1971 to 1989. The relative odds of dying by occupation at 1971 are shown in Table 8 controlling for age and period of follow up (1976- 1989). This Table also presents the estimates from the extended model controlling for material circumstances in terms of car access. The results of these analyses are also shown in Figure 1. Housing tenure was considered as a second indicator of material circumstances but proved to be highly collinear with car access in this sample of only employed women. Seven occupations are shown in model 1 to have significantly (p<0.05) raised mortality ratios compared to the base group clerical workers: all the manual groups, shop assistants (as expected), 14 and rather unexpectedly, the intermediate non-manual group, who, on the whole, have higher pay and status than clerical workers. Nurses are the only occupational group to suggest a lower risk of dying than clerical workers, although the relative odds are slight and not significant. The range of occupational mortality risks is moderate. The worst is scarcely more than 50% more than the best. In comparison SMRs among male occupations vary by a factor around 2 from worst to best among occupation orders I to XXVI (Goldblatt and Fox, 1990) or between male Social Classes I and V (Fox Goldblatt and Jones 1990). The effect of adjusting for car access is minimal. The adjusted estimates are plotted as crosses more or less alongside the unadjusted estimates and their confidence limits in Figure 1. The mortality differential for professionals, teachers and intermediate non-manual workers is slightly raised, remains the same for nurses, and reduces for all other occupations. When material circumstances were introduced into Goldblatt and Fox's (1990) model of male occupational mortality, the impact was very similar, a minor reduction in the excess mortality of manual occupations. In that study the level of living was proxied by housing tenure. All of the subsequent models were also estimated including car access but as the effect on mortality ratios was minimal they have not been presented. Only a very small part of the disadvantages of the manual occupations can be attributed in this way to lower household living standards, as proxied by car access. The variable itself has an independent and significant effect on mortality. Those without car access were, all else equal, 27% more likely to die than those with cars. Table 8 includes all women who were in an occupation in 1971. A number of these women would be short term incumbents of these occupations and may be distorting any mortality effects associated with that occupation. Evidence presented above showed that women who were working fill time in 1971 had spent a higher proportion of their time in paid employment in the occupation they were recorded in 197 1 than part time workers. Confining the sample to the 64% employed fill time (in Table 9) only makes a slight difference to the estimates compared with all women. The gap between worst and best has opened - as we expected given our exclusion of the more occupationally unstable part-timers, but it has hardly opened very wide, from 57% to 70%. 15 Table 4 showed that women who did not have a dependent child were more likely to be occupationally stable than women who did. This sub-sample is heterogenous, it includes women who have not started child bearing, women whose chddren are independent and women who have remained childless. Table 9 also shows the mortality analysis for women who did not have a child under 15 in 1971. Considering this sub-group of women has only a modest effect on the mortality ratios compared with all women, although, as with hll time workers, the mortality ratios tended to increase for occupations that had a significantly raised mortality ratio. The excess of worst over best occupation is 66%. There is of course considerable overlap of the two sub samples considered in Table 9. 80% of the workers with no dependent children were hll-timers. 77% of the fill-timers had no dependent children. An analysis (not shown) of those who satisfied both criteria did not sharpen the estimated mortality differentials, despite the fact that women working hll-time without dependent children were the most occupationally stable group in Table 6. 2. Women in same type of jobs, 1971 & 1981 Table 9 attributes occupational stability to the women in the LS on the basis of inference from the Women and Employment Survey. The sub-groups that were identified as likely to be stable have only slightly dif5erent occupational mortality patterns from all women. It could be that this is not a good measure of prospective occupational stability, so we turn to the alternative indicator of stability after the 1971 Census. Table 10 and Figure 2 show the mortality ratios for women who were recorded in the same occupation in 1971 and 1981 controlling for age and period of follow-up. Unlike the previous tables the period of follow-up is limited to 1981-1989, with 1981-85 the base time period. The effect of using this indicator of occupational stability is to reduce the overall number of occupations showing a significant mortality differential to just two, semi-skilled factory workers (odds ratio 1.77) and other semi-skilled workers (1.85). The range from best to worst occupation is now over 2, the same range as found among men's occupations, though the low parameter is not well determined. The consistently high mortality ratios shown by semi-skilled factory workers is consistent with other evidence discussed below. Two proxies for occupational stability are combined in Table 11, taking women who were in the 16 same occupation in 1971 and 1981 and who worked full time in 1971. In comparison with Table 10, two