Job autonomy and leisure-time physical activity across Europe: a multilevel perspective

Abstract

Prior research has widely recognised the relevance of job autonomy for workers’ leisure-time physical activity. This paper is, however, the first to assess this relationship with a cross-national focus. Apart from investigating the relationship between individual job autonomy and leisure-time physical activity within the European context, we more specifically study how this relationship is moderated by a country’s labour market security and country-level perceived availability of physical activity opportunities. We answer our research question with multilevel regression analyses on comparative cross-national data from the European Social Survey 2014 on 14,956 respondents in 18 European countries. Our multilevel models show that across Europe, workers with more job autonomy are more physically active in their leisure time. Perceived PA opportunities in a country do not affect the relationship, but in countries with greater labour market security, the positive influence of job autonomy on leisure-time physical activity is reduced–workers with low job autonomy seem to benefit relatively more from the labour market security in their country.

Keywords:

• Job control
• sports participation
• physical activity facilities
• labour market security
• country comparison

Introduction

It is widely recognised that leisure-time physical activity (PA) promotes people’s physical and mental health and advances social connectedness among a country’s population (Coenders et al., 2017; Dishman et al., 2021; Fialová, 2004; Holtermann et al., 2012; Penedo & Dahn, 2005; Wankel & Berger, 1990). In Europe, however, still 55% of all workers indicate to seldom or never exercise or play sports¹ (European Commission, 2018). Accordingly, the most prevailing public health issue in developed countries is the limited involvement in leisure-time PA (Cavill et al., 2006; Van Tuyckom & Scheerder, 2010).

People’s PA behaviour is often theoretically and empirically related to their day-to-day work conditions by scholars and policymakers (e.g. Biswas et al., 2020; Kirk & Rhodes, 2011; Van der Put et al., 2020). This connection between PA and employment is generally approached from the work/non-work interface perspective (Geurts & Demerouti, 2003), which advocates that circumstances at the workplace not only affect a person’s activities at work, but also those in a person’s leisure time, among which people’s PA behaviour. Building upon this reasoning, we here focus on a key psychosocial feature of people’s job, namely job autonomy. The autonomy that workers experience depends on the quantity and quality of job resources that enable a worker to effectively manage work activities, non-work activities and personal development (Bakker & Demerouti, 2007; Grönlund, 2007). In this study, we expect job autonomy to positively affect workers’ leisure-time PA by creating increased intention and possibility for PA.

The importance of job autonomy for workers’ well-being and satisfaction has been supported for decades (Bakker & Demerouti, 2007; Breaugh, 1985; Karasek, 1979). Prior single-country research has consistently shown that job autonomy enhances people’s PA in leisure time (i.e., Biswas et al., 2020; Choi et al., 2010; Hellerstedt & Jeffery, 1997; Kouvonen et al., 2005). Within the European context, Fransson et al. (2012) and Heikkilä et al. (2013) showed with pooled results from cohort studies that, overall, low job autonomy is associated with physically inactive behaviour and high job autonomy is associated with more physically active behaviour. In their studies, a direct comparison of differential country effects was, however, impossible because the measures of job autonomy and PA varied across the countries under study. In this research, we therefore intend to bring about a more in-depth understanding of the relationship between job autonomy and leisure-time PA in the European context with high-quality comparative data from the European Social Survey round seven (ESS7).

In addition, our approach allows us to assess to what extent the relationship between job autonomy and leisure-time PA depends on a country’s features. Various authors have voiced that contextual features affect the work-health relationship (Bambra et al., 2005; Grönlund, 2007; Pisljar et al., 2011). This notion is in line with the elementary sociological idea that context affects individuals’ behaviour (Coleman, 1994; Inglehart, 1997). The context of a person can be the family or neighbourhood, the organisation one works in, or the country of living (Burgers et al., 2021; Gehrmann & Wicker, 2022; Studer et al., 2011). The latter is the focus of this paper.

In our attempt to explain country-level variation in the relationship between job autonomy and leisure-time PA, we focus on two relevant country-level characteristics, namely the labour market security in a country and a country’s average perceived PA-friendliness of the direct living environment. There are good reasons to expect that both features affect the relationship between job autonomy and leisure-time PA. For instance, working in a country in which workers are concerned about the labour security of their job may give rise to elevated feelings of powerlessness and stress (Anderson & Pontusson, 2007; Sverke & Hellgren, 2002). Moreover, a favourable environment for PA may facilitate workers to be physically active (Bauman et al., 2012) and especially workers with limited job autonomy may be supported by a favourable environment for PA. Overall, our study is guided by the following two research questions: To what extent is, within the European context, a worker’s job autonomy associated with leisure-time PA? And: To what extent do a country’s labour market security and perceived PA opportunities moderate the relationship between job autonomy and leisure-time PA?

