Gender differences in the relationship of partner's social class to behavioural risk factors and social support in the Whitehall II study

ARTICLE IN PRESS

Social Science & Medicine 59 (2004) 1925–1936

Gender differences in the relationship of partner’s social class to behavioural risk factors and social support in the Whitehall II study M. Bartleya,*, P. Martikainena,b, M. Shipleya, M. Marmota a

Department of Epidemiology and Public Health, International Center for Health and Society, University College London, 1-19 Torrington Place, London WCIE 6BT, UK b Department of Sociology, Population Research Unit, University of Helsinki, Finland, UK

Abstract In most countries health inequality in women appears to be greater when their socio-economic position is measured according to the occupation of male partners or spouses than the women’s own occupations. Very few studies show social gradients in men’s health according to the occupation of their female partners. This paper aims to explore the reasons for the differences in social inequality in cardiovascular disease between men and women by analysing the associations between own or spouses (or partners) socio-economic position and a set of risk factors for prevalent chronic diseases. Study participants were married or cohabiting London based civil servants included in the Whitehall II study. Socioeconomic position of study participants was measured according to civil service grade; socio-economic position of the spouses and partners according to the Registrar General’s social class schema. Risk factors were smoking, diet, exercise, alcohol consumption, and measures of social support. In no case was risk factor exposure more affected by the socioeconomic position of a female partner than that of a male study participant. Wives’ social class membership made no difference at all to the likelihood that male Whitehall participants were smokers, or took little exercise. Female participants’ exercise and particularly smoking habit was, in contrast, related to their spouse’s social class independently of their own grade of employment. Diet quality was affected equally by the socio-economic position of both male and female partners. Unlike the behavioural risk factors, the degree of social support reported by women participants was in general not strongly negatively affected by their husband or partner being in a less advantaged social class. However, non-employment in the husband or partner was associated with relatively lower levels of positive, and higher negative social support, while men with non-working wives or partners were unaffected. Studying gender differences in health inequality highlights some of the problems in health inequality research more broadly. We are brought face to face with the fact that the development of conceptual models that can be applied consistently to aetiology in both men and women are still at an early stage of development. Closer attention is needed to the different processes behind material power and ‘emotional power’ within the household when investigating gender differences in health and risk factors. r 2004 Elsevier Ltd. All rights reserved. Keywords: Gender; Social support; Social class; Behavioural risk factors; UK

Introduction *Corresponding author. Tel.: +171-391-1707; fax: +171813-0242. E-mail address: [email protected] (M. Bartley).

Social gradients in cardiovascular disease, the differences between those in more and less favourable situations with respect to occupation, income and

0277-9536/$ - see front matter r 2004 Elsevier Ltd. All rights reserved. doi:10.1016/j.socscimed.2004.03.002

ARTICLE IN PRESS 1926

M. Bartley et al. / Social Science & Medicine 59 (2004) 1925–1936

prestige, might be expected to be similar in men and women. However, the evidence is conflicting. Many European and US studies show that socio-economic position measured according to the individual’s own occupation is less strongly related to health in women living with male partners than in men (Dahl, 1991; Siskind, Najman, & Veitch, 1992; Koskinen & Martelin, 1994; Stronks, van de Mheen, van den Bos, & Mackenbach, 1995; Harding, Bethune, Maxwell, & Brown, 1998). Furthermore, in partnered women, steeper health gradients are often seen when socioeconomic position is defined according to the occupation of the partner than according to the woman’s own occupation. Few studies investigate health inequalities in men according to the social position of their female partners. Those that do, tend to find little or no health gradient. Several studies have reported that household level variables, such as income (Luoto, Pekkanen, Uutela, & Tuomilehto, 1994), asset ownership (Moser, Pugh, & Goldblatt, 1988; Koskinen & Martelin, 1994), or combined measures of occupational social class level of both partners (Krieger, Chen, & Selby, 1999) are better predictors of women’s health than social class based on the woman’s own occupation alone. Differences between men and married or cohabiting women in health inequality are not found universally, however. In Finland, where male and female labour force participation rates are almost equal, health gradients are more or less the same in men and women, when socio-economic position is measured according to each individual’s own occupation (Martikainen, 1995). The differences between men and women in the relationship of socio-economic position to cardiovascular disease, and the fact that these differences themselves vary between national contexts, alert us to some important issues. Although differences in health between social classes, income or status groups have been described over and over again, less effort has been expended on developing explanations for these regularities that acknowledge and account for gender, ethnic and cultural differences in the distribution of risk factors. Gender differences in health inequality are often treated as discrepancies, rather than forming the focus for improving our explanatory frameworks. Recent papers have called for a greater attention to explanations focused on the social context rather than the individual (Diez-Roux, 2000; Brixi & Lan, 2000; Ribet et al., 2001). When thinking of risk factors as arising within a social context, it is less surprising that the features likely to be relevant for inequalities in behavioural and psycho-social risk factors differ according to gender. To take the example of smoking, social explanations for inequality in men have tended to focus on high stress and low rewards, mainly arising from the employment situation (Siegrist, 2000). In contrast, explanations for social inequality in smoking in women

have highlighted the burden of caring in situations of financial hardship (Graham, 1987). The most immediate social context for the great majority of individuals is the household within which they live. Household influences on health and risk factors are only just beginning to receive attention in the literature on social determinants of health (Chandola, Bartley, Wiggins, & Schofield, 2003a). The theoretical basis for research on the relationship of health to household conditions and relationships is therefore less elaborated than that for workplace or geographical area contexts. An example of this problem is that there is no clear understanding of the ways in which the socioeconomic position of a spouse or partner might be expected to influence health, either in men or in women. We have no very clear ideas why the social position of the male partner in a couple often seems to have so much more of an effect on health than that of the female partner. Relevant considerations would include the relationship of gender to access to the labour market, the power balance within households, domestic division of labour, and allocation of household resouces. Do women and men have more or less confiding relationships with, or receive more or less social support from partners in more privileged occupational groups? High degrees of responsibility at work do not seem to exempt women from the ‘double shift’ of formal and domestic responsibilities (Rubery & Tarling, 1988; Joshi, 1998). Women in professional and higher management positions who nevertheless find themselves expected to carry the full burden of performing, or at least planning and supervising, the domestic responsibilities of home and child care, may be less able or less inclined to provide high levels of emotional support to their spouses. Studies that find a ‘high flying’ female partner may be a risk factor for cardiovascular disease in men are consistent with the idea that a higher level of education or of occupational prestige in female partners might reduce the power of men within the household (Suarez & Barrett Connor, 1984; McDonough, Williams, House, & Duncan, 1999; Egeland, Tverdal, Meyer, & Selmer, 2002). Existing empirical and theoretical work suggest five main processes at work within households in producing the different ‘risk profiles’ of men and women. In countries where men have a stronger relationship to the labour market, occupation may be weaker as a measure of married or co-habiting women’s standard of living than is it is for men. When both members of a couple are in paid employment, the woman’s wage may be less than half of the total family income as she will tend to be paid less than her male partner. On the other hand, when out of work, she will have access to more money than solely the level of social security benefit that would be available to a man (or single woman). At these times, the woman’s living standard will depend on the occupation of the