more occupation groups, skilled manual workers and semi-skilled domestic workers, show significantly higher mortality ratios than clerical workers. In addition other intermediate occupations have an odds ratio above 2, although this is based on just 6 deaths. Table 11 also shows the mortality ratios for women who were in the same occupation in 1971 and 1981 and who did not have a dependent child in 1971. Again the occupational differentials are sharpened compared to those shown in Tables 8, 9 and 10. This time it is other semi-skilled occupations which show a relative odds ratio above 2 (based on 16 deaths). The estimates of Table 11 suggest greater, though less well-determined differentials than Tables 7 and 10. The high upper confidence limits associated with the high estimates meant it was not possible to plot the whole set on the common scale with Figures 1 and 2. Permanent Sickness ‘Thus far there does not appear to be a great deal of clearly patterned variation in the mortality of women by occupation. The identification of occupationally stable sub-groups increases the range of estimated mortality differentials to something like that found among men, but not in a wholly consistent or very well determined manner. It is possible that mortality analysis may fail to detect any health hazards of a particular occupation that lead to general ill health without being life threatening. There is partial evidence on morbidity in the LS in the form of responses on permanent sickness in the Census question on economic activity. The identification of permanently sick women in the 1981 Census only refers to women who have not been described as housewives or retired. Neither does it apply to anyone who, though in poor health, is still in employment. While the number of women who have retired will be small in this sample, under 60 in 1981, the exclusion of housewives who have left the labour force due to ill health is likely to understate poor health and may bias associations that can be detected. Overall, 1.3% of women who were recorded in an occupation in 1971 were permanently sick in 1981. This ranged fiom none of the 393 professional women to 2% of the 4300 semi-skilled factory workers. ‘Table 12 and Figure 3 show the results fiom a logistic regression with permanent sickness in 1981 as the outcome variable. Differentials are somewhat more marked than in the mortality analysis, 17 the ratio of worst to healthiest being 2.4. Manual occupations showed significantly higher chances of recording permanent sickness than clerical and most other non-manual groups, by factors approaching 2 (2.3 for semi-skilled factory workers). Thus far, a similar and slightly more marked pattern than for mortality. The odds ratio of 2 for nurses' permanent sickness contrasts with their relatively low mortality Semi-skilled factory workers showed an odds ratio above two. Salesworkers were on a par with 'other intermediates,' above clerical workers, but not significantly so. Professional women had to be excluded from the model because none of this small group was recorded as permanently sick in 1981. Table 13 shows the estimates of being permanently sick in 1981 for two occupationally stable sub-samples: full time workers and women with no dependent children. These sub-groups show similar estimates to all women. This suggests that the patterns of permanent sickness by 1971 occupation cannot have been systematically different for the excluded groups presumed to have mixed occupational histories behind them. DISCUSSION This paper attempts to overcome two problems in the study of female occupational mortality. We adopt a classification that had been developed specifically for women's occupations in the Women and Employment Survey. We use that survey itself to identify sub-groups of women that were occupationally stable. The LS was used to consider the occupational mortality of all employed women and for identifying occupationally stable sub-groups of women. One third of all employment histories up to 1971 had been in occupations other than the one then current. As this rate varied across demographic and labour market groups, it was possible to identifL both relatively stable occupations and sub-groups of women who were likely to have spent a large proportion of their employment experience in the occupation recorded at Census. Professionals, teachers and nurses appeared to be stable categories; full time workers and women without dependent children were occupationally stable. Three occupations were consistently unstable, other intermediate workers, semi-skilled factory workers and other semi-skilled workers. The mortality analysis suggested that differentials between women's occupations are modest, and little affected by household economic resources. They may, however, approach those recorded 18 for men's among the sub-groups identified as occupationally stable. Semi-skilled factory workers, who had been shown to be occupationally unstable, were one of the occupations that consistently showed significantly high mortality relative to clerical workers. For other categories estimates were not always consistent or well determined. Saleswomen did not emerge as being particularly endangered. On a health hierarchy we would place them on a par with skilled manual workers, though on a pay hierarchy they would be placed beneath them. Moreover, there was some evidence of a greater occupation effect for women becoming permanently sick rather than dying. We find some support for the idea that occupational instability obscures underlying differentials in mortality associated with women's occupation, though we have also found that 'unstable' occupations (intermediate, semi-skilled factory and other