We utilise unique high-quality cross-national data from the ESS7 on 14,956 workers in 18 European countries. The ESS7 data encompass comparable measurements of leisure-time PA behaviour, work conditions and relevant individual background characteristics. There are previous studies that also applied a multilevel or cross-national perspective to explain leisure-time PA in Europe (see, e.g. Van Tuyckom, 2011). Our contribution is unique because it assesses the relationship between job autonomy and leisure-time PA from such a multilevel perspective. To do so, we use a design with cross-level interactions to study variation in the job autonomy-PA association between countries. Consequently, our contribution adds to the knowledge of country-level features that affect the workforce’s daily work–non-work interface.

Theoretical background

Job autonomy and leisure-time physical activity

To begin, workers’ ability to decide how work is organised (i.e. job autonomy) may have positive effects on the intention to be active in their leisure time. To gain a better understanding of the positive association between job autonomy and PA, Häusser and Mojzisch (2017), in their physical activity-mediated demand-control model, combine a spill-over perspective with self-determination theory (Deci & Ryan, 2000). In this model, workers’ autonomy over work-related tasks is mirrored in subjective feelings of autonomy and decisional freedom in their leisure time. As such, feelings of autonomy are not restricted to the work domain but would trigger a general state of perceived self-determination (Abdel Hadi et al., 2021; Häusser & Mojzisch, 2017). Perceived self-determination is often associated with increased motivation, social development and personal well-being (Deci & Ryan, 2000). Moreover, elevated levels of perceived self-determination associate with higher levels of subjective vitality and active behaviour (Ryan & Frederick, 1997; Teixeira et al., 2012). To summarise, workers with more job autonomy are assumed to have the intention more often to be physically active in their leisure time.

Besides advancing feelings of self-determination, having job autonomy may also contribute to having the possibilities to engage in PA during leisure time, because key features of work autonomy are related to abilities for individual time management and scheduling (Bakker & Demerouti, 2007; Breaugh, 1985; Karasek, 1979). Workers with job autonomy have, for instance, more possibilities to work from home, to determine breaks, and to decide on work hours and order of work tasks. These job characteristics of workers with high job autonomy grant workers flexibility and thus facilitate possibilities to be physically active during a break, or before or after work. For example, participating in a morning workout, going for a bike ride during break time or joining a soccer match in the afternoon. Additionally, job autonomy is related to lower physical fatigue and may thus provide workers with sufficient energy to participate in leisure-time PA activities (Nijp et al., 2015).

In sum, a person’s job autonomy may not only increase the intention of a worker to be physically active but also provides the conditions to do so. This brings about our first hypothesis (H1): The more job autonomy a worker has, the higher a worker’s level of leisure-time PA. Figure 1 visually represents this hypothesis and our other hypotheses.

Cross-national variation in the job autonomy-physical activity association

As outlined in the introduction, we expect that characteristics of a country’s structural and cultural organisation may affect differences in leisure-time PA behaviour that are evoked by individual differences in job autonomy. Our reasoning resonates the elementary sociological idea that social context affects individual behaviour (Coleman, 1994; Inglehart, 1997); the social environment affects people’s beliefs and actions. This notion is also illustrated by Bronfenbrenner’s (1979) exemplary ecological system theory in which nested levels of influence are identified, among which are peers, neighbourhoods, physical environment, mass media and economic systems, which are important to understand human development. Our research on the job autonomy-PA association adopts this idea and investigates how two country characteristics, namely labour market security and perceived opportunities for physical activity affect this relationship.

A country’s level of labour market security

First, we argue that workers without job autonomy are supported in their PA by the structural condition of labour market security in their country of living. For starters, a country’s labour market security, as indicated by the number of permanent contracts in the labour market, may affect individuals’ feeling of security (Anderson & Pontusson, 2007; Shoss & Probst, 2012). Job security is important for workers’ private and public life as it is related to a sense of power and stability, whereas insecurity is related to powerlessness, stress, instability and lack of command. Labour market security in a country is thus expected to positively affect all workers’ well-being and job satisfaction (Anderson & Pontusson, 2007; Shoss, 2017; Sverke & Hellgren, 2002).

We advocate that a stable labour structure, resulting from labour market security in the country of living, also influences the association between individual job autonomy and leisure-time PA. Especially workers with limited job autonomy may benefit from security at the country level in their intention for PA. They obtain a sense of power and autonomy from labour market security in the country of residence which induces their leisure PA. Contrarily, we expect that such compensation may be smaller for workers who already have much autonomy in their work. Subsequently, we expect that in countries with greater labour market security leisure-time PA levels of workers without autonomy are substantially closer to the leisure-time PA levels of workers with high levels of job autonomy. This argumentation results in our second hypothesis (H2): The association between a worker’s job autonomy and leisure-time PA is weaker in countries with higher levels of labour market security.

A country’s perceived opportunity for physical activity

Second, we consider that ample availability of opportunities for PA may also reduce the PA gap between workers with and without job autonomy. Especially workers low on job autonomy are dependent on an extensive opportunity structure for leisure time PA, while workers with higher job autonomy are better able to adapt their work-related obligations to available (restricted) opportunities for leisure time PA. As such, having multiple options in the direct living environment to be physically active seems particularly valuable for those who experience more barriers to being active (Hoekman et al., 2016).

