Continuity and change in the transition to retirement: how time allocation, leisure practices and lifestyles evolve when work vanishes in later life

Abstract

With increased longevity and socio-structural as well as socio-cultural changes, ageing research has shown a growing diversity of patterns in retirement lifestyles (Scherger et al. in Ageing Soc 31:146–172, 2011. 10.1017/s0144686x10000577). The transition from work to retirement is of particular interest to the study of the everyday lives, leisure activities and lifestyles of older adults, as questions on the meaning of work and leisure, activity and productivity are re-negotiated. This paper addresses the questions: how are the everyday lives of older adults re-organised when work vanishes? Are there lifestyles that are more easily maintained through retirement, whereas others are more prone to change? And which patterns of social inequalities underlie these processes? Drawing on data from the GTUS, this paper discusses similarities and differences in the time allocation of older working and non-working adults aged 55+ years (matched sample). Results show that the time spent on work is primarily taken up by household chores, media use and personal activities. Hierarchical cluster analysis identifies four activity clusters resp. lifestyles among the 55+: (1) a passive leisure lifestyle, (2) an active leisure lifestyle, (3) a paid work-centred lifestyle and (4) a housework-centred lifestyle. None of the clusters, however, comprised exclusively working or non-working older adults. The active leisure cluster comprised an equal share of working and non-working persons, suggesting that this kind of lifestyle allows for stronger continuity across work and retirement. It was more easily obtained by higher educated women who live separated from their partners.

Introduction

In her 1972 work ‘The Coming of Age’, Simone de Beauvoir described the ways people imagine retirement as ‘either as a prolonged holiday or as a rejection’ (p. 236). This is clearly meant to be a simplification, but the question remains how people live and experience retirement today. With increased longevity and socio-structural and socio-cultural changes, ageing research has shown a growing diversity of retirement lifestyles (Scherger et al. 2011). Transiting from work to retirement, older adults need to find new projects and roles, build new identities, and, not least, allocate up to 8 h a day from paid work to other activities. This paper addresses the question regarding what those ‘other’ activities are. How are the everyday lives of older adults re-structured and organised when work vanishes? Which practices change and which don’t? Are there lifestyles that are more easily maintained through retirement, whereas others are more prone to change? And which patterns of social inequalities underlie these processes?

In the first section, the literature on later life lifestyles and social inequalities is reviewed, asking what role the transition to retirement plays for lifestyle change and continuity, and how social inequalities shape this process. A practice-theoretical perspective is chosen to guide the empirical research. In the second section, data from the German Time Use Survey (GTUS) a matched sample of working and non-working older adults (55+ years) are compared in regard to their time allocation in everyday life. In the last section, results are discussed, indicating the potential of practice-theoretical retirement research and pointing to future fields of research.

Later life lifestyles and the transition to retirement

Ageing research has observed a growing diversification of later life lifestyles over the past decades (Scherger et al. 2011). This diversification may have a number of underlying causes—increasing longevity and the differentiation between a third age and a fourth age (Laslett 1989), the socio-structural change of ageing (Tews 1993) and socio-cultural changes in Western postmodern societies in general. In this paper, however, I will focus on developments that do not affect older age in general, but are directly related to the retirement transition itself.

Talking about transitions from one life stage to the other, Martin Kohli’s model of the institutionalised life-course, and its de-standardisation, come in handy (Kohli 1985, 2007). In societies decreasingly bound by social or regional origin, the institutionalised life-course and its underlying system of chrononormativity (Freeman 2010) help to provide age-based norms on how individuals should live their lives. Originally formulated in queer theory by Elizabeth Freeman (2010), the concept refers to ideas about the ‘right’ time for particular life stages, like age-defined periods of compulsory education, the right to vote or marry and the ‘right time’ to retire. Today, however, the institutionalised life-course has become increasingly differentiated, de-standardised and individualised (cf. Mayer 2009). Whereas differentiation refers to a growing plurality of legitimate lifestyles in later life, de-standardisation implies a less tight coupling of life-course transitions to chronological age (Kohli 2007). Accordingly, many gerontologists have argued that the once predictable pattern associated with retiring is becoming increasingly differentiated by the age at which it occurs, if it is gradual of abrupt, voluntary or involuntary, early or late, etc. (Moffatt and Heaven 2017). The transition from work into retirement then becomes a highly personalised project (Moen et al. 2001) in which adults enter into a ‘roleless role’ (Burgess 1960). In today’s activating welfare states, differentiation and de-standardisation may result in the individualisation of transitional risks (Lessenich 2008)—a development that affects older adults with low levels of qualifications and ill health in particular (Phillipson et al. 2016).

