Inequalities in employment rates among older men and women in Canada, Denmark, Sweden and the UK

Many countries are developing policies to encourage older workers to remain in paid employment, thus delaying retirement [1], to reduce the burden on societies related to disability benefits or pension payments [2]. In the European Union, the old age dependency ratio (the number of older people as a proportion of those of working age) is expected to rise to 38% by 2030 [3]. Extending working lives may also improve the living conditions of older workers, through higher pension levels. While it is outside the scope of this paper to argue whether employment past 65 years of age is good or bad, we recognise that employment is one factor noted in the public health literature contributing to healthy ageing and a key priority of most governments [4].

Despite the policy push to work past 65 years and older, little evidence on inequalities exists about who remains in employment. We aim to address this gap by identifying inequalities in employment rates among older workers aged 65–75 years focusing mainly on those aged 65–69 years. While several factors may influence the decision to extend ones working life [3], an obvious factor is poor health. Poor health could force a person to exit employment even if they would prefer to continue working [5]. Indeed, evidence shows that health is a key determinant of early retirement [6]. Another important factor is sex. Studies show that, in general, employment rates are lower among women than men [7, 8]. This might indicate greater caring responsibilities for children and older relatives among women than among men. Finally, educational level is another major factor. For example, in Sweden, there are considerable differentials in the length of working life in different occupations. In fact, Nilsson et al. [9] show that workers with lower educational levels tend to leave employment earlier than high educated workers due to health reasons or unemployment. One explanation could be that those with lower levels of education typically have more physically demanding jobs than those with higher levels of education and therefore their work capacity is more affected by decreases in health than those with higher levels of education [10, 11]. Attempts at extending working lives may therefore potentially further aggravate social differentials in employment rates as well as in living conditions. However, a limitation of studies such as Nilsson et al. [9] is that the focus neglects workers older than 65 years. We found only two comparative studies that examine employment rates for those aged 65 years and older [12, 13]. A growing body of literature (see for example [14‐16]) about ‘unretirement’ – the process of returning to work after retirement and ‘bridge employment’ (also known as partial employment) – includes our age group of interest but focuses on other aspects such as the meaning of work (see for example [17‐19]), life satisfaction (see for example [20]) rather than inequalities. Our contribution to the literature is a comparative, public health focused examination of inequalities in employment rates past age 65 years. To the best of our knowledge, this is the first study to take such an approach.

This paper is an introduction to a larger project called Tackling Health Inequalities and Extending Working Lives (THRIVE) funded by the Joint Programming Initiative–More Years Better Lives. The THRIVE collaboration builds on previous comparative work that identified inequalities in the general working age population (see for example [7, 21, 22]) between Social Democratic and Liberal welfare regimes. As such, the selection of two Social Democratic welfare regimes–Denmark and Sweden and two Liberal welfare regimes–Canada and the United Kingdom (the UK) are based on these previous collaborations, extending research on employment and health to the older age groups. The following briefly explains the THRIVE theoretical framework.

Decisions to retire are multifaceted and structural factors related to policies can influence extending working lives. For example, the design of the pension system (e.g. earlier retirement could lead to lower pension levels), active labour market policies (ALMPs) (e.g. re-training older workers for local labour market demands) and employment protection policies (e.g. anti-discrimination laws, more permanent (and protected) work) [23]. Diderichsen, et al. [24] provides a theoretical framework for analysing how these policies may mitigate the effects of social disadvantage on the risk of disease and poor health and subsequent adverse social and economic consequences at older ages.

As illustrated in Fig. 1, policy entry points for intervention may target mechanisms at various levels operating between social disadvantage and limiting longstanding illness [24]. Social stratification (entry point A) involves the allocation of wealth and power in society (e.g. via education) and can impact on the path into social positions. Social protection policies (e.g. social insurance and social assistance) can help alleviate the income-related inequalities that exist due to social stratification. Policies also influence the exposure and the effect of being exposed to risk or protective factors (entry point B and C). For example, work environment policies may protect workers from health-hazardous exposures. Entry point D encompasses policies to mitigate the negative social and economic consequences of the disease (e.g. employment protection legislation, rehabilitation and work training to bring back people to work). For example, Social Democratic welfare regimes in the Nordic countries provide stronger social protection, family policies and employment protection policies than Liberal welfare regimes [25], resulting in higher employment rates among persons with low education and health problems in Sweden compared to the UK [7, 26, 27]. In THRIVE, we hypothesise that policies aimed at increasing employment among the general working age population (i.e. policies in Social Democratic welfare regimes) would facilitate higher employment rates also among older workers. The prevailing level of unemployment in the country could also affect the choices and chances of employment of older persons. In 2013, the unemployment rates in the different countries were 5.9% (CAN), 7.1% (DK), 8.1% (SE), and 7.5% (UK), respectively [28].