ARTICLE IN PRESS M. Bartley et al. / Social Science & Medicine 59 (2004) 1925–1936

male partner. How much of the household income a woman will have access to will depend on financial strategies and the power balance between partners (Winkler, 1997; Heimdal & Houseknecht, 2003). It would be safe to say, however, that for women with intermittent work histories due to caring responsibility, the income of the male partner will be the main influence on housing quality and the affordability of major items such as household appliances, holidays and cars (Fox & Adelstein, 1978; Moser et al., 1988). Secondly, legislation in most industrial nations excludes women from several of the most hazardous occupations. So, knowing an individual woman’s social class based on her own occupation will not predict hazard exposure as well as it would in a man. Thirdly, occupational social class may not reflect social status as well in married women as it does in men. A secretary may have a different ‘standing in the community’ if she is married to a doctor and works in his office than if she were living independently. This difference may be relevant within a framework that regards health behaviours as influenced by processes of social distinction. In many Northern European nations, ‘healthy’ lifestyles have become a marker of high social status, so that in a household whose highest-earning member belongs to a high-status group, there may be strong pressure for other members to adopt similar behaviours. Fourthly, there is now some debate around the effect of workplace stressors in men and women. Low control and high demands at work, or low rewards relative to efforts, have been shown to be related to health in many studies (Johnson, Stewart, Hall, Fredlund, & Theorell, 1996; Peter, Geissler, & Siegrist, 1998; Theorell et al., 1998) (Peter et al., 1998; Kuper, Singh, Siegrist, & Marmot, 2002; Kivim.aki et al., 2002; Kuper & Marmot, 2003). However, psychosocial conditions at work may be less salient in married women who must balance concerns for work and caring, and by the same token, those in the home may be more salient (Griffin, Fuhrer, Stansfeld, & Marmot, 2002; Chandola, Kuper, Singh-Manoux, Bartley, & Marmot, 2003b). Fifthly, there may also be important gender differences in both the major sources of ‘effort’, and the most salient forms of ‘reward’. Because of the differences in labour force attachment, and in power and role expectations within the household, we might expect to find gender differences in the relationship between social class (defined in terms of occupation) and a range of material, behavioural and psychosocial risk factors for cardiovascular disease. Insofar as research has shown a strong relationship between material deprivation and health behaviours (Lynch, Kaplan, & Salonen, 1997; Graham, 1998; Pickering, 1999), this would lead us to expect that women’s risk behaviours may be more influenced by the socio-economic position of their male partners than the other way around. However, the relationship of risk

1927

behaviours to living standards is itself under-theorised (Bartley, 2003). Why should living standards be related to the risk of nicotine addiction? Two possible pathways have been suggested. The first of these may be regarded as a stress pathway (Siegrist, 2000), and assumes a psycho-biological relationship of stressful circumstances to addictive behaviours. In a household with at least one high-earning partner, the stresses of caring will be reduced by space, labour saving equipment and the ability to buy-in domestic and child care assistance. The second possibility, based on the work of Bourdieu, relates health behaviours (in Northern Europe and North America at the present time) to social status aspirations. In these societies, ‘healthy’ behaviour has come to be seen as a form of ‘distinction’, in Bourdieu’s sense of a claim to membership of high-prestige social groups (Wilkinson, 1996, p. 171; Lindbladh et al., 1996). If ‘healthy’ behaviours in household members are a condition of acceptance in high prestige social circles, we might expect the economically dominant partner in a relationship to be in a position to ‘enforce’ such behaviours. In a low-strain environment, opportunity and motivation for ‘healthy’ behaviours are also more likely to be present. In this paper we examine the relationships of male and female social position defined according to individuals’ own occupation to patterns of behavioural and psychosocial risk factors in their spouses. The research sets out with six hypotheses 1. Health related behaviours will be more strongly related to socio-economic position (SEP) based on the individual’s own occupation in men than in women. 2. Health behaviours in women will be related to the SEP of their male partners. 3. Health behaviours in men will have no, or only a weak relationship to the SEP of their female partners. 4. The quality of social support received will be more strongly related to SEP based on the individual’s own occupation in men than in women. 5. The quality of social support received by women will be related to the SEP of their male partners. 6. The quality of social support received by men will have no, or only a weak relationship to the SEP of their female partners.

Sample and methods Sample The Whitehall II study (Marmot et al., 1991) is a prospective study of men and women aged 35–55 years at time of recruitment, working in the London offices of 20 Civil Service departments. The overall response rate

ARTICLE IN PRESS 1928

M. Bartley et al. / Social Science & Medicine 59 (2004) 1925–1936

was 73%, although the true response rate is likely to be higher because around 4% of those listed as employees were not eligible as they had moved before the study. Altogether, 10,308 men and women filled in a selfadministered health questionnaire and attended screening examination in 1985–1988. After initial participation at phase 1, a further postal questionnaire was carried out in 1989 (phase 2), 1991–1993 (phase 3 which also included a screening examination), 1995 (phase 4), and 1997–1999 (phase 5, also with a screening examination). The analyses presented in this paper are cross-sectional, being based only on phase 1 data.

following three categories: none or light, moderate (more than an hour of moderate physical activity per week) and vigorous (more than an hour of vigorous physical activity per week). A healthy diet variable combined information from frequency of eating fruit and vegetables and the type of bread and milk usually consumed. From these data a five point score, reflecting the number of healthy aspects of the diet, was compiled. This ranged from a score of four for those who ate fruit and vegetables daily, and usually consumed skimmed/ semi-skimmed milk and wholemeal/granary bread, to a score of 0 for those whose diet was poor in all four respects.

