Towards understanding causality between work engagement and psychological capital

The South African and international business environments currently demand much more from employees than during any previous time in history (Rothmann, 2003). Modern organisations expect their employees to take a proactive approach, show initiative, develop a sense of responsibility and be committed to the execution of high performance standards (Bakker, Schaufeli, Leiter & Taris, 2008). Organisations therefore require employees who feel energetic and are dedicated to and absorbed by their work, that is, who are engaged with their work (Bakker & Schaufeli, 2008). To encourage engagement in organisations today, personal resources such as optimism, self-efficacy and resilience could be employed, as it is suggested that these personal resources facilitate work engagement (Bakker & Demerouti, 2008). Such resources fall under the rubric of employee psychological capital (PsyCap) (Luthans, 2002a; Luthans, Luthans & Luthans, 2004; Youssef & Luthans, 2007). However, the temporal order between PsyCap and engagement has not been extensively researched, although some authors have theorised a likely reciprocal relationship (Bakker, Schaufeli, Demerouti & Euwema, 2007; Sweetman & Luthans, 2010). In other words, it is still unknown whether engagement leads to, is the consequence of or reciprocally interacts with employees’ PsyCap. This study reports on an investigation into this sequential ordering of PsyCap and work engagement.

highlighted by empirical evidence which proposes that engagement is positively linked to positive organisational outcomes, including job performance (Bakker & Bal, 2010;Halbesleben & Wheeler, 2008), client satisfaction (Salanova, Agut & Peiro, 2005), financial returns (Xanthopoulou, Bakker, Demerouti & Schaufeli, 2009) and positive organisational behaviour, such as personal initiative and learning (Bakker & Demerouti, 2008;Sonnentag, 2003). Given the meta-analytic relationship between employee engagement and indicators of performance such as customer satisfaction, turnover, safety and productivity (Harter, Schmidt & Hayes, 2002), organisations are realising the importance of employee engagement in contributing to the sustainment of their competitive edge in the global market (Schabracq & Cooper, 2000).
Positive organisational behaviour (POB) is defined as: 'the study and application of positively oriented human resource strengths and psychological capacities that can be measured, developed, and effectively managed for performance improvement in today's workplace' (Luthans, 2002a, p. 59). Luthans and colleagues have offered evidence that dimensions of POB are indeed open to development and, importantly, related to performance (Luthans, Avey & Patera, 2008;Luthans, Avey, Avolio & Peterson, 2010;Peterson, Luthans, Avolio, Walumba & Zheng, 2011).  have advanced that such employee positive psychological resources should buffer against the effects of stress, whilst Avey, Luthans, and Jensen (2009) provide some empirical evidence for this combative effect. Xanthopoulou, Bakker, Demerouti & Schaufeli (2007) also illustrate that personal resources could mediate between job resources and work engagement. Conceptually, Sweetman and Luthans (2010) advance that employee PsyCap, as an indicator of POB, can be thought of as a job resource that should help individual employees to obtain goals, buffer demands and facilitate personal growth. What is still lacking is a thorough understanding of the interaction of PsyCap and work engagement.

Research purpose
The focus on POB as a specific approach to employee management has a positive impact on human resource development and performance management; it is also an important means to equip today's employees with the personal skills to deal with the challenges of working life, because the investment in human capital seems to be vital to ensure organisational success (Luthans, Norman, Avolio & Avey, 2008). POB is deemed open to development with highly focused training interventions (Luthans, 2002a;, including taskmastery experiences, positive role modelling, goal setting, contingency planning and social support activities (Luthans, Avey, Avolio, Norman & Combs, 2006;Luthans, Youssef & Avolio, 2007;Luthans, Avey & Patera, 2008).
The focus of POB research and theory falls on four statelike psychological capacities that constitute a higher-order construct: psychological capital (PsyCap), consisting of hope, optimism, resiliency and self-efficacy . It has conceptually and empirically been shown that PsyCap is the underlying second order construct, with better predictive power than any of the individual constructs (Luthans, Avolio, Avey & Norman, 2007). Work engagement is considered by Bakker and Demerouti (2008, p. 209) to be 'a positive, fulfilling, work-related state of mind' predicted by job and personal resources (e.g. optimism, self-efficacy and self-esteem). The purpose of this research was to empirically investigate the relationship between PsyCap and work engagement, over time.

