Emotions are distinguished by dimensions such as valence (pleasantness), arousal (intensity) and cognitive appraisals (evaluations of control and certainty), according to scholars (e.g. Lazarus, 2006; Smith & Ellsworth, 1985). Anxiety, for instance, is marked by negative valence, high arousal and perceptions of uncertainty and low control. This reflects Spielberger’s (1985, p. 173) definition of anxiety as ‘feelings of tension, apprehension, and dread, and cognitions of impending danger’.

In addition, Spielberger (1972) distinguished anxiety from stress. He stated that stress should ‘be used exclusively to denote environmental conditions or circumstances that are characterized by some degree of objective physical or psychological danger’ (p. 488). According to him anxiety should ‘be used to refer to the emotional reaction or pattern of response [to perceived threat]’ (p. 488). Anxiety and stress are conceptually distinct from stressors. Podsakoff et al. (2007) defined stressors as challenging circumstances that can potentially lead to adverse effects on emotions, thoughts, behaviour, physiological health and overall well-being.

Within the TWA framework, WA is conceptualised as ‘feelings of nervousness, uneasiness, and tension about job-related performance’ (p. 537). Cheng and McCarthy (2018) posited that WA is influenced by both individual predispositions and environmental situations. Therefore, it manifests as both dispositional WA and situational WA. Dispositional WA is conceptualised as ‘individual differences in feelings of nervousness, uneasiness, and tension about job performance’ (p. 539). On the other hand, situational WA is conceptualised as a ‘transient emotional state reflecting nervousness, uneasiness, and tension about specific job performance episodes’ (p. 539). The TWA incorporates a dual-perspective approach to understanding the origins and manifestations of anxiety in the workplace. The dispositional aspect of this theory, grounded in trait-based perspectives, postulates that individual differences, such as demographics, core self-evaluations and physical health, contribute to dispositional WA. Conversely, the situational aspect of WA suggests that situational factors influence the immediate experience of situational WA. These situational factors include the emotional demands of labour, task requirements and broader organisational demands, together with job-specific characteristics such as job type, job demands and the degree of autonomy. This suggests that situational WA can fluctuate and is sensitive to the dynamic conditions of the work environment such as WFH. Finally, according to the TWA, both dispositional and situational WA ‘can exert negative and positive effects on job performance’ (Cheng & McCarthy, 2018, p. 538). The current study is focussed on situational WA because it sought to explore employees’ WA triggered by WFH job demands.

The TWA theorises that ‘situational characteristics and job characteristics are the core antecedents of situational workplace anxiety’ (Cheng & McCarthy, 2018, p. 546). Our study investigated the association between situational WA and four important aspects of a WFH environment: role ambiguity (RA), work overload (WO), job insecurity (JI) and work–home conflict (WHC). These were chosen because they are recognised as the most extensively examined job demands or stressors in the literature (Podsakoff et al., 2007). Within the framework of the job demands-resources (JD-R) model (Demerouti et al., 2001), they are defined as:

This can manifest in various forms of strain such as ‘burnout, depression, emotional exhaustion, fatigue, frustration, mental, psychological, and physical symptoms, and tension’ (Podsakoff et al., 2007, p. 442). Therefore, in the present study, WFH job demands encompass: (1) RA, which is uncertainty about required actions for role fulfilment; (2) WO, where time and resources are insufficient for meeting role obligations; (3) JI, the perceived probability of job loss based on one’s interpretation of the work environment and (4) WHC, the perceived clash between work and family demands.

Podsakoff et al.’s (2007) meta-analytical study, which examined 183 independent samples, found that job demands significantly exacerbated strain. The TWA is relatively nascent with limited empirical testing. Only one extant study conducted by Wang et al. (2023) has corroborated the predicted strong positive associated between job demands, specifically informational overload, and WA. However, the extant job demands literature findings are that WO (Alnazly et al., 2023), RA (Pretorius & Padmanabhanunni, 2022), WHC (Sanz-Vergel et al., 2011) and JI (Cheng & Chang, 2008) significantly positively predict anxiety. Consistent with the TWA and the preceding discussion, we propose the following hypothesis:

The TWA does not explicitly suggest moderation between situational and job characteristics (i.e., job demands) and WA. However, existing literature on job demands has identified potential moderators like gender, dependants, seniority, tenure, personality and leader–member exchange (for a review see Kwon & Kim, 2020). In this context, we hypothesise that gender moderates the relationship between WFH job demands and WA. This is supported by various theories explaining gender differences in areas such as cognitive performance, personality, social behaviours and psychological well-being (Hyde, 2014). Evolutionary theories (e.g. Buss & Schmitt, 1993) suggest that psychological gender differences are evolutionary adaptations with varying behavioural benefits for males and females. Cognitive social learning theory (Bussey & Bandura, 1999) views these differences as outcomes of distinct social reinforcements and cognitive processes like attention and self-efficacy. Social role theory (Eagly & Wood, 1999) attributes contemporary psychological gender differences to historical divisions of labour, leading to the development of gender-specific psychological traits.

These theoretical perspectives provide a foundation for understanding the empirical findings of gender-related research. For example, regarding psychological well-being, Hyde (2014) noted that adult women are twice as likely to experience depression as men, which is influenced by affective, biological and cognitive factors. While research directly addressing gender’s moderating role between WFH job demands and WA is scarce, Harlos et al. (2023) provided relevant insights. Their study, while not focussing on WA, examined gender as a moderator between job demands (i.e. role ambiguity, role conflict, role overload) and psychological strain (i.e. workplace bullying). They found a significant interaction between role conflict and gender, such that the effect of role conflict was stronger for women than men. Based on these discussions, we put forward the following hypothesis:

In the current study, we explore PsyCap as a personal resource that can be depleted by high levels of WA. Hobfoll (1989) defined personal resources as flexible capacities that reflect an individual’s perceived capability to impact and contribute positively to their work environment. Psychological capital ‘plays a decisive role in employees’ functioning at work’ (Xanthopoulou et al., 2007, p. 124) and was described by Luthans and Youssef-Morgan (2017) as a developmental state where an individual exhibits four key traits: (1) a belief in their own abilities to successfully tackle difficult tasks (self-efficacy); (2) an optimistic outlook on achieving success both presently and in the future (optimism); (3) the persistence to continue striving towards goals or to find alternative strategies when necessary (hope) and (4) the capacity to recover from adversity (resilience).

Conservation of resources theory posits that resource loss is a major component of stressful encounters, such as those experienced with WFH job demands. Following the TWA, we earlier hypothesised that WFH job demands positively influence WA, and from COR theory, we earlier argued that WA represents a state of resource loss. This implies that heightened WA could lead to resource depletion, diminishing PsyCap levels. Psychological capital, as a psychological resource, can also be understood through the broaden-and-build theory of positive emotions (Fredrickson, 2001). According to this theory, negative affect narrows an individual’s thoughts and behaviours, depleting their personal resources, while positive emotions broaden thoughts and actions, thereby building enduring personal resources. Consequently, negative affect associated with WA depletes personal resources such as PsyCap (Fredrickson, 2001). This line of reasoning was empirically supported by Zeidner et al. (2011) and Zeidner and Ben-Zur (2014) whose experimental studies demonstrated that stressful encounters cause significantly high levels of negative affect, like anxiety, and reduce psychological resources. Additionally, Cao et al. (2022) observed a robust negative relationship between negative affect and PsyCap, while Xie et al. (2021) and Yao et al. (2022) found a strong negative relationship between workplace stress and PsyCap. Given that workplace anxiety is positively associated with workplace stress (De Clercq et al., 2020), these findings support the hypothesised negative association between workplace anxiety and PsyCap. In alignment with the COR theory and the broaden-and-build theory, as well as the preceding discussions, we propose the following hypothesis:

Digital technologies are ‘combinations of information, computing, communication, and connectivity technologies’ (Bharadwaj et al., 2013, p. 471). These technologies have characteristics that reflect users’ perceptions of the technologies’ functional and non-functional features, such as perceived ease of use (PEU), perceived usefulness (PU) and technical support (TS) (Ayyagari et al., 2011). Perceived ease of use refers to the degree to which employees believe that digital technologies can be used without major effort, while PU refers to the degree to which employees believe that digital technologies support the accomplishment of tasks and enhance job performance (Ayyagari et al., 2011). Venkatesh et al.’s (2003) unified theory of acceptance and use of technology (UTAUT) suggests that functional characteristics such as PEU and PU influence how employees interact with digital technologies. When employees find digital technologies easy to use, their stress related to technology decreases, enhancing their psychological capital by raising their self-efficacy; additionally, when they perceive digital technologies as useful, they are more likely to see them as beneficial and integral to their work performance, thereby fostering optimism and hope (Venkatesh et al., 2003). Therefore, according to UTAUT, when employees perceive technologies as easy to use and useful, their beliefs in their ability to effectively use these technologies are strengthened, even in conditions of high WA associated with WFH.