Whereas some scholars see a growing individualisation of possible retirement lifestyles, others analyse the societal norms and expectations about a ‘good retirement life’ that limit those possibilities. Critical gerontology emphasises the role of powerful discourses about active, successful and productive ageing that frame certain lifestyles as ideal ways of allocating non-working time in later life today (cf. Boudiny 2013; Katz 2000). These ideals question the transition from work to retirement as a stage without work altogether, since they demand not only high levels of activity, like physical exercise, frequent cognitive training, consumption, but also high levels of productivity, like volunteering or care work (cf. Fernández-Ballesteros 2011; Caprara et al. 2013). Analysing Eurobarometer data, Vidovicová (2018) shows that the majority sees the role of older adults first and foremost as caretakers within the family (82%). Those societal expectations, however, do not necessarily match with older adults’ own role identifications or everyday practice. Analysing the German Transitions and Old Age Potential Survey (TOP), Mergenthaler et al. (2018) find that 40.5% of German retirees between the ages of 60–70 years belonged to a group they call ‘family disengagers’. This group conducts hardly any caregiving within the family, but also hardly any volunteering or other activities that are framed as productive. Similar results could be found in the Americans’ Changing Lives survey (ACL; Burr et al. 2007).

Critical gerontology has long emphasised that active ageing lifestyles are highly exclusive, discriminating against people with lower income, lower education, health restrictions, and women (Marshall and Katz 2016). Analysing SHARE data for Europe, Hank and Erlinghagen (2008) have found significant correlations between available resources and engagement versus non-engagement in later life. Based on the English Longitudinal Study of Ageing, Jivraj et al. (2012) found health and income to be the central predictors of social participation in retirement. Various studies have also found that the level of education influences the likeliness to volunteer or engage in educational activities in later life (cf. Burr et al. 2007; Morrow-Howell et al. 2014, for Germany, see Mergenthaler et al. 2018).

Beyond class, the literature suggests gender differences regarding retirement lifestyles. Time use studies have shown that the transition to retirement is accompanied by an increase in time spent on reproductive work, for example, domestic and care work, volunteering and informal help for both genders; however, significant gender differences persist depending on partnership status (cf. Klumb and Baltes 1999; Künemund 2006). Older women do not only spend more time on unpaid labour, but also inside the house, whereas older men take over volunteering or gardening and hence spend more time outdoors (Wanka 2017). This raises the question—posed by many feminist scholars beforehand—regarding the nature of paid and unpaid, as well as productive and reproductive work.

How retirement lifestyles emerge is not least dependent on the pathways people take to retirement. The kind of pathway a person can take towards retirement (e.g. direct retirement from full employment vs. indirect retirement from unemployment) is strongly dependent upon socio-economic and socio-demographic variables like gender, health or education (cf. Fasang 2012; Loretto and Vickerstaff 2015). Less qualified employees with lower education and income more often take an early retirement route in Germany (Radl 2007). The regional labour market and welfare state regime also play a role (Fasang 2010).

Retirement research has often resorted to continuity theory when trying to explain change and continuity across the retirement transition. Advanced by Atchley (1989), this approach regards retiring ‘as an opportunity to maintain social contacts and lifestyle patterns’ (Quick and Moen 1998, p. 45). Handling change by establishing continuity is thereby seen to be associated with increased wellbeing (cf. Wang and Shi 2014). Continuity theory has been applied in various qualitative (Nuttman-Shwartz 2004) and quantitative studies (Quick and Moen 1998; Von Bonsdorff et al. 2009; Wang 2007; Wang et al. 2008). Most longitudinal investigations detect a continuity of everyday life with declining outdoor activities (cf. Scherger et al. 2011). In contrast, strong work engagement may entail discontinuity and stress in the transition to retirement, affecting psychological wellbeing. Continuity theories, however, tend to naturalise and homogenise older adults, suggesting that all of them aim for continuity and suffer from change.

This paper aims to contribute to the current state of research by following a more explorative approach to changing lifestyles throughout the retirement transition. I argue that practice theories (Schatzki 1996; Reckwitz 2002; Shove et al. 2012) offer a more explorative and less functionalist approach towards the question how different later life lifestyles come into being throughout the transition from work to retirement. From a practice-theoretical perspective, retiring can be framed as a process that is assembled by social practices. Social practices are embedded, decentralised, incorporated, subconscious and routinised ‘temporally and spatially dispersed nexus[es] of doings and sayings’ (Schatzki 1996: 89). Consequently, ‘old’ is not something that we are, but something that we do—that we perform in our everyday lives (cf. ‘doing difference’: West and Fenstermaker 1995; ‘doing age’: Schroeter 2012).

However, it is not a homogeneous set of practices—there are different ways to ‘do age’. In ‘La Distinction’ (1979), Bourdieu describes how people make use of seemingly mundane practices to perform membership and distance themselves from members of other social classes. From gerontological studies of distinction (e.g. Bourdieu 1979; Gilleard and Higgs 2000), we learn that both pre- and post-retirement lifestyles are highly socially stratified. Which lifestyles older adults are able to and find desirable to engage in, hence, differs between different social groups and the sorts of capital they have available. But it also differs by the competencies people have acquired throughout their lives that enable them to engage in different social practices and exclude them from others (cf. Shove et al. 2012).

Hence, practice theories allow for a non-deterministic, non-functionalist and non-positivist way of accounting for power, social inequalities and continuity. A practice-theoretical perspective on life-course transitions suggests to ‘zoom in’ on the transition process itself, taking a closer look at the exact period in which people stop working. It shifts our focus to questions like: How are the everyday lives of older adults re-organised when work vanishes? Are there lifestyles that are more easily maintained through retirement, whereas others are more prone to change? And which patterns of social inequalities underlie these processes?

