No one is safe! But who’s more susceptible? Locus of control moderates pandemic perceptions’ effects on job insecurity and psychosocial factors amongst MENA hospitality frontliners: a PLS-SEM approach

Employee perception of job insecurity can result from general volatility of employment in the industry despite some organisational attempts to stabilise the rapid turnover issue by considering agility as a possible remedy [23]. For example, terms and conditions of services, as well as compensation, have been less desirable in hospitality than in many other sectors [24]. It is also well documented that work flexibility is rarely available to hospitality workers in the MENA region, with excessively burdening working hours, causing severe psychological and family issues [25, 26]. Moreover, due to the lack of alternative employment in some MENA countries and the weakness of trade unions, employees see as inevitable their poor working conditions and absence of viable options [27]. This is compounded by weak institutional frameworks that could guarantee enforceable minimum standards.

COVID-19 has caused the hospitality sector to shed millions of positions across the globe [28]. Moreover, the pandemic has exacerbated already poor job security perceptions in this industry as it has created increased fear of redundancies among service personnel [29]. We, thus, hypothesise that:

The relationships between job insecurity and psychosocial factors have been examined extensively. Such investigations include the meta-analysis work of [10, 30]. Much literature attempts to provide justification as to the linkages between job insecurity and adverse responses as far as psychosocial issues are concerned. Many researchers cite stress theory [31, 32] and the psychological contract violation [33]. The premise underlying stress theory is that job insecurity is a stressor that is difficult to cope with because it is not easy to control at the individual worker level. This leads to anxiety and psychological distress. Additionally, from a psychological contract theory perspective, workers exchange effort for pay and security of employment. A threat to job security is perceived as a violation of the latent psychological contract between the firm and the worker [33]. There is little question about the importance of hospitality workers given their direct association with customer satisfaction. Yet, workers assume arduous jobs with inflexible and long hours and have weak contract arrangements with limited union protections [34]. Hospitality work is almost by definition a precarious and fragile source of income [35], making it a precarious sector that affords workers little autonomy and low rewards [9]. Thus, the stress caused by work in the sector is further amplified during the COVID-19 pandemic, and we seek to understand the risk to hospitality workers health and well-being.

EB Faragher, M Cass and CL Cooper’s [36], which studied over 500 research papers, revealed the correlation between satisfaction with work and mental health conditions such as anxiety, depression and burnout. These correlations appear to be more pronounced than previously studied mental well-being aspects. In this perspective, job insecurity represents a work-life balance aspect that largely endangers the mental well-being of employees. We, therefore, anticipate that job insecurity can have clear links with negative psychosocial data [37]. These are significant parameters within the organisational commitment literature [38] and expectancy theory when workers contrast their old working conditions with current practices. We, therefore, hypothesise that:

We examine the possible moderating effect in our model based on individual differences by considering locus of control [15]. We reasoned that individuals might vary on how they experience the harmful effects of Covid-19 as a function of their own perceived control over situations. Individuals with internal locus of control will usually see themselves capable of managing externally-imposed events, using their own experience or self-efficacy. We then anticipate that model effects might be less pronounced among internals than individuals with external locus of control [39]. Scholarship into job insecurity (e.g. L Greenhalgh and Z Rosenblatt’s propositions) [40] stressed the moderating role of internal locus of control upon job insecurity.

Locus of control also moderates psychosocial reactions [41]. Hospitality employees’ job roles, moreover, afford them limited flexibility and autonomy due to the burden of supervision and menial nature of many of the tasks they are involved in. In this context, we suggest that internal locus of control may be associated with less pronounced adverse psychosocial reactions in comparison to external locus of control individuals. We, therefore, hypothesise that:

The Heterotrait-Monotrait Ratio of Correlations (HTMT) was computed and found to have values of less than .9, implying a satisfactory level of discriminant validity for all measures [58]. Furthermore, Table 1 indicates that all the constructs had average variance extracted (AVEs) higher than 0.5, composite reliability scores (CRs) between .78 and .90 satisfying the convergent validity and reliability criteria for all measures [51, 52]. Also, Table 2 shows that all Variance Inflation Factor values (VIFs) were less than 5, proving that collinearity is not a crucial issue [59].