Socio-economic position Socio-economic position of Whitehall II participants was assessed in terms of employment grade at Phase 1 in 1985–1988, and the position of the spouse in terms of social class. Employment grade was obtained by asking all participants for their civil service grade title. On the basis of this information three employment grades were defined: administrative (Grade 1: ‘‘High’’), professional & executive (Grade 2: ‘‘Middle’’) and clerical and support staff (Grade 3: ‘‘Low’’). These employment grades differ markedly in salaries; from an annual salary in 1987 of about d3000 to d6000 in employment grade 3 to about d18,000 to d62,000 in employment grade 1. Spouse’s social class was based on information from the participant on their spouse’s main current job. This information was only collected in versions 3 and 4 of the questionnaire that was administered to the last 7679 participants enrolled at phase 1 of the study. Spouses were then classified according to the Registrar General’s social classification as I=professional, II=managerial, IIINM=routine non-manual, IIIM=skilled manual, IV=semi-skilled manual and V=unskilled manual. For the purposes of this study classes I and II were combined, as were classes IIINM and IIIM, and classes IV and V. Behavioural risk factors The screening questionnaire contained sections on diet, alcohol consumption, smoking, and physical exercise. Four health related behaviours were included in this analysis. Current smoking was classified as ‘non smoker’ or ‘current smoker’. Alcohol consumption over the recommended limit was assessed with number of drinks in the past week obtained from questions on quantities of spirit, wine and beer in the past week. Among men more than 21 units and among women more than 14 units per week is above the recommended limit. Physical activity was assessed using questions on hours of mild, moderate and vigorous activity undertaken each week. These were aggregated into the

Social support Participants answered a ‘close persons questionnaire’ (Stansfeld & Marmot, 1992), which assesses support given and received from a maximum of four nominated persons, and the relation of these persons to the participants. From this may be derived whether or not a participant regarded his or her spouse or partner as their closest relationship. Not nominating the spouse as the closest person was assessed as a risk factor. Questionnaire items were included on the amounts of confiding support, and negative aspects of social support received from each of the four close persons. If the spouse or partner was not nominated as one of the ‘close persons’, the support questions were asked of this relationship separately. Both scales were calculated as sum of scores to responses to questions on different types of support with response alternatives ranging from ‘not at all’ (0) to ‘a great deal’ (3). Confiding/ emotional support questions were related to how much the respondent wanted to confide and actually confided in the closest person, how much the respondent trusted this person with personal worries and problems, how much he/she talked about his/her worries with the respondent, did they share interests and hobbies and did the closest person make the respondent feel good about themselves. Negative aspects of close relationships related to how much the closest person gave the respondent worries and problems, and whether talking to him/her made thing worse, and whether the respondent wanted to confide more and have more practical help from them. Variables which have continuous scores: that is, the measures of confiding and negative support, are grouped into tertiles for the cohort as a whole (men and women together) and the odds of being in the least favourable tertile (lowest for confiding and practical support and highest for negative support) are compared between groups. Because only married men and women are included in this analysis, the tertiles are not exact.

ARTICLE IN PRESS M. Bartley et al. / Social Science & Medicine 59 (2004) 1925–1936

Statistical methods Data analysis was carried out using logistic regression. Results are presented as odds ratios separately for men and women. Differences in each risk factor according to civil service grade and spouse’s social class were first shown adjusted for age only. Grade differences were then adjusted for spouse or partner’s social class, and differences according to the class of the spouse were adjusted for grade (i.e. the variables were mutually adjusted). The degree of attenuation in the odds ratios resulting from these adjustments allows a comparison to be made between the effects of Whitehall II participants’ own socio-economic position (as measured by their employment grade) and that of their spouse (measured by Registrar-General’s social class). We then tested whether the effect of spouse’s social class was different in male and female Whitehall participants by combining data for both sexes and fitting an interaction between sex and spouse’s social class. All calculations were carried out in SPSS version 9 r.

Results Socio-economic and sex differences in marital status, employment grade and class of spouse Table 1 shows the distribution of all male and female Whitehall II participants according to grade and marital status at the beginning of the study in 1985–1988. Whereas men in the higher civil service grades were more likely to be married than those in the lower grades, the opposite was the case for women. Throughout the

1929

following report we are analysing only married or cohabiting study participants. This sub-sample of married or cohabiting Whitehall participants contained 80% of men in the original sample but only 61% of women. Since the spouses’ social class could only be obtained from participants who received versions 3 and 4 of the phase 1 questionnaire, the numbers of married or cohabiting participants available was 5701 (4156 men and 1545 women). Of these, 219 participants (110 men, 109 women) did not provide information on the spouse or partner’s main current job. Table 2 shows the distribution of Whitehall II participants who were married at phase 1 according to their own employment grade and the Registrar-General’s social class of their spouse. Both men and women tended to be married to persons of similar socio-economic position. However this was far more strongly seen in women . Almost 86% of the married women in the highest civil service employment grade (grade 1) were married to men in social classes I or II, whereas this was true for only 41% of male Whitehall participants. This difference was made up for by non-working women: almost 33% of men, compared to only 7% of women in employment grade 1 (the highest grade) had non-employed wives or partners. Men and women in employment grade 3 (the lowest grade) were equally likely to be married to spouses in social classes IV or V. Behavioural risk factors Table 3 compares the social gradients in smoking, diet, alcohol consumption and exercise according to Whitehall participants’ own civil service grade and the

Table 1 Distribution of marital status(%) by employment grade in men and women Sex

Men

Grade

1 (High) 2 (Middle) 3 (Low) All

Women

1 (High) 2 (Middle) 3 (Low) All

Marital status Married or cohabiting

Single

Separated or divorced

All (N)

89.2 (2358) 78.1 (2808) 57.1 (364) 5530

6.7 (176) 15.9 (572) 32.7 (208) 956

4.1 (109) 6.0 (215) 10.2 (65) 389

100 (2643) 100 (3595) 100 (637) 6875

59.5 (226) 54.6 (727) 66.8 (1125) 2078

27.4 (104) 30.3 (403) 13.5 (227) 734

13.2 (50) 15.1 (202) 19.7 (331) 583

100 (380) 100 (1332) 100 (1683) 3395

ARTICLE IN PRESS M. Bartley et al. / Social Science & Medicine 59 (2004) 1925–1936

