Background
People’s lifestyle choices are shaped by their social environment. Partners, family members, friends, and neighbours have been shown to influence the extent to which people eat healthily and engage in physical activity [
1‐
8]. As a result, researchers have increasingly looked beyond the individual level and also examined the social environment to understand how people make (un)healthy choices [
9‐
11]. However, one relevant set of influential actors has received less attention: colleagues. This is surprising given that many adults spend the majority of their waking hours in the workplace, where they repeatedly encounter the same colleagues [
12,
13]. The workplace is a social arena in which people interact often and on a daily basis [
14], so that their shared time may surpass that spent with family and romantic partners [
3]. In other words, colleagues should be included in an inquiry into the influence of social environments on (un)healthy behaviours. The present study examines to what extent colleagues may have a part in one another’s lifestyle choices, specifically healthy diets and physical activity.
Similarity between people’s lifestyle behaviours has been attributed to various mechanisms: homophily, shared environments, or more direct forms of influence [
6]. In this article, we focus on the latter and argue for two potential ways in which colleagues can affect one another’s healthy behaviours: through encouraging each other to make healthy choices and by acting as role models. Firstly, via encouragement, colleagues can enhance each other’s motivation, increase their self-care and create a sense of shared responsibility for healthy choices [
1,
2,
15]. Previous studies examined generic social support and social capital at the workplace, focussing more on trust between colleagues and solidarity [
11,
13,
16,
17]. It is hard to see how this may enhance motivation for healthy choices, so unsurprisingly these studies found no association with healthy behaviours. By contrast, support specific to a behaviour has been suggested to be more predictive [
18], also when it concerns employee healthy behaviour [
19]. We therefore focus on encouragement from colleagues specific to healthy behaviours.
In serving as salient role models via their own lifestyle habits, colleagues can set norms and increase self-efficacy [
7,
14,
20,
21]. Norms are a powerful driver of behaviour while self-efficacy determines the extent to which people enact a given behaviour. Hence, we additionally examine the role of colleagues’ own healthy behaviours. Previous studies tended to focus on employee reports of what their colleagues do, rather than what these colleagues actually do. Additionally, the focus has been largely on healthy behaviours in the workplace [
9,
10,
20,
22] while we employ measures of eating and physical activity that also take place outside work. Both mechanisms, encouragement and role modelling, are expected to contribute making workplaces conducive to employee health and well-being, thereby creating a supportive environment for healthy behaviours [
23,
24].
Previous studies involving colleagues’ roles in shaping lifestyle choices tended to (a) solely model formal membership in organisations rather than measure close proximity and the chance to meet [
25], (b) incorporate group sizes too small to be conclusive [
4] or (c) aggregate individual-level measures to group-level variables [
13,
14], underestimating social influence from direct colleagues [
26]. To overcome this research gap, we use unique multilevel data from the European Sustainable Workforce Survey – ESWS [
27]. The ESWS comprise 4345 employees nested in 402 teams in 113 organisations in 9 European countries. Importantly, the data structure permits modelling which employees work together and thus have the potential to affect one another’s behaviours. We employ network autocorrelation models [
28] to account for the interdependence of employees’ healthy behaviours within a team. Network autocorrelation models enable the direct association of employees’ outcomes with those of their colleagues, offering a better test for finding out how colleagues’ healthy behaviours are related.
We examine two different behaviours: healthy eating and physical activity. Together, they have great potential to improve health, and are often paired in lifestyle-related recommendations [
21,
29]. Adults consume about a third of their daily calorie intake in the workplace and work impacts the extent to which employees engage in physical activity [
12,
30]. For example, many office workers spend most of their working day sedentary. Eating and physical activity behaviours share a social component, for example, lunch is often eaten together in the workplace [
17,
31], while differing substantially in exposure and ease of implementation. Since eating is typically more visible and occurs more frequently in the workplace than exercising, eating tends to be more prone to social influence [
19]. Studying the two comparable yet different activities of eating and physical activity is meant to provide nuanced insights into the role of the social workplace for healthy choices.
