Impact of job satisfaction on training motivation

This study aims to introduce new insights regarding factors influencing employees’ level of training motivation through investigating the impact of job satisfaction dimensions (pay, fringe benefits, contingent rewards, promotion, supervision, co-workers, operating conditions, nature of work and communication) on training motivation.Data were collected from 342 participants from six Jordanian ministries. The multiple regression technique was utilized to examine the predictive power of job satisfaction dimensions on training motivation. At the next stage, three sequential MR analysis rounds were conducted, each time using a different construct of training motivation (valence, instrumentality and expectancy) as a dependent variable.The results indicate that the dimensions of job satisfaction explain a low but significant variance of the overall training motivation model. Furthermore, it was found that only three dimensions of job satisfaction (nature of work, supervision and co-workers) respectively had a positive and significant impact on training motivation, while contingent rewards had a significant but negative impact.Regarding training motivation constructs, results indicate that the nature of work and supervision were the elements that have an impact on all constructs of training motivation. Finally, it was found that co-workers and contingent rewards had a significant impact on the training motivation constructs of expectancy (confidence in ability to learn) and valence (the perceived value of training outcomes), but no impact on instrumentality (rewards associated with learning).

In this context, Carlson, Bozeman, Kacmar, Wright, and McMahan (2000) assume that an individual's level of TM influences training's overall effectiveness. Zaniboni, Fraccaroli, Truxillo, Marilena, and Bauer (2011) claim that researchers have suggested that TM is a predictor of various training outcomes such as trainees' participation, preparation, affective and utility reactions, as well as knowledge and skills transfer. Moreover, it has been found that TM influences cognitive, skill training results and training transfer (Quinones, 1995 Empirical work on TM follows two approaches: the first approach deals with individual and situational factors' influence on TM. The other approach involves identifying TM predictors and their relationships with training outcomes, i.e. learning (Colquitt, LePine, & Noe, 2000;Medina, 2016).
Regarding the first approach, individual characteristics that influence TM involve first, demographic variables (e.g., Clarke & Metalina, 2000) and second, personality variables including: locus of control (e.g., Noe & Schmitt, 1986), achievement motivation (e.g., Mathieu Tannenbaum, & Salas, 1992;Carlson et al., 2000), anxiety (e.g., Webster & Martocchio, 1993), and self-efficacy (e.g., Noe & Wilk, 1993;Carlson et al., 2000;Switzer et al., 2005). Beier and Kanfer (2009) claim that the impact of individual antecedents on the choice to attend training is not consistent nor very strong. They added that this could be explained by the fact that these variables do not directly consider the content and the purpose of the training or the organizational environment. In this context, and although there have been several efforts to examine the influence of work-related factors on TM (e.g., Noe & Wilk, 1993;Tharenou, 2001;Adomaityte, 2013;Medina, 2016), Nguyen and Kim (2013) claim that compared to efforts devoted to understand the effect of individual characteristics on TM, the effort that has been made to articulate the effects of situational characteristics, especially environmental and organizational ones, is little. Bunch (2007) states that organizations spend billions annually on T&D; however, much of this investment appears fruitless and the role of organizational context has rarely been investigated. Similarly, Bell, Tannenbaum, Ford, Noe, and Kraiger (2017) who examined the evolution of T&D research over the past 100 years conclude that what happens prior and after training can seriously influence training effectiveness, therefore, it is important to take a systems perspective and consider the context within which training occurs. They added that contextual factors, for instance, managerial support and prior training experiences can influence whether employees choose to participate in non-mandatory T&D activities. Further they stress that the work context influences participation in and effectiveness of T&D activities, since it can increase or decrease employees' learning motivation and ability to apply what they have learned in training.
In view of organizational context, individuals' attitudes toward various organizations' context are generally referred to as job satisfaction. Luz, De Paula, and De Oliveira (2018) state that how much individuals experience pleasures in the organizational context is called job satisfaction (JS). Therefore, and considering the importance placed to organizational context on T&D effectiveness and motivation (e.g., Bunch, 2007;Nguyen & Kim, 2013;Bell et al., 2017), this study aims to answer one basic question, does JS influence employees' level of TM? The answer to this question should not be simplified and this question constitutes the current study's main objective and contribution. In this context, Bell et al. (2017) claim that organizations spend huge amounts of money on T&D, and almost every working person spends hours of their working times participating in T&D experiences. Therefore, there is a need to better understand how people learn at work and how to support T&D initiatives.

