Predicting Behavioural Intention of Manufacturing Engineers in Malaysia to Use E-Learning in the Workplace

This study aims to understand factors that affect the behavioural intention of manufacturing engineers in Malaysia to use e-learning in the workplace. Two hundred usable online questionnaires were collected from respondents who were engineers in Malaysian manufacturing companies. The data were analyzed using SPSS and Smart PLS version 3.2.6. Results supported all direct relationships except for the influence of prior experience in perceived ease of use. Interestingly, perceived usefulness and perceived ease of use fully mediated between computer self-efficacy and behavioural intention to adopt. The study provides theoretical implication to the technology acceptance model by confirming the mediating role of perceived ease of use and perceived usefulness in the context of a manufacturing setting in an emerging market. In practical terms, the study provides insights to guide organizations in designing e-learning systems that are well-received by employees at the workplace.


Introduction
In today's dynamic business environment, organizational learning is key to achieving sustainable competitive advantage. Through organizational learning, businesses are able to increase employees' competencies, enhance decision making, and fulfill needs such as cost effectiveness while addressing knowledge gaps (O'Brien, McCarthy, Hamburg, & Delaney, 2019). As such, organizations should focus on workplace learning processes where organizational learning effectiveness determines competitiveness.
Information technology has enabled knowledge to be created, saved, and shared in the workplace (Yoo & Huang, 2015). The evolution of digital technology has made a wide range of tools and applications available in the market to enhance teaching and learning processes (Werkel, Schmidt, Dikke, & Schwantzer, 2015) such that e-learning now greatly influences all learning that takes place in organizations. Batalla-Busquets and Pacheco-Bernal (2013) found European bank workers perceived elearning as a more adaptable and current training methodology. Teräs and Kartoğlu (2017) discovered that a well-designed e-learning process is similar to the way professional learning occurs in workplace settings. Ravenscroft, Schmidt, Cook, and Bradley (2012) recommended organizations effectively design social media for workplace learning.
Driven by the dynamism of organizations' human capital development in the digital era as well as the new norm of workplace training post-Covid-19, this study aims to address current research gaps and further explore the interplay of key determinants that may predict behavioural intention to use elearning in the workplace.
One of the Malaysian government's strategies under the Ninth Malaysia Plan was to build world-class human capital. E-learning was at the forefront of this strategy and the government took initiatives to evidence on e-learning in the workplace is scarce and outdated (Ong, Lai, & Wang, 2004;Hashim, 2008;Veloo & Masood, 2014). Therefore, this study aims to uncover the factors contributing to behavioural intention to use e-learning at the workplace among manufacturing engineers in Malaysia.

Significance of Study
Theoretically, this research integrated social cognitive theory (SCT) and the technology acceptance model (TAM) in predicting behavioural intention to use e-learning in the workplace, extending TAM with aspects of SCT that reflect personal, cognitive, and environmental factors. Furthermore, this study tested the mediating roles of perceived ease of use and perceived usefulness in furthering TAM.
This research is timely and contributes practically as e-learning has become crucial in developing employees' competencies to suit organizational needs in the current business dynamic. Uncovering the significant factors and addressing them allows organizations to ensure increased use of workplace elearning through promoting behavioural intention among employees.
Despite having been conducted in Malaysia, findings from this study can be applied to emerging markets in other regions as the data were collected mainly from manufacturing engineers working in multinational corporations (MNCs). Manufacturing firms operating in emerging countries are largely MNCs from developed countries that are more advanced in terms of using e-learning systems to train technical workforces such as engineers.