Data and methods

Data

Data used in this paper stem from the German Time Use Survey (GTUS) conducted between 2012 and 2013 by the German Federal Office of Statistics. A quota sample was obtained based on federal state, household type and the social position of the main income earner of the household. To prevent seasonal bias, the survey was carried out over the course of 12 months. It was administered to 5000 households and comprises data from more than 11,000 individuals aged 10 years and older. The respondents were asked to fill out an individual and household pen-and-paper questionnaire, as well as to keep time diaries on three days (two weekdays and one weekend day). In these diaries, respondents filled in main and side activities in 10-min intervals, as well as where, with whom and for whom the activity was taking place. To account for disproportionally biased samples, three weights—for households, persons and diary entries—were constructed on the basis of the representative German micro-census from 2012 and by generalised regression estimates. For the description of the sample, person weights are used; for the analysis of the diary entries, the weight for diary entries is used.

Variables

The raw diary information was coded into more than 230 categories according to the 2008 guidelines of the ‘Harmonised European Time Use Surveys’. These were then summarised into nine activity indices: (1) personal activities, including sleeping, eating and personal hygiene, (2) paid work, (3) education and training, (4) household work and family care, including childcare, cooking, gardening and shopping, (5) volunteer work, (6) social activities, including phone calls, cultural activities, and entertaining visitors, (7) leisure activities and hobbies, including sports, creative practices, and playing games, (8) media usage, including reading, watching television, or surfing the internet and (9) travel mentioned for the activities one to eight. The unit of analysis of the nine indices was recoded into hours, so a value of 1.5 for media usage implies that a person had spent 1 h and 30 min a day on average using different media.

Sample

37.2% of the survey population were 55 years and older and hence included in the data analysis. Sample characteristics are displayed in Table 1.

Table 1.

Socio-demographic characteristics, non-matched and matched sample 55+ years, German Time Use Data 2012/2013

	Working (all 55+)	Not working (all 55+)	Working (matched 55+)	Not working (matched 55+)
Average agea (years)	59.16	69.79	60.36	60.46
SD (years)	3.48	6.75	3.89	4.10
Gender (%)
Male	58.2	43.8	48.2	48.8
Female	41.8	56.2	51.8	51.2
Marital status (%)
Partnership	70.9	62.0	69.7	71.3
Widowed	3.3	16.4	3.9	3.5
Separated	17.6	15.6	19.4	18.2
Single	8.1	6.1	7.0	7.0
Education (%)
ISCED 0–2	3.7	11.7	4.9	5.5
ISCED 3–4	45.6	53.9	52.6	51.0
ISCED 5+	50.7	34.4	42.4	43.5
Occupational status (%)
Self-employed	19.7		23.8
Civil servant	14.3		12.8
White collar	48.4		46.7
Blue collar	17.6		16.7
Unemployed		4.5		19.5
Retired		82.7		40.6
Otherb		12.9		39.9
n (unweighted)	2.607	6.228	457	462

^aThe age distribution ranges from 55 years to 85+ years among the non-working group of the whole sample and from 55 years to 84 years among the working group of the whole sample and both groups of the matched sample

^bThe category ‘other’ comprises other non-working groups, particularly persons who are permanently unfit to work/disabled and domestic workers. In the unmatched sample, disabled persons make up 21.1% and domestic workers make up 71.5% of the category ‘other’ (7.4% are not working for yet other reasons). In the matched sample, 32.9% in the category ‘other’ are permanently unable to work/disabled and 61.3% are domestic workers (5.8% not working for other reasons)

Statistical procedure

This study poses the question what changes in older people’s everyday lives when work vanishes. It focuses on the transition period itself, hence the stage of life in which people usually retire, and thus early old age. The research question would ideally call for a longitudinal survey design which the German Time Use Survey does not offer. However, it allows for a comparison of working and non-working older adults. Yet, simply comparing working and non-working respondents would not only ignore questions of causality; it would also compare two very different social groups. Most obviously, non-working and particularly retired persons are on average older than the working population, and with this, less educated and more often female (Table 1). To account for both the lack of longitudinal data and the assumed bias, a matched sample of working and non-working persons aged 55 years and older was constructed by case–control matching.

Case–control matching is a statistical procedure used to construct quasi-experimental designs and hence interpret relations as quasi-causal. Matching confounding variables accounts for pre-existing differences, reduces selection bias and improves internal validity. Matching is generally used to account for selectivity in certain events and to answer questions about the conditionality of these events taking place. Retirement, however, affects a major part of the population, and this paper is not asking about the factors that lead to its occurrence. It does ask about the re-allocation of time, and hence, matching is used to construct quasi-causality. Constructing a matched sample of working and not-working adults aged 55 years and older allows for approaching the question which changes in everyday life practices might occur if the respondents who were still working by the time the survey had been taken would retire. Hence, matching is used to allow for a quasi-longitudinal analysis despite being limited to cross-sectional data (cf. Lewis and Michel 1990).

