Investigating Work Engagement of Highly Educated Young Employees through Applying the Job Demands-Resources Model

This study aims to investigate the applicability of the job demands and resources (JD-R)
model. While prior studies tend to focus on older employees, we investigate the potentially
effective job resources and demands specifically for the highly educated young individuals
that are the future of any organization. We collected primary cross-sectional data from 155
respondents in Indonesia with master’s or doctoral degrees, under 40 years of age, currently
working in public and private sectors. Participants completed an online questionnaire by
following a link distributed by email. PLS-SEM was used to analyze data regarding the
relationships between job resources (perceived organizational support (POS) and employee
voice), job demands (emotional demands), and work engagement. Our results indicated the
relationships for POS positively and emotional demands negatively with work engagement,
while employee voice was not significantly correlated with work engagement. Emotional
demands are considered to be perceived as stress rather than opportunity. Furthermore,
emotional demands did not moderate the correlation of POS and employee voice with work
engagement. A multi-group analysis found no significant differences between employees in
the private and public sectors. The results showed the JD-R model was partially applicable.
This study is one of few seeking to apply the JD-R model to highly educated young employees
in the private and public sectors. The specific results of this study will provide insight for the
organizations employing such individuals.

in human resource management (HRM) research (Arrowsmith & Parker, 2013). Work engagement is also a key factor in an organization's sustainability and competitiveness (Saks & Gruman, 2014), as it influences individual and organizational performance (Hoole & Bonnema, 2015). Engaged employees tend to show more positive emotions, optimism, and enthusiasm. Moreover, they can also use better their personal resources; have better health; and tend to transmit engagement to other people in their work environment (Bakker, 2011;Ferreira et al., 2020;Nguyen & Pham, 2020).
Although there is a lack of consensus among researchers regarding the most effective theory for analyzing work engagement, the job demands and resources (JD-R) model has been often used in the empirical studies of engagement (Saks & Gruman, 2014).  argued that this model is useful for HRM, as the organization can change job demands and resources. Furthermore, the JD-R model can reveal influencing factors and results of work engagement and is popular due to its flexibility (Bakker et al., 2014).
Since there is an increasing number of young employees to enter the workforce, particularly in Indonesia, there is a need to understand work engagement for young professionals. However, previous research focused largely on older employees (Avery et al., 2007;Kulik et al., 2016;Mauno & Minkkinen, 2020). Younger individuals tend to have lower engagement than older individuals (Park & Gursoy, 2012). There is a need to uncover factors that contribute to young employees' engagement. Thus, this study investigates the applicability of the JD-R model for younger employees. These employees have different tendencies in perceiving demands.
We focus on those with a minimum of a master's degree because we expect that people with a higher education level tend to have higher aspirations and different expectations in organizations based on their competitive advantages. At the same time, organizations may be interested in hiring and retaining highly educated employees for their innovation and sustainability. Thus, understanding how to improve work engagement in highly educated employees is crucial for organizations.
Employing the lens of JD-R model, we aim to understand how highly educated young employees tend to perceive their demands. Crawford et al. (2010) argued employees may perceive workplace demands as positive (challenges or opportunities) or as negative (stressful event), with varying effects on work engagement. Perceiving demands as stress is likely to negatively influence work engagement. On the other hand, perceiving demands as opportunities is likely to positively influence work engagement. This study aims to elucidate the effect of job demands on highly educated young employees, considering whether demands are perceived as opportunities or stress. Investigating this will be beneficial for future organizational practice on handling these types of employees.
Further, previous research focuses on work engagement in the private rather than the public sector (Vigoda-Gadot et al., 2013). These two sectors differ significantly (Byrne & MacDonagh, 2017;Knies et al., 2015), for example, in terms of their organizational structure and their employees' motivation and expectations (Perry et al., 2010). This study aims to address this gap by including employees in both the public and private sectors.