Results
The descriptive results in Table
1 show that on average, employees scored 4.73 on fruit consumption and 4.95 on vegetable consumption. This translates into eating fruit and vegetables about once per day. On average, employees engaged in physical activity three days per week.
Table 1
Descriptive statistics
| M | SD | M | SD | M | SD | Range |
Fruit consumption | 4.73 | 1.40 | | | | | |
Vegetable consumption | | | 4.95 | 1.12 | | | |
Physical activity | | | | | 2.92 | 2.21 | 0–7 |
Perceived encouragement eating | 2.22 | 1.22 | 2.22 | 1.22 | | | |
Perceived encouragement physical activity | | | | | 2.11 | 1.18 | 0–5 |
Female | 0.58 | | 0.58 | | 0.57 | | |
Age | 43.97 | 11.40 | 44.00 | 11.38 | 43.64 | 11.37 | 19–77 |
Education in years | 13.72 | 3.52 | 13.70 | 3.53 | 13.79 | 3.48 | 3–21 |
Health | 3.86 | 0.72 | 3.86 | 0.72 | 3.86 | 0.72 | 1–5 |
Partner | 0.74 | | 0.74 | | 0.74 | | |
Children | 0.50 | | 0.50 | | 0.50 | | |
Work hours per week | 39.63 | 9.77 | 39.60 | 9.80 | 39.65 | 9.74 | 0–60 |
Physical work demands | 3.17 | 1.45 | 3.17 | 1.45 | 3.15 | 1.45 | 1–5 |
Tenure in years | 8.81 | 8.67 | 8.84 | 8.70 | 8.63 | 8.63 | 0–49 |
Working from home | 1.73 | 1.37 | 1.73 | 1.38 | 1.74 | 1.38 | 1–7 |
WHP | | | | | | | |
Not available | 0.46 | | 0.46 | | 0.44 | | |
Available but not used | 0.33 | | 0.33 | | 0.42 | | |
Available and used | 0.21 | | 0.21 | | 0.14 | | |
Team size | | | | | | | |
10 employees or less | 0.39 | | 0.38 | | 0.39 | | |
11–20 employees | 0.23 | | 0.25 | | 0.25 | | |
21 employees or more | 0.38 | | 0.37 | | 0.35 | | |
Sector | | | | | | | |
Manufacturing | 0.32 | | 0.32 | | 0.32 | | |
Healthcare | 0.26 | | 0.26 | | 0.25 | | |
Higher education | 0.22 | | 0.22 | | 0.23 | | |
Transport | 0.09 | | 0.09 | | 0.09 | | |
Financial services | 0.06 | | 0.06 | | 0.06 | | |
Telecom | 0.05 | | 0.06 | | 0.06 | | |
Country | | | | | | | |
UK | 0.05 | | 0.05 | | 0.05 | | |
Germany | 0.06 | | 0.06 | | 0.06 | | |
Finland | 0.03 | | 0.03 | | 0.03 | | |
Sweden | 0.09 | | 0.09 | | 0.09 | | |
The Netherlands | 0.21 | | 0.21 | | 0.22 | | |
Portugal | 0.05 | | 0.05 | | 0.05 | | |
Spain | 0.04 | | 0.04 | | 0.04 | | |
Hungary | 0.18 | | 0.18 | | 0.18 | | |
Bulgaria | 0.29 | | 0.29 | | 0.27 | | |
N | 3111 | | 3145 | | 2992 | | |
We first examined hypotheses 1a, 1b, and 1c, which associated more perceived encouragement by colleagues with increased fruit and vegetable consumption and physical activity. We find support for all three of these hypotheses as seen in Table
2. Perceived encouragement was positively correlated to fruit consumption (B = 0.071,
p < 0.001), vegetable consumption (B = 0.052,
p = 0.001), and physical activity (B = 0.086,
p = 0.016). We furthermore found no significant spill-over for any of the three outcomes (fruit consumption: B = 0.034,
p = 0.058, vegetable consumption: B = 0.016,
p = 0.110 and physical activity: B=-0.029,
p = 0.055). This means that behavioural changes due to higher perceived encouragement did not spill-over to other colleagues. These findings support our first set of hypotheses, associating perceived encouragement of healthy habits by colleagues with greater fruit and vegetable consumption and physical activity.