To obtain a matched sample, employment status was dummy-coded into ‘working’ (1, including all types of paid work) and ‘not working’ (0, including retired, unemployed and other non-working persons1). Age, gender, marital status, and education were defined as matching criteria.2 Matching criteria should consider variables that are, based on the literature (see above), most likely to influence the outcome in question; in this case, retirement lifestyles. The more matching criteria are used, the more accurately do the sub-samples reflect each other. However, the more matching criteria are defined, the more the sample size is being reduced, which requires a considerate weighting of number of matching criteria and necessary sample size. With the four matching criteria defined above, the matching procedure resulted in a sample of 919 persons aged 55 years and older, 457 of them working (66.8% full-time, 33.2% part-time) and 462 not working. Table 1 summarises the matched sample’s key characteristics in comparison with the overall sample of adults aged 55 years and older.

How are the everyday lives of older adults re-organised when work vanishes? Are there lifestyles that are more easily maintained through retirement, whereas others are more prone to change? And which patterns of social inequalities underlie these processes?

To transform the results from single activities to activity clusters or lifestyles—and to hence, investigate further the interplay between different activities, a hierarchical cluster analysis based on Ward’s method was conducted with the nine activity indices. Ward’s method employs the sum of the squared Euclidean distance as an indicator for heterogeneity and maximises the variation between the cluster centres (Schendera 2010). Based on the elbow criterion (Backhaus et al. 2003), a cluster solution of four was suggested, which was validated by discriminant analysis and logical consistency. Analyses of variance were conducted to determine statistical significance regarding differences in activity indices (between groups or clusters). Similar to the other comparable analyses, we defined the level of statistical significance as α = 5%.

To investigate which socio-demographic variables determine cluster membership, and hence answer the third question, ‘Which social inequalities underlie changes in time allocation in the retirement transition?’, a multinomial logistic regression was calculated, with cluster membership as dependent variable (and Cluster 2 as reference category), and the socio-demographic variables age, sex, occupational status, education and marital status as independent variables. This method is often applied to determine relevant characteristics for group membership in three or more mutually exclusive groups (Mergenthaler et al. 2018). The independent variables were entered into the regression models simultaneously. Three types of measures were used to assess the model fit: − 2 log likelihood, Chi² statistics and Nagelkerke’s pseudo-R². All statistical analyses were carried out using SPSS 22.

Results

Time allocation among working and non-working 55+

Respondents included in the matched sample spent on average 10.76 h a day on personal activities (sleeping, eating/drinking, personal hygiene), 3.3 h on household chores and family care, 3.29 h for media use, 3.02 h for paid work, 1.24 h for social activities, 1.22 h for travelling, 0.74 h for leisure activities, 0.39 h for volunteering and informal work and 0.03 h for education.

Significant differences exist between working and not-working respondents of the matched sample for all activity indices (p < 0.001). Time spent working among the working group (4.8 h) is re-allocated to household chores and family care (+ 1.8 h among not-working older adults), media use (+ 1.3 h) and personal activities (+ 1.1 h; Fig. 1).

Fig. 1.

Time allocation of working and non-working persons aged 55+ years by activity indices in hours (x-axis); German Time Use Survey 2012/2013

Deploying ANOVAs with Bonferroni post hoc tests shows significant differences in hours spent for different activities between all categories of working and non-working older adults (p < 0.001; see Appendix Table 5). This shows already that we need not only to differentiate between working and non-working persons, but also take different work conditions into account when we talk about the retirement transition.

Table 5.

Activity indices in hours by occupational status with Bonferroni-adjusted post hoc tests; matched sample 55+ years; German Time Use Survey 2012/2013

