During the COVID-19 pandemic people experienced dramatic changes to their lives, social relationships and work: many people lost their jobs, some even lost their profession, many were furloughed, while many others had to radically change the way they worked (De Witte et al. 2021; Rudolph et al. 2021). The pandemic, it has been argued, can be understood as a shock—an uncontrollable, unforeseen event—that upset routinized ways of doing things and questioned established understandings of the world and ones’ place in it (Akkermans et al. 2020). The pandemic also made it clear that in absence of draconic enforcement, the success of any large-scale governmental intervention (like lockdowns, social distance regulation) depends on the creation of a sense of shared identity among citizens, a feeling of a shared fate with a wider larger collective, to ensure citizens’ continuing motivation to adapt their behavior (see Abrams et al. 2021; Reicher & Stott 2020; Van Bavel et al. 2020). Research on public (health) compliance coming forth since showed that if people feel they are part of a public collective, they are more likely act in the interest of the common good. This sense of social identity does not develop in a vacuum—it is affected by public messaging, public leadership (Bonell et al. 2020; Van Bavel et al. 2020)—and as we propose in this paper—peoples’ own experiences, also at work.
The present paper focusses on work-related experiences during the pandemic and investigates how impactful changes to work due to COVID-19 relates to peoples’ sense of identity as members of the working population. This is important to investigate, as previous research has shown that work-related experiences (e.g., an increase job insecurity) can potentially affect a person’s identification with wider societal entities, and through that affect wider societal attitudes and behavior (e.g., Selenko & De Witte 2021). Also, work is one of the foremost societal institutions, where people spend most time in their adult lives. Political scientists, economists and sociologists understand very well that the experiences that people make at work can affect people’s societal participation (e.g., Adman 2008; Elden 1981) but in work psychology this relationship is still under-investigated. For future public health planning, but also for policy makers interested in the broader societal consequences of working conditions, it is important to understand the role of work-related experiences for public behaviors outside work.
Drawing on the social identity approach (Tajfel 1982; Turner et al. 1987), and prior research on changes to work and their effect on identity (e.g., Ashforth & Schinoff 2016; Selenko et al. 2017; Selenko & De Witte 2021; Strich et al. 2021; Thatcher & Zhu 2006), we postulate that pandemic-related work changes plausibly affect and relate to citizens’ identity as members of the working population. Some of these changes were widely shared (e.g., the dramatic decrease in social contact at work), but others were more varied (e.g., changes in workload or autonomy; Syrek et al. 2022), and we anticipated that the specific nature of these changes can imply a potential threat—or boost—to work-related identity. According to the social identity approach, social category memberships offer norms that guide individual behavior along shared values (Tajfel 1982; Turner et al. 1987). Identification with a certain social group is expressed by showing empathy, support and mutual trust to fellow group members, for example in trying to protect them by adhering to compliance regulations (Reicher & Stott 2020).
Notably also, since social interactions were limited, most points of social transmission of the virus occurred though spaces of work: Either at one’s own (if not working at home), or at the working spaces of others, such as shopkeepers, transport officials, hospital workers or others. Being a member of the working population was hence more relevant for viral spread, than being ‘just’ a member of society. Consequently, feeling a sense of collective identity as a member of the general working population is likely to have impacted on people’s willingness to comply with measures that protect others and follow regulations that attempt to curb the spread of the virus.
This study aims to make several contributions. It will be the first to investigate a variety of COVID-19 related work changes and their effect on people’s identification with the working population. Understanding better which work-related changes foster or threaten this aspect of identity, prepares the ground for more effective public health messaging that aims to evoke collective identities (Bonell et al. 2020). Secondly, this study gathers empirical evidence for the link between a particular form of collective, social identity (identification as a member of the working population) and compliance with coronavirus mitigation measures. ‘Identification with the working population’ describes that part of people’s social identity where they define themselves as a member of the most inclusive work-related group, the working population. Much of the theoretical and policy recommendations on behavioral change in the pandemic drew on social science research conducted under different circumstances at a different time (e.g., in laboratory studies) and remained untested in the context of the pandemic (e.g., Cruwys et al., 2020; Jetten et al., 2020; Van Bavel et al., 2020). Even COVID-19 compliance studies that include individual and socio-economic factors (e.g., minority background, Goren et al. 2021, working-memory capacity, Xie et al. 2020, personality, trust in the government and health beliefs, Clark et al. 2020) tend to ignore the individual’s employment experiences. This seems short-sighted, given that work plays a central role in people’s lives and often informs how people feel, think and act toward society (Mutz & Mondak 2006; Stanojević 2021). Harnessing people’s identification with work-related groups is an untapped avenue when it comes to calling on collective, overarching group identities in public health messaging.
Thirdly, beyond the pandemic contexts, this study brings new knowledge on the link between individually experienced work-related changes and wider societal behaviors, by highlighting the role of identity (e.g., see also: Selenko & De Witte 2021; Van Hootegem et al. 2022). Particularly now, with the next major work-related change just happening, the lessons from this study seem pressing: The rapid implementation of generative AI tools at work brings radical changes to work, to concepts of expertise, and to occupations (see Selenko et al. 2022). We believe this study, albeit conducted in a unique context, can still be very informative for situations of major work-related changes and understanding their societal implications. Work, as we understand it, is a social incubator: it is the place where people most likely intersect with different people of different social backgrounds; and experiences made at work can serve as sources of information about one’s place in society (Mutz & Mondak 2006).
COVID-19 Pandemic-Related Work Changes and the Identity as a Member of the Working Population
Work is an important part of who we are. It provides access to important social functions (e.g., Jahoda 1992, Warr 2013) and enables people to feel part of something bigger and gain an understanding of one’s place in the world (Ashforth & Schinoff 2016; Hulin 2002). According to the social identity approach to work the work arena consists of numerous social categories, that can inform a person’s social identity (Selenko et al. 2017). People define themselves as members of an organization, a team, an occupation, and even, most inclusively, as members of the working population (encompassing everyone of working age).