In previous research, it is well acknowledged that a person’s physical environment provides cues and opportunities to be active (Atkinson et al., 2005; Hallmann et al., 2011; Lim et al., 2011). Accessible formal (e.g. swimming pool, gym) and informal (e.g. public open space, physical activities routes in parks) PA facilities offer people appropriate and readily available options to be active in their leisure time (Giles-Corti & Donovan, 2002; Thibaut et al., 2022). This alleged stimulating effect of sports facilities and other opportunities to be physically active fits within the socioecological rationale in which different environmental systems influence individual behaviours (Bronfenbrenner, 1979).

Earlier we expected that having little job autonomy is associated with limited PA in leisure time, but this negative relationship may be buffered by having sufficient PA opportunities in a country. Workers with low autonomy, who lack a positive spill-over from work to leisure, thus might be compensated through the ample availability of sports and PA facilities in their surroundings. Contrarily, we expect workers scoring high on job autonomy to be already more likely to be physically active regardless of the facilities in their environment. We thus expect that, especially for workers with little job autonomy, perceived opportunities for PA facilitate them to be more physically active, whereas the added value of perceived opportunities for PA is expected to be less apparent for workers with high job autonomy. Therefore, our third hypothesis reads (H3): The association between a worker’s job autonomy and leisure-time PA is weaker in countries with higher levels of perceived opportunities for PA.

Data and measurements

Data: European Social Survey 2014

To test our hypotheses we employ data from the academically driven cross-national European Social Survey 2014 (ESS7; Norwegian Centre for Research Data, 2018). This dataset is unique in that it includes information on job features in combination with information on PA in several European countries.² These ESS7 data are thus perfectly suited to answer our research question. For this survey, a representative random probability sample of Europeans was interviewed face-to-face. We have enriched the ESS7 data with secondary country-level information on labour market security and perceived PA facilities from the European Working Conditions Survey (Eurofound, 2015) and the Eurobarometer (European Commission, 2014). We excluded Israel, Norway and Switzerland because of unavailable country-level information (N = 34,655).

Our study focuses on the relationship between job autonomy and leisure-time PA. For this reason, we only selected respondents who were in paid work for at least 20 h a week. This threshold was adopted to ensure exposure to work and to make job autonomy a relevant work condition. This criterion caused us to remove 9,405 respondents without paid work as their main activity and 1,506 respondents that worked less than 20 h. Furthermore, we selected respondents between the ages of 18 and 65 to ensure people were part of the labour population (N = 15,693). Next, we applied a list-wise deletion of respondents with missing information (4.7%)³. Our final sample consists of 14,956 respondents in 18 countries, aged 18-65 years and working at least 20 h a week (see Table 1).

Table 1. Weighted country-level descriptive statistics.

	Country means		Country-level features
	LTPA	Job control	Job security	Perceived PA opportunity
Austria	3.23	6.01	.74	.88
Belgium	2.92	7.12	.73	.87
Czech Republic	2.88	4.86	.68	.73
Denmark	3.52	7.83	.79	.92
Estonia	3.75	6.66	.78	.78
Finland	3.61	7.70	.70	.85
France	2.21	7.23	.74	.89
Germany	3.62	7.47	.76	.89
Hungary	2.18	4.31	.73	.58
Ireland	3.87	5.98	.58	.78
Lithuania	3.23	5.91	.80	.77
Netherlands	3.30	7.39	.59	.95
Poland	3.24	5.58	.53	.70
Portugal	1.94	7.20	.55	.62
Slovenia	2.70	7.05	.69	.84
Spain	3.01	6.89	.55	.75
Sweden	3.36	7.88	.77	.90
United Kingdom	3.44	7.37	.73	.76
Mean	3.11	6.71	.69	.81

Source: European Social Survey (2014). Eurofound (2015). European Commission (2014). N = 14,956.

Measurements

Our dependent variable leisure-time PA is measured with the item ‘On how many of the last seven days did you walk quickly, did sports or other physical activity for 30 minutes or longer?’. By mentioning sports and quickly walking in the question, this measure likely refers to leisure-time PA over general PA. Leisure-time PA is treated as a continuous variable running from 0 to 7, representing days per week. Multinomial sensitivity analyses with categories of leisure-time PA produced virtually identical results (results available on request from the corresponding author).

A respondent’s job autonomy is measured with the item ‘Please say how much the management at your work allows you to decide how your daily work is organised?’. Answer possibilities ranged from 0 (I have no influence) to 10 (I have complete control).