Employees' job satisfaction
JS is a positive feeling concerning one's job, resulting from evaluating its characteristics (Robbins & Judge, 2013 Alegre, Mas-Machuca, and Berbegal-Mirabent (2016) identify three different lanes to clarify JS: first, teamwork, identification with the strategy, and employees' work-family imbalance; second, employees' work-family balance, autonomy and identification with the strategy; and third, manager support and identification with the strategy. Angbetic and Adelaine (2016) investigate the perceived organizational justice impact (distributive, procedural, and interactional) on employees' level of JS with regard to fairness. The results revealed that distributive and interactional justice positively affect employees' JS.
Some of the most accepted JS measures are the Minnesota Satisfaction Questionnaire (Weiss, Dawis, England, & Lofquist, 1967), the Job Descriptive Index (Smith, Kendall, & Hulin, 1969), and the Job Satisfaction Survey (Spector, 1985). This study will utilize the Job Satisfaction Survey (JSS), since it is one of the most universally used instruments (e.g., Liu (Liu et al., 2004) and Turkey (Yelboga, 2009). JSS measures nine organizational aspects including: pay, fringe benefits, contingent rewards, promotion, supervision, co-workers, operating conditions, nature of work and communication. Carlson et al. (2000) indicate that TM is one's desire to participate in training initiatives and completely embrace the training experience. Colquitt et al. (2000) followed Kanfer's (1991) definition of TM as the intensity, direction, and determination of learning-directed performance in training contexts. However, Zaniboni et al. (2011) claim that there is still vagueness in the definition and measurement of TM. While TM has been conceptualized based on phenomenological aspects, like interest, desire, and involvement in the learning process, other perspectives have conceptualized TM based on prospective behavioral outcomes, like goal intention, quantity and determination of the effort to learn (Zaniboni et al., 2011). These variations of views have created different approaches to measuring TM. For example, Noe and Schmitt (1986) built an eight-item scale to appraise learning motivation. Later, Noe, and Wilk (1993) developed a seventeen-item scale to measure the degree to which individuals perceive training as a useful and vital opportunity. Warr and Bunce (1995) created a twelve-item scale to measure distal and proximal forms of pre-TM. Warr et al. (1999) developed a six-item scale to assess trainees' motivation. Machin and Fogarty (2004) built a nine-item measure to assess the strength of trainees' desire to acquire new skills and trainees' intentions during training. Vroom's (1964) expectancy theory assumes that motivation is a function of three variables: expectancy, instrumentality and valence. Thereafter, Mathieu et al. (1992) and Tharenou (2001) studied TM using the valence-instrumentality-expectancy (VIE) approach. Adomaityte (2013) argues that Vroom's (1964) expectancy theory model is the most used and has been proven most useful for studying TM (e.g., Mathieu Mudor and Tooksoon (2011) state that human resources management practice, i.e. supervision, pay, and training, positively and significantly cor-relates with JS. Although Altarawneh (2005) states that training does not improve employees' satisfaction and/or commitment, several studies acknowledge the positive influence of T&D on employee satisfaction (e.g., Schmidt, 2007 In contrast, although the direct relationship -job satisfaction's impact on employees' motivation for training -has not been investigated sufficiently, the authors found a logical reason to propose this. Facteau et al. (1995) argue that the organizational commitment, intrinsic and compliance incentives, training reputation, as well as top management, supervisor and subordinate support, were found to be predictors of pre-TM. In a similar context, Egan, Yang, and Kenneth (2004) claim that JS is associated with organizational learning culture. Interesting findings were presented by Tsai, Yen, Huang, and Huang (2007), who investigated the content of JS in organizations that adopted downsizing strategies. They found that satisfaction significantly influenced the remaining employees' commitment to learning. From another angle, Chang and Lee (2007) suggest that both organizational culture and leadership have a significant positive effect on the operation of organizations' learning. Further, it was found that the process of learning in organizations has a positive and significant effect on JS. Jehanzeb, Rasheed, and Rasheed (2013) found that motivated workers had positive perceptions about the training initiatives offered by their organization. Ensour (2013) argues that employees' unwillingness to do training could be linked to their dissatisfaction with the managerial style in their organizations. Therefore, employee JS is assumed to result in a higher employee motivation for training.