semi-skilled) tend to show raised mortality even before the likely mobile cases have been discarded. The findings here of some occupational differentials in women's mortality are more significant than in earlier results fiom the LS (Moser, Goldblatt and Pugh, 1990, Sloggett et a1 forthcoming). The former used a finer occupational classification, the latter a broader one than used here. A possible reason for the different findings is that neither of the other studies had as much evidence on mortality, ten and fifteen years of follow-up respectively as opposed to nearly nineteen. A reason to be cautious about comparison is that this analysis is confined to employed women under 5 1 in 1971. Once non-employed women are considered, the importance of household characteristics appears to gain salience. We found fairly consistent high mortality and sickness rates among women doing semi-skilled factory work. This is itself consistent with Moser and colleagues' finding of significantly raised mortality in engineering and allied trades (Moser et al, 1990). It is also consistent with evidence such as that in the Women and Employment Survey which showed a positive correlation between stress and factory work (Joshi, 1984). Our finding of relatively high mortality among the nonmanuals in the 'other intermediate group' is unlikely to be explained by our finding that much of the experience behind these workers was in lower ranked, mainly clerical occupations. It is more likely to have a common explanation with that of the relatively raised longstanding morbidity reported by Arber (1989) for women in SEG 2. This category comprises Employers and Managers who are classed within W S category 'other intermediate'. Such jobs are often not 19 traditionally female. If it is indeed such jobs responsible for the higher risk, it could take more time to accumulate the evidence. Analysis of self-assessed health in the 1984 Health and Lifestyle Survey found that all manual groups and sales women were more likely to consider themselves in poor health (Macran et al, 1994, Macran et al, 1993). However the latter paper finds that occupational patterns are not consistent across types of health indicator. Counts of illness symptoms were relatively high among those with factory, unskilled or intermediate occupations. The unskilled were generally in the worst health, and shop workers emerged as being between manuals and other non-manuals on most counts, much as they do here. Why have we not found stronger evidence of specific occupational hazards? One answer could be that there are indeed hazardous occupations which are covered up by the classification used here. Another is that women do not generally get dangerous types of work. Occupational hazards may expose a worker to accidents which may result in death or they may expose a worker to substances which are detrimental to health in the long term, for example asbestos. Women rarely work in many occupations where the risk of accidents is high, for example, mining or building. Unequal access to the labour market, and conventions of what work is suitable for women seems to protect most of them against many types of hazard. A third possibility is that intermittent job histories may protect many women fi-om prolonged exposure to noxious substances over time. We have shown the usefblness of the classification of occupations used here, derived specifically for considering women's occupations, and also its limitations. The occupation groups comprise a wide range of jobs. It is possible that using a finer breakdown of individual jobs would result in a more notable effect of occupation on mortality. There may, for example, be hidden hazards for the small minority of women in 'mens jobs' obscured by their small number. Work summarizing the occupational mortality ofwomen fi-om the 1970-72 Decennial Supplement reports that 63 per cent of the occupational units (jobs) considered had at least one significantly raised mortality ratio among the causes of death examined (Roman, Beral and Inskip, 1985). A hrther lunitation of our use of this classification of occupations is the low proportion of women 20 who could be classified. The Census only effectively collects an occupation for women currently in the labour force. The 1971 Census does not record, for 51% of women aged 15-64 whether they had ever been employed, nor what their last, or most usual, occupation was. If there is an occupation effect on morbidity or mortality it is plausible that some of their victims would have left the labour force (witness the analysis of permanent sickness presented here). Such possible occupation effects cannot be examined without knowledge of previous occupations, or even longer follow-up in the LS. Data collection techniques for the recording of women's occupations could be improved. Until then the effect of occupation on women's mortality remains, at lease partially, obscured. There is also scope for better techniques of collecting data on womenls health. We can reach some conclusions beyond recommending better data. The indications of this study are that females as well as males in manual jobs face higher mortality than clerical workers. For neither women nor men is the occupational risk attributable to household circumstances. This strongly suggests some health disadvantage attaches to the occupation rather than the incumbents. There is not yet much long term evidence on women entering non-traditional occupations, but it should be particularly important to monitor their health, as well that of women employed longterm in manual work. 21 References Arber, S, (1989) 'Gender and class inequalities' in Fox, A.J.