1930

Table 2 Distribution of social class of spouse (%) by employment grade in men and women Sex

Men

Grade

I/II

IIIN+M

IV+V

Not employed

All (N)

41.2 (674) 31.3 (667) 19.8 (55)

22.6 (370) 33.9 (722) 43.3 (112)

3.5 (57) 7.4 (157) 14.0 (39)

32.7 (535) 27.5 (586) 25.9 (72)

100 (1636) 100 (2132) 100 (278)

All men

1396

1204

253

1193

4046

1 (High)

85.7 (102) 59.1 (308) 26.8 (213)

6.7 (8) 27.1 (141) 49.2 (392)

0.8 (1) 4.4 (23) 13.4 (107)

6.7 (8) 9.4 (49) 10.6 (84)

100 (119) 100 (521) 100.0 (796)

623

541

131

141

1436

1 (High) 2 (Middle) 3 (Low)

Women

Social class of spouse

2 (Middle) 3 (Low)

All women

social class of their partners. The most important influence on smoking in married male Whitehall II study participants was their own grade of employment in the civil service. The odds ratios for smoking in men were 5.58 in the lowest compared to the highest grade after adjustment for the social class of the spouse. The size of the gradient according to their own rank in the civil service was far greater than the gradient according to their spouse or partner’s social class. In married or cohabiting female Whitehall II Study participants however, the difference in smoking rates between the highest and lowest civil service grades was smaller than the difference between women with partners in social classes I and II compared to those with partners in classes IV and V (overall test for interaction between sex and spouse’s social class, p=0.04). Diet quality was related to civil service grade in both men and women. This risk factor was also significantly related to the social class of the spouse or partner in both men and women, even after adjustment for each participant’s own civil service grade. In both men and women, those with a spouse or partner in RG classes IV or V were significantly more likely to have a poor diet than those with spouses in classes I and II. Men in civil service grade 3 (the lowest grade) were less likely than those in grade 1 to drink alcohol over the recommended amount, and adjustment for the social class of the spouse made little difference to this. Alcohol consumption in men was not significantly related to the social class of their spouse, but men with non-employed wives or partners were significantly less likely to consume over the safe limit of alcohol, independent of

their employment grade. Women in the highest civil service employment grades were much more likely than those in other grades to consume over the safe limit. Similarly, women with spouses or partners in RG classes I or II were more likely to consume over the recommended amount than those with spouses in lower RG classes. This was still true even after adjustment for class of spouse. The gender differences in the effects of spouse’s class on alcohol consumption were reflected in a significant interaction (p=0.02). A low level of exercise was far more common in men in the lowest employment grade than in those in the highest grade, with an odds ratio of 8, even after adjustment for the spouse or partner’s social class. There was no relationship between exercise levels in men and the social class of their female partners. Women in the lowest employment grade also had about two and a half times the odds of doing little or no exercise compared to those in the highest grade, and this was not greatly reduced by adjustment for class of spouse. In women, there was also a significant gradient in exercise according to the social class of the husband or partner. Social support Table 4 compares the relationship of own employment grade and that of the spouse or partner’s social class to three measures of social support: positive support, negative support and whether or not the spouse is regarded as the closest person to the respondent. The amount of confiding and emotional support received from their partners was not significantly related to civil service grade in married or cohabiting male

ARTICLE IN PRESS M. Bartley et al. / Social Science & Medicine 59 (2004) 1925–1936

1931

Table 3 Odds ratios (95% CI) of current smoking, poor diet quality, no or low exercise and alcohol consumption above ‘‘safe limits’’ by sex, own employment grade and social class of spouse Men Prevalence (%)

Women Age adjusted OR (95% CI)

Mutually adjusted OR (95% CI)

Prevalence (%)

Age adjusted OR (95% CI)

Mutually adjusted OR (95% CI)

Current smoking (N=4030 men/1428 women) Grade 1 (High) 9.0 1 2 (Middle) 15.4 1.80 (1.4–2.2) 3 (Low) 36.5 5.77 (4.2–7.7)

1 1.75 (1.4–2.1) 5.58 (4.1–7.5)

11.8 19.7 28.7

1 1.75 (0.9–2.6) 2.70 (2.0–4.6)

1 1.49 (0.8–2.7) 1.91 (1.0–3.5)

Spouse’s social class I/II III IV/V Not employed

13.3 16.6 18.2 12.3

1 1.10 (0.8–1.3) 1.07 (0.7–1.5) 0.86 (0.6–1.0)

16.3 28.8 39.2 25.7

1 1.97 (1.4–2.6) 3.06 (2.0–4.6) 1.62 (1.0–2.5)

1 1.73 (1.2–2.3) 2.64 (1.7–4.0) 1.51 (0.9–2.3)

Poor diet (N=3979 Grade 1 (High) 2 (Middle) 3 (Low)

22.0 28.6 38.5

1 1.44 (1.2–1.6) 2.22 (1.6–2.9)

1 1.40 (1.2–1.6) 2.07 (1.5–2.7)

12.8 17.0 26.5

1 1.42 (0.7–2.5) 2.56 (1.4–4.5)

1 1.33 (0.7–2.4) 2.17 (1.1–3.9)

23.7 28.1 36.8 26.3

1 1.25 (1.0–1.5) 1.87 (1.4–2.4) 1.15 (0.9–1.3)

1 1.15 (0.9–1.3) 1.64 (1.2–2.2) 1.12 (0.9–1.3)

17.8 24.8 32.8 19.3

1 1.53 (1.1–2.0) 2.28 (1.4–3.5) 1.11 (0.6–1.8)

1 1.25 (0.9–1.7) 1.80 (1.1–2.8) 1.00 (0.6–1.6)

High alcohol consumption (N=4020 men/1427 women) Grade 1 (High) 17.8 1 1 2 (Middle) 16.9 0.89 (0.8–1.1) 0.90 (0.8–1.1) 3 (Low) 12.3 0.64 (0.4–0.9) 0.65 (0.4–1.0)

28.6 15.3 3.4

1 0.46 (0.3–0.7) 0.11 (0.1–0.2)

1 0.49 (0.3–0.8) 0.11 (0.1–0.2)

Spouse’s social class I/II III IV/V Not employed

1 0.77 (0.6–0.9) 0.95 (0.7–1.3) 0.56 (0.5–0.7)