A THEORETICAL MODEL
Main hypothesis: Employees' job satisfaction has a significant positive impact on employee motivation for training: • Job satisfaction and valence Employees perceptions and satisfaction with their contextual factors are perceived to have an impact on valence. Beier and Kanfer (2009) state that valence is the value individuals place on the outcome associated with training. Tracey et al. (2001) claim that the professional and informal relationships between supervisors and their subordinates can send explicit messages regarding the value of training. Although Zaniboni et al. (2011) found that contextual elements like job support were related to expectancy but not to instrumentality or valence. However, Beier and Kanfer (2009) state that perceptions of the organizational environment and the current work influence the valence that an individual places on T&D opportunities. Accordingly, the first sub-hypothesis is proposed: First sub-hypothesis: Employees' job satisfaction has a significant positive impact on the training motivation construct of valence: • Job satisfaction and instrumentality Tharenou (2001) found that employees' participation in T&D is greater if they expect that the skills and knowledge gained from training are instrumental for gaining extrinsic outcomes. Later, Egan et al. (2004) claim that organizational learning culture and employees JS are significant in determining employees' motivation to transfer training. Bell et al. (2017) indicate that supervisor and peer support, opportunities to apply learned skills on the job, and organizational culture and practices can determine the extent to which newly acquired competencies are applied on the job. They added that the work context influences participation in developmental activities and the effectiveness of those activities, since it can influence employees' learning motivation and ability to apply what they have learned in training. Given that instrumentality refers to the belief that training would lead to successful job performance, it is logical to assume that employees satisfaction with their work context would have an impact on instrumentality.
Second sub-hypothesis: Employees' job satisfaction has a significant positive impact on the training motivation constructs of instrumentality: • Job satisfaction and expectancy Employees satisfaction with their contextual factors is perceived to have an impact on expectancy, i.e. employees' assurance in the ability to learn or gain skills and knowledge through training. For example, Cohen (1990) claimed that workers who have supportive supervisors, participate in training activities with stronger beliefs that they would be useful.
Tracey, Hinkin, Tannenbaum, and Mathieu (2001) claim that supervisors' expression of their support for learning positively influences employees' confidence in acquiring knowledge and skills, thus can motivate that person to participate in training. Zaniboni et al. (2011) state that job support, including supervisors' and co-workers' support, was related to expectancy. They added that jobs that are designed to support personal and constant development have an impact on expectancy. Given the mentioned contextual factors that influence TM, it is logical to assume that: Third sub-hypothesis: Employees' job satisfaction has a significant positive impact on the training motivation constructs of expectancy.

RESEARCH METHOD
According to the research objectives, the deductive approach was adopted, and a structured questionnaire was developed for data collection (see appendix A). The research questionnaire combined two scales, with the first scale measuring JS. As mentioned earlier, JSS was utilized to measure employee satisfaction. This measure was originally developed by Spector (1985). The researcher contacted Paul Spector through ResearchGate and receivedwith appreciation -the original and the translated (Arabic) version of JSS. JSS measures nine organizational dimensions with each dimension consisting of four items; overall, JSS comprises 36 items.
The second scale measures TM. This study utilized Zaniboni et al.'s (2011) scale, which is based on three constructs (valence, instrumentality and expectancy). The TM scale was translated from English to Arabic by two colleagues. Thereafter, a comparison, revision and emendation were made for the two translated versions to issue a first draft of the Arabic questionnaire. This draft was translated back into English by two different colleagues to compare the authors' version with the original version. This process was done in order to assure its validity.
Both scales are ranked according to the Likert scale from 1 -Strongly Disagree to 5 -Strongly Agree as the Likert scale is broadly used for measuring attitudes (Pallant, 2005).

