As mentioned previously, job resources lead to work engagement and positive organisational outcomes (Bakker & Demerouti, 2017). Demerouti, Bakker, Nachreiner and Schaufeli (2001) describe job resources as the organisational aspects of the job that are instrumental in achieving work goals and may also reduce job demands. The resources used in this study as determinants of work engagement and performance are leadership behaviour (as a supervisory support resource), team effectiveness (as a social/collegial support resource) and technology and flexibility (as an organisational support resource).

Leadership behaviour was used in this study to gauge the perceptions of the participants of their leaders in response to the change that was necessary because of the lockdown. Specifically, this was based on work by Prewitt and Weil (2014) where they explain that leadership can change organisational stress using technical (quick solutions, setting procedures for known problems) or adaptive techniques (addresses fundamentals and dysfunctional dynamics), but crisis leadership realises that this is insufficient in the long-term without an adaptive response. Leadership has been shown to impact motivational states such as work engagement and outcomes such as performance (Caulfield & Senger, 2017; Stander, De Beer, & Stander, 2015). In this study, performance is looked at from the perspective of the positive work role behaviours described by Griffin, Neal and Parker (2007); specifically, proficiency (fulfils the prescribed or predictable requirements of the role), adaptivity (copes with, responds to and supports change) and proactivity (initiates change, is self-starting and future directed; Griffin et al., 2007). Work engagement has also been shown to affect performance (Gauche & De Beer, 2018).

Team effectiveness was a collegial resource used in this study and indicates support from colleagues in work – in addition to considering the effectiveness of the teams in reaching work goals. De Beer, Rothmann and Pienaar (2012) found job resources that contained colleague support as a positive predictor of work engagement and buffer for burnout. In terms of the resource technological flexibility which indicates support from organisations to workers in order to have remote working arrangements or be able to work remotely, Felstead and Henseke (2017) found that detachment of work was a growing trend, and that remote working is related to increased job satisfaction and work-related well-being, but that those advantages evoke work intensification and an increased inability to switch off.

Therefore, the stated exploratory model indicates a potential mediating effect of work engagement in the relationship between the antecedents (leadership, behaviour, team effectiveness and technological flexibility) and the performance variables (Zhao, Lynch, & Chen, 2010). Indeed, research has shown work engagement to be an important mediating variable in this regard (cf. Bakker & Demerouti, 2017; De Beer et al., 2012).

A cross-sectional quantitative survey design was used, and the sample consisted of 229 participants (n = 229). Although the survey is open for the duration of the lockdown, the data from 16 April up to 26 April 2020 was used for this study. Therefore, at that time, the lockdown was experienced for a total of 20 days (about 1 work month). Specifically, an online survey was created with a free licence account on the QuestionPro platform and shared on social media outlets (i.e. Facebook and LinkedIn) by inter alia the Society of Industrial and Organisational Psychology of South Africa. Therefore, participants were completely free to decide whether they would like to participate in the study or not, as such this sample is a non-probability sample. The inclusion criterion for the study was that participants needed to be working during the lockdown.

The mean age of the participants was 38.08 years (standard deviation [SD] = 10.11) and the majority identified as female (n = 159; 69.43%) and the rest identified as male. Furthermore, 68 participants (29.69%) indicated that they were employees who could be classified as essential service workers and 111 (48.47%) indicated that they had dependent children at home or dependents older than 60 (n = 30; 13.10%). In terms of relationship status, 122 (55.02%) participants indicated that they were married.

The researchers considered carefully the wording of the survey instructions to avoid any unnecessary words such as ‘crisis’ to avoid influencing the mind frame of employees but attempted to be as neutral as possible. Furthermore, all scales were adapted in such a manner that the participants would answer thinking about the last few weeks of lockdown. This entailed an instruction in some instances (e.g. ‘During the last few weeks’) and changes in tenses to suit the above adaption.