For our second set of hypotheses, we expected that the more an employee’s colleagues showed healthy behaviours, the more this employee would behave in healthy ways. We tested this using the network effect ρ, which related employees’ behaviours to that of their colleagues. The results on the network effects ρ in Table
2 suggest an association between colleagues’ fruit consumption (ρ = 0.329,
p = 0.002) and vegetable consumption (ρ = 0.238,
p = 0.024). If all colleagues raised their food consumption by one unit on average, this would lead to an increase of 0.329 in the employee’s fruit consumption and 0.238 in the employee’s vegetable consumption. Against our expectations, we found a significant negative network effect for physical activity (ρ=-0.449,
p = 0.009). If colleagues were more physically active by one day on average, employees would decrease their own activity by about half a day.
Some of the results for the control variables are worth noting. Physical work demands were positively related to total physical activity (B = 0.074, p = 0.016). Additionally, WHP appeared to play no role in the extent to which employees eat fruits and vegetables, but was related to physical activity. Employees that had WHP aimed at physical activity (either sport facilities at work or a financial contribution towards a sport activity outside) available and used this, reported higher total physical activity (B = 0.327, p = 0.038).
Table 2
Unstandardised coefficients (B) and standard errors (SE) from the network autocorrelation models on health behaviours
| B | SE | B | SE | B | SE |
Perceived encouragement | 0.071*** | 0.020 | 0.052** | 0.016 | 0.086* | 0.036 |
Network effect ρ | 0.329** | 0.108 | 0.238* | 0.106 | -0.449* | 0.172 |
Female (ref = male) | 0.237*** | 0.054 | 0.299*** | 0.043 | -0.155 | 0.091 |
Age | 0.010*** | 0.003 | -0.000 | 0.002 | 0.013** | 0.004 |
Education in years | 0.015 | 0.009 | 0.037*** | 0.007 | -0.023 | 0.015 |
Self-rated health | 0.251*** | 0.034 | 0.173*** | 0.027 | 0.498*** | 0.058 |
Partner (ref = unpartnered) | 0.159** | 0.057 | 0.151** | 0.045 | -0.130 | 0.098 |
Children (ref = no) | 0.019 | 0.050 | -0.042 | 0.039 | -0.238** | 0.086 |
Work hours per week | -0.001 | 0.003 | -0.004* | 0.002 | -0.008 | 0.004 |
Physical job demands | -0.008 | 0.018 | -0.004 | 0.014 | 0.074* | 0.031 |
Tenure in years | 0.004 | 0.003 | 0.002 | 0.003 | 0.001 | 0.006 |
Working from home | -0.000 | 0.020 | -0.002 | 0.015 | 0.052 | 0.033 |
WHP (ref = not available) | | | | | | |
Available, not used | -0.078 | 0.061 | -0.044 | 0.048 | -0.064 | 0.107 |
Available, used | 0.064 | 0.072 | 0.057 | 0.057 | 0.327* | 0.158 |
Team size (ref = 10 employees or more) | | | | | | |
11–20 employees | 0.067 | 0.065 | 0.037 | 0.050 | -0.060 | 0.108 |
21 employees or more | -0.011 | 0.059 | 0.016 | 0.046 | -0.128 | 0.102 |
Sector (ref = manufacturing) | | | | | | |
Health care | 0.018 | 0.078 | -0.032 | 0.061 | 0.575*** | 0.143 |
Higher education | 0.039 | 0.088 | -0.048 | 0.068 | 0.427* | 0.154 |
Transport | -0.117 | 0.104 | -0.054 | 0.082 | -0.070 | 0.173 |
Financial services | -0.094 | 0.113 | -0.124 | 0.089 | 0.503* | 0.206 |
Telecom | -0.033 | 0.119 | -0.041 | 0.093 | -0.351 | 0.205 |
Country (ref = Netherlands) | | | | | | |
UK | 0.043 | 0.131 | 0.155 | 0.108 | 0.913*** | 0.235 |
Germany | -0.229 | 0.128 | -0.077 | 0.097 | 0.205 | 0.208 |
Finland | -0.111 | 0.160 | 0.479** | 0.141 | 1.206*** | 0.318 |