Before examining the path and multigroup analyses, we ran Common-Method Bias (CMB) tests, which is considered a necessary procedure when using perceptual, self-report measures from a single survey [60]. The values of the inner variance inflation (VIFs) were all less than 3.3 (See Table 3). Consequently, we concluded that there were no multicollinearity or CMB issues found [61].

Using SPSS version 26, the majority of the sample were male (55%), millennials (45%) educated to a university degree (39%), single (51%), and with external locus of control (59%). Table 4 shows the descriptive statistics of the variables under investigation clustered into locus of control groups. It suggests that those with an external locus of control reported higher levels of COVID-19 perceptions, job insecurity, alienation, anxiety, depression, and burnout.

Given the reflective nature of the latent variables in our model [9], we performed Consistent-PLS Algorithm, followed by Consistent PLS Bootstrapping run at 5000 sub-samples [62] in order to analyse the hypothesised path model.

COVID-19 perceptions are found to positively predict job insecurity (β = .363, P < .001, f² > .15, P < .001), anxiety (β = .353, P < .001, f² > .15, P < .001), alienation (β = .268, P < .001, f² > .02, P < .01), depression (β = .285, P < .001, f² > .02, P < .01) and burnout (β = .268, P < .001, f² > .02, P < .01). Consequently, we judge H1 and H2 as fully supported (See Table 5).

As SRMR was .034 < .08, we elected to suggest that our hypothetical model was an excellent fit for the data [63]. First, the Q² values of all predictors were found to be higher than null, suggesting clear predictive relevance (See Table 6). Second, we compared the root mean squared error (RMSE) and the mean absolute error (MAE) values with the recommended naïve benchmark produced by the PLSpredict method [64] for each of the dependent constructs’ indicators. Since the PLS-SEM analysis (compared to the LM) yielded higher prediction errors in terms of RMSE (or MAE) for the minority number (see Table 7), we concluded that our model had a medium predictive power. Additionally, the R² values of job insecurity (.132), anxiety (.125), depression (.082), burnout (.073) and alienation (.073) were higher than null, which meant that our formulated model had significant predictive validity, according to J Cohen [56].

We assess the path model invariance across the two groups of locus of control (i.e., internal vs external) through running a multigroup analysis (MGA). However, before that, we need to ensure that measurement invariance is established as a prerequisite to MGA [65].

J Henseler, R-JBJ R. Sinkovics, R Daekwan Kim, CM Ringle and M Sarstedt [65] contend that running group comparisons through PLS-SEM can be “misleading” unless the equivalency of their measures is confirmed [66]. By utilising the “Measurement Invariance of the Composite Models” (MICOM) technique, this condition may be fulfilled [65, 66]. Therefore, before conducting any multigroup investigations (in our particular instance: “with no data pooling”), both configural equivalency and compositional invariance need to be established and verified [59, 65‐67]. Because we are using a PLS-SEM technique, we are able to achieve measurement configural invariance as a matter of course [59, 66]. Then, as a result of that, we ran a permutation test. The results are that all this study’s variates/constructs have their “Permutation P-values” more than 0.05; thus, we consider the null hypothesis to be accepted—meaning that the initial correlations of these constructs are non-substantially “different from 1” [66, 67] (See Table 8). This result offers valid evidence of compositional equivalency, implying a feasible multigroup assessment [59, 66].

With regards to the MGA, the t-values linked to the multiple comparisons and shown in parametric tests are analysed. All the paths are significant for any of locus of control groups (significance level = .01). It was nonetheless the case that the paths: COVID-19 perceptions == > anxiety, COVID-19 perceptions == > depression and COVID-19 perceptions == > alienation are significantly non-equivalent between the two groups. From Fig. 1, the findings that customer service employees with external locus of control are more vulnerable to develop anxiety (β_int. = .253 < β_ext. = .408, t_{ext. vs int.} = 2.756, P < .01), depression (β_int. = .122 < β_ext. = .365, t_{ext. vs int.} = 3.973, P < .001) and alienation (β_int. = .139 < β_ext. = .361, t_{ext. vs int.} = 3.342, P < .01) as consequence of their perceptions of COVID-19. Because all the paths are not moderated by the locus of control, it was concluded that H3 is supported only partially (See Table 9).