Trends from the research literature
Work engagement is aligned with POB, as both engagement and the facets of POB are considered to be state-like positive psychological capacities . Engagement is, however, considered to be more stable and longer lasting (Hallberg & Schaufeli, 2006). Engaged employees use resources such as optimism, selfefficacy, resilience and an active coping style to assist them to manage and influence their work environment with more success (Bakker & Demerouti, 2008;Luthans, Norman, Avolio & Avey, 2008). Both PsyCap and engagement are also known to have a relationship with and impact on organisational behaviours and outcomes (Bakker & Demerouti, 2008;Stajkovic & Luthans, 1998). In their meta-analysis, Avey, Reichard, Luthans and Mhatre (2011) illustrate that PsyCap relates positively to attitudes such as job satisfaction, organisational commitment and psychological well-being at work, and negatively to employee cynicism, turnover, stress and anxiety. It also relates positively to employee behaviours such as organisational citizenship and negatively to deviance. Engagement has been shown through meta-analysis to relate to indicators of performance such as customer satisfaction, turnover, safety and productivity (Harter et al., 2002). Moreover, in research conducted by Bakker, Gierveld and Van Rijswijk (2006), employees who used their resources optimally scored the highest in engagement; the researchers concluded that optimism, self-efficacy and resilience contribute specifically to engagement. Luthans (2002a) makes it clear that PsyCap is measurable, is based on sound theory and research and is clearly differentiable from populist positively orientated personal development approaches. Its developmental nature requires the PsyCap construct to be potentially state-like and therefore rules out the fixed trait-like personality, attitudinal and motivational variables associated with traditional organisational behaviour. This positive approach could therefore be applied to organisational behaviour as it supports a theory and research-driven point of view and methodology about old as well as new organisational behaviour concepts such as confidence, hope, optimism, happiness and resiliency (Luthans, 2002a;2002b).

Research objectives
There is a well-established notion that investment in human capital is crucial for organisational success in the competitive modern business environment . Luthans (2012) requires psychological capital to be open to development (i.e. state-like rather than traitlike), and to have a positive impact on especially employee performance, but also attitudes and behaviours. Similarly, engagement is conceptualised as being more state-like, and open to development (Schaufeli, Bakker & Salanova, 2006). Both PsyCap and engagement are constructs that in their aggregate form are made up of positively orientated statelike constructs, which can be developed and could contribute to positive work outcomes. However, establishing the temporal order of these variables has not previously been possible with cross-sectional data. The objectives set for this research are therefore to establish if there is:

Contribution of the study
Research to determine the relationship between PsyCap and engagement is important for both theoretical development and the management of quality of working life of employees.
Meeting the research objectives could contribute to training programmes aimed at improving employees' work engagement and psychological capital. From a POB perspective, the investigation is important as Luthans (2002b, p. 698) argues that 'a proactive positive organisational behaviour approach' is what is needed for contemporary and global business to survive. From an engagement perspective, Brim and Asplund (2009) point out that research has indicated that customers experience poor service when they are served by disengaged employees. Salanova et al. (2005) point out that for service workers, work engagement forecasts a service climate, which in turn is indicative of performance by the employees and loyalty of customers. Engagement can therefore make a true difference to employers as it contributes to organisational effectiveness and may present a competitive advantage . The question of how PsyCap and engagement relate to each other has remained mostly theoretical. The question concerning causality however has practical importance too, as this has a direct implication for future human resource development efforts (i.e. should we train to enhance PsyCap and facilitate engagement, or does engagement lead to PsyCap in the long term?). The objective of this study was to investigate the causal relationship and temporal order in the relationship between PsyCap and engagement with a longitudinal survey that tested the crosslagged effects between two measurements.