Leadership behaviour was measured with a 9-item scale created by the researchers using the work of Prewitt and Weil (2014) as a guideline on a scale from 1 (strongly disagree) to 5 (strongly agree). This scale measured the perceptions of participants regarding their leader’s behaviour during the past few weeks (see Table 1). An example item from this scale was ‘During the last few weeks my manager… demonstrated courage and active responsibility for dealing with the change’. Cronbach’s a coefficient for this scale was 0.94 and, therefore, considered highly acceptable.

Team effectiveness was measured with a 5-item scale which was developed by the researchers using Patrick Lencioni’s Team Effectiveness model as the guideline (cf. Hamlin, 2008). The scale ranged from 1 (strongly disagree) to 5 (strongly agree). An example item from this scale was ‘During the last few weeks my team members … accepted personal accountability to deliver their share of work to achieve the defined team goal’. Cronbach’s a coefficient for this scale was 0.89 and therefore considered acceptable.

Technological flexibility was measured with a 3-item scale developed by the researchers, ranging from 1 (strongly disagree) to 5 (strongly agree). An example item from this scale was ‘My organisation’s technology infrastructure enables remote and flexible working’. Cronbach’s a coefficient for this scale was 0.88, which is also highly acceptable.

Work engagement was measured with the 3-item ultrashort Utrecht Work Engagement Scale (Schaufeli, Shimazu, Hakanen, Salanova, & De Witte, 2019) on a scale of 1 (never) to 7 (always). An example item is ‘When working, I felt bursting with energy’. Cronbach’s a coefficient for this scale was 0.82, which is acceptable.

Performance was measured with an adapted and shortened 9-item version of the Work Role Performance scale (Griffin et al., 2007). For each component, one item was used for the individual, team and organisation, respectively, on a scale from 1 (strongly disagree) to 5 (strongly agree) and therefore consisted of three items for each of the three components: (1) proficiency (α = 0.72; e.g. ‘Over the last few weeks, you… carried out the core parts of your job well’), (2) adaptivity (α = 0.87; e.g. ‘Over the last few weeks, you…actively focused on learning new skills to better deal with the change at hand’) and (3) proactivity (α = 0.82; e.g. ‘Over the last few weeks, you…came up with ideas to improve the way in which your core tasks are done’).

Finally, two single-item non-demographic control questions were asked for use in the analysis about the participant’s general levels of optimism and pessimism from the Scale Optimism–Pessimism-2 (SOP2; Kemper, Wassermann, Hoppe, Beierlein, & Rammstedt, 2017) on a scale of 1 (not at all pessimistic/optimistic) to 7 (very optimistic/pessimistic). In the current study, the items were used individually as research has shown that optimism and pessimism are not necessarily perceived on a continuum (see Hummer, Dember, Melton, & Schefft, 1992).

With Mplus 8.4 (Muthén & Muthén, 2019), confirmatory factor analysis (CFA) was applied with the robust maximum likelihood estimation method – which is robust against the possibility of non-normality in the data. Importantly, given the context in which data collection was done and the potential biases of participants in answering the survey, all the items were regressed on the observed scores for optimism and pessimism to lessen any potential biases. Figure 1 is the conceptual CFA model which was provided by Mplus. Thus, in the CFA model controls were estimated at the item level in creation of the measurement model. Standard fit measures were considered for the models: comparative fit index (CFI), Tucker–Lewis index (TLI), root mean square error of approximation (RMSEA) and the standardised root mean square residual (SRMR).

Furthermore, correlations were also considered amongst the classic guidelines of 0.30 and above for a medium effect size and 0.50 and above for a large effect size.