Characteristics of the sample
The male respondents represented 54% of participants, and females represented 46%. Having such a finding implies consistency with labor gender distribution in the Jordanian labor market as males dominate the labor market according to the 2017 Ministry of Labor report. Almost 52% of respondents had more than ten years of work experience. Such findings reinforce the idea that the Jordanian government has deactivated the recruitment processes in the public sector. Moreover, 19% of respondents had less than five years of experience. This finding is consistent with the age distribution in the Jordanian labor market.
According to the Ministry of Labor's 2017 report (2017), the lowest age category in the labor market was the age group of 20-24 years as they represented 13.5% of the labor market. The percentage increased slightly to 17% in the age category of 25-29 years; the highest figure, 29.7%, was in the 30-39 age group, followed by the 40-49 group, with a percentage of 23.3%. Overall, 81% of research respondents had experience of more than five years, which shows the difficulties of entering the labor market for the younger generation. Regarding participants' qualifications, more than 71% of respondents held a first degree (Bachelor Degree) and more than 14% of respondents held Master's and Ph.D. degrees.

Validity and reliability
The research instrument used in this study was tested using face, content, construct, convergent and discriminant validity. The first type of validity, "face validity", was conducted by piloting the research instrument by several academic staff and managers at the Jordanian public sector in order to check the instrument in terms of relevance and appropriateness. Based on the face validity test, some minor changes have been made such as the demographic information. Regarding the content validity, the key issue was using scales and dimensions that have been developed, used and tested in the previous empirical and theoretical studies in the relevant literature. Accordingly, the current study adopted and implemented dimensions used and tested before for the two scales (JS and TM) (e.g., Spector, 1985;Zaniboni et al., 2011).
Regarding construct validity, the confirmatory factor analysis technique (CFA) is used as suggested by Hair et al. (2006). The CFA test aims to confirm whether the instruments and items used in this study generate the same loading and number of dimensions as suggested by the original developers. Table 1 confirms the same loading and factors of the constructs under investigation.
To establish the convergent validity, Table 1 indicates that items of the two scales show significant factor loadings. However, some items that did not reach the required threshold .60 were deleted. Moreover, Table 1 shows relatively high average variances extracted (AVE) as suggested by Bagozzi (1980) for measuring convergent validity. Moreover, Table 1 shows acceptable fit indices for the two variables under investigation in this study.
As Table 1 shows, deletion for several items was due to having weak factor loading and high errors. At the same time, deleting the above items has helped in improving the CFA model fit indices. Moreover, it is worth mentioining that the instrument used for measuring JS includes 19 negative items and reserachers reversed them according to the required steps and procedures.
For the purpose of discriminant validity, it is evident in Table 2 that the used and adopted instruments for the two variables are truly distinct from each other and reflect the supposed phenomena that other measures do not. Moreover, Table 2 shows the average variance extracted (AVE) from each construct is higher and more than squared correlations and shared variance. In other words, Table 2 informs that all the squared correlations and shared variance between each pair of variables are less than the variables AVEs which offers an empirical support for the discriminant validity among constructs and dimensions.
Thereafter, the reliability was checked. In this regard, Cronbach's alpha test was used to measure the reliability of each scale (JS and TM) to ensure that the adopted scales were purified and suitable for measuring constructs under investigation in new settings and contexts.  The Cronbach's alpha findings for all dimensions revealed acceptable as well as unacceptable outcomes. In other words, Cronbach's alpha values and item-total correlation scores were below as well as above the recommended alpha value of .60. Accordingly, the researchers utilized the results, which emerged in the column entitled "the Cronbach's alpha if item deleted" to improve the scales' reliability.
Better and more acceptable Cronbach's alpha and item-total correlation values were gained as a result of deleting the low correlated items. Specifically, the 'operating conditions' dimension, which is one of the JS scales, was dropped and removed in this study from any further analysis due to having low and poor reliability scores (Cronbach's alpha value for the scale: .339; Cronbach's alpha value if the item was deleted: 406).
According to the purification process results, the dimensions of JS were reduced to eight instead of the initial nine dimensions as suggested by Spector. Moreover, the initial set of items for measuring all dimensions of JS was reduced from 36 items to 26 items. In other words, 10 items from the JS variable were deleted, as recommended by the reliability test based on the results of Cronbach's alpha if an item was deleted. On the other hand, only one item was deleted from the first dimension of TM. Table 3 shows the final Cronbach's alpha scores for all dimensions after the purification process: At the construct level, Table 3 shows that the Cronbach's alpha values for the two scales (JS and TM) exceeded the recommended criterion of Cronbach's alpha value (JS = .806; TM: 806). At the dimensional level, Cronbach's alpha values for some dimensions were below .60 but more than .50 even after the purification process mentioned earlier. However, the researchers decided to keep such dimensions due to the idea that a Cronbach's alpha value of more than .50 was still acceptable and the item deleting option would not have enhanced the Cronbach's alpha score. Therefore, it was confirmed that those dimensions and scales were reliable for measuring JS and TM.