Review of the literature
Business environments are changing globally with subsequent changes in the psychological contract between employees and employers (Rothmann & Joubert, 2007). South Africa is not excluded from these challenges and South African organisations experience increasing pressure to improve their performance and sustain their competitiveness (Coetzee & De Villiers, 2010). Maslach, Schaufeli and Leiter (2001) and  point out that more is expected of employees in terms of time, effort, knowledge, skill, innovation, flexibility and speed-to-market, whilst job security, career opportunities and lifetime employment are deteriorating.
Although work is an economic activity, most people regard it as more than just an activity to provide n their daily livelihood. Recent studies (Fields, Wilder, Bunch & Newbold, 2008) suggest that especially younger (Generation X and Millennial) employees seek more than just a pay cheque. In addition to work being an important source of people's economic livelihood, being employed can also be seen as contributing significantly to people's identity (Ibarra, 2002). Although employees are therefore looking to do well and thrive and want to be completely engaged in their work (Loehr & Schwartz, 2003), research indicates that only 31% of employees worldwide are engaged, whilst 17% are actually disengaged (Blessingwhite Research, 2011).

Work engagement
Research has now clearly established the energy and identification dimensions of employees' work experience as described by the burnout and engagement phenomena (Schaufeli & Bakker, 2013;Demerouti, Mostert & Bakker, 2010;Gonzaléz-Roma, Schaufeli, Bakker & Llorets, 2006;Xanthopoulou, Bakker, Kantas & Demerouti, 2012).  and Demerouti et al. (2010) concluded that exhaustion and vigour likely represent separate energy constructs, whilst cynicism and dedication can rather be thought of as opposite ends of the continuum.
Most researchers agree that engaged employees reflect high levels of energy and a strong identification with their work . Engagement is seen as developing from a perspective of positive psychology as it also focuses on human strengths and optimal performance rather than on weaknesses and malfunctioning (Seligman & Csikszentmihalyi, 2000). It is also regarded as a positive organisational behaviour construct (Bakker & Demerouti, 2008).
According to Hallberg and Schaufeli (2006), engagement stresses the notion of positive attachment and optimal performance in the work environment in terms of well-being, with high levels of energy, involvement and commitment invested in one's work. Engagement is thus a positive, workrelated state of well-being or fulfilment, where engaged employees have high levels of energy, are enthusiastic about and show strong identification with their work Maslach et al., 2001). Engagement is thus a positive experience in itself (Schaufeli, Salanova, González-Romá & Bakker, 2002) and able to facilitate job and personal resources Hakanen, Schaufeli & Ahola, 2008). It is further associated with employees who are strongly attached to their work roles, by being physically involved, cognitively vigilant and performing their tasks with total emotional involvement (Coetzee & De Villiers, 2010), and who are prepared to go above and beyond typical in-role performance (Macey & Schneider, 2008). Furthermore, engaged employees view themselves as capable of handling their job demands successfully , contributing to higher levels of productivity and profitability, increased safety, greater attendance and retention (Fleming & Asplund, 2007). Luthans (2002aLuthans ( , 2002b initially applied ideas emanating from positive psychology to the workplace in his conceptualisation of POB, and refined his ideas into the more specific PsyCap soon thereafter Luthans & Youssef, 2004;Luthans, Youssef & Avolio, 2007). Positive organisational behaviour is reconcilable with positive psychology, because positive psychology emphasises the study of human strengths and virtues, with the aim of understanding and facilitating positive developmental outcomes (Seligman & Csikszentmihalyi, 2000). POB is intended to focus on a positive approach to developing and managing human resources in the modern work environment (Gardner & Schermerhorn, 2004;Luthans, Avolio, Avey & Norman, 2007).