Peoples’ work-related identities have been explored from a variety of different theoretical perspectives, proposing different mechanisms of identity formation: according to the social identity approach the cognitive perception of group memberships stands central for identity formation (e.g., Turner et al. 1987). Management research then also added that actual social interactions, regularly enacting work tasks, getting appreciation for them and building self-narratives around those enactments are equally influential for identity formation (Ashforth & Schinoff 2016; Pratt et al. 2024). We believe the COVID-19 pandemic and its work-related changes are likely to have affected the cognitive and behavioral processes that are relevant for identity formation.
In the work arena, pandemic-induced governmental and organisational measures affected work at the task level (e.g., skill variety, autonomy), at the level of the organization of work (e.g., role clarity, demands), at the interpersonal level (e.g., relationships and interactions with co-workers, supervisors, clients), and the organisational level (e.g., pay, contract, job insecurity) (see, e.g., Rudolph et al., 2021; Bakker et al., 2003 for a similar structuring). In this study, we focus on all four of these levels by looking at changes in autonomy, changes in demands, changes in the social interactions at work, and perceived job insecurity at the time. Workload, social interaction, decisional freedom and job insecurity are key factors in predicting stress at work (Karasek & Theorell 1990; Van der Doef & Maes 1999), and were also impacted by the pandemic (see, e.g., Rudolph et al. 2021; Syrek et al. 2022). These work-related changes are likely to have affected multiple work-related identities; in this study we are specifically interested in their relationship with the broad, societally related identity, people’s identification with the working population, as we deem this to be a central link to societally relevant health behaviors.
Changes in social interactions
One of the most pronounced shared experiences during the pandemic were changes in social interactions at work. As was their intention, the implemented lockdown and social distancing measures limited the physical spaces for social interactions. Representative surveys for example show that about half of all US workers worked from home during the pandemic (Brynjolfsson et al. 2020); the same was the case in other countries (e.g., Office for National Statistics 2021). For office workers, working from home limited social interactions to pre-scheduled meetings on virtual platforms, taking away numerous daily serendipitous encounters in places of work and associated places (e.g., with colleagues from other teams, fellow commuters, people working in coffeeshops).
Seen through an identity lens, many people had fewer chances to enact certain work-related identities (e.g., their identities as commuters, as office mates, etc.). People’s identities are informed by what they regularly do, what they enact, the appreciation they receive for what they are doing, the narratives they build around that enactment, which creates schemas and self-understandings over time (e.g., Ashforth & Schinoff 2016; Pratt et al. 2024). Working from home hence loosened peoples’ ‘conventional moorings’ with an organization, which makes it more difficult to keep the identification with one’s organization and occupations salient in one’s mind (see Ashforth 2020, 1764). Weakened organisational and occupational identities might have encouraged people to search for meaning elsewhere, which has been argued to have led to the ‘great resignation’ following in the aftermath of the pandemic (Xu et al. 2023).
Cognitively, these changes to social interactions at work constituted a shared national experience: Governmentally enforced social distancing measures (such as the closing of offices, the restriction put on social interactions at work), were imposed upon the entire working population within a country; the objective legal framework was the same for everyone (although how it was implemented varied across professions and organizations). Research on disasters shows that the shared experience of a national emergency or a disaster can evoke strong feelings of unity among the affected people and act as a powerful source of an ‘emergent sociality,’ a feeling of togetherness (Drury et al. 2009). In addition, the national public messaging at that time very much stressed the element of a ‘shared experience,’ people were made aware that they were ‘in it together.’
So paradoxically, while social distancing measures are likely to have reduced identifications with organizations, occupations and work roles (through affecting their enactment); they might have fostered the identification as a member of the broader, general working population (through the perception of a shared experience). It is even possible that in times of weakened organisational and occupational identities, which can be understood as ‘sub-group’ identities to the group of the working population, people turn to identifying with the ‘super-ordinate’ group (in this case: the working population), as experimental studies in social psychology show (Jung et al. 2019).
We hence propose Hypothesis 1: People who experienced a decrease in social interactions at work because of the pandemic, will report a stronger identification with the working population.
Changes in workload
The pandemic also affected workload: Many people saw their workload increase dramatically (e.g., health-care workers in ICUs), while others saw it dramatically lower (e.g., taxi drivers), while others again saw milder changes or specific shifts in tasks (e.g., more of the one task, less of the other) (Rudolph et al. 2021), and in many professions workload went initially down to go up again (Syrek et al. 2022). These changes were often augmented by pandemic-induced changes to family life: Remote working parents for example, had to cope with changes to children’s education as well, which created work-life balance struggles and contributed to exhaustion (Calderwood et al. 2022).
While there was no shared pattern of workload change, it is likely that an increase or a decrease of workload will have different meanings for work-related identity. An increase in workload requires more psychological involvement and a repeated re-enactment of certain work tasks. This helps with the internalization of associated occupational and organisational identities and strengthen them—the more you do something, the more you become it (e.g., Ashforth, Harrison, & Corley 2008; Ashforth & Schinoff, 2016). From an identity formation perspective (Pratt et al. 2024), working more means someone does more of the one activity that is the defining characteristic of the working population (namely, working), irrespective what the specific work looks like. Also, a growth in workload carries cognitive meaning: it signals that one’s job is in high demand in the organization. More widely, it also signals that someone is needed and valued in the working population, in contrast to others—who were furloughed or became unemployed. An increase in workload is therefore likely to strengthen the identification with the working population. Indeed, there is evidence that shows that in-role work performance is positively related to the identification as a member of the working population (Selenko et al. 2017).