Supporting Theories
Social cognitive theory (SCT) focuses on knowledge acquisition and internal mental structures (Bower & Hilgard, 1980) and emphasizes what learners know and how they come to get knowledge from the perspective of the interrelationship between the personal, behavioural, and environmental (Jonassen, 1991). The knowledge acquired potentially alters learners' subsequent behaviour. Among the important factors examined within SCT studies are computer self-efficacy and prior experience. The technology acceptance model (TAM) examines perceived usefulness and perceived ease of use as two unique variables which are theorized to explain users' behavioural intention toward a technology.
Perceived usefulness investigates the degree to which users think that the technology can enhance an outcome, while perceived ease of use concerns the degree to which users think that minimal effort is needed to adopt a technology (Davis, 1989).
As SCT is commonly used in studies involving learning, knowledge management, human resource development, and career development (Kim & Park, 2018) and TAM is fundamental in exploring determinants of information technology usage (Choudhury & Pattniak, 2020;Zheng & Li, 2020), integrating these theories is appropriate to investigating behavioural intention to use e-learning. This study adopted computer self-efficacy and prior experience from SCT as well as perceived usefulness and perceived ease of use from TAM to form an integrated research model.

Behavioural Intention to Use E-Learning in the Workplace (BI)
E-learning refers to education or learning via digital technology to acquire new skills and knowledge, and its success depends on how it's used. Behavioural intention to use an e-learning system refers to an individual's perceived likelihood of having the intention to use an e-learning system (Rezaei Dolatabadi, Ranjbarian, & Zade, 2012) and can be affected by social, technological, and organizational factors (Choudhury & Pattniak, 2020).
Behavioural intention towards e-learning was found to be positively influenced by perceived usefulness (Zheng & Li, 2020) and negatively influenced by attitude (Lee, Hsieh, & Chen, 2013) or positively influenced by both perceived usefulness and attitude (Boateng, Mbrokoh, Boateng, Senyo, & Ansong, 2016). Perceived usefulness and perceived ease of use were found to increase behavioural intention to use e-learning (Ndubisi, 2004). Lee, Hsieh, and Ma (2011) examined the influence of individual, organizational, and task characteristics, as well as subjective norms as determinants of behavioural intention to use e-learning in an organizational context. Garg and Sharma (2020) found ease of use influenced user satisfaction which resulted in users' intention to continuously use e-training in the workplace.

Perceived Ease of Use (PEU)
Perceived ease of use (PEU) refers to an individual's degree of belief that using a system is effortless (Boateng et al., 2016). PEU and perceived usefulness (Zheng & Li, 2020) positively impact both behavioural intention to use e-learning and actual usage (Rahmawati, 2019). Nevertheless, it has been found that PEU positively influences behavioural intention only through perceived usefulness and a positive attitude toward e-learning (Lee et al., 2013).

Perceived Usefulness (PU)
predicts perceived credibility negatively, which, in the end, positively influences behavioural intention to use an e-learning system (Ong et al., 2004). CSE has marginal influence on perceived behavioural control, and the perceived behavioural control strongly predicts the behavioural intention of e-learning usage (Ndubisi, 2004). This relationship is also mediated through PEU (Boateng et al., 2016). Kim and Park (2018) found CSE strongly predicted the behaviour intention for both learners and instructors, either as a mediating or direct variable.
When individuals have higher CSE, they are more likely to feel that a computer is easy to use, which causes them to perceive that an information system is also easy to use. Past studies have found that CSE has a positive influence on PEU (Boateng et al., 2016;Hsia et al., 2014;Ong et al., 2004).
When individuals have CSE, their ability to use an information system increases and influences their perception of an e-learning system's usefulness. Past studies have revealed that CSE has a positive relationship with PU (Zogheib, Rabaa'i, Zogheib, & Elsaheli, 2015;Chen & Tseng, 2012). Thus, we postulate that: Hypothesis 1: There is a positive and significant relationship between CSE and PEU.
Hypothesis 2: There is a positive and significant relationship between CSE and PU.