We control for personal and work-related aspects that (may) affect both a person’s leisure-time PA and job autonomy (Dobbin & Boychuk, 1999; Hellerstedt & Jeffery, 1997; Kouvonen et al., 2005; Kraaykamp et al., 2013; Mohammad Ali & Lindström, 2006; Umberson, 1987; Yoon & Bernell, 2013). We indicate respondents’ gender with male (0) and female (1). Age is measured linearly. Respondents’ health is assessed by asking respondents if they are hampered by their health in their daily activities (0/1). To enable cross-national comparison, educational attainment is measured in years of completed full-time education. Family features of living with a partner and having a child living at home are coded yes (1) and no (0). We categorised the place where respondents reside as either urban (1) or rural (0), and we also control for the number of contracted working hours, being self-employed (0/1), having a permanent contract (0/1), and working in the public sector (0/1). We measure a person’s occupational status with dummy variables: high-skilled white-collar, low-skilled white-collar and blue-collar (see Eurofound 2010). Table 2 gives an overview of the descriptive statistics.

Table 2. Weighted individual-level descriptive statistics.

	Min	Max	%	Mean	S.D.
Leisure time PA	0	7		3.11	2.806
Job control	0	10		6.71	3.453
Female (v. male)	0	1	49
Age	18	65		43.06	12.421
Hampered by health (1 = yes)	0	1	16
Education in years	4	20		14.01	3.832
Live with a partner (1 = yes)	0	1	65
Child living at home (1 = yes)	0	1	72
Urban (v. rural)	0	1	55
Occupational status
High skilled white-collar	0	1	47
Low skilled white-collar	0	1	24
Blue collar	0	1	29
Public sector (1 = yes)	0	1	23
Working hours	20	50		40.56	8.091
Permanent contract (1 = yes)	0	1	73
Self-employed (v. employee)	0	1	12
Degree of job security	0	1		.69	.105
Perceived PA opportunity	0	1		.81	.122

Source: European Social Survey (2014). Eurofound (2015). European Commission (2014). N = 14,956.

At the country level, we measure both labour market security and perceived available PA opportunities. First, a country’s level of labour market security is determined by aggregating respondent’s scores from the European Working Conditions survey 2015 regarding the statement ‘What is your employment status?’ (Eurofound 2015). We use the proportion of workers with a permanent contract as an indication of a country’s labour market security. Labour market security is lowest in Poland (53%) and highest in Lithuania (80%) (see Table 1).

Second, the perceived availability of PA opportunities in a country is derived from Eurobarometer data referring to 2013 (European Commission, 2014). It refers to the proportion of citizens that agree or totally agree with the statement ‘The area where you live offers you many opportunities to be physically active’. The perceived availability of PA facilities is lowest in Hungary (58%) and highest in the Netherlands (95%).

The correlation between countries’ level of labour market security and perceived PA opportunities is relatively low (r = .323, p = .000, N = 18), indicating that collinearity is limited.

Analytical approach

We employ multilevel regression modelling in Stata to test our hypotheses. This technique assumes a hierarchical structure, which in our data are individuals (level 1, N = 14,956) nested in countries (level 2, N = 18). Multilevel models account for variance in leisure-time PA at the individual and the country level, and the interdependence between level-1 units due to the shared country context. Due to the limited number of level-2 units in our data, we will interpret our outcomes with care and reflect on this in the discussion paragraph (Hox, 2010; Stegmueller, 2013).

We weighted the ESS7 data by design weights to account for differential selection probabilities within countries and sample weights to ensure equal numbers of respondents in each level-2 unit (Mehmetoglu & Jakobsen, 2022). We apply sample weights because we are not primarily interested in a country-specific description of job autonomy and leisure-time PA but in a test of theoretical mechanisms underlying our models and cross-level interactions.

The first model in Table 3 represents the null model (Model A). Model B portrays the unadjusted effect of job autonomy on leisure-time PA and Model C includes individual-level control variables. In Model D, we add the work-related aspects. In Table 4, we add country-level variables in Model E (main effects) and separately test our moderation hypotheses through cross-level interactions in Model F, G and H.

Table 3. Multilevel linear regression estimating leisure-time PA.

	Model A			Model B			Model C			Model D
	B		SE	B		SE	B		SE	B		SE
Intercept	3.111	***	.131	2.930	***	.171	3.376	***	.266	3.140	***	.254
Individual level
Job control				.029	*	.011	.033	**	.011	.035	***	.011
Female							−.171	*	.082	−.165	**	.064
Age							−.005		.003	−.006		.003
Hampered by health (ref. = no)							−.227	**	.080	−.240	**	.078
Education in years							.004		.012	.013		.011
Live with a partner (ref. = no)							−.178	**	.054	−.162	**	.055
Child living at home (ref. = no)							−.179	***	.053	−.178	***	.053
Urban (ref. = rural)							.084		.068	.107		.067
Occupational status (ref. = high skilled white-collar)
Low skilled white-collar										.129	*	.064
Blue-collar										.247	**	.075
Public sector (ref. = no)										.248	***	.046
Working hours										.001		.005
Permanent contract (ref. = no)										−.170	*	.084
Self-employed (ref. = no)										.089		.114
Variance components and fit
Level 1 variance (individuals)	5.997		.171	5.977		.170	5.933		.168	5.912		.167
Level 2 variance (countries)	.284		.090	.348		.101	.337		.100	.336		.102
Random slope job control				.001		.000	.001		.000	.001		.000
−2 loglikelihood	70,652			70,619			70,506			70,451

Source: European Social Survey (2014). N_individual = 14,956; N_country = 18. *p < .05; **p < .01; ***p < .001 (two-tailed).