Hypothesis 2 consequently states: People who report an increase in workload because of the pandemic, will report more identification with the working population than people who report a decrease.
Changes to autonomy
The pandemic also affected decisional freedom at work. People working from home reportedly experienced more flexibility in terms of scheduling their own work, although this was limited to how co-dependent work tasks were, and varied over time (Syrek et al. 2022). Surveys of people working from home show that some experienced an increase in their autonomy, while others noticed a decrease (De Witte & Van Hootegem 2020). Again, a prototypical pattern was difficult to make out.
Depending on whether autonomy increases or decreases, different consequences for the identity as a member of the working population are expected. Having more autonomy in a job is likely to be an identity enhancing experience: if someone has decisional freedom over work-processes, they are more likely to perceive a sense of ownership over their work, which in turn is closely related to work-related identification (Avey et al. 2009). Interestingly, political scientists have suggested (and found evidence) for a relationship between autonomy at work and political participation outside work (see Pateman 1970).
Also, decisional freedom comes with power, and an awareness that the organization respects a person’s individual capacities to take decisions. This will not only make an individual feel that they belong more to their work, but also it allows them to act more in according to their own personal needs and plans—in line with who they are, to their own work-related identity. In other words, autonomy might satisfy the identity motive of belonging and also offer more opportunity for identity enactment (Ashforth & Schinoff 2016). Research among hospital physicians confirms that more autonomy in decision making is associated with stronger identification with the hospital (Salvatore et al. 2018).
Hypothesis 3 therefore proposes: A perceived increase in work autonomy due to the pandemic is associated with more identification with the working population.
Job insecurity
One predominant concern during the COVID-19 pandemic was people’s worry about the continued existence of their job (e.g., Wilson et al. 2020). In social identity terms, job insecurity entails the worry to lose the membership of a valued group—the group of the employed people, and become a member of a stigmatized group—the group of unemployed people. Both these instances can be understood as a type of identity threat, to the identity as an employed person, to organisational identity and probably many more forms of work-related identity (Branscombe et al. 1999). Although unemployed people are still members of the working population, they are generally reporting to have low status, feeling looked down upon, feeling as if ‘on the scrapheap’ (Jahoda 1982). Job insecurity turns this unwanted reality into a potential future for employed people; they might feel already as ‘half-way out the door,’ no longer being a typical member of an (otherwise well) organization or a typical member of the working population. There is already some empirical support for this notion, from longitudinal survey studies among employed people: job insecurity negatively affects peoples’ feeling of identification with the working population over time, even when controlling for previous time points (e.g., Selenko et al. 2017; Selenko & De Witte 2021). In line with the existing evidence, we expect quantitative job insecurity to be negatively related to the identification with the working population.
Hypothesis 4 therefore states: People with more quantitative job insecurity will feel less identification with the working population.
Identification with the Working Population as a Predictor of COVID-19 Compliance
In this study we argue that the identification as a member of the working population plays a significant role for health compliance. We draw on the most recent literature in public health research, which recognizes a sense of overarching, collective identity as being of central importance for citizens’ willingness to comply with governmentally imposed coronavirus regulations (Van Bavel et al. 2020; Neville et al. 2021). Such so-called ‘shared’1 identities entail a concern for others of the same group, which makes it likely that people comply with protective measures of the group and individuals that are part of the group (Khan et al. 2015; Reicher & Stott 2020). Moreover, shared identities entail shared values, making it more likely that people who share an identity will conform to the norms of that identity group (Neville et al. 2021). Furthermore, being part of a social group also makes people feel capable and in control, which in addition might be associated with more compliance behavior (Greenaway et al. 2016).
In this study we propose that the identification as a member of the working population is a prominent avenue toward this overarching identity. Not only is ‘work’ the activity adults spend the most time with, it is also the place where people most likely meet and interact with people different to themselves or their friends and family (Hulin 2002; Mutz & Mondak 2006). From a more sociological point of view, work is understood as a social institution (Jahoda 1992) that structures peoples’ position in society. Very plausibly hence feeling part of this social institution will reflect people’s sense of identification with the working population in general. In times of the pandemic, people more likely define themselves as member of that all-inclusive superordinate identity, which will inform their actions. This leads to the fifth hypothesis:
Hypothesis 5: People who identify more with the working population will report more compliance with COVID-related regulations imposed by the government.
Taking these mechanisms together, we propose that work-related changes due to the COVID-19 pandemic will indirectly relate to compliance with health measures aimed at curbing the virus, through the identification with the working population. A perceived decrease in social contacts, and a perceived increase in workload and autonomy will be related to more identification with the working population, which in turn would relate to more compliance with COVID-19 regulations. We predict the opposite for job insecurity: previous research has shown that an increase in job insecurity can lead to less identification over time (Selenko et al. 2017). We expect to find this effect also reflected cross-sectionally, in that people who are more job insecure would report less identification and consequently less compliance. Job insecurity and health compliance have been negatively connected in the pandemic: In a study across different US federal states, Probst and colleagues (2020) found that job and economic insecurity were negatively related to compliance with the US Center for Disease Control’s COVID health guidelines. They also found this relationship to be moderated by state-level variations in the generosity of furlough and unemployment benefits. Although their study did not investigate any explanatory mechanisms, Probst et al. (2020) proposed that job and economic insecurity would cause cognitive overload and mental tunnelling, which would undermine the capacity to adhere to health guidance. We take a different approach in this study, by suggesting that the willingness to comply to public health guidance is not solely due to the individual’s cognitive capacities, but rather a decision taken in a social context.