Within an organisational setting, job resources play a critical role. Job resources encompass the physical, social and/or organisational elements of a job that serve one or more of the following functions: aid in accomplishing work objectives; alleviate job demands and their related physical and mental burdens and foster individual growth, learning and development (Bakker & Demerouti, 2007). Examples of job resources include career development opportunities, task variety, job autonomy, social support from co-workers and organisational support, such as the provision of material resources and supervisory support. In the context of WFH employees, TS for the digital technologies provided by the employer represents a non-functional characteristic of the technologies and an important job resource. Technical support refers to the assistance provided to employees to help them to effectively deal with technical issues related to digital technologies (Ayyagari et al., 2011). According to the COR theory, the provision of TS through training, assistance from experts and chatbots can mitigate the adverse influence of WA associated with WFH on PsyCap. In the literature, some evidence supports the buffering effect of DTC on workplace stress (Pullins et al., 2020; Tarafdar et al., 2014) and WA (Rangarajan et al., 2022). In alignment with the COR theory and the discussions above, we propose the following hypothesis:

Figure 1 represents the conceptual model developed for the study.

An online survey was created and pretested with 10 individuals to ensure that the questions were clear and appropriate. Subsequent minor revisions led to a final survey comprising an introduction; demographic details, WFH job demands, DTC, WA and PsyCap. The survey, designed to take between 20 min and 25 min, was specified as being for full-time employees who had transitioned to working entirely from home because of the COVID-19 pandemic. The survey was disseminated through personal contacts, social media platforms (WhatsApp, Facebook and LinkedIn) and snowballing techniques from 21 July 2022 to 05 September 2022, yielding 162 complete responses with no missing data. To determine the sample size required for this study, we used ‘10 times rule’ for studies using the partial least squares structural equation modelling (PLS-SEM) approach, which states that the:

The analysis indicated that a minimum sample size of 80 participants is required. Our sample size of 162 exceeds this requirement. The data were analysed using Statistical Package for the Social Sciences (SPSS) 29 (IBM Corp., Armonk, New York, U.S.) and SmartPLS GmbH (Oststeinbek, Schleswig-Holstein, Germany) statistical packages, with no missing data across all variables.

Table 1 illustrates the demographic profile of the participants. Two-thirds (66.67%) of the respondents were female, and most of the respondents were highly educated. Over half (55.56%) possessed postgraduate qualifications, were junior in their roles (58.64%), had dependent children (80.25%), were married (54.94%) and earned an annual income of between R700 000.00 and R1 500 000.00 (56.17%). Almost half (48.77%) of the respondents were aged 35 to 44 years old.

The survey instrument for this study was carefully adapted from established measures found in existing literature. For all variables except demographic ones, we employed 7-point Likert-type scales. These scales allowed respondents to express their level of agreement, ranging from 1 (strongly disagree) to 7 (strongly agree). The psychometric properties of these scales have demonstrated robustness in previous studies. Moreover, in the current study, each scale displayed satisfactory internal consistency and reliability.

Work-from-home job demands: Participants evaluated four specific job demands, each measured by distinct items from the scale developed by Rangarajan et al. (2022). For WO, a representative item was, ‘WFH creates many more requests, problems or complaints in my job than I would otherwise experience’. Role ambiguity was assessed with an item like, ‘I am unsure whether I have to deal with my WFH problems or with my work activities’. Job insecurity was gauged through statements such as, ‘Work from home will advance to an extent where my present job can be performed by a less skilled individual’. Lastly, WHC was measured by items including, ‘I do not get everything done at home because I find myself completing job-related work because of WFH’ The responses to these items were averaged to generate a single composite score for each job demand, with higher scores indicating greater job demands.

Workplace anxiety: We used the 8-item scale developed by McCarthy et al. (2016). Two sample items were: ‘I often feel anxious that I will not be able to perform my job duties in the time allotted’ and ‘I feel nervous and apprehensive about not being able to meet performance targets’. The responses to these items were averaged to generate a single composite score, with higher scores indicating greater WA.