Prior Experience (PE)
Prior experience (PE) refers to previous e-learning involvement and has been found to: (a) positively influence e-learning effectiveness (Haverila, 2011); (b) improve perceived behavioural control (Ndubisi, 2004); and (c) affect perceived usefulness and perceived ease of use, which, in turn, influence behavioural intention (Lee et al., 2013). Lee et al. (2013) argued that PE has a positive and significant influence on both PEU and PU; as individuals gain more experience on how to use an e-learning system, they become more familiar with that system and then tend to perceive that the system is easy to use. PE shows that users who have previous exposure to e-learning already have some knowledge on how to use a new e-learning system.
Hence, it will be easier to use e-learning. Lee et al. (2013) provided evidence that PE has a positive relationship with PU. Thus, we postulate that: Hypothesis 3: There is a positive and significant relationship between PE and PEU.
Hypothesis 4: There is a positive and significant relationship between PE and PU.

PEU and PU
Past studies have indicated that when an individual has PEU of an e-learning system, he/she will also have PU towards that e-learning system (Zogheib et al., 2015). Furthermore, several studies have proven that PEU has a positive and significant relationship with PU (Boateng et al., 2016;Lee et al., 2013;Ong et al., 2004). Thus, we postulate that: Hypothesis 5: There is a positive and significant relationship between PEU and PU.

PEU, PU, CSE, PE, and BI
When individuals find that e-learning is easy to use, their intention to use the e-learning system will increase. Al-Gahtani (2016) discovered that PEU positively influences behavioural intention. Most studies have supported the idea that PEU has a positive and significant relationship with BI (Lee et al., 2013;Garg & Sharma, 2020).
When individuals find an e-learning system useful, they will have the intention to use the e-learning system. Lee et al. (2013) highlighted that PU has a positive and significant relationship with behavioural intention to use e-learning. Other studies (Ong et al., 2004;Al-Gahtani, 2016) have also supported this finding.
However, Yuen and Ma (2008) found no significant direct relationship between CSE and BI, while Md Ariff, Yeow, Zakuan, Jusoh, and Bahari (2012) found only an indirect relationship between CSE and BI, which was through PU and PE.
Prior experience was found to positively influence behavioural intention to use e-learning in education (Lee, et al., 2013;Ndubisi, 2004) as well as in the workplace (Wang & Lin, 2014). Thus, we postulate that: Hypothesis 6: There is a positive and significant relationship between PEU and BI.
Hypothesis 7: There is a positive and significant relationship between PU and BI.
Hypothesis 8: There is a positive and significant relationship between CSE and BI.
Hypothesis 9: There is a positive and significant relationship between PE and BI.

Perceived Ease of Use (PEU) as Mediator
Past studies have indicated that PEU is a good mediator (Venkatesh, 1999). Furthermore, positive PE of e-learning was revealed to have positive influence on the adoption of online learning (Burton-Jones & Hubona, 2006). Haverila (2011) highlighted that PE is very important to determine e-learning outcomes. Kim and Park (2018) found CSE directly and indirectly influences adoption of e-learning.
Thus, we postulate that: Hypothesis 10: PEU mediates the relationship between CSE and PU.
Hypothesis 11: PEU mediates the relationship between CSE and BI.
Hypothesis 12: PEU mediates the relationship between PE and PU.
Hypothesis 13: PEU mediates the relationship between PE and BI.

Perceived Usefulness (PU) as Mediator
As a mediator, PU positively influences the relationships of motivation, computer anxiety, Internet selfefficacy (Hashim, 2008;Joo et al., 2017), PEU, and attitude (Boateng et al., 2016) with behavioural intention. Several studies have indicated that PU is a good mediator (Venkatesh, 1999;Burton-Jones & Hubona, 2006). Wei (2009) claimed that PU mediated the relationship of perceived compatibility of a technology system and BI. Huynh and Thi (2014) found PU mediates the relationship between PEU and BI. Thus, we postulate that: Hypothesis 14: PU mediates the relationship between CSE and BI.
Hypothesis 15: PU mediates the relationship between PE and BI.
Hypothesis 16: PU mediates the relationship between PEU and BI.
Therefore, we developed a research model as follows (see Figure 1).