14.8 6.3 3.1 7.9

1 0.41 (0.3–0.6) 0.20 (0.1–0.6) 0.55 (0.3–1.1)

1 0.49 (0.2–1.0) 0.42 (0.1–1.2) 0.73 (0.4–1.5)

No or low exercise (N=4037 men/1421 women) Grade 1 (High) 6.1 1 2 (Middle) 8.3 1.52 (1.1–1.9) 3 (Low) 33.0 7.76 (5.6–10.7)

1 1.53 (1.1–1.9) 8.03 (5.7–11.2)

16.8 20.8 38.5

1 1.18 (0.7–2.0) 2.54 (1.5–4.2)

1 1.15 (0.6–1.9) 2.34 (1.3–1.8)

Spouse’s social class I/II III IV/V Not employed

1 1.06 (0.8–1.4) 0.66 (0.4–1.1) 1.05 (0.7–1.4)

23.9 33.0 43.4 35.7

1 1.41 (1.0–1.8) 2.05 (1.3–3.0) 1.43 (0.9–2.1)

1 1.09 (0.8–1.4) 1.53 (1.0–2.3) 1.25 (0.8–1.8)

1 1.31 (1.0–1.6) 1.45 (1.0–2.0) 0.91 (0.7–1.1)

men/1391 women)

Spouse’s social class I/II III IV/V Not employed

20.8 16.1 19.0 13.0

8.0 10.7 8.3 8.9

1 0.74 (0.6–0.9) 0.90 4(0.6–1.3) 0.56 (0.4–0.7)

1 1.32 (1.0–1.7) 1.03 (0.6–1.6) 1.13 (0.8–1.5)

Whitehall participants. There was no sign that men in lower grades were more likely to receive low levels of emotional support; if anything the opposite was the case. There was a relationship to the social class of the

spouse or partner however. Men whose partners were in classes IV or V were significantly more likely to report low emotional support than those with spouses in classes I and II, independently of the man’s own civil service

ARTICLE IN PRESS M. Bartley et al. / Social Science & Medicine 59 (2004) 1925–1936

1932

Table 4 Odds ratios (95% CI) of low confiding and emotional support, high negative support, citing spouse as closest person by sex, own employment grade and social class of spouse Men Prevalence (%)

Women Prevalence (%)

Age adjusted OR (95% CI)

Mutually adjusted OR (95% CI)

Low confiding and emotional support (N=3932 men/1327women) Grade 1 (High) 35.0 1 1 2 (Middle) 35.3 1.03 (0.9–1.1) 1.01 (0.8–1.1) 3 (Low) 30.2 0.81 (0.6–1.0) 0.77 (0.5–1.0)

29.5 31.6 39.5

1 1.08 (0.6–1.7) 1.48 (0.9–2.3)

1 1.05 (0.6–1.6) 1.39 (0.8–2.2)

Spouse’s social class I/II III IV/V Not employed

1 1.05 (0.8–1.2) 1.40 (1.0–1.8) 0.97 (0.8–1.1)

31.9 38.0 36.2 44.6

1 1.27 (0.9–1.6) 1.16 (0.7–1.7) 1.63 (1.0–2.4)

1 1.15 (0.8–1.5) 1.02 (0.6–1.5) 1.55 (1.0–2.3)

High negative support (N=3924 men/1332 women) Grade 1 (High) 29.9 1 2 (Middle) 36.7 1.33 (1.1–1.5) 3 (Low) 46.8 2.05 (1.5–2.6)

1 1.36 (1.1–1.5) 2.14 (1.6–2.8)

32.1 29.2 33.0

1 0.89 (0.5–1.3) 1.09 (0.7–1.6)

1 0.93 (0.6–1.4) 1.28 (0.8–2.0)

Spouse’s social class I/II III IV/V Not employed

1 0.92 (0.7–1.3) 1.01 (0.7–1.3) 0.97 (0.8–1.1)

1 0.85 (0.7–1.0) 0.88 (0.6–0.1) 0.95 (0.8–1.1)

32.7 30.5 17.9 43.4

1 0.92 (0.7–1.1) 0.46 (0.2–0.7) 1.63 (1.0–2.4)

1 0.83 (0.6–1.1) 0.41 (0.2–0.6) 1.55 (1.0–2.3)

Spouse is not closet (N=3961 men/1376 women) Grade 1 (High) 5.8 1 2 (Middle) 7.1 1.25 (0.9–1.6) 3 (Low) 10.6 1.94 (1.2–3.0)

1 1.27 (0.9–1.6) 2.02 (1.2–3.1)

8.9 16.2 20.8

1 1.99 (0.9–3.9) 2.74 (1.3–5.4)

1 1.84 (0.9–3.7) 2.50 (1.16–4.76)

Spouse’s social class I/II III IV/V Not employed

1 0.87 (0.6–1.1) 0.77 (0.4–1.3) 0.82 (0.6–1.1)

14.1 21.0 17.2 27.0

1 1.63 (1.1–2.2) 1.28 (0.7–2.1) 2.29 (1.4–3.7)

1 1.40 (1.0–1.9) 1.08 (0.6–1.8) 2.10 (1.3–3.3)

34.3 35.3 41.5 33.5

35.4 33.3 35.6 34.8

7.3 6.9 6.4 6.2

Age adjusted OR (95% CI)

1 1.03 (0.8–1.2) 1.36 (1.0–1.7) 0.97 (0.8–1.1)

1 0.94 (0.6–3.9) 0.87 (0.5–1.5) 0.84 (0.6–1.1)

Mutually adjusted OR (95% CI)

grade. In women, there was little indication of a relationship of either civil service grade, or the spouse’s social class, to low emotional support. The main difference between men and women in the relationship of partner’s social class to emotional support was in the effect of non-employment. Men with non-employed partners had a slightly lower risk, even than those with partners in the most advantaged social classes I and II. In contrast, women with non-employed male partners were significantly more likely to receive low social support than those with partners in classes I and II. Men in each lower civil service grade were more likely to report high levels of negative support than those in the grade above them. Adjustment for the class of their spouse did not explain this: if anything it increased the

strength of the relationship. In women, their own civil service grade had little effect. Whereas in men the partner’s social class had no effect, in women this was significantly related to negative social support. This relationship was in a somewhat surprising direction however. Women with husbands or partners in classes IV or V were significantly less likely than those with spouses in higher social classes to report high levels of negative support, although those with non-working partners were more likely to do so. The apparent difference between the effects of spouse’s class in men and women was supported by a significant interaction between sex and spouse’s class (p=0.002). Civil service grade was significantly related to the likelihood that the spouse or partner was not regarded

ARTICLE IN PRESS M. Bartley et al. / Social Science & Medicine 59 (2004) 1925–1936

as the closest confidante in both men and in women. This relationship was independent of the spouses’ social class. In men, the spouse’s class had no effect on whether or not they were cited as the closest person. However, in women, spouse or partner’s class was related to the likelihood that they were cited as the closest confidante: in particular, non-employed male spouses were over twice as likely not to be regarded as the closest person in a woman’s social network regardless of her civil service grade. The interaction between sex and social class of spouse was significant (p=0.01), confirming the stronger relationship between husbands’ than wives’ socioeconomic position and the likelihood of their being regarded as the closest confidante.