The major implications of the study in the MENA region show that hospitality workers who are in direct contact with customers have a higher perception of job insecurity, anxiety, depression, burnout and alienation as a result of COVID-19, and this will, as several empirical research investigations show, adversely impact the well-being, attitudes, and behaviours of employees. Our study has offered evidence suggesting positive relationships between COVID-19, job insecurity and psychosocial factors; this suggests that organisations should place emphasis on those initiatives that are within their control as COVID-19 is largely beyond the scope of a management team to address on a corporate level. However, as has been shown in prior research, managers are able to manage, to some degree, COVID-19’s adverse effects on hospitality workers through a statement about their compliance with statutory health guidance and establish the necessary supervisory arrangements for employees whose roles expose them more critically to COVID-19.

Moreover, as recommended by D Bangwal and P Tiwari [77], who considered environmental factors that influence job satisfaction and subsequent intention to quit, we too suggest an important role for management which should offer resources such as masks, vaccination, sanitiser, and other direct firm-level supports, including supervisory support, to address burdens introduced by the spread of COVID-19 on hospitality employees.

The effects of COVID-19 on employee mental well-being and job security are significant. It is, therefore, timely for organisations to develop plans to attempt to remedy the associated psychological distress. Another key finding is related to the moderating role of locus of control. We wanted to see if dispositions might alter the anticipated effects. We anticipated that workers with external locus of control would react poorly to the shock of the pandemic, and that is exactly what we learned. This implies that employees who begin with poorer self-efficacy and confidence in their capability to shape external events are going to respond poorly to public health crises such as COVID-19. External locus of control, therefore, raises significant challenges to management to enhance employee reactions to COVID-19 since locus of control is dispositional before attitudinal and thus hard to change. What we concluded is that COVID-19 is largely outside of any individual, business, or governmental entity’s control, so it is hard to suggest how management, independently, might shelter employees who view the world as externally controlled from the negative reactions to COVID-19.

We acknowledge some limitations in our investigation. The first is linked to the broad nature of the sample; however, it is selective regionally as a result of the data being from the MENA region only. Authors such as H Kabasakal, A Dastmalchian, G Karacay and S Bayraktar [78] contend that the broad socio-cultural practises in the MENA context are thought to have high power distance scores implying high levels of expectations and acceptance of unequal power distribution in the region. The region also scored high on collectivism (i.e., the degree of individual expression of pride, cohesiveness and loyalty to their social groups). Future research could replicate this work in other regional contexts or undertake cross-cultural research using our model to establish the ways in which the results of our study might vary in different settings. Since our study did not record the participants’ country of work or residence, future research might consider this by replicating our study with the country of work or residence counted as a potential moderator. Secondly, our investigation was based on employees in the hospitality sector because they are more impacted by COVID-19 due to their in-person involvement in service provision during the pandemic than employees who can work from home. Such limitations show that more research is required to broaden the sample and job categories. Thirdly, although hospitality employees are likely to be underemployed, our choice of recruiting participants through LinkedIn might be associated with a bias towards educated people, especially with 39% of our sample were educated to a degree level. Fourthly, We needed to shorten the survey aiming for a better response rate [79] and minimized respondent fatigue [80] whilst bearing in mind the aim and objectives of this study; therefore, we purified the scales as to what items to eliminate was made based on the opinion of a panel of four research experts in the area of occupational health and psychology—a common practice in similar survey-based investigations in social science [10, 81]. Although the psychometric properties were sufficiently established for all the measures employed in this study, further research, however, might replicate our study with the whole items of each measure included in the survey.

Finally, a cross-sectional design was used for testing our proposed model, so it can be limited in terms of definitively establishing causal relationships between variables. Further research is required to capture longitudinal data to study in more depth the impacts and causal relationships. However, using a longitudinal design to reflect causality, as argued by PE Spector [82], has been exaggerated and that it only presents limited benefits over the cross-sectional design in most instances in which it is employed. Also, according to the findings of P Tharenou, R Donohue and B Cooper [83], research conducted using a cross-sectional investigation may not necessarily lack interpretability and validity provided they are supported by sound theoretical foundations. Cross-sectional research is also suitable for crises such as the coronavirus pandemic [e.g., 84]. What is needed for future research is to go further than just the use of locus of control as a variable. Future research could deploy more moderators encapsulating personality and demographics and work-life balance issues in the MENA context. This will aid managerial understanding and assist in operationalising the research findings.