Methodology
This quantitative study used non-probability purposive sampling and collected two hundred effective responses from engineers working in Malaysian manufacturing companies via an online survey conducted in November 2019. The unit of analysis was the individual. The first two sections of the questionnaire explored the organizational background, e-learning system used, and demographic information of the respondents. The third section included questions related to the predictors and mediating variables. The last section included questions on the dependent variable. A Likert-type scale was used to rate the items in each variable. Measurement items were adapted from Lee et al. (2013), Kim and Park (2018), and Hsia, Chang, and Tseng (2014). The survey instrument was reviewed by two experts to determine its validity and clarity.
Statistical Package for Social Science (SPSS) and Smart PLS 3.2.6 were used to analyse the data. SPSS was used to do frequency and descriptive analysis as well as testing for common method variance (CMV). Smart PLS was used to analyse the measurement and structural models. The measurement model measured the outer model for validity and reliability, including discriminant validity, convergent validity, composite reliability, construct validity, and reliability analyses. The structural model measured the inner model on regression analysis and tested the hypotheses. Path coefficients were obtained in terms of direct and indirect effects. T-values and p-values were also acquired to determine whether the hypotheses should be rejected or accepted. The confidence level was set at 95%, with a significance level of 0.05 under a one-tailed test. In this study, PLS was used to perform a nonparametric bootstrap procedure based on the assumption that data were not normally distributed.

Profile of Respondents
The majority of respondents were male (61%), within the age group of 26 to 30 years old (40.5%).
Furthermore, 54.5% were single and possessed a bachelor's degree (68%). Most (72%) worked in large manufacturing organizations, and 79% of the respondents' companies had developed in-house elearning systems. Slightly more than half of the companies (56.5%) did not force their employees to use e-learning systems; rather, it was on a voluntary basis.

Common Method Variance
Data were examined using SPSS; factors were unrotated, and the extraction method was principal component analysis. The results showed that six extracted factors explained 73.65% of the total variances. The first factor explained 45.13% of the variance extracted, which was less than the required criteria of 50%. Hence, there was no single factor representing a majority of the total variance extracted, and the samples were therefore considered to be free from common method bias (Podsakoff, MacKenzie, Lee, & Podsakoff, 2003).

Assessment of the Measurement Model
Hair, Hult, Ringle, and Sarstedt (2016) recommended a cut-off value of 0.50 for significant loadings.
All the individual items of each construct consisted of the highest loadings, and all the cross loadings were lower than the main loadings respectively (see Table 1). Hence, the constructs were valid.
For a variable to demonstrate convergent validity, the average variance extracted (AVE) must be more than 0.50 (Hair, Hult, Ringle, & Sarstedt, 2016). All AVE values were above the 0.50 threshold, indicating the model showed convergent validity.
Cronbach's alpha, composite reliability and Rho_A must be higher than 0.70 to demonstrate the reliability of the research model (Hair et al., 2016). Cronbach's alpha is said to have limitations as it assumes all indicators are equally reliable and have equal loadings on the construct but Rho_A is loading oriented (Hair et al., 2016). Cronbach's alpha, composite reliability and Rho_A values of each variable were higher than 0.70 (see Table 1). As such, the model was considered reliable.

Standardized Root Mean Square Residual (SRMR)
To eliminate model misspecification, the standardized root mean square residual (SRMR) was calculated. The SRMR should have a value of less than 0.10 for reasonable model fit. The SRMR in this study was 0.074 indicating acceptable model fit.

Assessment of Structural Model
To assess the structural model, a one-tailed test was used with a significance level of 0.05, indicating a 95% confidence level (see Table 3). Note. BI = behavioural intention; CSE = computer self-efficacy; PE = prior experience; PEU = perceived ease of use; PU = perceived usefulness. **p < 0.001. *p < 0.05. Table 4 shows the detailed results of mediation analysis. Of the seven hypotheses tested, four were supported as the indirect paths were significant. The lower and upper limits of confidence interval bias did not straddle '0', indicating there was mediation (Preacher & Hayes, 2008).
If the direct relationship between the independent variable and the dependent variable is significant and the indirect relationship is also significant, then partial mediation occurs; however, if the direct relationship is not significant while indirect is significant, then full mediation occurs (Zhao, Lynch, & Chen, 2010).