Discussion The paper began with an attempt to combine theoretical approaches to class and gender differences in health risk, which generated six hypotheses regarding the distribution of risk factors in men and women. We hypothesised that because of gender differences in labour market attachment, and associated differentials of power within the household, socio-economic position based on the individual’s own occupation would tend to have a stronger influence on behavioural risk factors and social support in men than in women, and the SEP of male partners a stronger influence on the risk profiles of women. Our first hypothesis was that health related behaviours would be more strongly related to socio-economic position based on the individual’s own occupation (civil service grade) in men than in women. This was confirmed for smoking and exercise, though not for diet or alcohol consumption. Civil service grade was significantly related to the risk of alcohol consumption over the ‘safe limit’. However, it was men and women in the highest grade who were most at risk in this respect, and the relationship was stronger in women than in men. In the case of diet, socio-economic position based on civil service grade had a similar effect in men and women. The second hypothesis was that health behaviours in women would be related to the socio-economic position (as measured in terms of social class based on occupation) of a male spouse or partner, after allowing for the woman’s own civil service grade. This was once again clearly seen in the cases of smoking, diet and exercise, which remained significantly related to spouse’s social class after adjustment for the women’s own civil service grade. The social class of the spouse was not significantly related to alcohol consumption in either men or women. We expected to find that health behaviours in men would have no relationship to the SEP of a female

1933

spouse or partner. Once again, this was clearly confirmed only for smoking and exercise. The social class of the female partner did remain significantly related to men’s diet quality after adjustment for the male partner’s civil service grade. Men with nonemployed partners were less likely to have ‘risky’ levels of alcohol consumption. Because our measure of social support is focused on the relationship to the ‘closest person’ who is most often the spouse, our hypotheses in relation to this psychosocial risk factor were also governed by ideas about interactions in the home. We expected that, in households where the male partner, most closely connected to the labour market, was of a more privileged socioeconomic position, some sources of strain on these relationships would be minimized. Firstly, the quality of social support received was therefore expected to be more strongly related to SEP based on own occupation in men than in women. The data show that this was only the case for negative social support. Confiding support and the probability that the spouse was the closest confidante were both more strongly influenced by SEP (civil service grade) in female than in male Whitehall II participants. Secondly, we expected that the quality of social support received by women participants would be influenced by the SEP (occupational social class) of their partner. Once again this was only confirmed for negative support, but in a surprising direction. After adjustment for their own civil service grade, women with partners in social classes IV and V were significantly less likely to report high levels of negative support than those with partners in the most privileged social classes. Finally, under the heading of psychosocial risk, we expected that social support received by male Whitehall II participants would have no, or only a weak relationship to the social class of their spouses or partners. This was confirmed for negative social support and for reporting the spouse as closest confidante. However, men (unlike women) with partners in social classes IV and V were significantly more likely to report low confiding and emotional support, compared to men whose partners were in the most advantaged classes I or II. Although the prediction that less advantaged social class position in the male partner would be linked to lower levels of social support received by women was not confirmed, non-employment in the husband or partner was associated with relatively lower levels of positive and higher negative social support. Women with non-employed partners seem to have had the least favourable psychosocial situation. In contrast, men with non-working wives or partners were unaffected. Because of the wider situation of male dominance in the labour force, giving easier access to high paid employment (and higher pay in many cases even in the same type of

ARTICLE IN PRESS 1934

M. Bartley et al. / Social Science & Medicine 59 (2004) 1925–1936

employment), lack of employment in the male partner may have more severe consequences for the household in financial terms. In terms of conventional power relations within households, it may still be the case that where the woman is employed (as are all women Whitehall II participants in this study) lack of paid employment in the male partner creates role incongruence. Do the findings help us to understand gender differences in inequalities in cardiovascular disease? In order to be consistent with gender differences in inequalities in cardiovascular disease, we would expect to see the social class of female partners having only a very small, if any, effect on risk factor exposure in men. There should be greater differences in risk exposure between married women according to the socioeconomic position (in this case, social class) of their male partners, than according to their own socioeconomic position (in this case, their civil service grade). Relationships of risk factors in women to the SEP of male partners would also have to be in the right direction: greater risk in women married to men in less advantaged social classes. The findings for smoking and exercise are consistent with observed differentials in disease and mortality, in that the class of the male partner had a strong relationship to these health behaviours in women, regardless of women’s own civil service grade. Social patterning of diet quality was ambivalent in this respect, and the gradients for alcohol consumption high enough to increase health risk were in the wrong direction. However, smoking and exercise are found in many studies to be the risk factors most closely implicated in eventual cardiovascular disease incidence. The findings for social support are more difficult to interpret. The findings that civil service grade was more strongly related to the risk that the spouse was not the closest confidante, and to low confiding and emotional support, in women than in men, do not accord with what we might expect if gender differences in health inequality were in part produced by the distribution of these risk factors. In contrast, negative social support, the measure that best predicts psychological ill-health (Stansfeld, Rael, Head, Shipley, & Marmot, 1997; Stansfeld, Fuhrer, & Shipley, 1998) was strongly associated with civil service employment grade in men but not in women. This is consistent with negative support as one reason for stronger social gradients in the health of men than women. Social support from the partner received by women, unlike behavioural risk factors, was in general not strongly negatively affected by their husband or partner being in a less advantaged social class. The results for this particular form of psychosocial risk factor were therefore less compatible than those for behavioural risk factors with observed gender differences in health inequality.