Positive organisational behaviour
According to Luthans (2002aLuthans ( , 2002b and , POB can be managed successfully to ensure an improvement in employees' performance. The principle contribution of POB furthermore lies in the fact that it is generative and contributes to optimal functioning , in the sense that the application of positive psychology in the workplace, as POB, emphasises the significance of a positive approach .

Theoretical positioning
The Conservation of Resources theory ([COR]; Hobfoll, 2001; sees the attainment and preservation of resources as the prime human motivation. Individuals seek to obtain resources, but also to apply them in life. The COR has been used to describe the functioning of PsyCap as a higher-order construct, consisting of its constituent dimensions of hope, optimism, resilience and efficacy (Avey et al., 2011). The COR also conceptualises 'gain spirals' (Hobfoll & Shirom, 2000), where positive reciprocal relationships exist between positively oriented individual states. This idea of gain spirals is also used to explain the better predictive power of the second order construct of PsyCap, compared to any of the individual constructs of hope, optimism, resilience and efficacy (Luthans, Avolio, Avey & Norman, 2007;Sweetman & Luthans, 2010).
Engagement is most often conceptualised within the Jobdemands Resources theory ([JD-R]; Demerouti, Bakker, Nachreiner & Schaufeli, 2001). Within this theory, job demands are seen as contributing to burnout, whilst job resources are seen as contributing to work engagement. In turn, burnout relates to negative individual and organisational outcomes, whilst engagement is seen as the antecedent of positive outcomes .
In line with the idea of gain spirals, Sweetman and Luthans (2010, p. 57) conceptualise the relationship between PsyCap and work engagement as being reciprocal. Specifically, they theorise that vigour will be relating to 'efficacy in motivating the effort, hope in providing the willpower and developing alternative pathways to achievement, optimism in expecting future success and resiliency in the continued pursuit of goals', and dedication to 'the efficacy related to involvement in one's work, optimism in attributions of significance and pride, hope in dedicated waypower and pathways, and resiliency in continuing in the face of challenging obstacles and adversity'. Absorption is seen to relate mostly to efficacy, optimism and resiliency (Sweetman & Luthans, 2010). Conversely, they also see efficacy and resiliency as contributing to all three dimensions of work engagement, optimism to dedication and absorption, and hope to vigour and dedication.  theorise that the relationship between specifically self-efficacy and employee well-being may be reciprocal.
The above has, to our knowledge, received limited attention from researchers. Xanthopoulou et al. (2007) illustrated that personal resources (in their case, self-efficacy, organisationalbased self-esteem and optimism) partially mediated between job resources and work engagement. Luthans (2012) and colleagues (Avey, Luthans & Mhatre, 2008;Avey et al., 2011;Sweetman & Luthans, 2010) have called for more research into the antecedents of PsyCap, and specifically longitudinal research to advance the understanding of the interaction between PsyCap and work engagement. The latter is precisely what we address in this investigation.

Research design
The research will investigate the hypothesised model shown in Figure 1.

Research approach
The researchers made use of a quantitative, longitudinal research design that tested the cross-lagged effects between two measurements. A longitudinal cross-lagged panel design requires the same participants to be assessed on the same variables over time. A cross-lagged analysis makes it possible to investigate the temporal precedence of the cause, which is a necessary condition for causal inference (Cook & Campbell, 1979). The temporal order between PsyCap and engagement was investigated by means of structural equation modelling. Primary data was gathered by means of a penand-paper survey, and variables were represented by items on Likert-type scales, with items construing dimensions, and dimensions summing to represent constructs.