Discussion and Implications
Hypothesis H1, which describes a positive relationship between computer self-efficacy and perceived ease of use, was supported with a T-value of 8.290 (see Table 3). Consistent with previous studies (Lee et al., 2013;Hsia et al., 2014;Boateng et al., 2016), individuals with computer self-efficacy were found to have higher perceived ease of use with e-learning systems in the context of workplace learning.
Computer self-efficacy exists if an individual can complete certain tasks using a computer. This individual perceives that the problems faced are not difficult although others might feel otherwise (Hsia et al., 2014). With computer self-efficacy, an individual feels more comfortable when using an e-learning system (Lee et al., 2013).
Hypothesis H2, which postulates that there is a positive relationship between computer self-efficacy and perceived usefulness, was accepted with a T-value of 3.301 (see Table 3). This confirms that individuals with higher computer self-efficacy have higher perceived usefulness towards an e-learning system (Zogheib et al., 2015) in a workplace. Computer self-efficacy was a crucial factor to be considered because it had a significant and positive influence on perceived usefulness, consistent with previous findings (Chen & Tseng, 2012;Rabaa'i, 2016). Computer self-efficacy positively predicts perceived usefulness because an individual with high computer self-efficacy also has higher learning effectiveness.
Computer self-efficacy improves knowledge transfer, which leads to better practical usage of technology systems.
Hypothesis H3, which suggests a positive relationship between prior experience and perceived ease of use, was not supported, with a T-value of 0.351 (see Table 3). This means that prior experience had no influence on perceived ease of use which contradicts several past studies discussed earlier. However, this finding may also indicate that, for an individual, prior exposure to or practice with a technology system is not a significant factor that influences the perceived ease of use (Adewole-Odeshi, 2014).
Every individual has different learning characteristics which indirectly influence their thoughts towards ease of use of an e-learning system (Haverila & Barkhi, 2009). Prior experience gained by anyone may be unique. Most of the respondents in this study, being engineers, would have graduated from higher education institutions and therefore would have had experience with e-learning systems.
Hypothesis H4, which assumes a positive relationship exists between prior experience and perceived usefulness, was not supported, with a T-value of 0.832 (see Table 3). Despite having more experience with e-learning systems, in both this study and in Lee et al. (2013), engineers' perceptions of usefulness were not influenced by prior experience.
Hypothesis H5, which postulates there is a relationship between perceived ease of use and perceived usefulness, was supported, with a T-value of 5.866 (see Table 3). This is consistent with previous findings (Chen & Tseng, 2012;Lee et al., 2013;Punnoose, 2012) that have shown that an individual who perceives that e-learning is easy to use also directly perceives that the e-learning system is useful.
Furthermore, this supports the findings of studies that have demonstrated that perceived ease of use has a significant and direct effect on perceived usefulness in e-learning adoption (Al-Gahtani, 2016;Huynh & Thi, 2014;Zogheib et al., 2015).
Hypotheses H6 and H7 predict that behavioural intention to use e-learning at the workplace will be positively and significantly influenced by perceived ease of use and perceived usefulness. Both hypotheses were supported with T-values of 2.974 and 4.071 respectively (see Table 3). The findings revealed that perceived ease of use and perceived usefulness were antecedents to the intention to use an e-learning system at the workplace. Since perceived ease of use, perceived usefulness, and behavioural intention to use e-learning at the workplace are the main constructs in TAM, which is a well-established research model, many past studies have provided evidence of this significant relationship (Chen & Tseng, 2012;Huynh & Thi, 2014;Al-Gahtani, 2016). It is believed that when an elearning system is perceived to be easy to use and provides advantages at the workplace, the probability of intention to use will be higher.
Hypotheses H8 and H9 predict that behavioural intention to use will be positively and significantly influenced by computer self-efficacy and prior experience respectively. Both hypotheses were not supported as the T-values were 0.343 and 0.244 (see Table 3). Although previous studies (Lee et al., 2013;Wang & Lin, 2014;Ndubisi, 2004) have found significant relationships, this study suggests that prior experience does not influence behavioural intention. Surprisingly, computer self-efficacy did not directly influence behavioural intention. Similar observations have been made in previous studies such as Yuen and Ma (2008) and Md Ariff et al. (2012).
The mediating effect of the TAM beliefs (perceived ease of use and perceived usefulness) were examined in hypotheses H10 to H16. Full mediating effects were found for hypotheses H11 and H14, but H10 and H16 showed only partial mediation (see Table 4). This is consistent with past studies (Venkatesh, 1999;Burton-Jones & Hubona, 2006;Svendsen, Johnsen, Almås-Sørensen, & Vittersø, 2013) that have demonstrated that perceived ease of use and perceived usefulness are able to act as mediating variables in TAM. Interestingly, hypotheses H12, H13, and H15 were not supported (see Table 4), which contradicts previous findings (Haverila, 2011;Burton-Jones & Hubona, 2006). In this study, the relationship between prior experience and perceived ease of use was not significant in the context of e-learning in the workplace. Consequently, perceived ease of use did not show a mediating effect between prior experience and perceived usefulness, and between prior experience and behavioural intention.