	N	Mean	SD	Minimum	Maximum	Sig. mean difference (< 0.001)
Personal activities/regeneration
Self-employed (A)	92,297,436	10.72	1.83	4.67	16.83	B C D E F G
Civil servant (B)	49,478,538	10.54	1.92	6.67	14.67	A C D E F G
White collar (C)	180,851,537	10.09	1.73	4.50	14.83	A B D E F G
Blue collar (D)	64,455,931	9.99	1.43	7.00	13.17	A B C E F G
Retired (E)	135,585,144	11.25	1.73	4.83	16.00	A B C D F G
Unemployed (F)	65,392,905	10.96	1.98	4.83	15.33	A B C D E G
Othera (G)	132,333,652	11.65	2.35	3.00	22.67	A B C D E F
Total	720,395,143	10.77	1.98	3.00	22.67
Paid work
Self-employed (A)	92,297,436	5.80	3.62	0.00	14.00	B C D E F G
Civil servant (B)	49,478,538	4.27	4.28	0.00	13.00	A C D E F G
White collar (C)	180,851,537	4.91	3.87	0.00	18.83	A B D E F G
Blue collar (D)	64,455,931	5.82	3.93	0.00	12.50	A B C E F G
Retired (E)	135,585,144	0.55	1.80	0.00	9.67	A B C D F G
Unemployed (F)	65,392,905	0.79	2.01	0.00	12.00	A B C D E G
Othera (G)	132,333,652	0.15	0.71	0.00	5.00	A B C D E F
Total	720,395,143	2.99	3.87	0.00	18.83
Education
Self-employed (A)	92,297,436	0.02	0.13	0.00	1.17	B C D E F G
Civil servant (B)	49,478,538	0.02	0.13	0.00	0.83	A C D E F G
White collar (C)	180,851,537	0.02	0.38	0.00	6.50	A B D E F G
Blue collar (D)	64,455,931	0.02	0.12	0.00	0.83	A B C E F G
Retired (E)	135,585,144	0.04	0.26	0.00	3.00	A B C D F G
Unemployed (F)	65,392,905	0.04	0.21	0.00	1.50	A B C D E G
Othera (G)	132,333,652	0.04	0.26	0.00	2.67	A B C D E F
Total	720,395,143	0.03	0.26	0.00	6.50
Housework/care
Self-employed (A)	92,297,436	1.98	1.89	0.00	8.17	B C D E F G
Civil servant (B)	49,478,538	2.41	1.77	0.00	6.67	A C D E F G
White collar (C)	180,851,537	2.70	2.14	0.00	10.50	A B D E F G
Blue collar (D)	64,455,931	2.53	2.38	0.00	9.00	A B C E F G
Retired (E)	135,585,144	3.78	2.15	0.00	10.50	A B C D F G
Unemployed (F)	65,392,905	4.04	2.11	0.00	9.00	A B C D E G
Othera (G)	132,333,652	4.87	2.44	0.00	21.00	A B C D E F
Total	720,395,143	3.30	2.38	0.00	21.00
Volunteer work
Self-employed (A)	92,297,436	0.29	0.91	0.00	6.00	B C D E F G
Civil servant (B)	49,478,538	0.36	0.85	0.00	3.50	A C D E F G
White collar (C)	180,851,537	0.42	1.19	0.00	8.33	A B D E F G
Blue collar (D)	64,455,931	0.12	0.58	0.00	3.83	A B C E F G
Retired (E)	135,585,144	0.61	1.62	0.00	10.33	A B C D F G
Unemployed (F)	65,392,905	0.58	1.45	0.00	8.83	A B C D E G
Othera (G)	132,333,652	0.25	0.70	0.00	4.17	A B C D E F
Total	720,395,143	0.39	1.16	0.00	10.33
Social activities
Self-employed (A)	92,297,436	0.99	1.54	0.00	8.50	B C D E F G
Civil servant (B)	49,478,538	1.22	1.56	0.00	8.50	A C D E F G
White collar (C)	180,851,537	1.17	1.31	0.00	6.50	A B D E F G
Blue collar (D)	64,455,931	0.96	1.31	0.00	7.00	A B C E F G
Retired (E)	135,585,144	1.39	1.53	0.00	8.50	A B C D F G
Unemployed (F)	65,392,905	1.34	1.27	0.00	4.83	A B C D E G
Othera (G)	132,333,652	1.49	1.49	0.00	8.50	A B C D E F
Total	720,395,143	1.25	1.44	0.00	8.50
Leisure activities
Self-employed (A)	92,297,436	0.65	1.38	0.00	8.50	B C D E F G
Civil servant (B)	49,478,538	0.71	1.10	0.00	4.00	A C D E F G
White collar (C)	180,851,537	0.56	1.05	0.00	5.17	A B D E F G
Blue collar (D)	64,455,931	0.39	0.77	0.00	3.17	A B C E F G
Retired (E)	135,585,144	1.16	1.39	0.00	8.33	A B C D F G
Unemployed (F)	65,392,905	0.65	1.00	0.00	4.83	A B C D E G
Othera (G)	132,333,652	0.87	1.44	0.00	10.00	A B C D E F
Total	720,395,143	0.74	1.24	0.00	10.00
Media use
Self-employed (A)	92,297,436	2.42	1.83	0.00	8.50	B C D E F G
Civil servant (B)	49,478,538	2.80	1.96	0.00	8.67	A C D E F G
White collar (C)	180,851,537	2.83	1.68	0.00	8.83	A B D E F G
Blue collar (D)	64,455,931	2.77	2.04	0.00	9.50	A B C E F G
Retired (E)	135,585,144	4.00	2.30	0.00	10.00	A B C D F G
Unemployed (F)	65,392,905	4.56	2.31	0.00	11.67	A B C D E G
Othera (G)	132,333,652	3.68	2.28	0.00	9.50	A B C D E F
Total	720,395,143	3.30	2.16	0.00	11.67
Travel
Self-employed (A)	92,297,436	1.13	1.06	0.00	8.17	B C D E F G
Civil servant (B)	49,478,538	1.67	1.69	0.00	9.00	A C D E F G
White collar (C)	180,851,537	1.30	1.09	0.00	7.50	A B D E F G
Blue collar (D)	64,455,931	1.40	1.31	0.00	7.83	A B C E F G
Retired (E)	135,585,144	1.22	1.73	0.00	13.33	A B C D F G
Unemployed (F)	65,392,905	1.03	1.77	0.00	13.33	A B C D E G
Othera (G)	132,333,652	1.02	1.19	0.00	7.83	A B C D E F
Total	720,395,143	1.22	1.39	0.00	13.33

^aThe category “other” comprises other non-working groups, particularly persons who are permanently unfit to work / disabled and domestic workers. In the unmatched sample, disabled persons make up 21.1% and domestic workers make up 71.5% of the category “other” (7.4% are not working for yet other reasons). In the matched sample, 32.9% in the category “other” are permanently unable to work / disabled and 61.3% are domestic workers (5.8% not working for other reasons)

Comparing different sub-activities within the indices, two bundles of practices stick out in particular: food-related practices and practices of media usage. Non-working persons spend on average 1 h a day more on eating and related practices (0.4 h more shopping, 0.4 h more preparing food and 0.2 h more eating and drinking). Also, non-working persons spend about 48 min a day more watching television than working persons, with the unemployed watching significantly more television (3.1 h on average a day) than retirees (2.5). It is, however, not just ‘old’ media that retirees use—they also spend more time using the computer or smartphone for non-working purposes (0.5) than all working groups (0.3). Conversely, blue-collar workers (0.4) and unemployed persons (0.5) spend the least amount of time reading, while retirees spend the most time on it (0.9).