Digital technology characteristics: Participants evaluated three specific DTC, each measured by distinct items from the scale developed by Rangarajan et al. (2022). For PU, a representative item was, ‘Use of DT enhances my job effectiveness when I WFH’, while PEU was assessed with an item like, ‘I find our digital technologies for WFH to be easy to use’. Lastly, TS was measured by items including, ‘The training provided for digital technologies is complete and sufficient’. The responses to these items were averaged to generate a single composite score for each characteristic, with higher scores indicating greater satisfaction.

Psychological capital: We used six items from the scale created by Luthans and Youssef-Morgan (2017) to represent the four facets of PsyCap: hope, self-efficacy, resilience and optimism. Two sample items were: ‘I believe that I can bounce back from any setbacks that have occurred’ and ‘I expect good things to happen in the future’. The responses to these items were averaged to generate a single composite score, with higher scores indicating greater PsyCap.

The following demographic information was also gathered and, in line with the literature, was used as control variables: gender (male, female), age in years (18 to 34, 35 to 44, 45 to 54, 55 to 64); education (less than degree, degree, postgraduate); marital status (never married, married, divorced); dependent children (yes or no); organisational level (junior, middle, senior, executive) and annual income (less than R400K, R400K to R700K, R700K to R1.5 million, more than R1.5 million).

We used several procedures to address potential common method bias because of self-reports from a single survey. Firstly, on the cover page of the survey, we reminded participants that their responses would remain anonymous and that there were no right or wrong answers. Secondly, we conducted two tests to check for the presence of common method bias. Harman’s single-factor test found that 25% of the variance was attributable to a single factor, which is below the threshold value of 50%. Additionally, Kock’s (2015) full collinearity test found that the largest variance inflation factor (VIF) for the latent variables in our study was 2.46, below the threshold value of 3.3. Taken together, these results suggest that common method bias was not a concern in our dataset.

An application for ethical approval was made to the Gordon Institute of Business Science, University of Pretoria and ethical approval was received on 19 July 2022. The study was conducted in accordance with the 1964 Helsinki Declaration. Informed consent was obtained from all subjects involved in the study.

We first checked the Mardia test for assessing multivariate normal distribution. The results were significant (p < 0.001), indicating that the assumption of multivariate normality was violated. Because of the complex structural model, relatively small sample size (N = 162) and non-normal data in our study, we followed Hair et al.’s (2019) recommendation and tested our conceptual model using the PLS-SEM method, which consists of the measurement and structural models. If the measurement model or confirmatory factor analysis (CFA) describes the relationships between the latent variables and their indicators within acceptable thresholds, the structural model can then be tested by assessing the hypothesised relationships between the latent variables using linear regression (Hair et al., 2019).

We used Cronbach’s alpha values to assess internal reliability, the measure of consistency across items. Table 2 shows that all constructs had values that exceeded the 0.7 threshold, confirming internal reliability (Hair et al., 2019). We assessed convergent validity, the degree to which observed variables account for a latent construct using factor loading, average variance extracted (AVE) and composite reliability (CR). Convergent reliability is considered established when factor loadings exceed the 0.70 threshold (Hair et al., 2011). As shown in Table 2, all factor loadings met this criterion, except for JI3 (0.69), JI4 (0.68), WHC1 (0.65) and PsyCap1 (0.66). Although these specific loadings are slightly below 0.70, internal reliability is still supported because a threshold of 0.50 is also recognised as acceptable (Bagozzi & Yi, 1988). In addition, Table 2 indicates that for all constructs, the AVE values ranged from 0.50 to 0.79, while the CR values ranged from 0.85 to 0.96, both surpassing the recommended thresholds of 0.50 and 0.70 (Hair et al., 2019), respectively. Thus, all the study’s constructs demonstrated satisfactory convergent validity.

Table 2 presents the descriptive statistics of the focal constructs of the study. In examining the descriptive statistics of our study, a nuanced understanding of the various facets of working from home emerged. Job demands or stressors, conceptualised as WO, RA, JI and WHC, presented relatively lower mean scores, suggesting low perceptions of the WFH job demands as stressors. On the other hand, WA showed a slightly higher mean score of 3.22 (standard deviation [SD] = 1.59), pointing to a moderate level of anxiety experienced by employees. In contrast, the characteristics of digital technology used for WFH – PU, PEU and TS – demonstrated high mean scores, suggesting strong perceptions of the efficacy and supportive nature of the WFH digital technology. Additionally, PsyCap had the highest mean score of 5.86 (SD = 1.07), suggesting very high levels of psychological resources among the respondents.