Table 4. Multilevel linear regression estimating leisure-time PA with the country-level and moderation effects.

	Model E			Model F			Model G			Model H
	B		SE	B		SE	B		SE	B		SE
Intercept	.217		1.276	−.244		1.36	−.178		1.35	−.477		1.360
Individual level
Job control	.034	**	.011	.204	**	.062	.172	*	.087	.260	***	.087
Country level
Degree of job security (JS)	.862		1.188	1.656		1.251	.847		1.184	1.538		1.233
Perceived PA opportunity (PAO)	2.900	**	1.033	2.794	**	1.063	3.414	**	1.113	3.193	**	1.125
Cross-level moderations
Job control * JS				−.247	**	.092				−.207	*	.096
Job control * PAO							−.174		.110	−.105		.106
Variance components and fit
Level 1 variance (individuals)	5.911		.167	5.912		.167	5.911		.167	5.911		.167
Level 2 variance (countries)	.237		.091	.230		.085	.234		.090	.228		.086
Random slope job control	.001		.000	.000		.000	.001		.000	.000		.000
−2 loglikelihood	70,445			70,439			70,442			70,438

Source: European Social Survey (2014). N_individual = 14,956; N_country = 18. *p < .05; **p < .01; ***p < .001 (two-tailed). Note: The models include the control variables that are included in Table 3 Model D.

Results

First, from the null model in Table 3, the intraclass correlation shows that (.284/(.284 + 5.997)=) 4.5% of the variance in leisure-time PA is located at the country level which justifies multilevel modelling. The intercept of this model refers to a grand mean (b = 3.111) and implies that, on average, workers in Europe are physically active for about three days a week.

In model B, we observe that workers with more job autonomy report being more physically active in their leisure time (b = .029). Including individual-level control variables in Model C and D slightly increases the estimate of job autonomy which confirms that job autonomy is independently associated with a worker’s PA. Based on Model D, the average marginal effect at variable means of workers with complete autonomy is 3.24 days of physical activity per week, whereas workers without job autonomy are only active for 2.89 days per week. This is a confirmation of our first hypothesis.

The control variables included in Model D indicate that leisure-time PA is higher among male (b = −.165) and healthier (b = −.240) workers and workers who live with a partner (b = −.162) and have a child at home (b = −.178). Blue-collar (b = .247) and low-skilled white-collar (b = .129) workers are both more active than high-skilled white-collar workers. Finally, leisure-time PA is higher among those who work in the public sector (b = .248) and do not have a permanent contract (b = −.170).

Model E in Table 4 shows the main effects of a country’s level of labour market security and perceived PA opportunity structure on individuals’ leisure-time PA. The proportion of perceived PA opportunities associates positively with workers’ leisure-time PA. As expected, in general workers are more physically active in countries with more perceived PA opportunities (b = 2.900). The degree of labour market security does not directly affect a worker’s leisure-time PA.

In Models F and G, our moderation hypotheses are tested with cross-level interactions. The degree of a country’s labour market security weakens the positive relationship between job autonomy and leisure-time PA (b = −.247). This significant interaction term is visualised in Figure 2. The slope of high autonomy declines and the slope of low autonomy increases as far as the slopes and confidence intervals converge around a job security proportion of 0.7. In other words, in countries with higher levels of labour market security, those with low and high job autonomy do not substantially differ in their level of leisure-time PA. As such, higher levels of a country’s labour market security lower the differences in leisure-time PA by job autonomy. This confirms Hypothesis 2.

Figure 1. Conceptual model.

Figure 2. Visualisation of the effect of job autonomy on leisure-time PA by the degree of labour market security.

We also expected that perceived PA opportunities in a country would particularly help workers low on job autonomy to be physically active. We however do not find a significant interaction and therefore cannot confirm our third hypothesis.

Additional analyses

To determine the generalisability of our results we performed robustness analyses. We, first, follow the many studies on work characteristics that run separate analyses for men and women and age groups (e.g. Kouvonen et al., 2005). Additionally, we differentiated our analyses for individuals with and without children. Tables are found in Appendix A.

Overall the sensitivity analyses indicate that the effect of job autonomy on leisure-time PA is apparent for all workers. The effect is somewhat stronger for male workers and workers older than 40, and those without children in the home. Additionally, the interaction between job autonomy and the country-level degree of labour market security is robust among all groups of workers, except among female workers. Tables A2–A4 (see Appendix A) also indicate that for female workers, older workers and workers with children a country’s PA facilities do matter. In these groups, workers with low job autonomy benefit relatively more from living in a country with a high degree of perceived PA opportunities than workers with higher levels of job autonomy. Thus, for female and older workers and workers with children, a higher country’s degree of perceived PA opportunities mitigates the differences in leisure-time PA by job autonomy.