Research participants
During the first measurement (Time 1), surveys were distributed to all 1003 permanent employees working in all departments across all units of the particular business unit, which forms part of a larger chemical factory. A total of 908 completed questionnaires were returned and only 892 of the questionnaires could be used, representing a return rate of 88.9%. During the second measurement (Time 2), 915 questionnaires were distributed to the same study population as in the first measurement. A total of 358 completed questionnaires were returned which represented a response rate of 39.1%. When matching participants from the pre-to the post-measurement was done, only 163 respondents could be matched, representing a response rate of 17.8% (compared to the original 915 potential respondents) of useable data for the longitudinal study.
The results presented below are based on longitudinal data for 163 respondents. The characteristics of these participants are reported in Table 1. Table 1 indicates that the majority of participants represented in the study population were men, which reflects the demographics of the organisation very well, given the fact that the technical, maintenance and operational areas of the organisation are still dominated by men. The race groups are also representative of the actual population of the organisation with the majority of the participants represented in the White group. The average age of respondents was 40.4 years (SD = 11.16 years) and with regard to the years of service, the average tenure was 13.4 years (SD = 10.35 years). The majority of the participants (91.4%) were on supervisory and non-managerial levels.
Responses are on a five-point Likert scale, ranging from 1 ('strongly disagree') to 5 ('strongly agree'). A high score reflects high levels of engagement. A typical question for the vigour dimension, which has six questions, is: 'I am bursting with energy in my work'. A typical question for the dedication dimension, which has five questions, is: 'My job inspires me'. The alpha coefficients for the three subscales varied between 0.80 and 0.90 (Schaufeli & Bakker, 2003). Storm and Rothmann (2003)  A measure of PsyCap, derived from the four constructs (hope, optimism, self-efficacy and resilience) that form the PsyCap Scale (Luthans, 2002a), was constructed and validated in the current sample (De Waal, 2011). The three items that represented the respective individual scales best, based on factor analytic results, were taken to form the new scale. A reliability coefficient of 0.69 was obtained in the South African sample, and the factor structure was established with confirmatory factor analysis. Response options were also on a five-point Likert scale, ranging from 1 ('strongly disagree') to 5 ('strongly agree').
A biographical questionnaire was developed to gather information regarding the demographic characteristics of the participants. Information gathered included age, gender, race, job level and specific work function in the business under investigation, as well as years employed in the organisation.

Research procedure
For practical reasons, data gathering for the first measurement (Time 1) took place once a week over a period of 14 months

Statistical analysis
Current debate exists about more traditional, 'frequentist' approaches to statistical analysis, especially in psychology (Bem, Utts & Johnson, 2011;Lee & Wagenmakers, 2005;Wagenmakers, Wetzels, Borsboom & Van der Maas, 2011, Van de Schoot, Hoijtink & Jan-Willem, 2011. Wagenmakers et al. (2011) argue for the use of a Bayesian statistical approach, whilst other authors (Bem et al., 2011) are more moderate in their view, considering both approaches as applicable. However, said authors also note the intrinsic subjectivity in statistical model building, and consider specifying a prior distribution informative. Van de Schoot et al. (2011, p. 1) have called for a move from the traditional approach of rejecting the null hypothesis to the 'informative' hypothesis. This is also termed the 'specified experimental hypothesis' (Bem et al., 2011, p. 716). This approach hails from Bayesian statistical methodology and, in layman's terms, allows researchers to account for what is already known. A Bayesian approach allows for accounting for existing knowledge in model-building, or alternatively stated, 'data-driven model building' (Vrieze, 2012, p. 233). The Bayesian view is in fact that model parameters should be considered as random variables themselves, but with a probability distribution that may be informed by the true unknown fixed parameter values (cf. Vrieze, 2012). Especially with small sample sizes, as is the case in our analysis, there is less certainty that confidence intervals will have good coverage and that point distributions are unbiased (Asparouhov & Muthén, 2010). The Bayesian approach can effectively deal with any sample size (Lee & Wagenmakers, 2005). In combination with equality constraints, this approach allows for a parametric bootstrap imposed on regression coefficients. Van de Schoot and Strohmeier (2011) have illustrated this approach to yield gain in statistical power. The most commonly used model selection procedures are the Akaike information criterion (AIC), Bayesian information criterion (BIC) and the deviance information criterion ([DIC]; . Here, we have opted to employ the BIC. Vrieze (2012) has presented some evidence that, especially in small samples, and with less complicated models, the BIC may be more appropriately used.
The Mplus 7.11 was used for the analyses . The Bayesian estimator as implemented in Mplus was used for estimation. After estimation, parameter trace plots and kernel density plots were inspected to confirm mixing (Asparouhov & Muthen, 2010). Descriptive statistics (e.g. means, standard deviations, skewness and kurtosis) and inferential statistics (e.g. correlation analyses) were used to analyse the data. Cronbach alpha coefficients were employed to determine the internal consistency, homogeneity and unidimensionality of the measuring instruments (Clark & Watson, 1995). Pearson's product-moment correlation coefficients were used to specify the relationship between the variables. Effect sizes (Steyn, 1999) served to decide on the practical significance of the findings.
A cross-lagged model was used to examine the temporal order in the relationship between PsyCap and engagement. Cross-lagged models enable the researcher to examine the temporal order in the relationships between the variables (Cook & Campbell, 1979