To transform the results from single activities to activity clusters or lifestyles—and to hence, investigate further the interplay between different activities—a hierarchical cluster analysis was conducted with the nine activity indices (see above). A cluster solution of four was suggested: Cluster 1 comprises 37.5%, Cluster 2 18.1%, Cluster 3 34.5%, and Cluster 4 9.9% of the sample. Clusters differ significantly in regard to all activity scales (p < 0.001). Work takes up the least time in Cluster 1 (0.015 h a day on average) and most time in Cluster 3 (6.3). Activities in Cluster 1 are centred on personal activities, media use and social activities. Cluster 2 is characterised by social activities that take place at different locations—as the time spent for travel suggests—little media use, little housework and little paid work. Cluster 4, finally, is structured by housework and some paid work (Fig. 2). As a consequence of these results, Cluster 1 will be referred to as ‘passive leisure lifestyle’, Cluster 2 as ‘active leisure lifestyle’, Cluster 3 as ‘paid work-centred lifestyle’, and Cluster 4 as ‘housework-centred lifestyle’.

Fig. 2.

Lifestyle clusters of working and not-working persons aged 55+ years by activity indices in hours (x-axis); German Time Use Survey 2012/2013

Social inequalities and time allocation in later life

Do the identified clusters differ by socio-demographic variables? Analysis shows that socio-demographic variables structure cluster membership to a different extent: first, there are rather clear-cut differences between people who are working (among whom the paid work-centred lifestyle is most dominant) and people who are not working (among whom the passive leisure lifestyle is most dominant). Cluster 3 is mainly characterised by economically active older adults (46%), whereas non-working older adults fall predominantly into Cluster 1 (53%). These clusters can thus be assumed to be more intensely structured by work—or its absence. Other than in regard to the occupational status, persons that constitute these two clusters share similar characteristics (more men, higher educated persons, similar marital situations). Clusters 2 and 4, however, combine a nearly equal share of working and non-working persons. Both are constituted by more women than men; however, they differ in regard to education and marital status. The housework-centred lifestyle engages more persons with a lower educational attainment and persons who live in a partnership than the other clusters (Table 2).

Table 2.

Lifestyle clusters by socio-demographic characteristics, matched sample 55+ years, German Time Use Data 2012/2013

	Passive leisure (Cluster 1)	Active leisure (Cluster 2)	Paid work-centred (Cluster 3)	Housework-centred (Cluster 4)
Average age (years)	60.62	61.49	59.64	60.14
SD (years)	3.79	5.27	3.42	3.26
Gender (%)
Male	53.1	43.6	51.0	29.9
Female	46.9	56.4	49.0	70.1
Marital status (%)
Partnership	68.8	73.0	68.1	79.2
Widowed	4.1	3.5	3.5	3.2
Separated	19.1	20.8	20.5	9.8
Single	8.0	2.7	8.0	7.8
Education (%)
ISCED 0–2	5.6	4.2	3.6	10.9
ISCED 3–4	50.2	46.8	53.5	60.2
ISCED 5+	44.1	48.9	42.8	28.8
Occupational status (%)
Self-employed	7.4	18.0	16.4	11.9
Civil servant	7.4	5.3	7.9	4.1
White collar	16.4	22.4	32.9	35.9
Blue collar	3.1	8.1	14.4	14.0
Unemployed	12.9	6.2	5.7	11.4
Retired	27.1	22.5	8.4	16.6
Othera	25.7	17.6	14.3	6.1
n (unweighted)	330	174	308	105

^aThe category ‘other’ comprises other non-working groups, particularly persons who are permanently unfit to work/disabled and domestic workers. Disabled persons make up 9.7% of Cluster 1, 6.1% of Cluster 2, 3.3% of Cluster 3, and 0% of Cluster 4. Domestic workers make up 14% of Cluster 1, 9.8% of Cluster 2, 10.8% of Cluster 3, and 6.1% of Cluster 4

A multinomial logistic regression model using age, sex, occupational status, education and marital status as predictors for cluster membership confirms the results of the bivariate analyses (Table 3). The explanatory power of the model is highly significant with the predictor variables explaining 22.7% of variance (Nagelkerke’s R²; p < 0.001). Non-working persons are, obviously, significantly less likely to engage in paid work-centred lifestyles than white-collar workers, but civil servants and self-employed persons are nearly twice as likely to engage in passive (civil servants) or active (self-employed) lifestyles than white-collar workers. Persons falling into the ‘other non-working’ category (of whom a majority are domestic workers) are less likely to lead a housework-centred lifestyle. This result suggests that the latter lifestyle is not dominated by unemployed housekeepers, but rather women working part-time and engaging in housework beyond that. Women are significantly more likely to engage in either an active leisure or housework-centred lifestyle than men, who are more likely to have a work-centred lifestyle. Persons with lower education are more likely to conduct passive leisure activities or household chores and the higher educated are more likely to lead an active leisure life. Persons without a partnership (widowed, single or separated/divorced) are less likely to engage in active leisure or housework-centred lifestyles than married persons.