(ed) Health Ineaualities in European Countries, Gower, Aldershot, 250-279. Dale, A., Gilbert N. and Arber, S. (1985) 'Integrating women into class theory' Sociolom Vol 19 no 3, 384-409. Davies, H., Joshi, H. and Clarke, L. (1993) Is it cash the deprived are short of! BSPS Conference paper, Newcastle, September 1993. Dex, S . (1987) Women's Occupational Mobility, Macmillan Press, London. Employment Gazette. (1992) 'Women and the labour market: results fiom the 1991 Labour Force Survey', Employment Gazette, 433-459, HMSO,London. Fox A.J. and Collier, P.F. (1976) 'Low mortality rates in industrial cohort studies due to selection for work and survival in the industry'. British Journal of Preventative and Social Medicine, 30, 225-230. Fox A.J., Goldblatt P. and Jones D. (1 990) ' Social Class Mortality Differentials: Artifact, Selection of Life Circumstances?' Goldblatt, P. (eds) Longitudinal Study: Mortalitv and Social Organisation, LS Series no 6, London: HMSO 100-108 Goldblatt, P. (1990) eds Longitudinal Studv: Mortality and Social Orpanisation. LS Series no 6, HMSO, London. Goldblatt, P. and Fox, A.J. (1990) 'Mortality of men by occupation' In Goldblatt, P. (eds) 1,ongitudinal Study: Mortality and Social Organisation. LS Series no 6, London: HMSO 110-129. Hakim, C. (1981) Job segregation: trends in the 1970s. Employment Gazette December, 521-533. Joshi, H. and Newell, M. L. (1987) Job downgrading after childbearing. In M Uncles (ed) London Papers in Regional Science 18. Longitudinal Data Analysis: Methods and Applications. London: Pion 89- 102. Joshi, H. (1984) Women's participation in paid work: further analysis of the Women and Employment Survev, Research paper no. 45, Department of Employment. Joshi, H. and Owen, S. (1987) 'How long is piece of elastic? The measurement of female activity rates in British censuses, 1951-1981', Cambridge Journal of Economics, Vol 11 no 1, 55-74. McDowall, M. (1983) 'Measuring women's occupational mortality', Population Trends no 34, 25-29, HMSO, London. Macran, S., Clarke, L. and Joshi, H. (1993) Women's Health: Dimensions and Differentials. Paper presented at the XXII General Population Conference, IUSSP, Montreal. 22 Macran, S., Clarke, L., Sloggett, A. and Bethune, A. (1994) Women's Socio-economic status and selfassessed hea1th:identifjring some disadvantaged groups, Sociolop of Health and Illness, Vol. 16 No 2, 182-208. Martin, J. and Roberts, C. (1984) Women and employment: A lifetime perspective, Department of Employment/OPCS, HMSO, London. Moser, K., Goldblatt, P. and Pugh, H. (1990) 'Occupational mortality of women in employment' In Goldblatt, P. (ed) Longitudinal Studv: Mortalitv and Social Oryanisation. LS Series no 6, 129-144, HMSO, London. Moser, K., Pugh, H. and Goldblatt, P. (1990) 'Mortality and the social classification of women' In Goldblatt, P. (eds) Longitudinal Study: Mortality and Social Organisation, LS Series no 6, 145-162, HMSO, London. OPCS (1978) Occupational Mortality Decennial supplement 1970-72, DS no. 1, HMSO, London. OPCS, (1986a) Occupational Mortality: The Registrar General's Decennial Supplement for Great Britain 1979-80.1982-83 London :HMSO OPCS, (1986b) General Household Survey 1984, HMSO, London Registrar General (1908), Supplement to the sixty-fifth annual report of the Registrar General. Part 11, HMSO, London. Registrar General (1 93 l), Decennial Supplement: England and Wales 193 1. Part 1 1a, HMSO, London. Roman, E., Beral, V. and Inskip, H. 'Occupational Mortality among women in England and Wales' British Medical Journal, 292, pp 194-196. Sloggett A., Clarke L., Joshi H. E. and Eames (forthcoming) Ecological and Individual Factors in Women's Mortality in England. in Clarke J and Zaba B (eds) Environment and Population Change. Liege: Ordina (in press). Stevenson, T.H.C. (1927) Decennial Supplement, England and Wales, 1921, HMSO, London. Stewart, M. and Greenhalgh, C. (1984) Work history patterns and the occupational attainments of women. Economic Journal, Vol. 94,493-519. Roberts, H. and Barker, R. (1986) The social classification of women, LS working paper no. 46, Social Statistics Research Unit, City University, London. Weatherall, R. (1992) A comparison of mortality measures in the OPCS Longitudinal Study, LS User guide no. 8, Social Statistics Research Unit, City University, London. Weatherall, R., Joshi, H. and Macran, S. (1993) 'Double burden or double blessing? Employment, Motherhood and Mortality in the Longitudinal Study of England and Wales', Social Science and Medicine. 285-297. 23 Table 1 Women's Occupational Groups 1 . Professional occupations Barrister, solicitor, chartered and certified accountant, university teacher, doctor, dentist, physicist, chemist, pharmacist, dispensing optician, qualified engineer, architect, town planner, cMl servant (Assistant Secretary and above). 2. Teachers Primary and secondary school teacher, teachers in further and higher education (not universities), head teacher, nursery teacher, vocational and industrial trainer. 3. Nursing, medical and social occupations SRN, SEN, nursing auxiliary, midwife, health visitor, children's nurse, matron, dental nurse, dietician, radiographer, physiotherapist, dispenser, medical technician, houseparent, welfare occupations (inc social workers), chiropodist, occupational therapist. 4. Other intermediate non-manual occupations Civil Servants (Executive Officer to Senior Principal and equiv. in central and local govt), computer programmer, systems or 0 & M analyst, librarian, surveyor, personnel officer, manager, self-employed farmer, shop keeper, publican, hotelier, buyer, company secretary, writer, journalist, artist, designer, windowdresser, entertainer, musician, actress. 