Literature Review Work Engagement
Most research use the definition of work engagement from the study of Schaufeli et al. (2002), measuring the characteristics of work engagement that consist of vigor, dedication, and absorption. Vigor denotes a "high level of energy and mental resilience while working" (p. 74); dedication refers to an experience of having a "sense of significance, enthusiasm, inspiration, pride, and challenge" (p. 74); and absorption signifies a "state of being fully concentrated and deeply engrossed in one's work" (p. 75). According to Saks and Gruman (2014), research on engagement in the workplace mainly focuses on measuring it, including its determinants and consequences. Prior research revealed that workplace engagement positively affected work outcomes such as performance at both organizational and individual levels. Considering its importance, prior research also focused on the antecedents of work engagement. One major line of theoretical argument on the antecedents of engagement in the workplace is the JD-R model.

Job Demands and Resources Model
Drawing upon the theory of conservation of resources theory (COR) (Hobfoll, 1989), individuals attempted to pursue, maintain, and preserve available resources and accumulate all those with continuous positive effect, which is likely to have both positive organizational and individual wellbeing. Hence, individuals who have more experience of resources positively, are likely to stimulate better relationship between individual and organizational well-being.
The JD-R model is a popular theory for investigating the process toward engagement, with the working conditions divided into two aspects: demands and resources Byrne & MacDonagh, 2017). Both job demands and job resources may include physical, social, or organizational aspects. The difference is job demands require employees to invest extra energy into dealing with the stress related to it. In contrast, job resources offer support in organizational support and positive feedback (Schaufeli et al., 2006).
The JD-R model proposes that job demands and resources influence work engagement (Saks & Gruman, 2014). Bakker et al. (2011) argued that job resources assist work engagement and that insufficient resources may lead to lower engagement, regardless of job demand level (Hakanen et al., 2008). This suggests that adequate perceived resources produce engaged employees. Crawford et al. (2010) also argued job resources were positively related to work engagement. Saks and Gruman (2014) argued that job demands might negatively influence mental and physical resources, leading to disengagement (p. 162). However, Crawford et al. (2010) suggested job demands were positively or negatively related to work engagement, depending on whether they act as an opportunity or a hindrance. Some research based on the JD-R model suggested that the influence of job resources on work engagement is stronger when employees encounter high job demands Bakker et al., 2011).

Applicability of the JD-R Model to Highly Educated Young Employees
There has been still inconclusiveness on the definition of young or old workers (James et al., 2011;Smyer & Pitt-Catsouphes, 2007). Consequently, this study follows James et al. (2011) who defined young workers as anyone under 40 years of age. They argued that differences in how job conditions are perceived are linked to engagement in different age groups, suggesting that the JD-R model's applicability is different for specific age groups. We predict that young individuals can increase their work engagement by utilizing job resources, but they need more resources than older individuals to engage themselves. Specifically, those who are new to entering a workplace need more time to adapt to the workplace, build new relationships, etc. Hence, even though they are provided the same resources, young individuals' engagement will remain lower compared with older individuals.
According to Park and Gursoy (2012), younger individuals tend to place higher value on leisure than work and have lower work centrality. It implies that younger individuals will actively seek working environments that give them more flexible time for leisure, and their jobs are not their central priority. Hence, they may be more affected by insufficient job resources. Similarly, Park and Gursoy (2012) found that young individuals are generally less engaged than older individuals. Their results implied that younger individuals are less likely to allocate their job resources to work since they have low work centrality. Furthermore, young employees had to make essential vocational decisions, form new relationships, and adopt new roles (Akkermans et al., 2013;Elfering et al., 2007;Savickas, 1999). These, particularly for employees entering the workplace after graduation or starting their career with little or no prior experience, are likely to make them disengaged because they have less experience facing the new situation. Moreover, in terms of the interaction of job resources and demands, sufficient resources are likely to facilitate engagement as job resources enable employees to cope with job demands. Fewer resources will hinder individuals' ability to cope with job demands, leaving them less engaged.
Among young employees, the highly educated have some specific characteristics compared with the less educated. Organizations generally prefer to hire younger individuals due to assumptions about their flexibility and adaptability to new technologies (Korsakienė et al., 2019). These assumptions may partly be because young workers are more likely to possess higher education in emerging countries like Indonesia. Moreover, highly educated individuals may feel an obligation to prove their capabilities based on their education level while also believing that they deserve a better job. This is similar with "the sense of competence" (Akkermans et al., 2013). We expect that such a sense of competence influences employees' perceptions of workplace demands. The sense of competence of highly educated young employees is likely to make them perceive demands as stress, which can make them less engaged.