The results presented here do not support the idea that women’s employment, even in a high-prestige job, is harmful to male partners. In no respect were men with partners in the highest social class at greater risk than other men. However, non-employment in the male partner is clearly related to high levels of psychosocial risk factors in women. Men in less privileged social classes are known to experience far more frequent periods of unemployment (Moser, Goldblatt, Fox, & Jones, 1987; Fox & Shewry, 1988; Arber, 1996; Kasl & Jones, 2000) and long-term work disability as well as lower income when in work. In households headed by men in these groups, there will be, over time, a combination of chronically low income from work, combined with more frequent and longer periods unemployed or outside the labour force (Westergaard, Noble, & Walker, 1989). Over time, therefore, women living with men in less advantaged occupational social classes are more likely to accumulate both material and psychosocial hazards of living in a household with a non-employed partner. The women participants in the present study were all in employment, at phase I of the Whitehall II Study. This makes them a suitable sample for the comparison of the effects of own versus partners’ socio-economic position. It may, however, mean that the importance of women’s own occupational position is greater in these data than might be found in a general population sample, in which a higher proportion of women would have to be classified on the basis of a past occupation. The composition of the sample does have some disadvantages, most importantly that participants in the Whitehall II study come from a relatively narrow occupational range. Although pay levels at the time of this survey did vary widely, job content was less representative of the working population, in that there were no skilled and few semi-skilled manual workers, and a higher proportion of routine clerical workers. The specific hypotheses addressed here do not involve differing levels of knowledge or culture as causes of social differences in health risk, but rather focus on differences in economic resources and power within the household. Therefore, civil service grade and social class are the most relevant measures of socio-economic position.

Conclusion Studying gender differences in health inequality highlights some of the problems in health inequality research more broadly. We are brought face to face with the fact that the development of conceptual models that can be applied consistently to aetiology in both men and women is still at an early stage. This paper has used a combination of theoretical approaches adopted from

ARTICLE IN PRESS M. Bartley et al. / Social Science & Medicine 59 (2004) 1925–1936

different areas of the social sciences and epidemiology to generate a set of hypotheses. These have been based on consideration of gender differences in access to more secure and privileged positions in the labour force combined with imbalances of power over household roles. Gender differences in access to privileged employment are strikingly illustrated in the different distribution of men and women amongst the civil service grades. Differences in power between household roles are equally vividly evident in the enormous difference in the proportions of high grade men and women with nonemployed partners. When confronted with data from the Whitehall II study on social differences in cardiovascular risk factors, only some of the hypotheses received confirmatory evidence. Whereas smoking, diet and exercise in women were influenced as much or more by their partners’ social class, social support was more strongly linked to women’s own socio-economic position. Closer attention is needed to the different processes behind material power and ‘emotional power’ within the household when investigating gender differences in health and risk factors.

Acknowledgements The Whitehall II study has been supported by grants from the Medical Research Council; British Heart Foundation; Health and Safety Executive; Department of Health; National Heart Lung and Blood Institute (HL36310), US, NIH: National Institute on Aging (AG13196), US, NIH; Agency for Health Care Policy Research (HS06516); and the John D and Catherine T MacArthur Foundation Research Networks on Successful Midlife Development and Socio-economic Status and Health. MM is supported by an MRC Research Professorship. We also thank all participating civil service departments and their welfare, personnel, and establishment officers; the Occupational Health and Safety Agency; the Council of Civil Service Unions; all participating civil servants in the Whitehall II study; and all members of the Whitehall II study team.’’

References Arber, S. (1996). Integrating nonemployment into research on health inequalities. International Journal of Health Services, 26, 445–481. Bartley, M. (2003). Commentary: Relating social structure and health. International Journal of Epidemiology, 32, 958–960. Brixi, O., & Lan, T. (2000). Comportements. In A. Leclerc, D. Fassin, H. Grandjean, M. Kaminski, & T. Lang (Eds.), Les Inegalites Sociales de Sante (pp. 391–402). Paris: INSERM/ La Decouverte.

1935

Chandola, T., Bartley, M., Wiggins, R., & Schofield, P. (2003a). Social inequalities in health by individual and household measures of social position in a cohort of healthy people. Journal of Epidemiology and Community Health, 57, 56–62. Chandola, T., Kuper, H., Singh-Manoux, A., Bartley, M., & Marmot, M. (2003b). The effect of control at home on CHD events in the Whitehall II study: Gender differences in psychosocial domestic pathways to social inequalities in CHD. Social Science & Medicine, 58, 1501–1509. Dahl, E. (1991). Inequality in health and the class position of women: The Norwegian experience. Sociology of Health and Illnesses, 13, 491–505. Diez-Roux, A. V. (2000). Multilevel analysis in public health. Annual Review of Public Health, 21, 171–192. Egeland, G. M., Tverdal, A., Meyer, H. E., & Selmer, R. (2002). A man’s heart and a wife’s education: A 12 year coronary heart disease mortality follow-up in Norwegian men. International Journal of Epidemiology, 31, 799–805. Fox, A. J., & Adelstein, A. M. (1978). Occupational mortality: Work or way of life? Journal of Epidemiology and Community Health, 32, 73–78. Fox, A. J., & Shewry, M. (1988). New longitudinal insights into relationships between unemployment and mortality. Stress Medicine, 4, 11–19. Graham, H. (1987). Women’s smoking and family health. Social Science & Medicine, 25, 47–56. Graham, H. (1998). Promoting health against inequality. Health Education Journal, 57, 292–302. Griffin, J. M., Fuhrer, R., Stansfeld, S. A., & Marmot, M. (2002). The importance of low control at work and home on depression and anxiety: Do these effects vary by gender and social class? Social Science & Medicine, 31, 783–798. Harding, S., Bethune, A., Maxwell, R., & Brown, J. (1998). Mortality trends using the longitudinal study. In F. Drever, & M. Whitehead (Eds.), Health inequality (pp. 143–155). London: HMSO. Heimdal, K. R., & Houseknecht, S. K. (2003). Cohabiting and married couples’ income organization: Approaches in Sweden and the United States. Journal of Marriage and the Family, 65, 525–538. Johnson, J. V., Stewart, W., Hall, E. M., Fredlund, P., & Theorell, T. (1996). Long-term psychosocial work-environment and cardiovascular mortality among Swedish men. American Journal of Public Health, 86, 324–331. Joshi, H. (1998). The opportunity costs of childbearing: More than mothers’ business. Journal of Population Economics, 11, 161–183. Kasl, S. V., & Jones, B. A. (2000). The impact of job loss and retirement on health. In L. F. Berkman, & I. Kawachi (Eds.), Social epidemiology (pp. 118–136). Oxford: Oxford University Press. Kivim.aki, M., Leino-Arjas, P., Luukkonen, R., Riihim.aki, H., Vahtera, J., & Kirjonen, J. (2002). Work stress and risk of cardiovascular mortality: Prospective cohort study of industrial employees. British Medical Journal, 325, 857. Koskinen, S., & Martelin, T. (1994). Why are socioeconomic mortality differences smaller among women than among men. Social Science & Medicine, 38, 1385–1396. Krieger, N., Chen, J. T., & Selby, J. V. (1999). Comparing individual and household based measures of social class to