Analysis tools
Descriptive statistics techniques such as mean, standard deviation, skewness and kurtosis were implemented to assess the answers of respondents and to make sure that their responses were normally distributed.
Furthermore, MR was performed to examine the impact of JS on TM. The main aim of MR is to examine the predictive power of the independent variables (eight dimensions of JS) on the dependent variable (TM). Thereafter, MR analysis was conducted three times sequentially, each time using a different construct of TM as a dependent variable in order to give more insights for researchers, practitioners and readers.

DATA ANALYSIS AND FINDINGS
The mean values for JS dimensions ranged between 2.53 (fringe benefits) and 3.66 (nature of work); standard deviations ranged between .716 (nature of work) and .942 (supervision) (see Table  4). Having a moderate mean score for the dimension of fringe benefits indicates that public sector employees believed they deserved better benefits. This could be attributed to limited resources allocated for the public sector in Jordan. This area needs more studies to explore the reasons behind moderate employee satisfaction in the Jordanian public sector regarding the pay dimension with a mean of 2.76, promotion dimension 2.62, fringe benefits 2.53 and contingent rewards 2.76.
Regarding the TM scale, the highest mean was for the expectancy dimension with a mean of 3.89, indicating that employees have the confidence in their abilities to learn and improve their skills, whereas the lowest was for instrumentality with learning with a mean of 3.49. In terms of overall scales, Table 4 indicates a positive and more than moderate assessment for JS with a mean of 3.11 and TM with a mean of 3.75.
Regarding normality, two statistical approaches were used for exploring data normality, skewness and kurtosis. Skewness refers to the symmetry of the data collected. A skewed finding implies a variable whose mean is not in the center of the dis-tribution, while kurtosis measures the peakedness of the collected data, meaning that the distribution of the data collected can be too peaked with short and thick tails, or too flat with long and thin tails. In this regard, the most acceptable and wellknown rule of thumb to assume normality indicates that scores for skewness and kurtosis should not fall outside the range of 1 and -1. Thankfully, the findings shown in Table 4 demonstrate that all variables and dimensions in this study are normally distributed since such findings did not violate the accepted rule of thumb of normality.

Model assessment and hypotheses testing
As mentioned earlier, MR aims to examine the predictive power of the independent variables (eight dimensions of JS) on the dependent variable (TM). Thereafter, this study conducted MR analysis three times sequentially, each time using a different construct of TM as a dependent variable.

Assumptions and parameters of multiple regressions
Before running and using MR, researchers should make sure that the estimated errors are at the minimum level and do not violate the findings of the study (Hair et al., 2006;Tabachnick & Fidell, 2007). Table 5 summarizes the key assumptions of MR and provides answers for the associated parameters. Note: * M -mean, S.D. -standard deviation, S -skewness, K -kurtosis, N -number of items on the scale. ** The cut-off point between -1 and 1. Table 5 shows that none of the MR assumptions were violated in the data gathered for this study. Firstly, the sample size (N = 342) was well above the rule of thumb, which relies on the number of independent variables. Secondly, findings confirm the normality of data as none of the independent variable dimensions violated the rule of thumb for skewness and kurtosis, since none of the kurtosis and skewness results fell outside the range of 1 and -1. Thirdly, multicollinearity was measured using two tests (tolerance and the Variance Inflation Factor (VIF), and the findings presented in Table 5