Results
The first step in our analysis focused on finding the best fitting measurement model. Based on the larger Time 1 sample, we tested competing measurement models with confirmatory factor analysis, with a Bayesian estimator. Three models were compared using the BIC. Results are presented in Table 2.
The BIC values reported in Table 2 indicate that the model in which both engagement and PsyCap are represented by one-factor models best represents the data, given fit and parsimony. Subsequent analyses are based on this model. Table 3) we report item loadings for the various items that represented the model.

Next, (in
The factor loadings and intercepts reported in Table 3 became the informative priors in the longitudinal measurement model. Residual variances were specified with noninformative priors (Mplus default). Loadings, intercepts and variances were parameterized to be equal across time points. Thus, the measurement model was assumed to be invariant across time points. This was assumed and not tested, because of the small sample size.
Descriptive statistics, Cronbach's alpha coefficients and the correlation coefficients between PsyCap and work engagement dimensions (n = 163) obtained at both measurements (Time 1 and Time 2) of the longitudinal study are reported in Table 4.

Model BIC
One-factor engagement, one-factor PsyCap 46695.517 One-factor engagement, four-factor PsyCap 46746.365 Two-factor engagement, four-factor PsyCap 46772.332 PsyCap, psychological capital. Table 4, the work engagement measure shows positive kurtosis at Time 1 and Time 2, whilst the PsyCap measure shows positive kurtosis at Time 2. The PsyCap measure presents with lower reliability scores, whilst work engagement indicates excellent reliability. Time 1 and Time 2 data indicate a positive practically significant correlation of medium effect between PsyCap and work engagement.

As indicated in
The final step in our analysis was to estimate the most appropriate structural model, given the available data. We compared models in a cross-lagged fashion (see Table 5).
It is seen in Table 5 that the structural model that best fits the data is the model that allows stability as well as significant effects from work engagement at Time 1 to PsyCap at Time 2, based on the BIC values. To follow, in Table 6, we report descriptive values for the best fitting model. Table 6 that the work engagement (T1) > PsyCap (T2) model has a standardised beta value of 0.195. The credibility interval remains positive.