5. Clerical occupations. Typist, secretary, shorthand writer, clerk, receptionist, personal assistant, cashier (not retail), telephonist, office machine operator, computer operator, punch card operator, data processor, draughtswoman,tracer, market research interviewer, debt collector. 6. Shop assistant and related sales occupations People selling goods in wholesale or retail establishments, cashier in retail shop, check-out and cash and wrap operator, petrol pump attendant, sales representative, demonstrator, theatrelcinema usherette, programme seller, insurance agent. 7. Skilled occupations Hairdresser, manicurist, beautician, make-up artist, cook, domestic and institution housekeeper, nursery nurse, travel stewardess, ambulance women, van driver and deliveries, baker, weaver, knitter, mender, darner, tailoress and dressmaker (whole garment), clothing cutter, milliner, upholsterer, bookbinder, precision instrument assemblers, laboratory assistant, driving instructor, policewoman. 8. Semi-skilled factory work Assembler, packer, labeller,grader, sorter, inspector, machinist, machine operator, people wrapping, filling or sealing containers, spinner, doubler, twister, winder, reeler. 9. Semi-skilled domestic work Waitress, barmaid, canteen assistant, childminder, playground or playgroundsupervisor, nanny, au pair, home help, care attendant, ward orderly, housemaid, domestic worker. 10. Other semi-skilled occupations Agricultural worker, groom, kennel maid, shelf-filler, bus conductress, ticket collector, postwoman, mail sorter, laundress, dry cleaner, presser, mail order and catalogue agent, market and street trader, collecting saleswoman, traffic warden, telephone operator. 11. Unskilled occupations Cleaner, charwoman, kitchen hand, labourer, messenger. Source:Martin and Roberts (1984) 24 Table 2 Distribution of occupation,Women and Employment Survey and OPCS LS, 1971 Women aged 16-59 in 1980 I Occupation Survey WES N=2509 OPCS LS Survivors to 1980 N=496501 Professional 0.4 0.9 1.o Teacher 4.8 5.3 5.3 Nursing 6.6 6.6 7.3 Other Intermediate 4.2 5.9 6.0 Clerical 29.5 32.1 31.8 Shop Assistant 11.7 8.2 10.6 Skilled occupation 7.8 9.8 8.1 Semi-skilled factory 1.7 10.8 9.8 22.6 9.9 9.9 Other semi-skilled 7.3 4.1 4.0 Unslulled 3.4 6.4 6.3 Semi-skilled domestic - OPCSLS N=49088 Source: Women and Employment Survey, LSHTM analyses (WES)and OPCS Longitudinal Study (LS) as analysed at LSHTM 'The sample size is larger for h s calculation because it does not exclude women who have missing data on the other co-variates. 25 Table 3 Percentage of time in paid work in the occupation current in 1971, by age group and 1971 occupation, women in employment in 1971 15-24 25-34 - n 2 79 3. Nursing 4. Other intermediate 35-44 - 74 42 75 29 84 121 39 37 67 39 37 37 60 44 87 305 77 165 73 137 81 740 130 55 69 66 50 59 66 73 29 55 66 293 32 54 64 195 143 149 11 41 16 31 42 149 69 105 75 141 73 568 164 32 42 42 41 56 41 184 171 44 19 59 17 62 21 57 84 172 76 77 1 65 56 1 63 573 67 2509 146 31 91 53 79 56 68 68 23 80 46 5. Clerical 91 334 6. Shop 7.Skilled occupations 8. Semi-skilled factory 9. Semi-skilled domestic 76 82 91 79 150 74 116 10. Other semi-skilled 41 28 35 11. Unskilled 54 22 All 81 833 Months 49 79 Average months per employed woman n 10 n 6 1 . Professional 2. Teaching n 5 All ages under 51 n 2 - ?40 45-5 1 Y O 5 62 YO - 80 7 126 202 TO 268 Source: WES 26 YO YO 96 117 117 168 32 146 Table 4 Percentage of time employed in the same occupation as that observed in 1971, by selected descriptions of fertility history Never had a child Ever had a child (Age of youngest child) 0-4 - n 6 YO - n 4 83 79 79 3. Nursing 4. Other intermediate 73 49 93 5 . Clerical 6. Shop 7.Skilled occupations 8. Semi-skilled factory 9. Semi-skilled domestic 5-9 All with no dependent children 10-14 - n 2 YO - n 1 42 92 10 74 15 5 73 11 16 57 28 18 52 75 33 26 6 66 40 14 88 456 75 284 86 60 80 84 67 75 119 50 174 36 60 95 45 57 53 100 49 62 49 6 30 36 80 202 71 366 67 71 70 118 10. Other semi-skilled 49 31 39 153 36 34 31 11. Unskilled 63 30 55 54 38 12 77.5 1176 56.4 1333 60.9 294 Y O 1 . Professional 2. Teaching All Y O Yo 64 81 n 111 5 70 48 72 66 83 530 50 57 36 168 22 60 70 9 49 12 62 66 78 3 13 48 36 31 48 71 55 12 59 18 61 42 60.2 399 59 270 72 1546 18 9 - 12 Source: WES 27 24 YO n 1 6 91 71 131 Table 5 Percentage of time in paid employment in same occupation as 1971, by occupation and whether worked full or part time in 1971 Occupation in 1971 Full Time Part Time YO n YO n - 8 - 2 Teachers 83 101 74.7 19 Nursing 70.9 115 66.3 49 Other intermediate 43.4 96 - 10 Clerical 82.3 591 77.1 147 Shop Assistant 59.2 172 49.5 116 Skilled occupation 68.4 195 58.9 41 Semi-skilled factory - 11 28 30 Semi-skilled domestic work 75.9 33 1 70.4 115 Other semi-skilled 44.6 59 38.8 123 Unskilled 62.3 55 47.5 29 All 71.4 1690 60.2 79 1 Professional Source: WES 28 Table 6 Percentage of time in paid employment in same occupation as 1971, by occupation, full time employment and no dependent child 1971 Occupation in 1971 Full time OR no dependent child Full time AND no dependent child YO n Y O n - 6 74.5 8 Teacher 80.7 82 82.7 110 Nursing 71.0 115 70.2 131 Other Intermediate 47.2 69 43.8 97 Clerical 83.8 487 81.8 627 Shop Assistant 67.2 125 56.5 213 Skilled occupation 71.4 8 68.6 165 Semi-skilled factory 54.7 114 43.1 14 Semi-skilled domestic 79.6 239 76.0 394 Other semi-skilled 46.3 35 46.9 94 Unskilled 63.5 55 60.1 58 All 74.6 1298 70.3 191.1 Professional Source: WES 29 Table 7 Proportion of women by occupation in 1971 who were recorded in the same occupation in 1981 Occupation 1971 n % in same occupation 1981 Professional 159 61.9 Teachers 1663 89.6 Nurses 1736 74.6 113 11.4 Clerical 6952 77.1 Shop Assistant 1137 49.6 Shlled Manual 1244 42.1 Semi-skdled factory 1444 45.2 Semi-Skilled domestic 1388 43.5 Other Semi-Skilled 383 32.7 Unskilled 830 41.1 17049 58.2 Intermediate All Source: OPCS Longitudinal Study 30 N c c 0 d 3 P 2 I x d r. N 1 N x 2 r. 1 01 W 9 Y 2 I- 2 n z n 0 2 01 Y 3 n r. c1 0 9 r. 3 N $ r. $ I 2 d 2 d 0 Y 3 2 I= N W o n m 0 o t01 N o s o 0 W 2 I I- -? e4 Y 2 N 3 In 3 I 0 9 4 r! 3 I- 00 0 0 I z 0 2 W 0 n 0 F d n ? 