The final hypothesis (Hypothesis 6) therefore states: Identification with the working population will act as a mediator between job related changes (changes to social interactions, autonomy, and workload) and job insecurity on the one side and compliance behavior on the other.
Method
A multi-country online survey among a sample of employees, collected in Belgium, the UK and the US in July 2020 serves as the basis for the hypotheses tests. All three countries were strongly affected by the pandemic, with lock down measures in place in 2020 and fairly comparable health guidance. The period when the data of this study was collected fell in July 2020, just between what turned out to be lockdown one and lockdown two (in autumn 2020). In July, however, lockdown measures in all three countries eased a bit, and smaller gatherings between people outside the private household were being allowed: In the US, states varied to the degree that they reopened (Centers for Disease Control and Prevention, CDC, 2023), in the UK most lockdown restrictions were lifted in July 2020, with hospitality reopening and small gatherings of people allowed (Brown & Kirk-Wade 2021), in Belgium also small gatherings were allowed and sporting and cultural events reintroduced (Wikipedia 2022). In all three countries mask mandates were still widely in place, people were strongly encouraged to work from home where possible, social distancing rules (e.g., no handshakes) were still in place.
Respondents were recruited with the help of a panel-survey company (Respondi), who targeted employed adults, over 18 years old in all three countries. The survey was hosted on an independent platform (Qualtrics) by the authors of this paper; separate but identical surveys were designed for each of the three countries. The survey in Belgium was conducted in the Flemish-speaking part of this country, and the survey was translated into Dutch by the authors. The other two surveys were conducted in American- and British English, respectively. Participation in this study was entirely voluntary and anonymous and the data collection was ethically approved for in each country by the universities of the author team.
In total 453 (428) employees from the UK, 518 (474) from the US and 460 (420) Belgian employees participated (numbers in brackets indicate the final sample size without missing values on the variables of interest in this study; which is also the sample described subsequently). We do not have any reason to assume that the hypotheses would work differently in the three countries, hence for the purpose of this analysis, we merge these samples for more power, while controlling for nation of origin. To ensure this is sensible to do, we tested all measurement instruments for measurement equality across the three countries (see below).
Across the countries, 47.9% of respondents were women and on average 46.50 years old (SD = 11.74). Most respondents held only one job (93.2%) and reported to work full-time (74.6%), on a permanent contract (93.3%). People worked on average 36.60 hrs a week (SD = 10.27). They worked in a variety of organizations, with private organizations (53.9%) being most frequently mentioned, followed by public sector (excluding central or local government) (19.7%) and central and local government (9.3%). As for job roles measured according to the ILO’s ISCO-88 industry classification system, 50.9% of respondents worked in high skilled white collar professions (managerial roles, professionals or technicians and associate professionals), 31.8% worked as low-skilled white collar professions (e.g., clerical support workers, sales and service workers), 5.9% worked in high-skilled blue collar professions (e.g., skilled agricultural worker, craft or trades worker) and 8.9% were in low-skilled blue collar professions (e.g., plant or machine operators, and elementary occupations). There were also nine people who worked in the armed forces. People worked on average 12.71 years (SD = 10.57) in their job.
Measures
COVID-19 related work changes. We assessed perceived change in social interaction, workload and decisional freedom in their job with a single item each. Each of the item was introduced ‘Since the COVID-19 crisis started…’ and then participants had to indicate on a five-point scale ‘the social interaction in my job has…’, ‘my workload has…’ and ‘the extent to which I can make decisions in my job’ (1 decreased greatly; 2 decreased slightly; 3 stayed the same; 4 increased slightly; 5 increased greatly). Table 1 provides the means and standard deviations of these three items. While workload change and decision-making change were relatively normally distributed, social contacts was quite skewed. In line with what could be expected, almost two thirds of all respondents indicated that their social contacts at work either decreased greatly (37.3%) or slightly (27.8%), whereas only very few said that it increased slightly (4.4%) or greatly (1.8%)—indicating that a decrease in social contact was indeed a prototypical, shared experience, that people probably were aware of.
Table 1
Pearson correlations of work-changes due to COVID, shared identity and complains with COVID-19 health regulations.