Conclusion and discussion

Our study provides a unique cross-nationally comparative insight into the relationship between job autonomy and leisure-time PA in Europe. More specifically, we answer the question whether a country’s degree of perceived PA opportunities and level of labour market security is related to the association between job autonomy and leisure-time PA. In doing so, our study innovatively adds a country-comparative element to the investigation of the work/non-work interface perspective on physical activity. With physical inactivity levels increasing and 55% of European adults being part of the labour force, knowledge about the interaction between work features and the country context is crucial in improving leisure-time PA within Europe.

Our results show that workers with high job autonomy are more physically active in leisure time, regardless of their educational attainment, occupation and other control variables. This finding is in line with previous studies on job autonomy and PA in Europe (e.g. Fransson et al., 2012; Heikkilä et al., 2013; Kouvonen et al., 2005). Our results thus underscore that in the European context job autonomy is a principal factor for workers to be physically active. Likely a positive carry-over of feelings of autonomy affects the desire to engage in PA during non-work time (Häusser & Mojzisch, 2017; Ryan & Frederick, 1997). Additionally, the autonomy over scheduling and time coming with job autonomy provides the possibilities to better facilitate leisure-time PA (Nijp et al., 2015) and may leave workers with sufficient energy to engage in leisure-time PA.

In addition, this study innovatively indicates that the association between leisure-time PA and degree of job autonomy is related to countries’ labour market security. A higher degree of labour market security curtails the influence of job autonomy on being active. That is, an environment in which workers are confident about the continuity of their labour may generate general feelings of power and lowers stress among workers (Sverke & Hellgren, 2002), and this mitigates the relevance of having job autonomy for leisure-time PA. With these findings, we add to the existing body of knowledge and provide insight into the context-dependency of job autonomy and leisure-time PA.

A few implications of our findings may be noted. First, our results highlight the importance of PA opportunity for workers. Having autonomy in one’s work provides more possibilities to be physically active (see also Grubben et al., 2022). Attention to aspects of job autonomy, such as control over time scheduling and order of tasks, is therefore essential to increase opportunities for and levels of PA among the workforce. For those with little job autonomy, it is relevant to decrease the barriers to becoming active. As such, we recommend a multilevel view that approaches this barriers problem from the individual, organisational and country perspective. For instance, workplace physical activity interventions targeted particularly at workers with low job autonomy and country-level measures that enhance the sense of labour market security.

Second, in addition to the importance of job autonomy, our study especially underscores the importance of labour market security to facilitate sustainable leisure-time PA levels, especially for workers with limited job autonomy. Consequently, we advise policymakers intending to increase levels of leisure-time PA among the workforce to pay due attention to aspects of labour market security. For instance, we recommend considering the consequences for PA and health in designing labour market policies (e.g. European Union’s employment strategy).

Third, our study investigated the importance of opportunities to be physically active in the area people live to facilitate leisure-time PA. We thus suggest, in line with the policy recommendations of the World Health Organisation, to uphold and intensify the attention within national and European policy for creating PA-friendly environments. Moreover, we encourage future researchers to dig deeper into the scientific puzzle for which groups a developed PA opportunity structure is especially relevant. These are the social groups that struggle more with creating a sustainable work-life balance, namely workers with children and women (Emslie & Hunt, 2009; Sirgy & Lee, 2018). Furthermore, workers over 40 years of age with low job autonomy seem to be supported more by PA facilities in their environment (Bonaccorsi et al., 2020). This issue requires future research that combines a focus on the work-life balance for specific social groups with a cross-national perspective.

Besides assets and innovative elements, our study surely has limitations. First, using data on 14,956 workers in several countries comes at the cost of detail. Future research may want to assess the various dimensions of job autonomy that are expressed in empirical research, such as method or scheduling autonomy (Nijp et al., 2015). Also, leisure-time PA is currently measured as days per week which might explain the relatively small effect. A more detailed measure of leisure-time PA (e.g. the international PA questionnaire by Craig et al., 2003) may deepen our understanding of the relationship. Additionally, our sample of quite homogeneous countries may have affected the estimates of cross-level interaction effects. Our findings could be strengthened by future studies covering a larger and more heterogeneous set of countries. Second, our measure of a country’s PA structure is subjective and may be dependent on a priori activity levels of respondents. Nevertheless, research by Hoekman et al. (2016) showed that judgments of one’s environment are more important to individuals in their PA behaviour than objective features. Third, social learning processes in one’s youth greatly influence PA levels as well as that PA levels may affect work-related outcomes (Burgard & Lin, 2013; Pacheco et al., 2014). We are not able to completely control for such reversed causality in this paper. We, therefore, interpret our results with care and want to point out the importance of elaborate longitudinal studies to verify the results of cross-sectional studies such as this one. Nonetheless, our study comprehensively illustrates the relationship between job autonomy and leisure-time PA, and the country-level influence of labour market security and perceived PA opportunities. We conclude that job autonomy is a relevant resource for the European labour force to engage in leisure-time PA and that countries’ labour market security and perceived PA opportunities influence differences in leisure-time PA related to job autonomy.