ARTICLE IN PRESS 1936

M. Bartley et al. / Social Science & Medicine 59 (2004) 1925–1936

assess class inequalities in women’s health. Journal of Epidemiology and Community Health, 53, 612–623. Kuper, H., & Marmot, M. (2003). Job strain, job demands, decision latitude, and risk of coronary heart disease within the Whitehall II study. Journal of Epidemiology and Community Health, 57, 147–153. Kuper, H., Singh, M. A., Siegrist, J., & Marmot, M. (2002). When reciprocity fails: Effort-reward imbalance in relation to coronary heart disease and health functioning within the Whitehall II study. Occupation and Environmental Medicine, 59, 777–784. Lindbladh, E., Lyttkens, C. H., Hanson, B. S., Ostergren, P., Isacsson, S.-O., & Lindgren, B. (1996). An economic and sociological interpretation of social differences in health related behaviour: An encounter as a guide to social epidemiology. Social Science & Medicine, 43, 1817–1827. Luoto, R., Pekkanen, J., Uutela, A., & Tuomilehto, J. (1994). Cardiovascular risks and socio-economic status—differences between men and women in Finland. Journal of Epidemiology and Community Health, 48, 348–354. Lynch, J. W., Kaplan, G. A., & Salonen, J. T. (1997). Why do poor people behave poorly? Variation in adult health behaviours and psychosocial characteristics by stages of the socio-economic lifecourse. Social Science & Medicine, 44, 809–819. Marmot, M. G., Smith, G. D., Stansfeld, S., Patel, C., North, F., & Head, J., et al. (1991). Health inequalities among British civil servants—the Whitehall II study. Lancet, 337, 1387–1393. Martikainen, P. (1995). Mortality and socio-economic-status among Finnish women. Population Studies—A Journal of Demography, 49, 71–90. McDonough, P., Williams, D. R., House, J. S., & Duncan, G. (1999). Gender and the socio-economic gradient in mortality. Journal of Health and Social Behaviour, 40, 17–31. Moser, K., Goldblatt, P. O., Fox, A. J., & Jones, D. R. (1987). Unemployment and mortality. British Medical Journal, 294, 509. Moser, K. A., Pugh, H. S., & Goldblatt, P. O. (1988). Inequalities in women’s health—looking at mortality differentials using an alternative approach. British Medical Journal, 296, 1221–1224. Peter, R., Alfredsson, L., Hammar, N., Siegrist, J., Theorell, T., Westerholm. (1998). High effort, low reward, and cardiovascular risk factors in employed Swedish men and women: Baseline results from the WOLF study. Journal of Epidemiology and Community Health, 52, 540-547. Peter, R., Geissler, H., & Siegrist, J. (1998). Associations of effort-reward imbalance at work and reported symptoms in different groups of male and female public transport workers. Stress Medicine, 14, 175–182.

Pickering, T. (1999). Cardiovascular pathways: Socio-economic status and stress effects on hypertension and cardiovascular function. Annuals of the New York Academy of Sciences, 896, 262–277. Ribet, C., Lang, T., Zins, M., Bingham, A., Ferrieres, J., & Arveiler, D., et al. (2001). Do cardiovascular risk factors in men depend on their spouses’ occupational category? European Journal of Epidemiology, 17, 347–356. Rubery, J., & Tarling, R. (1988). Women’s employment in declining Britain. In J. Rubery (Ed.), Women and recession (pp. 100–132). London: Routledge and Kegan Paul. Siegrist, J. (2000). Place. Social exchange and health: Proposed sociological framework. Social Science & Medicine, 51, 1283–1293. Siskind, V., Najman, J. M., & Veitch, C. (1992). Socioeconomic-status and mortality revisited—an extension of the Brisbane area analysis. Australian Journal of Public Health, 16, 315–320. Stansfeld, S. A., Fuhrer, R., & Shipley, M. J. (1998). Types of social support as predictors of psychiatric morbidity in a cohort of British Civil Servants (Whitehall II Study). Psychological Medicine, 28, 881–892. Stansfeld, S., & Marmot, M. (1992). Deriving a survey measure of social support: The reliability and validity of the close persons questionnaire. Social Science & Medicine, 35, 1027–1035. Stansfeld, S. A., Rael, E. G., Head, J., Shipley, M., & Marmot, M. (1997). Social support and psychiatric sickness absence: A prospective study of British civil servants. Psychological Medicine, 27, 35–48. Stronks, K., van de Mheen, H., van den Bos, J., & Mackenbach, J. P. (1995). Smaller socio-economic inequalities in health among women: The role of employment status. International Journal of Epidemiology, 24, 559–568. Suarez, L., & Barrett Connor, E. (1984). Is an educated wife hazardous to your health? American Journal of Epidemiology, 119, 244–249. Theorell, T., Tsutsumi, A., Hallquist, J., Reuterwall, C., Hogstedt, C., & Fredlund, P., et al. (1998). Decision latitude, job strain, and myocardial infarction: A study of working men in Stockholm, The SHEEP study group Stockholm heart epidemiology program. American Journal of Public Health, 88, 382–388. Westergaard, J., Noble, I., & Walker, A. (1989). After redundancy: The experience of economic insecurity. Cambridge: Polity Press. Wilkinson, R. G. (1996). Unhealthy societies: The afflictions of inequality. London: Routledge. Winkler, A. E. (1997). Economic decision-making by cohabitors: Findings regarding income pooling. Applied Economics, 29, 1079–1090.