Theoretical Implications
This study confirms the mediating role of the TAM beliefs (perceived ease of use and perceived usefulness). Both perceived ease of use and perceived usefulness were found to mediate the relationship of the antecedents (except for prior experience) with behavioural intention to use e-learning at the workplace. This finding contributes to TAM as past studies have neglected these two beliefs as mediators. This study has also validated the integrated framework of SCT and TAM in predicting the behavioural intention to use e-learning in the workplace. Past studies were conducted mainly on students in universities, while limited studies were carried out within the industrial or workplace context, adopting either one of these theories independently.

Practical Implications
Findings from this study on the determinants of behavioural intention to use e-learning at the workplace provide guidelines to organizations on significant factors to prioritize when establishing an e-learning system. For instance, it is recommended that organizations ensure computer self-efficacy among employees before implementing an e-learning system. Computer usage culture must be instilled among the employees. Organizations should also ensure the e-learning system is both easy to use and useful for the employees. Organizations can set up a group of pioneer users before officially rolling out the new e-learning system. The pioneer users can try the e-learning system and provide feedback especially in areas of ease of use and usefulness. Organizations can then use this feedback to improve the e-learning system to fit the needs and expectations of employees.

Conclusion and Limitations
Organizational learning is essential to achieving sustainable competitive advantage in the current volatile business environment. Digital transformation in organizations has enabled e-learning at the workplace to drive organizational learning. In gauging behavioural intention to use e-learning, this study developed a research model that integrated SCT and TAM and addressed gaps in previous studies.
In this study, we advanced current understanding by predicting behavioural intention to use e-learning at the workplace among manufacturing engineers in an emerging country. Computer self-efficacy was found to positively influence perceived ease of use and perceived usefulness, while indirectly predicting behavioural intention. While prior experience did not affect perceived ease of use, it influenced perceived usefulness. Perceived ease of use influenced perceived usefulness, and both positively impacted behavioural intention. Most importantly, the roles of perceived ease of use and perceived usefulness as mediators have been confirmed in the context of this study.
The scope and model of this study address the importance of individual differences. The R 2 value of the model is 0.615, where factors explained 61.5 % of behavioural intention to use e-learning systems. Thus, approximately 39 % of variances were not explained. Future studies could consider the role of other factors such as system, environment, and expectation in behavioural intention to use e-learning in the workplace. Furthermore, this study focused on the behavioural intention to use e-learning in organizations. Future studies may extend this model to assess the effectiveness of an e-learning system after usage. This study targets engineers of manufacturing organizations in Malaysia. Future studies could collect data from other industries and countries or regions to explore various contexts.