| M | SD | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 12 | 13 | 14 | 15 | 16 | 17 | 18 | 19 | |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 1. Age | 46.50 | 11.74 | — | ||||||||||||||||||
| 2. Gender | .48 | .50 | –.14** | — | |||||||||||||||||
| 3. Org D1 | .09 | .29 | .03 | –.03 | — | ||||||||||||||||
| 4. Org D2 | .25 | .43 | –.03 | .17** | –.18** | — | |||||||||||||||
| 5. Org D3 | .60 | .49 | .04 | –.14** | –.39** | –.71** | — | ||||||||||||||
| 6. Occ D1 | .51 | .50 | –.04 | –.04 | .05 | .13** | –.12** | — | |||||||||||||
| 7. Occ D2 | .32 | .47 | .01 | .17** | .02 | –.05 | .00 | –.73** | — | ||||||||||||
| 8. Occ D3 | .06 | .24 | .04 | –.12** | –.08** | –.09** | .13** | –.27** | –.18** | — | |||||||||||
| 9. Full-time | .75 | .44 | –.07* | –.09** | .10** | .00 | –.06* | .11** | –.10** | –.01 | — | ||||||||||
| 10. Contract type | .93 | .25 | .16** | –.02 | –.00 | –.08** | .07** | .01 | .01 | –.04 | .19** | — | |||||||||
| 11. Tenure | 12.71 | 10.57 | .45** | –.09** | .08** | .05 | –.08** | .05 | –.08** | .05 | .08** | .17** | — | ||||||||
| 12. Hrs worked | 36.60 | 10.27 | .00 | –.08** | .01 | –.05 | .06* | .07* | –.05 | .01 | .33** | .08** | .03 | — | |||||||
| 13. Job change | .08 | .28 | .12** | –.04 | –.01 | –.04 | .06* | .02 | –.03 | .00 | .05 | .20** | .17** | .03 | — | ||||||
| 14. Quant. JI | 2.09 | .98 | –.12** | .04 | –.10** | –.09** | .11** | –.05 | .03 | .03 | –.06* | –.16** | –.14** | –.02 | –.13** | — | |||||
| 15. Qual. JI | 2.72 | 1.04 | –.12** | .08** | –.07* | .10** | –.07** | –.01 | –.01 | –.01 | .02 | –.11** | –.05 | –.01 | –.11** | .64** | — | ||||
| 16. COVID social contacts change | 2.06 | 1.00 | –.04 | –.07* | –.09** | –.02 | .08** | –.11** | .05 | .07** | .02 | –.04 | .00 | .00 | –.04 | –.05 | –.07** | — | |||
| 17. COVID workload change | 3.04 | 1.13 | –.11** | .04 | –.04 | .04 | .00 | .07** | –.02 | –.08** | .19** | .03 | –.02 | .10** | .00 | –.08** | –.02 | .27** | — | ||
| 18. COVID decision making change | 2.94 | .72 | –.05 | –.11** | –.02 | –.07* | .07* | .07* | –.05 | –.03 | .08** | .06* | –.03 | .02 | .00 | –.08** | –.14** | .35** | .30** | — | |
| 19. Identification w the working pop | 4.08 | .72 | .13** | .03 | .03 | .01 | .02 | .04 | –.01 | –.02 | .10** | .10** | .06* | .09** | .07** | –.14** | –.11** | –.06* | .05 | .04 | — |
| 20. Compliance w COVID measures | 3.37 | .54 | .10** | .10** | .02 | –.02 | .02 | .09** | –.02 | –.03 | –.03 | .02 | .06* | .00 | .02 | .04 | .04 | –.22** | –.05 | –.05 | .14** |
[i] Note. Gender (1 = woman, 0 = man), Org D1 = local government employee (1 = yes, 0 = no), Org D2 = public sector employee (1 = yes, 0 = no), Org D3 = private sector employee (1 = yes, 0 = no), Occ D1 = high skilled white collar job (1 = yes, 0 = no), Occ D2 = low skilled white collar job (1 = yes, 0 = no), Occ D3 = high skilled blue collar job (1 = yes, 0 = no), full-time ‘works full-time’ (1 = yes, 0 = no), Contract type (1 = permanent, 0 = not permanent), Job change ‘Job change due to the pandemic’ (1 = yes, 0 = no).
*p < .05, **p < .01.
Identity a member of the working population was measured with an adapted scale by Doosje et al. (1995) (see also Selenko et al. 2017; Selenko & De Witte 2021). This measure consists of four items, asking people to indicate their disagreement or agreement with items like ‘I see myself as a part of the working population’ on a five-point Likert scale (higher values indicating higher identification). This scale showed good reliability, Cronbach’s alpha = .87.
Compliance with COVID-19 public health measures. We used a shortened version of Nivette and colleagues’ (2021) checklist of public health recommendations by the World Health Organisation (WHO) and CDC. Specifically, we asked respondents to indicate to which degree they followed five public health recommendations, that were the same across all three countries: (1) adhere to social distancing, (2) avoid contact with people at risk, (3) avoid groups, (4) don’t shake hands, and (5) stay at home. These guidelines were the same in Belgium, the UK and the US. People had to indicate how often they followed these recommendations (1 = Never to 4 = All of the time), the five items scale showed good reliability, Cronbach’s alpha = .76.
Job insecurity
To measure quantitative job insecurity people had to indicate their agreement or disagreement on a five-point Likert scale to four items, such as ‘I feel insecure about the future of my job’ taken from Vander Elst et al. (2014). The scales was highly reliable, Cronbach’s alphaquantitative JI = .90.
Control variables
We control for pandemic-related job change because this is presumably the most dramatic change a person could undergo and entails all kind of work-related changes. People were asked to indicate whether they started a different job in a different organization, since the COVID-19 crisis started. This was the case for 8.7% of all respondents.
We also control for qualitative job insecurity, which concerns peoples worries about unwarranted changes to aspects of their job. Naturally, pandemic related changes to work are likely to alert people to further potential changes to their work in the future, potentially worrying them. In order to ensure that we are truly assessing job change, but not the worry about future job changes, we included qualitative job insecurity as a control variable. For qualitative job insecurity, we used the four-item scale by Fischmann et al. (2022), which includes items like ‘I think my job will change for the worse.’ Cronbach’s alphaqualitative JI = .92.
We also include trust in the organization as an instrumental variable, to allow us to test our results against endogeneity biases (Antonakis et al. 2010; Semadeni et al. 2014). Trust was measured with four items of the scale by Robinson (1996), to which participants had to give their agreement on a five-point scale (1 = strongly disagree, 5 = strongly agree). An example item would be ‘I believe my employer has high integrity.’ The scale was highly reliable, Cronbach’s alpha = .89.
Analysis strategy
The analysis proceeds in three stages. First, measurement equivalence is tested across the three groups. This is to ensure that the measures function in the same way across the three countries and two language groups and that combining the data is warranted. Secondly, the hypotheses were tested using Structural Equation Modelling including a bootstrapping of indirect effects in MPlus. Finally, to test for the robustness of our effects and rule out endogeneity effects and control for reverse causality, a two-stage last-squares (2SLS) instrumental variable estimation is employed in SPSS (Antonakis et al. 2010).
