Development of job demands, decision authority and social support in industries with different gender composition – Sweden, 1991–2013

This study aims to explore the development of job demands, decision authority and social support within and between industries with different gender composition in Sweden. The specific research questions are whether high job demands, low decision authority and poor social support among women and men: i) have changed over the study period within the industries; ii) differ between industries over the study period.

The respondents are employees living in Sweden, aged 16–64 years, who were not absent, due to sickness or other types of leave, for at least 3 months prior to the data collection. Data were derived from the Swedish Work Environment Survey (SWES) conducted biennially since 1989 by the Swedish Work Environment Authority in collaboration with Statistics Sweden. SWES participants are sampled from the biennial Labour Force Survey (LFS). More than 20,000 persons are randomly drawn from the whole Swedish population, and are stratified by county, sex, citizenship and employment status. Further on, a random sub-sample of gainfully employed people from LFS are sent self-completion SWES questionnaires [33]. These questionnaires contain, among other, questions about the psychosocial work environment.

The current study uses cross-sectional data from 12 waves of SWES (1991 to 2013) and comprises in total 109,698 respondents. The response rate has decreased from 87.2% in 1991 to 58.7% in 2013, with a mean response rate of 75.1% over the 12 waves [3]. The response rate by year of SWES is provided in the additional file 1. Participation across waves has been lower for men and among individuals with low education, low income, and with foreign background. Furthermore, it has also been lower among the self-employed and among individuals with part-time or short-term employment contracts.

SWES data were linked to registry data from the Longitudinal integration database for health insurance and labour market studies (LISA) which annually integrates data on the labour market, educational and social sectors on all individuals that are more than 16 years of age registered in Sweden. In the current study, the labour market industry that the respondent was employed in at the time of the data collections, as well as the participants’ highest degree of education, were drawn from LISA.

The psychosocial work factors used in this study were drawn from SWES 1991–2013: job demands, decision authority and social support.

The variables job demands and decision authority were based on four items each whereas social support was derived from two separate questions. These variables serve as indicators of the demand-control-support model dimensions [34] and have been used in previous studies [34‐36]. According to the above procedure, respondents were classified as having high job demands if they reported at least two of the following: “have to skip lunch, work late or take work home” at least once every week; at least half of the time “does not have time to talk or even think of anything other than work”; “the work requires undivided attention and concentration” nearly all the time; and agree fully or to some extent to “having far too much to do” (Chronbach’s alpha 0.64). Respondents were classified as having low decision authority if they reported two of the following: “possibility to set the work tempo” half of the time or less; mostly not or never “possibility to decide when various tasks are to be done”; never or mostly “not involved in planning the work”; and agree fully or to some extent to “having too little influence at work” (Cronbach’s alpha 0.63). Respondents were classified as having poor social support if they reported that mostly not or never “received support” from the superiors and/or colleagues “when the work becomes troublesome” (Chronbach’s alpha 0.64). The original response items and scales used to compose each of the variables can be found in the additional file 2.

Age and education of the participants were derived from registry data. Age was categorized in five groups: 16–25, 26–35, 36–45, 46–55, 56–64. The highest degree of education of the respondents was classified into five categories: compulsory education, two years of upper secondary school, three or four years of upper secondary school, university or equivalent less than three years, and university or equivalent three years or more.

The proportion of respondents reporting to be exposed to the studied psychosocial work factors between 1991 and 2013 in industries with different gender composition were calculated. Due to changes in the SNI classification system over time, the analyses were conducted separately for periods using the same industry classification: 1991 to 2007 and 2009 to 2013. Because of decreasing response rates, the use of weighted data based on sex and occupation, provided by Statistics Sweden, was considered. However, the variation between the proportions was lower than 3% between weighted and unweighted data. As the use of weighted data restricted the statistical analyses that were possible to perform, unweighted data was used. To estimate if the adverse psychosocial work factors varied between the years when comparing to the first year of the study (1991), logistic regressions were conducted, adjusted for age. SWES data were pooled in 5 groups: 1991 (the reference category), 1993 to 1995, 1997 to 2001, 2003 to 2007 and 2009 to 2013. Estimates are reported as Average Marginal Effects, which can be interpreted as the average change in the probability (0 to 1) of reporting exposure to high job demands, low decision authority and poor social support in each of the group years compared with the reference group. The SWES wave of 1991 was used as the reference category, as it is the year before the Swedish economic crisis, when work conditions were expected to be better than in later years [37]. Estimates show the discrete change from this category with 95% confidence interval. To estimate if the psychosocial work factors differed between industries and between the earlier and later periods of the study, SWES data was pooled into four groups. The first group comprises data from the beginning of the study period (1991, 1993 and 1995), the second group and third group uses data from the middle of the study period (1997, 1999 and 2001; 2003, 2005 and 2007, respectively). The fourth group uses data from the end of the study period (2009, 2011 and 2013). Logistic regressions were conducted, adjusted for age, education and year of data collection. Estimates are reported as Average Marginal Effects, which can be interpreted as the average change in the probability (0 to 1) of reporting exposure to high job demands, low decision authority and poor social support in each industry compared with the reference group. Knowledge intensive services was used as the reference category, as it is a gender-mixed industry and employs highly skilled workers in the private sector, often associated with good psychosocial work conditions. Estimates show the discrete change from this category with 95% confidence interval. All analyses were stratified by sex and performed in STATA version 15.1 for Mac™. The study is reported in adherence with the STROBE guidelines.