Hypothesis Development
Among many types of job resources, we focus on POS (perceived organizational support); and the ability to contribute to decision-making processes, referred to as employee voice; as these factors were demonstrated to be essential job resources (Byrne & MacDonagh, 2017;Conway et al., 2016). Organizational support involves action from each management level that indicates to the employees that their work is valued, the organization cares about their well-being, and that support is available for employees to engage with their work, job, or organization (Byrne & MacDonagh, 2017;Eisenberger et al., 1986). According to , all aspects, such as appreciation, feedback, autonomy, and performance feedback, affect engagement, although appreciation is the most salient. Byrne and MacDonagh (2017) found that POS has a significant positive correlation with engagement; thus, adequate POS is likely to produce engaged employees. This leads to our first hypothesis: Hypothesis 1: POS has a positive correlation with work engagement.
Employee voice refers to employees' ability to contribute ideas or opinions and feel heard by the management (Farndale et al., 2011). Employees who feel valued and included in organizational matters by being asked their opinions, ideas, and views are more likely to be engaged (Reissner & Pagan, 2013). Rees et al. (2013) found that employee voice was significantly positively correlated with engagement. Conway et al. (2016) indicated that the employee voice mechanism acted as a job resource, strengthening engagement. On the other hand, Byrne and MacDonagh (2017) found that employee voice did not significantly correlate with engagement. As the majority research suggests a positive correlation, our second hypothesis is: Hypothesis 2: Employee voice has a positive correlation with work engagement.
We focused on emotional demands that influence work engagement, as research has shown that positive emotion has a positive correlation with engagement. In contrast, negative emotion contributes to strain, resulting in burnout (Schaufeli et al., 2009). According to De Jonge et al. (2008), emotional demands can be defined as emotionally demanding work, such as controlling emotions and facing negative events in the workplace. While physical demands may not be present in all jobs, emotional demands are expected to exist in most workplaces. Furthermore, emotional demands might be perceived by highly educated young employees as an additional stressor, creating a burden. As such, this group is likely to react negatively to these job demands. According to , the negative relationship of job demands with work engagement may be because employees perceived such demands as stress. Therefore, our third hypothesis is: Hypothesis 3: Emotional demands have a negative correlation with work engagement.
Moreover, job resources and demands interact with each other in predicting well-being in the workplace (Bakker et al., 2014). The JD-R theory states that the relationship between job resources and work engagement is stronger under high job demands, including cognitive and emotional demands, which has been supported through empirical research (Bakker et al., 2014;Hakanen et al., 2005). Previous research indicates job demands have a moderating effect on the relationship, as job resources are crucial and are used to cope with the high level of job demands. Therefore, we hypothesize that: Hypothesis 4: Emotional demands positively moderate the relationship between employee voice and work engagement. Hypothesis 5: Emotional demands positively moderate the relationship between POS and work engagement.

Method Sample and Procedure
This study collected cross-sectional data through an online questionnaire sent to the awardees of the Indonesia Endowment Fund for Education (LPDP scholarship), a government scholarship for pursuing master's and doctoral degrees, for Indonesian citizens. We chose this group because this scholarship is open to all Indonesian citizens, already has alumni, and has a group email list that made the questionnaires more feasible. The inclusion criteria were individuals below the age of 40 currently working in the private or public sector, excluding self-employed individuals. The participants were invited to take the survey through emails sent to the mailing list of the awardees. Participation was voluntary, and all participants provided informed consent. This study was approved by the Research Ethics Board, Faculty of Psychology, of the Universitas Indonesia. Data were collected over two weeks during early 2020. In total, 155 respondents participated in the survey, exceeding the minimum required sample size as suggested if we use partial least squares structured equation modeling (PLS-SEM) to perform the data analysis. This study used PLS-SEM to conduct data analysis.