Test of measurement equivalence
Given that this study was carried out in three different countries and two different language groups, a test for measurement invariance is essential. To do so, several differently restrained measurement models were compared against each other, using MPlus. A configural model (separate factor loadings, separate intercepts and variances for each country) fit well to the data, Chi2 (336) = 689.972, p < .001, RMSEA = .047, CFI = .976, TLI = .971, SRMR = .043 (Vandenberg & Lance 2000); particularly after allowing two items of identification with the working-population correlate for all groups. This model got gradually worse by fitting factor loadings equal (metric invariance), ΔChi2 (26) = 90.254, p < .001, RMSEA = .049, CFI = .972, TLI = .968, SRMR = .053; and intercepts equal (scalar invariance), ΔChi2 (34) = 268.131, p < .001, RMSEA = .059, CFI = .956, TLI = .955, SRMR = .065. Although the differences between the configural, metric and scalar model were significant, they don’t offer reason for serious concern: The difference in CFI indices between configural and metric model is quite negligible (0.004) and offers no reason to reject the invariance hypothesis (Vandenberg & Lance 2000). The difference between scalar and metric model is larger (ΔCFI = 0.016) and can be explained with differences in intercepts in the Belgian sample. By freeing the intercepts of five items in the Belgian sample, the difference in model fit of the scalar model to the metric model became acceptable ΔChi2 (29) = 144.574, p < .001, RMSEA = .053, CFI = .964, TLI = .962, SRMR = .061. Not achieving full invariance is a rather common occurrence, and freeing intercepts to achieve partial scalar invariance is seen as an acceptable practice, as long as the majority of items on a factor are still invariant (Putnick & Bornstein 2016). In our case the freeing of those five items that there seem to be systematic differences in the Belgian responses, when compared to the US and UK sample, despite the same underlying latent trait. This might have to do with language differences, as the Belgian study was conducted in Flemish rather than English. In sum, these invariance tests confirm that items were understood similarly across the three country samples, giving reason to merge the three samples. To control for intercept differences, nation was included as a dummy variable in the analysis.
Results
Table 1 presents the correlations of the variables of interest and all demographic variables. It suggests an initial support of some of the hypotheses: the more people experienced a decrease in social contacts, the more they felt part of the working population. Also, identification with the working population correlated positively with compliance to COVID measures.
To test our hypotheses, first, a measurement model was constructed in a similar fashion as described in the measurement invariance tests, this time not differentiating between the country samples. This model which included compliance to COVID-19 regulations, identification with the working population and qualitative and quantitative job insecurity and fits well to the data Chi2 (112) = 326.066, p < .001; RMSEA = 0.037, CFI = 0.978, TLI = 0.974, SRMR = 0.036. It fits significantly better than a one factor model, ΔChi2 (6) = 4670.598, p < .001 or a two-factor model that merges qualitative and quantitative job insecurity; ΔChi2 (3) = 1594.199, p < .001.
The measurement model was then remodelled into a structural model by adding the single item measures of COVID-19 related work change (change in social interaction at work, change to workload, change in decision making capacity), the quantitative job insecurity measure and the two nation dummy variables as well as the job change and qualitative job insecurity control variable. COVID-19 related work change, job insecurity and the control variables were modeled to affect the latent identification variable. Identification with the working population was in turn modeled to affect the compliance variable; also direct effects of COVID job change, job insecurity and the control variables were included. Maximum likelihood was used as the estimation method. The model fit well to the data, Chi2 (202) = 662.436, p < .001, RMSEA = .041, CFI = .959, TLI = 0.951, SRMR = .042.
Hypotheses testing
Table 2 gives an overview over the estimated model coefficients predicting identification with the working population. As expected, reporting fewer interactions at work due to COVID-19 was related to a higher identification with the working population, thereby confirming Hypothesis 1. Also, experiencing an increase in workload was associated with more identification as postulated by Hypothesis 2, whereas people with higher job insecurity reported less identification with the working population (as predicted by Hypothesis 4). A change in autonomy was not related to identification, Hypothesis 3 was not supported. Of the control variables, being from Belgium (the comparison country of the two country variables) was associated with lower identification, qualitative job insecurity did not matter.
Table 2
Structural Equation Model unstandardized estimates of changes in work predicting shared identity with the working population and COVID Compliance controlling for nationality and job insecurity.
| IDENTIFICATION WITH THE WORKING POPULATION | COVID COMPLIANCE | |||||
|---|---|---|---|---|---|---|
| B | SEB | p | B | SEB | p | |
| UK | .183 | .034 | <.001 | .195 | .037 | <.001 |
| US | .179 | .031 | <.001 | .104 | .038 | .006 |
| Qualitative job insecurity | –.012 | .021 | .564 | .006 | .025 | .823 |
| Change of job | .028 | .037 | .460 | .055 | .049 | .260 |
| Change in social contact | –.033 | .014 | .020 | –.101 | .017 | <.001 |
| Change in workload | .034 | .012 | .005 | .004 | .014 | .775 |
| Change in decision making | .004 | .021 | .842 | .001 | .023 | .950 |
| Quantitative job insecurity | –.061 | .022 | .006 | .007 | .023 | .764 |
| Identification with the working population | .174 | .044 | <.001 | |||
| R2 | 0.062 | 0.099 | ||||
| Est/SE Est | 4.587** | 5.283** | ||||
[i] Note. **p < .001.
In relation to COVID-19 health compliance behavior, in line with Hypothesis 5, feeling a stronger identification with the working population was associated with more compliance. There was also a direct effect of reduced social interaction at work on the compliance with COVID-19 public health measures in the SEM (this effect, however, was not significant in the instrumental variable estimation or the normal OLS regression, and therefore not robust). In this study, respondents from the UK and US both showed more compliance than respondents from Belgium.