The results show that high job demands and low decision authority are particularly high in education and health and social care. The development of these factors has been worse in these industries, compared to others in the labour market, over the study period. This negative development is particularly pronounced among women.

The increase in the proportions of high job demands and poor decision authority was steeper in the 1990s, for workers in education and health and social care (Figs. 1 and 2). These results could reflect the organizational changes that occurred in these industries, following the economic crisis in the 90s and NPM, which can affect the development of negative health consequences, as suggested in a previous paper [18].

The highest proportions of workers reporting low decision authority were found in health and social care, for both men and women. Thus, poor decision authority appears to be experienced across occupations and genders in this industry, indicating that this is a psychosocial work environment problem to be targeted at the industry level. Low professional autonomy across occupations and positions in health and social care has been found to be a consequence of the introduction of new public management in previous studies [15, 18].

When analysing workers in the female-dominated industries, there are noticeable differences between genders. The proportions of women reporting high job demands have remained on similarly high levels between the end of the 1990s and the later data collections of this study. For men, the high proportions decreased again after the 1990s. Also, for low decision authority, the proportions reported by women have increased further in the later years of the study period, for those in female-dominated industries. The different patterns for women and men may be explained by the fact that, within female-dominated industries, men and women have different occupations (horizontal gender segregation within the industry) and occupy different positions (vertical gender segregation within the industry). Most men in this industry have at least 3 years of university education (see Additional file 1), whereas the majority of women are found in the lowest educational groups, in caring jobs with close contact with patients and clients. A deterioration of job demands and decision authority in caring jobs have previously been reported [18, 38].

The increased risks for mental ill-health in female-dominated industries observed in Sweden in the last years have been hypothesized to be due to poorer psychosocial work environment in these industries compared to others [21‐23]. Though no mental ill-health measures have been included in this study, poorer development of job demands and job control for both men and women in these industries has been observed.

The proportions of high job demands fluctuated between 1991 and 2007 among both genders in labour intensive services, and among men in knowledge intensive services. The highest proportions were observed in the second part of the 1990s, and the values were similar at the start and end of the study period. This finding is in line with previous research, showing increased job intensity in times of economic crisis also in high-skilled jobs [39]. In public administration, the average marginal effects were lower among men in all time periods, when compared with the reference category of knowledge intensive services.

The proportions of workers reporting low decision authority are the lowest for those in knowledge intensive services and public administration. This finding supports results based on data from European Union (EU) countries, showing that job control has increased in the last decades particularly in high-skilled jobs, and especially so in the Nordic countries [39].

The highest proportions of poor social support were reported by both men and women in machinery operations. This can be due to the nature of work in this industry, as it can have a somewhat individualistic connotation. For women, in recent years, those working in machinery operations report the highest values of poor social support. This can be interpreted as the gender in minority having worse social support than those in majority, in line with the tokenism theory [40], but further studies are needed, including exploring the dramatic increase in 2013 for women in machinery operations.

The primary focus of the present study was not to compare women and men. However, although the chosen strategy of gender stratified analysis does not allow to make actual gender comparisons, some evident patterns emerged. Many more women seemed to have lower decision authority than men in all industries in our study (Figs. 3 and 4). On the other hand, more men than women reported poorer social support (Figs. 5 and 6), with one exception: women in machinery operations in 2013. Thus, in addition to systematic differences between industries, there seems to be structural gender-related differences across the labour market.

We chose to analyse the development of the psychosocial work factors by industry because of its relevance for the organization of work as this level may be particularly suitable for interventions aimed at improving the psychosocial working conditions. However, our results lend support for that all variation in working conditions may not be captured at the industry level. The results regarding gender differences within industries suggest that there are also conditions that are specific to occupation and positions within industries, as discussed in a recent report by the Swedish Social Insurance Agency [4]. However, further exploration of the present data is needed in order to disentangle to what extent differences are due to variation between occupations and positions, to variations between genders, or both.