Discussion
The objective of this study was to conceptualise and investigate the causal relationship and temporal order in the relationship between PsyCap and work engagement with a longitudinal survey and cross-lagged research design.
As indicated by the correlations, the relationship between PsyCap and work engagement was positive for both Time 1 and Time 2. This finding is in line with other research (Simbula, Guglielmi & Schaufeli, 2011). The use of a longitudinal research design that tested the cross-lagged effects between the two measurements yielded an important new finding in the understanding of the relationship between PsyCap and engagement in this specific study population. Results revealed that PsyCap at Time 1 did not predict engagement   T1  T2  T1  T2  T1  T2  T1  T2  T1  T2  T1 Hakanen et al., 2008). Cordery (2007) also found engagement to be a strong predictor of hope, optimism and self-efficacy. Our results did not indicate a significant effect of PsyCap on work engagement, over time, which is in opposition to Bakker, Gierveld and Van Rijswijk (2006), who found that optimism, self-efficacy and resilience contributed to engagement.
Although PsyCap is associated with personal resources, and engagement with work-related phenomena, engagement predicted PsyCap, which could theoretically be understood with the idea of gain spirals, as postulated in the Conservation of Resources theory (Hobfoll, 2001;. Within this framework, the current results are taken to indicate that work engagement facilitates the building of PsyCap. Admittedly, the additional illustration of the interaction of PsyCap with work engagement over time would have been stronger proof of a reciprocal relationship. A possible interpretation of this finding is that individuals are more actively engaged in daily working life than in actively building personal resources (i.e. PsyCap). Taking a more pragmatic view, this also makes sense: work is a major factor through which individuals not only sustain their lives, but also construct their identities (Ibarra, 2002). Bakker, van Emmerik and Euwema (2006) have also described the crossover phenomenon, which implies that an individual's working life should not be considered as an isolated phenomenon, separate from the rest of their life experience. It is conceivable that positive, work-related resources such as work engagement should affect the individual over time to such an extent that it has an impact on their personal resources (PsyCap). We could advance here that based on these assumptions, the relationship over time from work engagement to PsyCap should at least be stronger, compared to the relationship over time from PsyCap to work engagement. This finding further highlighted the importance of investigating possible ways of promoting work engagement in the working environment.

Limitations
We found that one-factor models of our variables of interest (i.e. PsyCap and work engagement) best fit the data. Although this view on the variables is defensible, with PsyCap and work engagement both presenting second-order constructs construed of sub-dimensions, it did limit us in terms of our investigation of the relation between the respective subdimensions. What we thus represent here is a limited view on super-order level and intricacies of the relationship between sub-dimensions may have been overlooked.
Due to practical constraints, the time lag between our first and second data gatherings is unequal. The longest period from the first measurement to re-evaluation was 21 months, and the shortest period seven months. The extended period in which the data was collected and the differences in time lag could have influenced the results. Although the longitudinal design is also a strong point of this study, the influence of the variance from initial to reassessment has not been accounted for. Also, our choice of fit statistic (BIC) may be insensitive to smaller effects (cf. Vrieze, 2012).
The present study has been limited to participants employed in a specific division of a large chemical plant and the findings can therefore not be readily generalised to other occupational groupings in the particular manufacturing facility, or other organisational contexts. Indeed the mostly highly technical nature of the work performed may be deemed supportive of our finding that the relationship from work engagement to PsyCap is strongest over time, in the sense that mastering complex tasks may contribute directly to building personal PsyCap. Similar investigations would therefore have to be conducted in other organisations to verify these findings and to make more general conclusions concerning the relationship between constructs measured in this study.

Suggestions for future research
We must note that our view of the relationship between these variables was very limited, in the sense that we did not consider any antecedents or outcome variables. Although we took a very narrow focus on the relationship between PsyCap and work engagement, more work with larger models is indicated. Also, future researchers would do well to investigate cross-lagged effects between such antecedents and outcomes and the dimensions of PsyCap and work engagement.
Our results indicate that the relationship between work engagement and PsyCap over time may be stronger than the inverse relationship, at least in this sample. However, to truly illustrate this relationship, and specifically test for the possible reciprocal nature thereof, one would need to test these relationships with multiple data points.

Conclusion
The management of employee engagement is again highlighted here. Job resources such as social support from colleagues and supervisors, performance feedback, opportunities to apply a wide variety of skills and tasks, autonomy, learning and development opportunities, coaching and positive work experiences are all positively related to engagement (Bakker & Demerouti, 2008;Koyuncu, Burke & Fiksenbaum, 2006;Schaufeli & Bakker, 2003), and are noted as avenues for possible intervention in facilitating employee engagement. Closely related, PsyCap can be enhanced through task-mastery experiences, positive role modelling, goal setting, contingency planning and social support activities . Managers tasked with developing employees would do well to incorporate such activities into their employee training and development initiatives.