Table 3.

Multinomial logistic regression, reference Cluster 3—paid work-centred lifestyle, Exp(B), n = 465 (unweighted), significance levels: *p < 0.10; **p < 0.05; *** p < 0.001 (two-sided)

Independent variables	Passive leisure (Cluster 1)	Active leisure (Cluster 2)	Housework-centred (Cluster 4)
Age	1.08***	1.12***	1.06***
Sex: male (female = ref)	1.07***	0.50***	0.31***
Occupational status
(i) Self-employed	0.77***	1.53***	1.02***
(ii) Civil servant	1.94***	1.02***	0.67***
(iii) Blue collar	0.41***	1.01***	0.96***
(iv) Retiree	5.20***	3.54***	2.59***
(v) Unemployed	4.93***	2.34***	3.30***
(vi) Other (white collar = ref)	4.00***	1.81***	0.30***
Education
(i) ISCED 0–2	1.99***	0.88***	3.18***
(ii) ISCED 3–4 (ISCED 5–6 = ref)	1.16***	0.86***	1.59***
Marital status
(i) Single	1.08***	0.33***	0.77***
(ii) Separated	1.05***	0.96***	0.31***
(iii) Widowed (married = ref)	1.24***	0.69***	0.52***

Discussion

Whereas much research is concerned with describing and categorising retirement lifestyles, this paper deploys a practice-theoretical life-course perspective on retirement lifestyles, asking how such lifestyles come into being throughout the transition from work to retirement.

What changes occur in the everyday lives of older people when work vanishes? Results from a matched sample of GTUS data show that the average time adults aged 55 years and older spend working is mainly taken up by household chores and family care (food-related practices in particular), media usage and personal activities when they stop working. These results are similar for persons who become unemployed, retire or do not work for other reasons. Despite an increase in research on ageing and new technologies (cf. Marshall and Katz 2016), only few studies have focused on the role of ‘old’ or mundane media, like television or books, in the lives of older adults (Peine and Neven 2018). The same holds true for the study of eating, which is quite an elaborated field of research in sociology (cf. Warde 2016), but has not yet spread to ageing studies.

Beyond this re-allocation of time to single activities, four lifestyle clusters of persons aged 55+ could be identified: a paid work-centred, a housework-centred, an active leisure and a passive leisure lifestyles. Similar activity clusters can be found in the literature, for example, for Germany (Mergenthaler et al. 2018) or France (Guillemard 1972). However, most of those studies are concerned with lifestyles of retired persons only, whereas this paper is concerned with the linkages between pre- and post-retirement lifestyles. Which lifestyles are more likely maintained across work and retirement? First of all, none of the identified clusters are exclusively performed by either working or non-working older adults—there are, hence, no lifestyles exclusively performed by retirees. Applying the assumption of a quasi-longitudinal data set-up, two types of retirement trajectories can be identified: First, some experience a rapturous change in lifestyles throughout the retiring process, from a lifestyle centred on paid work (Cluster 3) to a more passive leisure lifestyle (Cluster 1). Paid work-centred lifestyles can hardly be maintained in retirement, and more passive leisure lifestyles seem to emerge only after people stop working. Second, others experience relative continuity between working stage and retirement stage lifestyles (Cluster 2, 4), supporting the assumption that these kinds of lifestyle are more easily maintained across work and retirement.

Multinomial logistic regression showed that social inequalities affect the likeliness to engage in one lifestyle over the other and that (former) occupational status plays a significant role in this. Self-employed persons are more likely to engage in active leisure lifestyles throughout their working lives already, and it hence might be easier for them to carry these lifestyles to retirement. Gender, education, and marital status also play a role: while older men might transition from strongly work-centred lifestyles to a more passive retirement stage and older, less educated and married women transition from (part-time) work to a retirement filled with household chores, a certain group of highly educated, independent women might be able to enjoy the promises of a late freedom (Rosenmayr 1983).

Now what are the new insights a practice-theoretical perspective brings to retirement research in particular? With social practices being ‘temporally and spatially dispersed nexus[es] of doings and sayings’ (Schatzki 1996: 89), the results must be discussed against the backdrop of their time–space structuration and, consequently, people’s abilities to allocate time and appropriate place (time–space competencies; cf. May and Thrift 2003). Active leisure lifestyles are constituted by practices that are very varied in terms of where, when and with whom they take place (this can also be seen by the relatively high share of time spent on travel within these lifestyles). (Formerly) Self-employed persons might be more likely to acquire these competencies throughout their working lives than blue- and white-collar workers, whose days are structured by a working time and working place that they didn’t chose. Self-employed persons often have to ‘make spaces into places’ (Rowles and Bernard 2013): They need to turn a room in the home, the train or a café into a workplace, and they need to establish their own working schedule. Also women—who more likely have to manage paid work, care work and domestic work at once, and allocate their daily routines accordingly—might be more time–space competent. As work autonomy is higher in higher occupational positions, these competencies are also related to social class (captured here through educational attainment).