The primary aim of this study was to investigate social and health differentials in employment rates among men and women aged 65–75 years and how this compares across Canada, Denmark, Sweden and the UK. Our specific research questions were:

Based on results from the research questions above, we plan to identify which inequalities (e.g. between men and women, different socio-economic groups or specific countries) to focus on in future studies.

We used nationally representative cross-sectional survey data from the the Canadian Community Health Survey (CCHS), Survey of Health, Ageing and Retirement in Europe (SHARE) in Denmark and Sweden and the English Longitudinal Study of Ageing (ELSA). In this paper, our population of interest are individuals aged 65–75 years.

The CCHS is an ongoing series of cross-sectional health interview surveys administered approximately every 2 years by Statistics Canada, the national statistics agency. Using a multi-staged, stratified sampling frame, the CCHS target population consists of household residents aged 12 and older who are living in private dwellings in all provinces and territories. The survey design features and core content have remained largely unchanged during the series of surveys starting in 2001. We used data from the 2012–2013 survey cycle. For more information on the CCHS see [42].

SHARE is a panel study with bi-annual follow-ups regarding health and socioeconomic conditions (www.​share-project.​org) (for more information on SHARE see [43]). SHARE data was collected as face-to-face interviews carried out by trained interviewers. We used data from the 5th wave collected in 2013/14. The number of participants (and participation rates) in Denmark and Sweden were 4146 (61.0%) and 4556 (33% for the refreshment sample and 59% for the panel sample), respectively. The numbers of interviewed in the age interval 65–75 were 1295 in Denmark and 1914 in Sweden (see Table 1).

ELSA is a nationally representative panel survey of people aged 50 and above in England which began in 2002 (for more information on ELSA [44]). The data used here is taken from the 2013 wave. SHARE is partially harmonised with ELSA to facilitate cross-national analysis. ELSA is a household survey and had a response rate of about 70% for at least one household member at wave 1. Attrition is low between waves with 86% of core sample members responding to wave 6. Refreshment samples are included and elicit a similar response rate to wave 1. Cross-sectional weights are available to correct nonresponse bias and attrition. The UK Data Service manages and licenses the use of the ELSA data.

We stratified employment rates by sex, educational level and health status (having limiting longstanding illness (LLI) or not). We generated descriptive summary tables for all four countries.

For the age group 70–75 years, stratified employment rates were harder to examine due to the lack of available data on this age group. However, we also calculated employment rates for data that was available in the surveys used for this study by sex, educational level and those with and without a LLI or not for those aged 70–75 years. Similarly, to the age group 65–69, employment rates for this older age group were lower for women (employment rates among men ranged from 27% to 5% compared to employment rates among women ranging from 14% to 1%). For low-educated women with a LLI, employment rates were lower than 5% across all countries.

Our comparative approach illustrates the variations in who continues to remain in employment. For example, employment rates vary little by educational level in the UK data. For UK men without an LLI the employment rate for low, medium and high educated is 30% in the UK. On the other hand, employment rates are much higher for highly educated persons compared to low-educated persons in Canada, Sweden and Denmark. To illustrate this point, our findings from Sweden show that the employment rate for highly educated women without an LLI is twice as high as the employment rate for low-educated women without an LLI. These results suggest that level of education is a stronger determinant for employment at older ages in Canada, Denmark and Sweden than in the UK. However, this is surprising given that compared to the other three countries, the UK has much more of a class structured society. Our previous studies [47, 48] have shown distinct class patterns in the UK in employment rates among men and women with LLI, compared to in Sweden. Although our present study shows lower overall employment rates in the UK than in Canada and Sweden, more research is needed to understand why there is no (or very little) social gradient among ageing workers in the UK.