A test of the indirect effects showed that an increase in workload had a positive indirect relationship with compliance with COVID-19 health measures (B = .006, SEB = .002, p = .017), and that a decrease in social contacts had a negative indirect relationship with COVID compliance through identification with the working population (B = –.006, SEB = .003, p = .036). There was also a negative indirect relationship between job insecurity (quantitative) and health compliance (BE = –.011, SEB = .005, p = .02), as would be expected. Hypothesis 6 (the mediation hypothesis) was hence supported for social contacts, quantitative job insecurity and workload, but not for autonomy.
Instrumental variable estimation of the identity to health compliance relationship
We conduct a 2-Stage-Least-Squares (2SLS) instrumental variable estimation to address endogeneity issues associated with one-wave data. This method can address possible common method bias and reverse causality issues that can arise from a one-wave study design. It thus is a statistical method to recover causal parameters and functions to validate the results of the other analyses (see e.g. Antonakis et al. 2010; Bollen 2012). The challenge of 2SLS estimations in general is to find suitable instrumental variables. In order to meet the criteria for a 2SLS estimation, instrumental variables need to be exogenous (uncorrelated with the dependent variable) and they need to be relevant (correlated with the independent variable) in a theoretical and statistical sense (Semadeni et al. 2014).
In this study we used trust in the organization as an instrumental variable. Organisational trust and organisational identification are conceptually closely related and understood to be different types of psychological attachment to the organization (Ng 2015). Given this close relationship, it is likely that high organisational trust might also spill over into stronger identification with the working population; however, there is no meaningful reason (or evidence) why organisational trust would be related to COVID compliance. Statistically, in our study organisational trust and identification with the working population were correlated significantly (r = .147, p < .001). It was also correlated with COVID compliance (r = 086, p = .002), but this became non-significant, once identification was taken into account (rpartial = .049, p = .076), indicating that the zero-order correlation was solely through its association with identification with the working population.
The 2SLS estimation happens in two stages: In stage 1, the endogenous (predictor) variable (here: identification with the working population) is regressed on the instrument (here: trust in the organization) and all covariates of the equation (here: all work-related changes, job insecurity, and nation dummies). This generates predicted values for identification with the working population. In the second stage, these predicted values are used as predictors of the dependent variable (compliance with health regulations), together with all covariates. As can be seen in Table 3, the instrumental variable predicted identification significantly (1st stage estimation) and the F-statistic being larger than 10 also suggests that it was a strong instrument, making the 2SLS reliable (Stock et al. 2002). Stage 2 of the regression shows that identification had still a significant relationship with compliance. This indicates that even when controlling for potential endogeneity, the effect is still there.
Table 3
Results of a 2SLS regression and OLS regression of identification with the working population on COVID compliance.
| 2SLS –1st STAGE | 2SLS –2nd STAGE | OLS | |||||||
|---|---|---|---|---|---|---|---|---|---|
| B | SEB | p | B | SEB | p | B | SEB | p | |
| UK | 0.278 | 0.047 | <.001 | 0.135 | 0.045 | .003 | 0.200 | 0.037 | <.001 |
| US | 0.209 | 0.046 | <.001 | 0.047 | 0.042 | .262 | 0.099 | 0.035 | .005 |
| Qualitative Job insecurity | 0.035 | 0.024 | .151 | 0.024 | 0.018 | .184 | 0.020 | 0.018 | .266 |
| Change of job | –0.159 | 0.068 | .019 | –0.012 | 0.054 | .826 | –0.046 | 0.052 | .378 |
| Change in social contact | –0.057 | 0.020 | .005 | –0.102 | 0.017 | <.001 | –0.116 | 0.016 | <.001 |
| Change in workload | 0.065 | 0.018 | <.001 | –0.002 | 0.015 | .898 | 0.010 | 0.014 | .478 |
| Change in decision making | 0.004 | 0.029 | .882 | 0.015 | 0.022 | .500 | 0.021 | 0.022 | .344 |
| Quantitative Job insecurity | –0.083 | 0.025 | <.001 | 0.026 | 0.021 | .206 | 0.007 | 0.019 | .720 |
| Identification with the working population | 0.296 | 0.088 | .001 | 0.086 | 0.021 | <.001 | |||
| Trust in the Organization (Instrument) | 0.190 | 0.022 | <.001 | ||||||
| R2 | .117 | .084 | .090 | ||||||
| F (9,1311) | 19.287** | 13.510** | 14.429** | ||||||
[i] Note. **p < .001.
In terms of size, it is notable that the coefficient for identification with the working population is larger in size in the second stage of the 2SLS than in the ordinary the OLS estimation (see last column in Table 3) or the estimation of the coefficient through SEM (see Table 2). This is unusual but not implausible: it indicates an underestimation of the effect due to endogeneity influences. For example, not included effects in this study, such as general concerns about the nation, might have affected the relationship between identification and health compliance, and the 2SLS controls for that.
Discussion
This study tested whether pandemic-induced work changes relate to a change in people’s identification with the working population, and through that, relate to their compliance with COVID-19 mitigation measures (Van Bavel et al. 2020). It is the first study to show that a higher identification with the working population was indeed associated with more compliance to COVID-19 health regulations. People who felt that they were more a part of the working population also behaved more in line with public health guidance and were therefore more protective of their fellow citizens. The importance of such an overarching identification with a collective for mutually supportive behaviors has been illustrated before in other contexts such as mass gatherings (Khan et al. 2015) or crowd solidarity after disasters (Drury et al. 2009). It is also assumed to have played a key-role in the adherence to COVID-19 guidance in the pandemic context (Bonell et al. 2020; Van Bavel et al. 2020), and the present study finally delivers empirical support for this assumption on the individual level in a sample of regular working people in three nations. This effect holds, even after controlling for endogeneity using a 2SLS equation. This finding is even more crucial now that countries have moved out of the pandemic, and great emphasis is put on individual’s choice when it comes to responsible health behavior, rather than policing or regulating.