This study may fuel new research questions. With this being the first paper to assess the relationship between job autonomy and leisure-time PA from a cross-national multilevel perspective, future research may build upon our work. For instance, our sensitivity analyses indicated that relationships differ for crucial social groups. Retaining a healthy workforce, therefore, requires additional theorising and empirical research into the relationship between job autonomy and PA for distinct groups of workers. Moreover, besides influences at the societal level, the work-health relationship may also be affected by a worker’s organisation. For instance, organisational physical activity programs or health checks may directly affect workers’ PA but may also impact the relationship between job autonomy and physically active behaviour (Van der Put et al., 2020).

To summarise, this study enhances our understanding of country differences in the relation between job autonomy and leisure-time PA and may inform policymakers to critically assess leisure-time PA promotion policies for the workforce. Special attention is required to increase levels of job autonomy, labour market security and the improvement of PA opportunities to be physically active to help counter health inequalities within the European labour force.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

This work was supported by the Dutch Research Council (NWO) under Grant NWA.1160.18.249

Notes

1 Seldom refers to respondents exercising or playing sports three times a month or less often (European Commission, 2018).

2 The 2014 wave of the European Social Survey is the most recent dataset available that has country comparative information on both job autonomy and leisure time PA.

3 The combination of weighing the data and multilevel modelling demanded us to use list-wise deletion instead multiple imputation techniques to acquire reliable regression estimates.

Appendix A.

 

Table A1. Multilevel linear regression estimating leisure-time PA differentiated by sex, age group and having a child at home.

	Sex						Age group						Having a child at home
	Male (N = 7,607)			Female (N = 7,349)			18 to 40 (N = 6,176)			41 to 65 (N = 8,780)			Yes (N = 7,454)			No (N = 7,502)
	B		SE	B		SE	B		SE	B		SE	B		SE	B		SE
Intercept	3.296	***	.388	2.666	***	.347	3.198	***	.350	2.574	***	.319	2.815	***	.393	3.139	***	.446
Individual level
Job autonomy	.042	**	.016	.032	**	.010	.030	**	.010	.042	**	.015	.030	*	.013	.042	**	.013
Female							−.291	**	.090	−.045		.073	−.223	**	.069	−.095		.093
Age	−.013	***	.003	.003		.004							.000		.005	−.008	*	.003
Hampered by health (ref. = no)	-.282	*	.110	−.195	**	.072	−.287	***	.068	−.217	*	.091	−.298	*	.137	−.179	*	.086
Education in years	.011		.018	.020		.013	.008		.012	.018		.014	.003		.016	.028	+	.015
Live with a partner (ref. = no)	-.147	*	.074	−.186	*	.082	−.205	*	.094	−.074		.060	−.156	*	.076	−.153	*	.062
Child living at home (ref. = no)	-.138	+	.074	−.205	***	.055	−.314	**	.109	−.079		.075
Urban (ref. = rural)	.072		.057	.143		.089	.149	+	.085	.087		.066	.170	+	.100	.013		.086
Occupational status (ref. = high skilled white-collar)
Low skilled white-collar	.164		.100	.091		.060	.109		.079	.142		.099	.192	*	.092	.049		.109
Blue-collar	.237	*	.109	.283	*	.114	.272	*	.115	.239	*	.102	.333	***	.093	.148		.092
Public sector (ref. = no)	.448	***	.076	.097	+	.053	.199	*	.084	.249	***	.049	.297	***	.051	.174	*	.082
Working hours	.002		.006	−.001		.005	.001		.007	.000		.004	.000		.006	.000		.007
Permanent contract (ref. = no)	-.154		.094	−.186	+	.112	−.228	+	.128	−.138		.091	−.112		.112	−.200		.127
Self-employed (ref. = no)	.192		.141	−.077		.164	.018		.221	.117		.146	.117		.173	.097		.199
Variance components and fit
Level 1 variance (individuals)	5.946		.186	5.802		.177	5.583		.175	6.107		.178	5.938		.180	5.860		.193
Level 2 variance (countries)	.342		.106	0.316		.109	.191		.340	.528		.200	.301		.113	.290		.098
Random slope job autonomy	.001		0.001	.000		.000	.000		.000	.002		.001	.001		.001	.001		.001
−2 loglikelihood	35,794			34,604			28,812			41,616			38,493			31,977

Source: European Social Survey (2014). N_individual = 14,956; N_country = 18. +p < .10; *p < 0.05; **p < 0.01; ***p < 0.001 (two-tailed).

Table A2. Multilevel linear regression estimating leisure-time PA differentiated by sex.