For the structural model, additional demographic control variables were added to the model in which all the latent variables were regressed on gender, essential services worker, and age, seeing whether these three demographic variables were statistically significant predictors of the variables, thereby also controlling for these variables in explaining variance. Essential service worker in this context indicated whether someone was an essential service worker as defined by the government but working from home. The focus was on the direction of the standardised b coefficients and the statistical significance associated with each of the parameters, which is set as a p < 0.05. For all potential indirect effects in the model, bootstrap resampling was used (10 000 draws), paying attention to the 95% confidence intervals (CI) of the estimates in order to ascertain whether the estimate crossed zero (Hayes, 2017).

The CFA showed that the measurement model (see Figure 1) was an acceptable fit to the data: CFI = 0.92; TLI = 0.90; RMSEA = 0.058; SRMR = 0.054. As can be seen in Table 3 all the factor loadings were statistically significant and acceptable. Moreover, the standard errors were relatively small, indicating accuracy of the estimation process.

Regarding the control variables of optimism and pessimism at item level, the following was evident: Optimism explained small, but statistically significant, variance in all of the leadership behaviour items except for item 2. For team effectiveness, pessimism had a statistically significant negative relationship with item 1. No statistically significant relationships were observed with technological flexibility’s items (p > 0.05). Optimism, but not pessimism, had a statistically significant with all of the work engagement items. In terms of the performance items, optimism had av statistically significant relationship with the items of proficiency (items 1 and 3), adaptivity (item 1) and proactivity (items 1 and 2). Finally, pessimism had statistically significant negative relationships with the items of adaptability (items 1 and 3). Interestingly, both optimism and pessimism had an effect on item 1 of adaptability. The reader can refer to Table 1 below to read the item text for each item corresponding to the specific number.

The correlations (see Table 2) showed that all the significant relationships between variables were positive and that most of the effect sizes could be described as medium. Specifically, leadership behaviour had a positive correlation with all of the variables with medium effect but not with proactivity (r = 0.21; small effect). The relationship between technological flexibility and proactivity was not statistically significant. The highest correlation was between proficiency and adaptivity (r = 0.82; large effect).

In line with the conceptual model in Figure 2, a structural model was created. The fit of this model was adequate: CFI = 0.92; TLI = 0.89; RMSEA = 0.056; SRMR = 0.062. The results are given in Table 3.

As can be seen, leadership behaviour predicted work engagement positively (β = 0.31; SE = 0.09; p < 0.001) but did not have any other direct relationship in the sample. Similarly, technological team effectiveness also predicted work engagement positively (β = 0.24; SE = 0.08; p = 0.003). Of the resources, technological flexibility was the only to predict a performance variable, specifically adaptivity (β = 0.26; SE = 0.12; p = 0.027). Lastly, work engagement did not predict proficiency (p = 0.201) but did predict adaptivity (β = 0.27; SE = 0.13; p = 0.031) and proactivity (β = 0.33; SE = 0.10; p = 0.002).

In terms of the control variables’ structural paths to all the factors, the overwhelming majority of the results were non-significant except for the positive relationship of age on work engagement – indicating that the higher the age of the participant, the higher the work engagement (β = 0.17; SE = 0.07; p = 0.014), but a negative relationship for age on technological flexibility (β = -0.33; SE = 0.10; p = 0.002) – indicating that as age increased the technological flexibility decreased. Furthermore, being male had a significant relationship with technological flexibility in the sample. Lastly, being a remote essential service worker had a significant positive relationship with proactivity (β = 0.14; SE = 0.07; p = 0.035).

Results from the bootstrapped resampling showed that there were four noteworthy indirect relationships in the model. The first two relationships were leadership behaviour to adaptability through work engagement (estimate = 0.11; 95% CI [0.03, 0.21]) and team effectiveness to adaptability through work engagement (estimate = 0.07; 95% CI [0.02, 0.15]). Similarly, in the final two relationships both of these indirect relationships also existed for proactivity, that is leadership behaviour (estimate = 0.15; 95% CI [0.08, 0.24]) and team effectiveness (estimate = 0.09; 95% CI [0.04, 0.17]) to proactivity through work engagement.