Measurements
This study used measurements validated through prior studies related to work engagement, job resources (POS and employee voice), and job demands (emotional demands). Questions on work engagement were adopted from the 9-item version of the Utrecht Work Engagement Scale, which includes vigor, dedication, and absorption (Schaufeli et al., 2006). POS was measured using a 9item scale developed by Eisenberger et al. (1986). Employee voice was measured using a 4-item scale developed by Conway et al. (2016). Perceived emotional demands were measured utilizing a 3-item scale from the Copenhagen Psychosocial Questionnaire (COPSOQ) developed by Kristensen and Borg (2003). All questions in the questionnaire were translated into Bahasa Indonesia (the local language) by a professional sworn translator and were back-translated to check the quality of the translation. All items are presented in Table 1.  Table 2 shows the demographics of the respondents. Respondents' average age was 30.19, with the majority in the 28-31 range (35.5%), with the smallest percentage in the 36-39 group (9.7%). The majority of respondents were women (59%). Most of the participants had a master's degree (95%), and 60% worked in the public sector. Furthermore, many of our samples were relatively new in their organization. Many of them were three years or less in their organizations (85.9%).

Testing the Model of Measurement
Based on previous research, this study analyzed the data utilizing the PLS-SEM using Software Smart PLS 3. We investigated the hypotheses and conducted a multiple group analysis to observe correlations across private and public sectors. PLS-SEM was utilized because it can accommodate small sample sizes with the non-normal distribution of variables, does not require many assumptions, and is used for theory confirmation and investigation of relationships between variables in formative constructs (Falk & Miller, 1992). Due to the relatively small number of samples, we used the t-test to check for the late response bias and found that the result was satisfactory.
We conducted the assessment of the measurement model and structural model (Chin, 1998), checking the measurement model by conducting tests of internal consistency/reliability, convergent validity, and discriminant validity (Hair et al., 2016;Ong & Bahar, 2019). The results are displayed in Table 3. As suggested by Chin (1998), the indicator loading value should be above .6 (the cut-off value). Therefore, items with the outer loading below .6 were removed from the main analysis. We assessed the remaining items for internal reliability with Cronbach's α and composite reliability, which should have a minimum value of .7 (Nunnally & Bernstein, 1994). All factors were above .7, excluding emotional demands (.68). We checked convergent validity by measuring the average variance extracted (AVE). All indicators were above .5, indicating that the constructs can support over half of its indicators' variance (Hair et al., 2016). We also measured the Fornell-Larcker, which compares the square root of AVE values of constructs with its correlation with any other construct (Fornell & Larcker, 1981). All square root values of AVE were higher than the correlations between constructs. Based on these tests, we justified the reliability and validity of the measurement model (Hair et al., 2011).
According to Hair et al. (2016), discriminant validity can be confirmed by cross-loading measurement and the Fornell-Larcker criterion (Fornell & Larcker, 1981). All indicators' outer loadings on the associated constructs were greater than all their loadings on other constructs. Therefore, discriminant validity for the constructs was established, as shown in Table 4. The collinearity of the constructs was tested to ensure the path coefficients were not biased (Ong & Bahar, 2019). We found that all values of variance inflation factor (VIF) were below 5, as indicated by VIF guidelines to ensure that the model has non-collinearity (Hair et al., 2016;Ong & Bahar, 2019).

Hypothesis Testing
To assess the structural model, as suggested by Ong and Bahar (2019), we conducted the bootstrapping procedure developed by Preacher and Hayes (2008) to estimate the significance of the path coefficient. The bootstrapping was conducted for 155 cases and 5,000 samples. This allows the standard error estimation to determine the significance of the path coefficients. To measure the accuracy of the model's prediction, we used R 2 (Figure 2). The value of R 2 was .42, indicating that the model was around a moderate level. Moreover, as shown in Figure 2, we also examined the model's predictive relevance with the Q 2 value (Geisser, 1974;Stone, 1974) derived from the blindfolding procedure. The value should exceed 0. Our results showed that the value of Q 2 was .28. According to Hair et al. (2011), this demonstrated that the model has prediction relevance. POS (β = .42, p < .001) and emotional demands (β = -.14, p < .05) were positively correlated and statistically significant as predictors of work engagement, while the other hypotheses were not supported (Table 5).  We conducted a multi-group analysis focusing on private and public sectors (Table 6). There were no significant differences in all paths analyzed between employees in private and public sectors. However, POS for private sector employees showed a significant positive relationship with work engagement, while for public sector employees, there was no significant correlation.