The second major contribution of this study is that it illustrates how work-related changes are related to wider societally relevant behavior, by highlighting the role of identity as a mechanism. What we found in this study in the context of the COVID-19 pandemic was that work-related changes mattered differently for people’s identification as members of the working population. A change in social contacts, a change in workload, and feelings job insecurity were found to be important for working population identification, and through that had an indirect relationship with health compliance behaviors. A change in autonomy, however, did not play a role. This is surprising, as political scientists have long suggested that autonomy at work would be central to political engagement and activity outside work (e.g., Pateman 1970). Perhaps this is due to the sample characteristics in this study: most people in this study were white-collar workers (81.3%), who typically have a higher degree of autonomy; and experiencing more autonomy on top of an already quite autonomous job, or slightly less autonomy might not make a noticeable positive difference.
The changes in social contacts and workload and their relationship with identification, as well as the relationship between identification and job insecurity were in line with the expectations: We found the paradoxical (but expected) effect that having less social contact made one feel more part of the working people in one’s country, which might signal the proposed shared experience effect (Drury et al. 2009). More research is needed to establish the true nature of this mechanism. Work psychological research already identified the relevance of organisational climate for work outcomes, our study seems to indicate that it is also the sharedness of the wider, societal context that matters. We also found that an increase in workload was related to more identification as a member of the working population, as proposed. Future studies might want to explore the specificities of this effect, as workload is a multifaceted construct, often also ‘reflecting the amount or difficulty of one’s work’ (Bowling et al. 2015, 96), which might have separate consequences on identity and well-being. Finally, as expected there was a negative relationship between job insecurity and identification with the working population, as has been found in previous studies also over time (e.g., Selenko et al. 2017; Selenko & De Witte 2021). Building on the findings by Probst et al. (2020), we found that the relationship between quantitative job insecurity and health compliance was partially mediated by identification with the working population, suggesting a plausible explanatory mechanism.
In sum, we see support for the link between work changes, identification with the working population and compliance with COVID-19 health measures. This supports the notion that if work-changes can trigger identity processes to do with wider societal belonging, they also will have consequences for wider societal behavior.
Limitations
The biggest caveat to the findings of our study is its unique context, set right between two major lockdowns at the height of the pandemic. There are likely to have been many other individual level and country level influences on work-related changes and compliance behavior than we included in this study. Another limitation of this study is the rather snapshot-like measurement of people’s multifaceted social identity. This study focused on the identification with the working population, as there is reason to suggest that this identity plays a role for societally relevant behaviors (e.g., Selenko & De Witte 2021). However, we do acknowledge that there are many more identities, personal and work-related, which will have been affected by the changes to life and work brought about by the pandemic. For example, working while parenting small children was a source of immense stress during the pandemic (e.g., Calderwood et al. 2022) and might have made people rather more aware of their parental identity and decreased the salience or centrality of work-related parts of their identity. Unfortunately, we did not assess the person’s family situation in our study, which is a limitation. Also, we did not include media-use or belief in conspiracy theories, which were widespread during the pandemic—and might have impacted identification with others in the working population as well as compliance with health guidance. Neither did we assess individual attributions or beliefs surrounding public health guidance. For example, it is likely to make a difference whether health guidance is attributed as primarily protecting business health or protecting public health. Also, the ability to act in a health-compliant way is very often not just up to individual choice, but determined by socio-economic factors (Farias & Pilati 2022; Probst et al. 2020). Finally, we need to point out that the findings of this study rest on a cross-sectional design. The questions on experienced work-changes were retrospective—they could hence have been affected by present identification with the working population, despite there being no theoretical reason to think so. The instrumental variable estimation does support the proposed direction of the relationship between identification with the working population and COVID-19 health compliance behavior and suggests that it is causally robust.
Practical Implications and Conclusions
There are several practical implications of our findings. First of all, for public health officials interested in implementing health policies, it is important to realize that (1) the identification with the working population plays a role and (2) that this is under the influence of work-experiences. If countries want their citizens to comply with public health guidance, it is important to acknowledge the wider, also work-related context that people are in and the changes they experience there, rather than just regard peoples’ demographics, living situation or personal motivations. To enhance compliance with public health measures, policy makers might want to be mindful to strengthen people’s identities as members of the working population in their public messaging.
For policy makers but also unions and organizations this study brings interesting evidence on the link between work changes and societally relevant behaviors and attitudes. Work is the place where people spend most of their waking day, where they learn about and interact with society. Changes to work thus can also have ‘societal meaning’ as our study shows, for the identification with the wider working population, and behaviors of relevance to the working population. The establishment and protection of good workplaces therefore is not only in workers and organizations interest but becomes an issue of importance for a functioning society.
One might speculate how the next, collectively experienced dramatic change to work (e.g., due to generative AI or industries transforming toward greener methods of production) might impact social cohesion and affect societally relevant behavior. Our study shows that it would be important to inspect work-related changes carefully for their identity affecting potential, to mitigate any risks of identity threat, and to ensure that workers can still feel part of the wider working population, even if their work is being transformed.
In conclusion, although our study was set in the pandemic, it offers a couple of generalisable lessons regarding the importance of work-related experiences for societal behavior. Most importantly it shows that the individual drivers of good citizenship are informed (also) by the employment context.
Notes
[4] Note that strictly speaking, a shared identity is still a perception assessed on the individual level: It is the individual feeling as ‘we,’ rather than as a distinct ‘I’ and indicates the feeling of belonging to a group, the self-categorization as a member of that group. Other group members might not share this sentiment.
Competing Interests
The authors have no competing interests to declare.