	Male (N = 7,607)									Female (N = 7,349)
	B		SE	B		SE	B		SE	B		SE	B		SE	B		SE
Intercept	.278		1.438	−.514		1.542	−.170		1.486	.442		1.268	.002		1.426	-.231		1.391
Individual level
Job autonomy	.040	*	.017	.294	***	.086	.182		.132	.031	**	.010	.097		.061	.133	*	.055
Country level
Degree of labour market security [LMS]	1.753		1.547	3.028	+	1.656	1.730		1.536	−.194		1.025	.430		1.217	-.133		1.031
Perceived PA opportunity [PAO]	2.255	+	1.208	2.154	+	1.271	2.850	*	1.266	2.925	**	.979	2.945	**	.994	3.728	***	1.112
Cross-level moderation
Job autonomy * LMS				−.366	**	.126							−.096		.085
Job autonomy * PAO							−.179		.163							-.130	+	.071
Variance components and fit
Level 1 variance (individuals)	5.944		.185	5.945		.185	5.944		.185	5.802		.198	5.801		.175	5.800		.174
Level 2 variance (countries)	.255		.010	.245		.083	.252		.098	.235		.561	.237		.210	.236		.187
Random slope job autonomy	.001		.001	.001		.001	.001		.001	.000		.000	.000		.000	.000		.000
−2 loglikelihood	35,789			35,782			35,787			34,599			34,598			34,597

Source: European Social Survey (2014). N_individual = 14,956; N_country = 18. +p < .10; *p < .05; **p < 0.01; ***p < 0.001 (two-tailed). Note: The models include the control variables that are included in Table 3 Model D.

Table A3. Multilevel linear regression estimating leisure-time PA differentiated by age group.

	18–40 (N = 6,176)									41–65 (N = 8,780)
	B		SE	B		SE	B		SE	B		SE	B		SE	B		SE
Intercept	1.465		1.124	.551		1.096	1.439		1.119	−1.202		1.626	−1.650		1.776	−1.873		1.790
Individual level
Job autonomy	.029	**	.010	.166	***	.040	.033		.060	.040	**	.015	.239	*	.112	.305	*	.127
Country level
Degree of labour market security [LMS]	.822		1.047	2.125	+	1.086	.825		1.043	.432		1.405	1.241		1.571	.439		1.383
Perceived PA opportunity [PAO]	1.444		.886	1.493		.909	1.476	+	.877	4.338	***	1.278	4.206	***	1.276	5.188	***	1.523
Cross-level moderation
Job autonomy * LMS				−.201	**	.064							−.288	+	.160
Job autonomy * PAO							−.005		.079							-.333	*	.162
Variance components and fit
Level 1 variance (individuals)	5.583		.173	5.580		.172	5.583		.178	6.107		.178	6.108		.178	6.107		.178
Level 2 variance (countries)	.158		.058	.162		.059	.158		.512	.336		.143	.319		.136	.313		.136
Random slope job autonomy	.000		.000	.000		.000	.000		.000	.002		.001	.002		.001	.001		.001
−2 loglikelihood	28,809			28,805			28,809			41,609			41,606			41,603

Source: European Social Survey (2014). N_individual = 14,956; N_country = 18. +p < .10; *p < .05; **p < .01; ***p < 0.001 (two-tailed). Note: The models include the control variables that are included in Table 3 Model D.

Table A4. Multilevel linear regression estimating leisure-time PA differentiated by having a child at home.

	Yes (N = 7,454)									No (N = 7,502)
	B		SE	B		SE	B		SE	B		SE	B		SE	B		SE
Intercept	.057		1.379	-.619		1.499	-.806		1.574	.542		1.183	-.035		1.274	.538		1.297
Individual level
Job autonomy	.028	*	.013	.231	*	.092	.253	*	.125	.040	**	.013	.161	*	.064	.041		.091
Country level
Degree of labour market security [LMS]	.815		1.205	2.158	+	1.300	.745		1.201	.592		1.183	1.372		1.377	.593		1.191
Perceived PA opportunity [PAO]	2.734	**	.971	2.443	*	1.020	3.891	**	1.262	2.717	*	1.135	2.770	*	1.153	2.721	*	1.349
Cross-level moderation
Job autonomy * LMS				-.295	*	.130							-.174	+	.105
Job autonomy * PAO							-.284	+	.154							-.001		.126
Variance components and fit
Level 1 variance (individuals)	5.935		.180	5.936		.179	5.935		.180	5.861		.192	5.860		.192	5.861		.192
Level 2 variance (countries)	.223		.097	.221		.086	.221		.092	.198		.082	.200		.081	.198		.082
Random slope job autonomy	.001		0.001	.000		.001	.001		.001	.001		.001	.001		.001	.001		.001
−2 loglikelihood	38,487			38,482			38,482			31,972			31,970			31,972

Source: European Social Survey (2014). N_individual = 14,956; N_country = 18. +p < .10; *p < .05; **p < 0.01; ***p < .001 (two-tailed). Note: The models include the control variables that are included in Table 3 Model D.