Sweden | -0.459*** | 0.124 | 0.082 | 0.083 | 0.531** | 0.194 |
Portugal | 0.307* | 0.129 | 0.046 | 0.095 | -0.598** | 0.222 |
Spain | -0.008 | 0.135 | -0.319** | 0.117 | 1.445*** | 0.281 |
Hungary | -0.487*** | 0.119 | -0.415*** | 0.093 | 0.324* | 0.143 |
Bulgaria | -0.500*** | 0.106 | -0.102 | 0.062 | 0.734*** | 0.161 |
Constant | 1.403** | 0.547 | 2.513** | 0.527 | 1.522* | 0.618 |
Pseudo R2 | 0.12 | | 0.14 | | 0.07 | |
N | 3111 | | 3145 | | 2992 | |
Sensitivity analyses
We performed several sensitivity analyses to gauge the robustness of our findings. This is especially relevant for the network effect, which is dependent on the construction of the weight matrix [
28]. Firstly, instead of creating separate samples and weight matrices for each of the three dependent variables, we re-ran our analysis using a single sample and an identical weight matrix across all three models. The single sample was reduced to N = 2922, as it excluded employees whose information was incomplete on any of the three dependent variables. The results turned out highly robust for all hypothesised associations.
Secondly, we re-ran our analysis without using row normalisation, which assumed that every colleague had the same influence on an employee’s behaviour. Arguably, small teams offer fewer interaction partners than large teams, allowing for more frequent contact with every colleague. The results without row normalisation remained the same for perceived encouragement (H1a-1c) and network effect for vegetable consumption (H2b). We found a marginally significant network effect for fruit consumption (H2a: ρ = 0.086, p = 0.071) and a significant positive network effect for physical activity (H2c: ρ = 0.233, p = 0.049).
Thirdly, to further assess whether some colleagues affected one another more than others, we reconstructed the network using nomination data. In the survey, each employee had named up to three colleagues whom they meet outside work, and up to 3 colleagues with whom they enjoy working. Employees likely had more contact with these colleagues and could thus have been more influenced by them. The results for perceived encouragement remained stable (H1a-1c), while all three of the network effects became insignificant (H2a-2c). However, this null effect, was likely caused by poor statistical power, as those networks were extremely sparse [
26].
Fourthly, to assess whether WHP may play a role in the extent to which colleagues affect each other’s behaviours, we re-ran our analyses without the control variable for availability and use of WHP. For fruit consumption and physical activity our results remained robust, but in the case of vegetable consumption we found a marginally significant network effect (p = 0.059).
Lastly, to assess whether results could be country- or sector-specific, we performed jack-knife procedures excluding either one country or one sector at a time [
51]. The findings stayed the same concerning perceived encouragement for fruit consumption, vegetable consumption and physical activity (H1a-1c). The findings also turned out similar for the network effects on fruit consumption (H2a). By contrast, the network effects remained weaker for both vegetable consumption and physical activity (H2b-2c): the effects remained unaltered in terms of direction but failed to reach statistical significance in some of the subsamples. This suggests the impact of colleagues’ behaviours may differ per sector and country.