We assessed discriminant validity, the degree to which constructs are empirically distinct, in three ways. Firstly, individual items primarily loaded onto their corresponding constructs with minimal cross-loading, supporting discriminant validity. Secondly, all heterotrait-monotrait ratio (HTMT) values were below the threshold of 0.85 set by Henseler et al. (2015) and ranged from 0.10 to 0.76, confirming discriminant validity. Finally, as shown in Table 3, the square root of the AVEs is greater than the correlations for all latent constructs (Fornell & Larcker, 1981), providing further evidence of discriminant validity.

We followed the recommendations of Hair et al. (2011) and Benitez et al. (2020) by using the bootstrapping function in SmartPLS to generate 5000 bootstrapped samples for model fit testing and structural model estimation. For the moderation hypotheses, we followed Matthews et al.’s (2018) recommendation for moderation analysis using SmartPLS, which suggests employing the product-indicator-method (PIM) for continuous moderators and multi-group analysis (MGA) for categorical moderators. Firstly, we confirmed the absence of multicollinearity by examining the VIF. All VIF values were below the threshold of 3.0, ranging from 1.28 to 2.46, thus supporting the absence of multicollinearity. Next, we assessed model fit using R² values, which were 0.36 for workplace anxiety and 0.19 for PsyCap, indicating adequate explanatory power. Additionally, we examined the standardised root mean squared residual (SRMR), and the model’s SRMR value was 0.071, below the guideline of 0.08 (Benitez et al., 2020), suggesting an acceptable model fit. Lastly, we evaluated the predictive relevance of the model using the Q² values derived from the blindfolding procedure in SmartPLS. The Q² values for workplace anxiety and PsyCap were 0.644 and 0.537, respectively, providing further evidence of the path model’s predictive relevance.

Table 4 shows the results of hypotheses testing. Surprisingly, while JI (β = 0.3154, t = 3.459, p < 0.01) is positively related to WA, other WFH job demands (RA, WO, WHC) were not significantly associated with WA. Consequently, only H1C was supported whereas H1A, H1B, and H1D were not supported. Furthermore, Table 4 also indicates that among the interaction variables, the only significant predictor of PsyCap was PEU × WA (β = −0.272, t = 2.046, p < 0.05). The moderation effect, quantified at 0.039, is classified as small. Therefore, H4B was supported, while H4A and H4C were not.

Simple slope analysis was conducted to probe the nature of the significant PEU × WA interaction. The results indicate that at low levels of WA, individuals with high PEU exhibit significantly higher PsyCap compared to those with low PEU. However, as WA increases, the PsyCap for individuals with high PEU decreases to levels comparable to those with low PEU. These results suggest that while high PEU is associated with higher PsyCap at low levels of WA, the buffering effect of PEU diminishes as WA increases. Therefore, H4B is not fully supported because high PEU does not consistently buffer the negative impact of high WA on PsyCap. Figure 2 illustrates the slopes.

To examine the moderating effects of gender on the association between WFH job demands (i.e. RA, WO, JI, WHC) and WA, we conducted MGA, given the categorical nature of our moderator (Matthews et al., 2018). Table 5 presents the results of this analysis. It is important to note that in MGA, a significant p-value represents a significant difference between the path coefficients for males and females in each of the investigated relationships (H2A to H2D). The MGA results show that gender moderated the association between WO and WA (∆β = 0.402, p < 0.05) and JI and WA (∆β = 0.493, p < 0.01). Therefore, H2B and H2C were supported, whereas H2A and H2D were not.

A simple slope analysis (Aiken & West, 1991) was conducted to probe the nature of the significant Male × WO interaction. The simple slope for females (β = 0.260, p < 0.05) was significant, while the simple slope for males (β = −0.142, p > 0.05) was not significant, showing that for females, an increase in WO is associated with an increase in WA. For the Male × JI interaction, the simple slope for females (β = 0.534, p < 0.001) was significant, while the simple slope for males (B = 0.041, p > 0.05) was not significant, showing that for females, an increase in JI is associated with an increase in WA. These findings underscore the conditional influence of gender on the WO–WA and JI–WA associations. Specifically, the impact of WO and JI on WA is stronger for females. Figure 3 and Figure 4 illustrate the slopes.