2 3 n Y N 2 Y 1 I 01 01 U d d x n 2 b d P N 01 c? n m 0 9 I 0 0 01 1 \4 2 N 0 8 8 0 0 m -? 1 0 W 01 I- 0 9 r. 2 t- m .. H0 0 U ?, L m Table 9 Mortality odds ratios by occupation adjusted for age in 1971 and duration of follow up: full time workers and women with no dependent children 1971. Full Time employed No dependent children N=31291 Occupation in197 1 Deaths 76-89 N=30066 O.R. t P Deaths 76-89 O.R. t P Professional 10 1.os 0.15 0.881 12 1.29 0.86 0.388 Teachers 79 1.23 1.64 0.1 57 1.12 0.81 0.42 Nurses 65 0.94 -0.42 0.676 61 0.96 -0.25 0.8 Intermediate 123 1.28 2.33 0.02 107 1.29 2.29 0.022 Clerical 330 1.oo 358 1.oo Shop Assistant 92 1.32 2.34 0.02 137 1.25 2.17 0.03 Skilled Manual 167 1.26 2.42 0.015 168 1.31 2.85 0.004 Semi-SkilledFactory 206 1.5 4.45 0 21 1 1.55 4.92 0 Semi-skilleddomestic 101 1.6 4.05 0 143 1.33 2.78 0.005 Other Semi-Slulled 77 1.6 3.66 0 78 1.59 3.63 0 Unskilled 46 1.43 2.25 0.025 105 1.54 3.81 0 Source: OPCS Longitudinal Study 32 Table 10 Women in the same occupation 1971 and 1981: Mortality odds ratios controlling for age and follow-up Total cases= I1 165 All Women Occupation in 1971 & 1981 Deaths 81-89 O.R. t P 4 1.12 0.21 0.83 1 Teachers 47 1.22 1.21 0.228 Nurses 30 0.79 -1.16 0.244 6 1.9 1.52 0.128 170 1.oo Shop Assistant 42 1.17 0.88 0.379 Slulled Manual 46 1.33 1.7 0.089 Semi-Slulled Factory 73 1.77 4.02 0 Semi-skilled domestic 59 1.27 1.55 0.121 other semi-Slulled 21 1.85 2.6 0.009 Unskilled 34 1.16 0.79 0.432 Professional Intermediate Clerical Source: OPCS Longitudinal Study 33 Table 11 Women in the same occupation 1971 & 1981: Mortality odds ratios adjusted for age and follow-up. Full time employed and women with no dependent children. Full time employed 1971 N=10769 Occupation in 1971 Deaths 81-89 t P O.R. No dependent children 1971 N= 10076 Deaths 8 1-89 O.R. t P Professional 3 1.26 0.39 0.699 2 1.09 0.12 0.908 Teachers 42 1.39 1.79 0.073 31 1.40 1.64 0.102 Nurses 16 0.68 -1.42 0.156 17 0.83 -0.71 0.479 Intermediate 6 2.22 1.86 0.063 5 2.00 1.49 0.137 Clerical 110 1.oo 119 1.oo Shop Assistant 18 1.21 0.72 0.468 33 1.34 1.46 0.146 Skilled Manual 39 1.51 2.17 0.030 36 1.46 1.96 0.050 Semi-Skilled Factory 54 1.92 3.86 0.000 54 1.94 3.95 0.000 Semi-skilled domestic 21 1.80 2.42 0.015 27 1.19 0.79 0.428 Other SemiSkilled 14 1.93 2.26 0.024 16 2.15 2.81 0.005 Unskilled 9 1.73 1.55 0.121 20 1.39 1.34 0.181 Source OPCS Longitudinal Study 34 Table 12 Odds ratios of being permanently sick in 1981 by occupation and age in 1971 All Women under 51 Total cases= 436 7 Occupation in 1971 Sick 81 O.R. t P Professional 0 na na na Teachers 15 0.95 -0.18 0.856 Nurses 39 2 3.61 0 Intermediate 28 1.32 1.26 0.206 Clerical 96 1 Shop Assistant 40 1.31 1.4 0.16 Skilled Manual 59 1.77 3.4 0 Semi-Skilled Factory 87 2.32 5.61 0 Semi-skilled domestic 65 1.63 3 0.003 Other Semi-Skilled 26 1.84 2.72 0.007 Unskilled 52 1.98 3.9 0 20-24 23 1 25-29 26 1.73 1.91 0.056 30-34 34 2.27 3.01 0.003 35-39 75 3.96 5.73 0 40-44 121 5.55 7.46 0 45-49 228 9.3 - 10.08 0 Age in 1971 Source: OPCS Longitudinal Study 35 Table 13 Odds ratios of being permanently sick 1981 by occupation 1971 by age for women who worked full time or had no dependent children in 1971, age standardized No dependent children N=265 16 Full Time employed N=27525 Occupation in 1971 Sick 81 t P Sick 81 O.R. O.R. t P Professional 0 na na na 0 na na na Teachers 13 0.8 -0.61 0.54 12 0.98 -0.1 0.94 Nurses 28 1.8 2.63 0.01 26 1.81 2.63 0.01 Intermediate 23 1 0.17 0.86 26 1.4 1.47 0.14 Clerical 81 1 87 1 Shop Assistant 19 1.2 0.6 0.55 29 1.17 0.72 0.47 Skilled Manual 45 1.5 2.11 0.04 43 1.48 2.08 0.04 Semi- Skilled 71 2.3 4.94 0 72 2.35 5.25 0 Semi-skilled domestic 21 1.5 1.64 0.1 40 1.68 2.65 0.01 Other SemiSkilled 17 1.5 1.47 0.14 21 1.84 2.47 0.01 Unskilled 14 2 2.28 0.02 27 1.83 2.67 0.01 Factory Source OPCS Longitudinal Study 36 Mortality by Occupation in 1971 Figure 1: All women employed in 1971, controlling for age and follow up x Adjusted for car access Relative Odds Ratio 4.0 3.0 2.0 1.o 0.0 Prof Teach Nurse Inter Cler Shop Skill Factr WES Occupational Group I 95% /C.I I Odds ratio Dom 0th ss Unsk Figure 2: Women in the same occupation in 1971 and 1981 Mortality by Occupation, controlling for age and follow up Odds relative to Clerical ~ 4.0 3 .O 2.0 'I T 1.o 0.0 - - Prof Teach Nurse Inter 1- Cler Shop Skill Factr WES Occupational Group +Odds Ratio I---- Dom Othss Unsk I I I I I L 4 0 I I I I I I I I I 1 I I I I I I I I I I I I I I I I I 0 cc; I I I I 1-b I I 1 I I I I I I I I t+l , I I I I I I I I 1 m m E a" I 0 0 I I I H 0 Pi 0 WORKING PAPERS USING OPCS LONGITUDINAL STUDY *We do not supply copies of papers which have been subsequently published, or for various other reasons have been discontinued No * 1 Title Author ( 8 ) Migration and health of the elderly. D. Dawkins 1982 * * 2 Housing tenure: an example of using record linkage to study differentials in cancer incidence, survival and mortality. 