Discussion
The aim of this study was to gauge colleagues’ effects, if any, on one another’s healthy behaviours. While previous studies demonstrated that partners, family members, friends and neighbours influence the extent of people’s healthy eating and exercising, the role of colleagues has remained understudied [
1‐
8]. This is surprising, considering the amount of time many adults spend at work, frequently interacting with the same colleagues [
12,
14]. We studied whether colleague encouragement and behaviour could be associated with the extent to which employees eat fruit and vegetables and engage in physical activity [
23,
24]. We used the European Sustainable Workforce Survey [
27], with data on employees nested in teams, to allow us to reconstruct which colleagues worked together and could thus directly affect one another.
We found that, as expected, the more employees perceived their colleagues to encourage them to behave in healthy ways, the higher the employee’s fruit and vegetable intake, and the more the employee was physically active. This result is in line with previous studies [
3,
29,
32,
36]. Perceived encouragement may relate to healthy behaviours in several ways – e.g. through positive reinforcement [
1], instilling a sense of responsibility [
15], and creating a sense of belonging [
16] – all of which may increase self-care and enhance motivation. Some studies looked at generic social support or social capital and found no effect [
11,
13,
16,
17] while our findings suggest that support specific to the behaviour does matter.
Our network models also showed that employees’ healthy eating may follow from colleagues’ fruit and vegetable consumption, as colleagues may represent salient role models whose behaviour sets the norm about what could be expected based on social learning theory [
39]. Moreover, observing behaviour from colleagues may increase the employee’s self-efficacy, by demonstrating the ability to, for example, bring a healthy snack to work rather than buying from the vending machine [
20,
21]. Since employees consume about a third of their daily calories in the workplace, this is an important setting to promote healthy eating [
12]. Previous studies mainly gauged healthy eating in the workplace and found similar results [
9,
10,
20,
22]. Our measure of healthy eating comprised fruit and vegetable intake in general, thus also outside the workplace. Our results suggest that colleagues’ influence extends to private and leisure settings.
As opposed to our expectations, we found a negative correlation between employees’ and colleagues’ physical activity. This result is not yet conclusive, since the sensitivity analysis indicated a positive trend when allowing for the influence of colleagues to vary with team size. Previous studies did find an association between employees’ and colleagues’ physical activity [
10,
22], but they focused mainly on employees’ perceptions of colleagues’ behaviour instead of their actual behaviours, which may arguably matter more. One explanation for our negative result may be that physical activity typically takes place outside work, where it is hardly visible to colleagues. Eating at work takes place daily and usually happens together with colleagues, whereas physical activity behaviours occurs less frequently, making it less prone to social influence [
19]. Some employees may participate in group sports classes at the workplace together with their colleagues, but this is arguably a small group. Additionally, in our sample WHP initiatives aimed at healthy eating were more often used than programmes promoting physical activity, which may also mean eating behaviour of colleagues is more visible. To be effective, norms should be specific to the situation [
52‐
54], hence norms on dieting can be supported more easily via observation of others in the workplace, a setting which often includes eating meals [
19]. Moreover, extremely athletic colleagues could demotivate others by giving them the impression that this level of physical activity is out of reach for them [
15,
30].
Of further interest is to note that employees with more physical work demands reported higher physical activity, which could be shared by colleagues who work in a similar job. These employees may already feel that they are active enough during the working day, although previous research has shown that occupational physical activity is no substitute for leisure-time physical activity [
47]. Finally, even though employees often engage in WHP together with their colleagues [
19], which could be one of the mechanisms through which colleagues’ lifestyle choices relate to each other, results showed the influence of colleagues extends beyond shared WHP use. The exception here concerned vegetable consumption, which we no longer found statistically significant when removing WHP from our analyses, suggesting that the impact of colleagues eating vegetables could take place during shared lunch in the healthy worksite cafeteria. Nevertheless, these findings show that colleagues also matter for employees that do not participate in WHP.