One explanation for the gender gradient could be that women are more likely to exit paid employment for informal caring responsibilities [49]. Such responsibilities also depend on prevailing policies and arrangements not least concerning the care of older persons. But we expected to see less of a gender gradient in employment rates in Denmark and Sweden given that both countries have strong gender equity policies and provide health and social benefits to alleviate caring obligations that may prevent women from working longer than men. More research is needed to understand the gender gradient in the 65–69-year age group in these settings.

Our study supports previous research that suggests health plays an important role in continuing to work [50]. While we do not know the exact reasons why those older than 65 years of age are not working, lower employment rates among those with a LLI compared to without a LLI suggest that health could be one factor. However, in this study, we were unable to identify which types of health conditions causing a LLI have higher employment rates than others. Future studies should explore this issue to identify what type of work accommodations are needed for older workers to continue working. Certain work accommodations may be easier to achieve in jobs for high educated persons than for those with low education. Similarly, the needs of persons with physical disabilities may be easier to accommodate than the needs of those with mental disabilities.

Our study provides a comparative perspective on who continues to work longer. Our descriptive analysis illuminates that context matters, aligning with results from Larsen and Pedersen [12]. We propose further questions for consideration by policy-makers when designing policies to extend working life. For example, are ‘flexicurity’ policies–strong social security and weak employment protection–effective at older ages?

Employment rates for Canada were surprising. We expected that employment rates would be more comparable to the UK, another Liberal welfare regime but Canadian employment rates were more comparable to employment rates in Sweden, a Social Democratic welfare regime. A greater understanding of Canada’s approach to extending working lives is needed.

Despite many countries’ push to extend working lives, little research has explored who continues working. Our study contributes to the literature by showing that people are working past 65 years of age but employment rates seem to be influenced by sex, educational level and having a LLI. Better data and reporting on ageing workers is needed to gain insight into why these differentials might exist and policy measures to reduce them. Qualitative interview studies are warranted on how persons reason about their decisions to extend or not extend their working lives, to inform policy making and better guide further quantitative studies.

Only two previous studies [12, 13] focused on the role of health, education and gender on employment rates for those aged 65 years and older across several countries. The social differentials identified in this paper can serve as a starting point for policy-makers and researchers in future policy adaptions and research studies.

While SHARE and ELSA were previously harmonsized, to our knowledge, this is, the first study to include the CCHS to study older persons and employment rates comparatively.

This study is descriptive, so while we can identify differences, we are unable to determine their significance or level of interaction. As such we cannot conclude whether educational differences in employment rates remain after controlling for confounders. Further analytical studies are needed to assess the association between paid work and LLI. Future multivariate analysis should also control for occupation type and income. Also, the sample sizes of the included surveys vary. The number of observations in the SHARE sample used for Sweden and Denmark is much smaller than the other countries, and therefore the estimates may be unstable. A shortcoming of the study is the modest size of the SHARE surveys. The relative high nonresponse rates might introduce a bias because the distribution of educational level among the participants does not reflect that of the population [51]. Combined with differential health status this might imply that illness is more underreported among people with low education than among people with more education. However, if this bias is of similar magnitude in all countries, the cross-country comparison will still be valid. Another limitation is that the wording of survey questions about LLI were phrased differently.

Due to unreliable estimates, employment rates by age do not provide a complete picture. However, our results indicate that people are working past 65 years and that there are social differentials in their employment rates. Future studies need larger sample sizes of older ages collect better employment data for those aged 65 years and older. Also, our study only provides a ‘snapshot’ in time so does not account for ‘unretirement’–the process of retiring and then re-entering work [16]. Platts et al. [16] found that 1 in 4 Britons unretire. However, the authors found that unretirement was more likely among men, persons in good health and those with higher educational attainment, all of which align with our study’s findings.

The definition of work (> 1 h per week in the last week) means that we do not know the intensity of work in different groups or at different ages. It may be that the actual number of hours worked per week varies between groups in ways we have not been able to measure. Further, in the UK, the individuals were asked whether they worked ‘in paid employment’. There might be some misclassification among those who were sick-listed only last week.

Educational attainment is only one way to measure socioeconomic status. Using other indicators (e.g. income or occupation) could have produced different results [52]. We chose educational attainment because this variable was most comparable across datasets.