A practice-theoretical perspective on emerging lifestyles in retirement, hence, emphasises the importance of time–space competence. Who has not acquired this competence by the time of retirement, or unlearnt it throughout their lives, needs to be provided with possibilities to (re-)acquire it. Practical implications arising from a practice-theoretical perspective target both employers and non-profit organisations and call for a process-oriented transition management. Such a kind of transition management may foresee a time period of ‘fading out’ (e.g., reducing hours, working from home) in which autonomous allocation of time and appropriation of places can be learnt, and/or include the establishment of retirement counselling groups as ‘communities of practice’ (Lave and Wenger 1991) who transition together.

However, the results discussed in this paper face methodological and data-based limitations. First, as the sample description elucidates, matching leads to a bias in both working and not-working persons’ socio-demographic characteristics as compared to the overall sample. This bias is deliberate in regard to the defined matching criteria: The age distribution between working and not-working persons in the matched sample is similar, hence excluding all (older) persons who are not in the retirement transition period; this also affects the number of widowed persons. The distribution of gender and educational attainment is more equal than in the overall sample, resulting in a higher percentage of women and a lower percentage of persons with high educational attainment in the ‘working’ group and in a higher percentage of men and a lower percentage of persons with low educational attainment in the ‘non-working’ group. Whereas the different occupational groups among working persons are hardly affected by the matching process, matching does lead to a lower percentage of retired persons in favour of unemployed persons, disabled persons and homemakers, which is both due to the matched non-working sample being younger and more female. Within the matched sample, significant differences exist between the different occupational groups in regard to age, gender, marital status and education (Appendix, Table 4; p < 0.001). This bias must be considered when interpreting the results.

Table 4.

Occupational groups by socio-demographic characteristics, matched sample 55+ years, German Time Use Data 2012/2013

	Retired (matched 55 +)	Unemployed (matched 55 +)	Domestic work (matched 55 +)	Disabled (matched 55 +)	Other non-working (matched 55 +)
Average age (years)	63.35	58.46	58.35	58.64	59.15
Gender (%)
Male	72.5	60.4	5.0	38.0	60.7
Female	17.5	39.6	95.0	62.0	39.3
Marital status (%)
Partnership	73.0	42.4	93.8	62.9	95.8
Widowed	6.1	1.0	2.9	0.0	4.2
Separated	14.3	40.9	3.3	27.6	0.0
Single	6.6	15.7	0.0	9.5	0.0
Education (%)
ISCED 0–2	3.8	2.8	9.8	3.6	24.9
ISCED 3–4	40.3	60.5	61.7	51.7	38.6
ISCED 5+	55.9	36.7	28.5	44.6	36.5
n (unweighted)	203	96	104	49	10

Second, time use data themselves have limitations. GTUS does not provide any information about health and physical functioning of the respondents. Even though the focus of this study is on early old age, health nevertheless plays a major role in people’s ability to participate in certain practices and lifestyles (cf. Scherger et al. 2011).Moreover, time use data are not longitudinal. Despite matching, data do neither allow to ‘look back’ nor to ‘look ahead’: it does not provide any information about former occupations of the group of retirees, and it does not give any clues about whether people change their lifestyles later in life. As Boudiny (2013) suggests, later lifestyles are dynamic and might change after the ‘honeymoon stage’ of retirement that this paper focuses on. Constructing not only quasi-causality through matching samples, but actually tracing practice change among the same persons over time would bring tremendous opportunities for lifestyle research in gerontology. This is why the project ‘Doing Retiring’3 that the author is involved in combines time use data with a longitudinal qualitative study that follows 30 older adults in Germany from before to 3 years after retirement.

Nonetheless, the simple exploratory analysis of German Time Use data as displayed in this paper already offers multiple points of connection for future research: first, the role of mundane practices, like the use of ‘old’ media and eating, and how they change across the life-course, is a field of research age studies are still developing. Second, with increasing flexibilisation of workplaces, the boundaries between work and retirement are increasingly blurring, which might make it easier to carry specific lifestyles into retirement. Focusing on subgroups with highly flexible working conditions, like the self-employed, artists, or academics, can provide innovative insights into the emergence of new retirement lifestyles. Third, gender plays a continuously significant role in the configuration of everyday lives in and throughout retirement, with new living arrangements creating spaces for a diversification of older women’s lifestyles. This calls for research focusing specifically on older single, separated or divorced women.

Appendix

See Tables 4 and 5.

Funding

This author is part of the Research Training Group ‘Doing Transitions’ funded by the Deutsche Forschungsgemeinschaft (DFG). The data file used in this paper is a scientific use file provided by the German Federal Office of Statistics for scientific purposes only.

Compliance with ethical standards

Conflict of interest

The author declares that there are no competing financial interests.

Ethical approval

An ethical approval was not required.