Limitations and suggestions for future research
Several limitations of our study are worth noting. Firstly, the cross-sectional research design was unable to separate selection from influence processes [
2]. However, it seems unlikely that employees choose to work in a formal work team based on the healthy behaviours of colleagues. Although the extent to which healthy behaviours are the norm may differ among occupations, we addressed this variation by including control variables for education and sector. Future research would benefit from using longitudinal data to examine influence processes over time. As argued by self-determination theory [
35], individuals internalise cues from their environment to shape their intrinsic motivation. It would thus be interesting to study how long it takes a new employee to adapt to current workplace health norms.
Secondly, as noted in the robustness analyses, the network effect is dependent on the construction of the weight matrix [
28]. We based the weight matrix on colleagues working in the same department. Not all employees within each department filled out the questionnaire and some employees were more likely to lack information on the eating and physical activity variables as we have shown. Especially in large departments, employees may not have had contact with all their colleagues. Data on complete networks would have been desirable. We addressed this potential shortcoming by examining several alternative specifications of the weight matrix, which provided mostly consistent results.
Thirdly, the measure of physical activity was very general and addressed any type of physical activity in the past week. This could range from moderate activities such as walking the dog or cycling to work to more extreme sports such as mountain biking or running half marathons. The fact that these activities may be very different could further explain why we did not find a network effect. Similarly, our measures for fruit and vegetable consumption could have been more informative, for example by highlighting how large a portion is or by following the WHO guideline to eat at least 400 g of fruit and vegetables per day [
55]. However, other studies, such as the European Social Survey [
43], used the same measures, making our findings comparable.
Fourthly, in focussing on the role of perceived encouragement and colleague behaviour, we left out other mechanisms that may explain how workplace social relations affect healthy eating and physical activity behaviours. For example, we have tended to focus on positive influences, but also processes of peer pressure, social control and stigmatisation (e.g. fat shaming) could impact the healthy choices employees make [
10,
56]. In addition, our measure for perceived encouragement was very general. A more detailed measure may provide richer insights. In order to paint a complete picture of the role colleagues may have in each other’s healthy habits, future studies should address this too.
Finally, some more information on the context and environment in which employees work together with their colleagues would have provided more insights. For example, it may be easier for colleagues to exercise together at work when there are showers present, or when the office is close to a park where they can go for a lunch walk together. We addressed this with the data we had available by including a control variable for WHP, but it would be good if future studies pay more attention to the environment. This may also provide insights in whether colleagues are more or less influential depending on the other options present for making healthy choices in the workplace.
Strengths and implications
Among our study’s strengths were the focus on both colleagues’ encouragement and actual behaviours, addressing encouragement specific to the behaviour rather than generic social support and examining behaviours that also take place outside the workplace. Furthermore, our study is one of the first to address the role of co-workers’ behaviours using a network approach incorporating direct colleagues. This allowed for a finer grained analysis than the aggregation of individual-level measures or relating employees who may not work in close proximity. This study thus represents an important first step, showing that it is promising for managers and public health policy makers to incorporate workplace social relations to promote healthy behaviours.
The implications of our study relate to the realisation that when designing health interventions, it is important to incorporate the social work environment alongside other social actors such as partners, family members, and friends. We showed that colleagues are relevant sources of social support when it comes to healthy behaviours and may act as role models. In stimulating employees to make healthy choices, organisations can make use of mentors or health champions, which are employees who have adopted a healthy lifestyle themselves and help their colleagues to do so too [
57]. For worksite health promotion activities, it is also important to draw on the positive influence colleagues could have in helping one another make healthy choices. Crucially, not only do colleague encouragement and behaviours contribute to creating a culture of health in the workplace [
23], they also indirectly support the entire work population, including those not using dedicated programmes in the workplace.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.