3 A review of the literature on migration in England and Wales. * 4 Social mobility around the time one has children. Now published in Lonsitudinal Studv No 2, 1985; OPCS, HMSO, London. * 5 Cause of death amongst people registered with cancer in 1971-75. D. Leon 1983 E. Hoinville 1983 A.J. Fox 1984 D. Leon A. Adelstein 1983 * 6 Bereavement and cancer: some results using data from the LS. Now published in British Medical Journal, 25 August 1984, 289: 461-464. D.R . Jones P.O. Goldblatt D.A. Leon 1984 * 7 From official health statistics to interactive epidemiology data. A.J. Fox D.R. Jones D. Leon 1983 * 8 LS mortality 1971-81 in regional heart study areas. Some preliminary notes on the relationship with region and water hardness. * 9 Water nitrates and stomach cancer: some 1984 P.O. Goldblatt mortality results from the OPCS Longitudinal Study. D.R. Jones 1984 - "10 Years ont1. Now published in: PoDulation Trends, 1984; A. Brown A . J . Fox 37: 20-22. 1984 Preliminary outline of projects to be covered I l l 0 years on" . A.J. Fox .E. M .D. Grundy *10 OPCS Longitudinal Study *11 D . R . Jones by *12 Cancelled '13 E.M.D. Grundy A.J. Fox Migration during early married life. Now published in Journal of EuroDean Association for PoDulation Studies. 1985, 1984 1: 237-263. K.A. Moser P. Goldblatt 14 Mortality of women in private and non-private households using data from the OPCS Longitudinal Study. 1984 H.S. Pugh 15 Estimating the extent of homeworking. 1984 A.J. Fox P.O. Goldblatt *16 Social class mortality differentials: artefact selection or life circumstances? Now published in Journal of Epidemioloqy and Communitv Health, 1985, 39: 1-8. 1984 *17 Cause of death in widow(er)s and spouses. Now published in Journal of Biosocial Science D.R. Jones P.O. Goldblatt 1987, 19; 1: 107-121. 1984 *18 Unemployment and mortality in the OPCS K.A. A.J. D.R. P.O. Longitudinal Study. Now published in The Lancet, December 1984: 1324-1329. Moser Fox Jones Goldblatt 1984 *I9 Divorce, widowhood, remarriage and geographic E.M.D. Grundy mobility. Now published in Journal of Biosocial Science, 1985, 17: 415-435. 20 1984 Some notes on the effects of level of aggregation D.R. Jones on analysis of mortality in the OPCS LS by area. 1984 *:21 Male socio-demographic mortality differentials from the OPCS Longitudinal Study 1971-81. A.J. D.R. K.A. P.O. Shortened version published in PoDulation Trends 1985, 40: 10-16. Full version published in Proceedings of the American Statistical Association Meetinq, August 13-16, 1984. f22 Fox Jones Moser Goldblatt 1984 D.R. Jones P.O. Goldblatt Mortality following widowhood: some further results from the OPCS Longitudinal Study. Now published in Stress Medicine, 1986, 2: 1984 129-140. D.R. Jones and others 23 Mortality 1971-81 and migration in England and Wales in 1966-71: some data from the OPCS Longitudinal Study. 1984 D.R. Jones 24 Introductory notes on regression models in the analysis of mortality data in the OPCS LS. *25 Change in death rates with length of follow-up. 1985 . A.J. Fox P.O. Goldblatt 1985 f26 Mortality of women in the OPCS Longitudinal Study: differentials by own occupation and household and housing characteristics. 27 An investigation of alternative methods of calculating person-years-at-riskin the OPCS LS. K.A. Moser P.O. Goldblatt 1985 D.R. Jones 1985 *28 Migration and fertility behaviour in England E.M.D. Grundy *29 A Longitudinal perspective on recent socio- A.J. Fox E.M.D. Grundy and Wales. Now published in Journal of Biosocial Science, October 1986, 17, 4: 415-435. demographic change. Now published in the Proceedinss of British Societv for PoDulation Studies' Conference on IIMeasurins SocioDemosraDhic Chanse", Occasional Paper 34 OPCS . f 3 0 Unemployment and mortality: further evidence from the OPCS Longitudinal Study. A shortened version published in The Lancet, February 1986: 365-367. 1985 1985 K.A. A.J. D.R . P.O. Moser Fox Jones Goldblatt 1985 Stress and heart disease: evidence of association between unemployment and heart disease from the OPCS Longitudinal Study. The Postsraduate Medical Journal, 1986, 62: 797-799. Proceedings of The Coronary Prevention Group Conference on 18-19 November 1985, tlDoesStress Cause Heart Attacks?": 123-130. 31 Socio-economic differentials in cancer. f32 D.A. Leon A review paper written for the International Agency for Research in Cancer. 1985 Socio-economic differences in cancer and heart disease. In "Health Inequalities in European Countries1I,Proceedings of ESF/ESRC Workshops. Gower Press 1988, ed. A.J. Fox. D.A. Leon R. Wilkinson f33 Mortality and deprivation: evidence from the OPCS Longitudinal Study. Edited proceedings of the Eugenics Society Symposium 1985 The Political Economy of health and Welfare. Ed W.M. Keynes. A.J. Fox D.A. Leon f34 Retirement migration and its consequences in England and Wales. Aseins and Societv, 7: 59-82, 1987. E.M.D. Grundy *35 Some preliminary notes on status incongruity B. Scott and mortality. 1986 1985 "36 D.R. Jones Heart disease mortality following bereavement. Now published in 1 Attacks? Proceedings of Conference of The Coronary Prevention Group 1 8 - 1 9 November 1 9 8 5 : 77-102. Heart disease mortality following widowhood: D. R Jones some results from the OPCS Longitudinal Study. Journal of Psvchosomatic Research, 1 9 8 7 , 3 1 ; 3 : 325-333. *37 *38 "39 Racial minorities in the London labour and housing markets: a longitudinal analysis 1971-81. To Appear as "Racial minorities and industrial change" In Misration. EmDlovment and the New Urban Order. Ed. M. Cross. Cambridge University Press, Comparative Ethnic and Race Relations Series. C. Hamnett W. Randolph C. Evans Preliminary notes on changes in male economic activity patterns, 1 9 7 1 - 8 1 . A.J. Fox Socio-demographic mortality differentials: new longitudinal perspectives. Paper written for Symposium "The Social Aetiology of I1l-Healthlf at French Ministry of Research and Technology, Paris, March 2 0 , 1 9 8 6 . Now published in the Revue d'eDidemiolosie et de sante Dublime, A.J. Fox 35: *40 41 *42 20-27, 1985 1986 1986 1987 Cancelled Using the OPCS Longitudinal Study to classify ethnic origin. J. Webster Social class mortality differentials