Discussion
This study has successfully answered whether JD-R can be applied to the work engagement of young and highly educated employees in both the private and public sectors. There are at least three contributions related to the JD-R model.
First, this study extends the JD-R model by demonstrating that the role of POS as a resource for work engagement is limited only in the private sector. One possible reason is the bureaucratic nature of the public sector (Boyne, 2002), which is more strict in public sectors than in private sectors. The strict rules and regulations in public sectors may constrain line managers to support their employees (Byrne & MacDonagh, 2017). Especially in the Indonesian context, national rules and regulations for a public servant are applied to every public servant in Indonesia, limiting the kind of support organizations may provide for their employees. POS may not be perceived as a resource for employees in public sectors.
Second, this study shows that employee voice is not perceived as job resources, and this result is consistent both in the private and public sectors. This result is surprising given employee voice is argued as access to decision making and thus provides employees with resources (Byrne & MacDonagh, 2017). Conway et al. (2016) confirmed that argument and demonstrated that employee voice is significantly related to work engagement. The results of this study challenge that assumption. This study shows that, for the current data context in Indonesia, employee voice has no relationship with work engagement. One possible reason is that many individuals in our sample may not feel secure speaking up or may feel unable to change the working environment due to the power imbalance found in Indonesia (Riantoputra et al., 2016). They may prefer not to speak up. Moreover, many of them do not have a lot of experience in their workplace, which may infer the relative power imbalance between these young workers and their senior counterparts. In other words, this study extends the JD-R model by demonstrating that what is regarded as resources may not be identical across cultures.
Third, this study contributes to the JD-R model by demonstrating that job demands are perceived more as stress (not as challenges). That is, highly educated young employees perceive emotional demands as a stressor rather than an opportunity. Previous research indicates that, compared to their older counterparts, young employees tend to have less emotional regulation in relation to high workload (Carstensen et al., 1999;Charles et al., 2001;Mauno et al., 2013). In the same vein, in other studies, young employees may experience more stress at work due to less coping resources and less working experiences than the older employees (Aldwin & Levenson, 2001;Barnes-Farrell et al., 2002;Diehl et al., 1996).
Furthermore, our results revealed that emotional demands did not moderate the relationships between job resources and engagement. This may be because the perception of emotional demands ranges within the moderate level for these groups of individuals. Prior research indicates that high demands may result in strong correlations between job resources and work engagement (e.g., .

Limitations and Suggestions for Future Research
This study had some limitations. In the present cross-sectional design, we were unable to test causality. Future research may want to employ other research design, such as longitudinal research or experimental study. In addition, the sample size was relatively small, which makes the discussion relatively limited. Furthermore, our analysis focused on only a few types of job resources and demands. Future research should incorporate other types of job demands and resources that potentially affect the engagement of highly educated young employees.

Implications
This study found that the types of job resources and demands that correlate with work engagement may vary for different individuals. We found that our results partially confirm the findings of prior studies, especially for employee voice. It implies that the applicability of this theory depends on the types of individuals or even types of workplaces. Therefore, it indicates that every organization needs to identify what types of resources their employees potentially need.
Our findings suggest that organizations need to support employees in order to produce engaged employees, including providing feedback and help, and also to support employees "to apply their full capabilities at work" . Furthermore, organizations should pay more attention to younger employees' ability to handle problems related to high workloads. In other words, organizations need to use different stress management interventions (e.g., part-time work, other forms of flexible scheduling, skill development, training) for older and younger employees and try to prevent breaches of the psychological contract by suiting human resource practices (Peterson & Spiker, 2005) to employees' age-related needs (Bal et al., 2008) and personal goals (Mauno et al., 2013)

Conclusions
This study confirms that the JD-R model is partially applicable to highly educated young employees in Indonesia. Specifically, POS and emotional demands are significantly related to work engagement. However, employee voice is not related to work engagement, indicating that employee voice is not regarded as a resource for young highly educated employees in Indonesia. The current study demonstrates that young and highly educated employees in Indonesia tend to perceive demands as stressors, not as opportunities or challenges to grow. This tendency is related to the extent to which employees are willing to engage in their work. Finally, the current study has advanced the JD-R model by demonstrating that it tends to be more applicable in private sectors than in the public sector in Indonesia.