To our knowledge, this study is one of the first of its kind exploring the development of the dimensions of the demand-control-support mode over a long period of time and between industries, considering their gender composition, in a representative sample of the whole Swedish working population.

However, there are several limitations that should be accounted for when interpreting the results. First, the response rate has been decreasing over time, and lower responses rates have been observed for men, individuals with low education, low income and foreign background, across the study period. Hence, it is with caution that these results can be generalized to the Swedish working population as a whole. Second, the data are based on self-reported measures, as most studies on psychosocial working conditions, which are vulnerable to response bias. The measures of the demands, decision authority and support dimensions that were used in the present study are not directly comparable with the standard measures of the demand-control-support model [24]. However, these measures have shown adequate internal consistency and have been used in previous studies, allowing for comparisons between groups and over time [34‐36].

The industry classification used in this study considers not only the gender composition but also the type of work of the industry, providing a more scrutinized and complete classification than if only numbers of men and women in each industry were taken into consideration. However, the official categorization of industries changed in 2007. Overall, the classification of industries is very similar for both periods. The two groups that changed between one classification to the other were knowledge intensive services and labour intensive services. This was mainly due to new and growing industries being identified and classified. To overcome this limitation, our analyses of development over time were made in separate time-periods. Although only a small proportion of the populations is classified differently in the two study periods, this should be considered when interpreting our results.

Knowledge intensive services was chosen as the reference category for the analyses of differences between industries. The reason was that knowledge intensive services is a gender-mixed industry employing individuals with high educational levels in high-skilled jobs, often associated with good working conditions. However, high job demands were rather common in this industry over the study period. Larger AMEs of reporting high job demands would have been observed in some industries, if the industry with the best conditions would have been used as the reference category. For interpretability of the results, we decided to use the same reference category for all work dimensions.

In this study, a seven-industry classification was used to explore the development of psychosocial working conditions. In forthcoming studies, this classification could be further explored. For example, as public administration and knowledge intensive services present similar distribution of educational and work exposure levels, they could be merged into one group in future studies. On the other hand, an even more detailed analysis of variation between industries could be explored if using the original 15 or 19 industries in the Swedish Standard Industrial Classification (SNI) system.

It should be noted that all employees at a production unit are classified according to its main activity. This means that the differences between male- and female- dominated industries in working conditions that we have shown, probably are underestimated for occupations in minority. These are administrators, service workers and managers, that are differently “gendered” than the workers performing the core production, and have other job tasks that might affect job demands, control and social support. However, it is noteworthy that in spite of this limitation, several significant differences between industries were found. To get more precise estimates, information on occupation could be used for stratification of workers performing the core tasks from other functions in the production units, for example.

To compare change over time between industries, data was pooled in four time periods, each comprising three waves of SWES. The period between 1991 and 1995 captures not only the early effects of the economic crisis of 1992 [41, 42] but also the early effects of new public management, which affected particularly the public sector, as discussed in previous sections. Using all periods allows for a better understanding of the development of psychosocial working conditions over time.

Given how working life is organized, studying the development of psychosocial work factors over time by industry contributes with applicable knowledge to the design of healthy workplace. This study contributes with an approach that opens future research to link the fields of production technology and work organization with the work environment and occupational health, as previously requested [43‐45]. If only gender, class or occupation are compared when exploring changes of work environment over time, it is not possible to disentangle whether changes in work conditions are due to structural changes in industries, effects of work environment improvements or due to changes of the occupational distribution of the labour force [46]. Industries tend to use similar technologies and management strategies. Hence, it is reasonable to believe that larger structural and technological changes in an industry, such as in health and social care during the 1990s, also lead to changes in job tasks and composition of occupations [47]. Future longitudinal research comparing units that are linked to the production process, such as industries, should advance knowledge on the pathways to healthy or hazardous work.

The present study also contributes to the research field of stratification, inequality and gendered processes in working life. The results indicate inertia of horizontal division of labour in industries with different types of operations or work objects. Since a few decades ago, industries with the core operation aimed at “handle material things” employ almost half of the male workforce in Sweden. On the other hand, almost half of the female workforce has “humans” as work objects, organized in two large welfare services. One mechanism to be explored in future research concerns how operations with different work objects are “gendered” and associated with different status, and how this affects the workers influence in periods of structural changes, as indicated in the present study.