of men aged 1 5 - 6 4 in 1 9 8 1 : a note on first results from the OPCS Longitudinal Study for the period P.O.Goldblatt 1986 1986 1981-83. (Updated version in Pomlation Trends no 51) f43 Unemployment and mortality, 1 9 8 1 - 8 3 : follow up of the 1 9 8 1 LS Census sample. A shortened version published in the British Medical Journal, 10 January 1 9 8 7 , 2 9 4 : 8 6 - 9 0 K.A. P.O. A.J. D.R. Moser Goldblatt Fox Jones 1986 .44 f.45 46 Labour market restructuring in Greater London 1 9 7 1 - 8 1 : Evidence from the OPCS Longitudinal Study. C. Hamnett W. Randolph Socio-tenurial polarisation in London: a longitudinal analysis, 1 9 7 1 - 8 1 . Shortened version in Urban Studies. C. Hamnett W. Randolph The social classification of women. (Now reprinted as SCOWW - see p. 7 ) . H. Roberts R. Barker 1986 1986 1986 47 48 *49 - Have inequalities in health widened? (Mortality differences in the 1 9 7 0 s and early A.J. Fox P.O. Goldblatt 1980s). 1986 Modelling socio-economic change: an application M.C. Shewry 1987 of generalised linear models to changes in the circumstances of individuals between 1 9 7 1 and 1 9 8 1 . Longitudinal insights into the ageing population. A.J. Fox In D. Evered and J.Whelan (Eds). Research and the 1 9 8 7 A CIBA Foundation Symposium, John Wiley & Sons, London. *50 New longitudinal insights into relationships between unemployment and mortality. Stress Medicine, Vol 4 , No 1, January-March 1 9 8 8 pp 1 1 - 1 9 . 51 A social classification scheme for women. A.J. Fox M.C. Shewry 1987 H. Roberts R. Barker 1987 *52 What are people doing when they grade women's work? m i 1 m , Vol 40 No. 1 March 1 9 8 9 . H. Roberts R. Barker 1987 53 Mortality differences at working ages: the use of generalised linear models to compare measures. P. Goldblatt Inequalities in women's health: developing an alternative approach. A shorter version published in 1 , 30 April 1 9 8 8 , K. Moser H. Pugh P. Goldblatt 2 9 6 : 1221-1224. 1987 Social class differences in mortality in Great Britain around 1 9 8 1 . Journal of Institute of British Actuaries. S. Haberman D. Bloomfield Socio-economic status and cancer: results from the OPCS Longitudinal Study. M. Kogevinas M.G. Marmot A.J. Fox 54 *55 56 1987 1987 1988 *57 f58 Inequalities in women's health in England and Wales: mortality among married women according to social circumstances, employment characteristics and life cycle stage Published in Senuq, Vol XLVI, pp 7 1 - 8 4 , 1990. K. Moser H. Pugh P. Goldblatt Mortality of employed men and women. A I I J I I J Vol 20, Part 3: pp 285-306, 1 9 9 1 . P. Goldblatt A.J. Fox D. Leon 1988 1988 59 Mortality of men by occupation. P. Goldblatt A.J. Fox 1988 60 Associations between unemployment and fertility among young women in the early 1980s. 61 The social and geographical mobility of South Asians and Caribbeans in Middle Age and later Working Life. *62 Smoking, class and lung cancer mortality among women. Social Science and Medicine, Vol. 32, NO. 10, pp 1105-1110. B. Penhale 1989 A. Stuart 1989 H. C. P. S. Pugh Power Goldblatt Arber 1989 - economic status and breast cancer in England and Wales: time trends in incidence, survival and mortality. 63 Socio M. Kogevinas P. Goldblatt H. Pugh 1989 64 The Longitudinal Study: households, families and fertility. (Now available as LS User Guide No. 1). 65 Family and demographic circumstances and mortality among married women of working ages B. Penhale 1989 A. Mercer P. Goldblatt H. Pugh 1989 66 Occupational Mortality of women aged 15-59 at death in England and Wales. K. Moser P. Goldblatt 1990 67 English and French households in historical perspective. Also in INSEE ( I n s t i t u t N a t i o n a l d e l a S t a t i s t i q u e e t d e s & t u d e s &conomiques) no 8, Paris: February 1991. 68 Living arrangements of young adults in France and England. R. Wall 1990 B. Penhale 1990 69 Residence patterns of the elderly in England and France. (Also in INSEE, as no 67 above) 70 Age difference asymmetry and a two-sex perspective. R. Wall 1990 M. Ni Bhrolchain 1990 *71 Geographical variations in female labour force participation: an application of multilevel modelling. *72 Measuring housing deprivation using the OPCS Longitudinal Study. (Updated version). C. Ward A. Dale 1991 M. Williams A. Dale 1992 WORKING PAPERS ARE AVAILABLE FREE ON REQUEST TELEPHONE: 071 477 8486 USER GUIDES TO THE OPCS LONGITUDINAL STUDY 1 Households, families and fertility. B. Penhale 1990 2 The measurement of ethnicity. A. Stuart 1990 3 The analysis potential of the LS. I. Plewis 1990 4 User guide to computing with the LS* (Second edition). Brian Dodgeon Modelling categorical data with GLIM. C. Ward 5 1992 1991 6 * M. Rosato Using the OPCS Stage I11 Epidemiological package. 1991 introduction to the area based variables in the LS. R. Creeser A comparison of mortality measures in the OPCS R. Weatherall Longitudinal Study 1992 7 . An a. 9. A guide to the fertility and infant mortality datastream 1993 R. Creeser 1992 10. An examination of the quality of OPCS Longitudinal P. Babb Study data for use in fertility analysis L. Hattersley 1992 11. Using SPSSX Statistical Procedures with the LS S. Gleave R. Creeser B. Dodgeon 1993 * Only available to those individuals who have made arrangements to work with the LS at OPCS. SCOWW: The Social Classification of Women's Work. R. Barker H. Roberts 1990 USER GUIDES ARE AVAILABLE FREE ON REQUEST TELEPHONE: 0 7 1 477 8486 ******** THE OPCS LONGITUDINAL STUDY USER MANUAL Order forms are available on request. The cost is f15, and includes the Manual, an accompanying Data Dictionary disk, and postage and packing. (The Manual is available without the Data Dictionary for a charge of f5.00 to cover administration, postage and packing).