In the post-COVID-19 pandemic world, the prevalence of WFH has increased. However, our understanding of WA and the influence of job demands on WA in this context remains limited. Additionally, in the context of WFH, the downstream impact of WA on personal resources, such as PsyCap, is not well understood. This study makes several theoretical contributions by addressing these gaps and enhancing our comprehension of WA and its effects. Firstly, it contributes to the literature by offering support for the differential impact WFH job demands have on WA. Secondly, it is one of the few studies that offers empirical validation for TWA, marking a significant step in the theory’s development, especially given the relative newness of the theory and the limited empirical evidence supporting it. Thirdly, by examining interactions within the TWA framework, the research uncovers nuanced understandings of how job demands associated with WFH relate to WA and how WA in turn influences PsyCap. Fourthly, our findings contribute to the limited understanding of the impact of WA (Cheng et al., 2023) by elucidating the WA–PsyCap relationship and its moderators in the context of working from home. It thereby broadens the scope of understanding of the predictors of PsyCap. Finally, by focussing on South Africa – a non-WEIRD context – this research responds to the call for more diverse samples in behavioural research (Henrich et al., 2010), thereby adding valuable insights from the developing world’s perspective.

Our study is among the few that provide valuable insights to managers and organisations with WFH employees regarding the role of job demands on WA. It also examines the influence of WA on PsyCap. It also explores the moderating effects of gender on the job demands–WA association and the moderating effects of DTC on the WA–PsyCap association. Firstly, our findings indicate that for employees working from home, job insecurity is a significant contributor to workplace anxiety. Therefore, managers should develop strategies to address this job demand to mitigate WA. To address job insecurity, managers should consistently communicate clear information about the organisation’s WFH policies and its overall performance. Furthermore, the significant interactions between work overload, job insecurity and gender highlight the necessity for managers to develop targeted strategies to manage WA among female employees. This is because the association between job demands and WA was found to be stronger for females.

Secondly, our unexpected finding that WA did not significantly influence PsyCap in employees working from home underscores the importance of considering the context in which employees operate. Managers should recognise that the home working environment might buffer the negative effects of WA. This is particularly important as companies design and implement their WFH policies for a post-COVID-19 pandemic work environment.

Finally, our finding of a significant interaction between WA and PEU of digital technologies indicates that in low WA environments, when employees perceive digital technologies as easy to use, this buffers the negative effects of WA on PsyCap. This underscores the importance for managers and organisations to adopt a proactive approach in training employees on the use of new digital technologies. By enhancing employees’ proficiency with these technologies and ensuring they are perceived as user-friendly, organisations can harness the buffering effects of PEU on the WA–PsyCap link.

This study has several strengths. Methodologically, the strength of our study lies in the use of PLS-SEM, which does not require data to follow a normal distribution and is well suited for analysing complex structural models even with relatively small sample sizes, enhancing the robustness and credibility of our findings (Benitez et al., 2020; Hair et al., 2011). Theoretically, our conceptual model (see Figure 1), which is firmly grounded in the TWA and JD-R models, elucidates the complex associations between WFH job demands, WA, DTC and PsyCap. Additionally, our study highlights the complex nature of WA by examining it in a WFH context. Finally, this study contributes unique insights into WA from South Africa, a non-WEIRD (Henrich et al., 2010) setting, thereby enriching the understanding of WA.

Despite these strengths, limitations exist that suggest useful future research. Firstly, the study’s cross-sectional design limits causal inference; therefore, future research should explore causal relationships using longitudinal or experimental designs. Secondly, reliance on self-reported data, while necessary, may introduce bias. However, it is also important to recognise that the constructs being investigated are inherently subjective, making self-reporting a necessary and valuable approach for capturing personal experiences and perceptions. Thirdly, this study, while comprehensive, did not account for several established predictors of WA (e.g. dispositional WA, emotional labour demands) and PsyCap (e.g. core self-evaluations, empowering leadership behaviours). Future research would benefit from incorporating these variables to provide a more holistic understanding of WA and PsyCap. Finally, we recognise that our research, conducted through surveys involving employees from various organisations, lends a degree of generalisability to our findings. Yet, it is important to note that concentrating on a single organisation could provide a more focussed empirical testing of the TWA through the conceptual model developed in this study.