Findings of model assessment and hypotheses testing
As mentioned before, the MR technique predicts the relative power, importance and contribution of all independent variables on the dependent variable in order to be able to answer the main hypothesis and sub-hypotheses.
In terms of the overall model fit (the main hypothesis), the results of the MR analysis indicate that the independent variable explains and predicts only 18.8% of variance of the dependent variable (training motivation). In other words, the low R 2 score (18.8%) indicates that the overall dimensions of JS explains a low but significant variance of the overall TM constructs at the significance level (p < .00). Table 6 presents findings of the MR analysis for all hypotheses developed in this study.
The MR analysis shows that four dimensions of JS (supervision, contingent rewards, co-workers and nature of work) had a significant impact on the overall TM model. The R 2 value for the eight dimensions was relatively low. Furthermore, it was found that three dimensions of JS (supervision, co-workers and nature of work) had a positive and significant Beta value (β = 0.178, P < 0.05; β = 0.165, P < 0.05; β = 0.286, P < 0.05).
The results show that nature of work scored the highest β value, assuming that employees' satisfaction with the nature of their job (meaningfulness of the job, sense of pride, enjoyment and enthusiasm for the job's activities) influenced their motivation to participate in training. This finding suggests that employees who enjoyed and loved what they do, who felt that their work was meaningful, had more enthusiasm to engage in training activities. In a similar context, Zaniboni et al. (2011) claim that the nature and type of job is likely have an influence on employees' TM. Orpen (1999) found that job involvement correlated significantly with TM. Tracey et al. (2001) state that individuals with high job involvement appreciate opportunities to take part in training activities to boost their job situation. Furthermore, highly in-volved workers are more likely to develop high levels of pre-training self-efficacy, particularly for job related training.
Supervision was found to have a positive and significant Beta value, assuming that employees' satisfaction with their supervisors (admiration, supervisors' competence, fairness and care about subordinates' feelings) influenced employee TM. This finding is consistent with the findings of Cohen (1990) and Tracey et al. (2001). In a similar context, Facteau et al. (1995) concluded that supervisors' support is positively related to pre-TM. Overall, Bunch (2007) stresses the importance of supervisors' support by stating that even well-prepared customer service training will not be transferred if supervisors assess only the number of transactions processed rather than customer satisfaction.
Employee satisfaction with their co-workers (liking co-workers, co-workers' competence, enjoyment working with co-workers, bickering and/or fighting) had the least but positive and significant influence on employees' motivation for training. Similarly, Facteau et al. (1995) claim that three social support elements (top management, supervisor and subordinate support) were found to be predictive of pre-TM.
The findings also show that contingent rewards had a negative and significant impact and contribution on the overall TM (β = -0.139, P < 0.05), assuming that contingent rewards (e.g., apprecia- Finding Accepted Accepted Accepted Accepted tion and recognition for good work) led employees to focus on current work accomplishments rather than long-term learning. Several efforts were made to examine goal orientation and the theory of TM (Zaniboni et al., 2011). In this context, Zaniboni et al. (2011) claim that performance goal oriented people are concerned with goals that reveal task competence and receive positive judgment, therefore, no significant relationships were found between performance goal orientation and the three dimensions of the TM. In this context, Colquitt and Simmering (1998) found a negative correlation between performance orientation and expectancy and TM. Chiaburu and Marinova (2005) found a significant relationship between the learning approach goal orientation and the TM, while no significant relationship was found between the performance goal orientation and the TM.
The findings also demonstrate that four dimensions (pay, promotion, fringe benefits and communication) of JS did not have a significant impact on the overall TM (P > 0.05), meaning that employee satisfaction with the mentioned variables had no relationship with TM. Perhaps pay, fringe benefits and promotion were perceived as a result of current efforts and not a long-term issue, resulting in employees' focusing on the current performance that organizations needed, thereby making training less attractive. In this regard, Orpen (1999) states that if the employees believe that there is 'something in training' for them, then they will react better to training. Therefore, managers should provide valuable training outcomes, dependent on employees making the required effort to benefit from the training (Orpen, 1999). However, Elsbach (2004) argues that perceptions regarding the possible benefits of training, e.g. promotion or pay increases, better predict the probability for training success.
To gain more insight into what enhances TM, another round of analysis was conducted for each construct of TM separately. Table 6 indicates that JS was a significant predictor of TM (p < .00) in the three models. However, it is clear that the predictive power of JS is low for the three models respectively (R 2 .13. R 2 .16, R 2 .16,).
For Model 2, four dimensions of JS (contingent rewards, supervision, co-workers and nature of work) had a significant influence on TM (valence).
Regarding the nature of work, it was found that positive attitudes toward the nature of work had a positive impact on valence, assuming that employees who enjoyed and loved what they do and who felt their work was meaningful found that the training outcome (valence) was more attractive.
In this context, Renta-Davids, Jiménez-González, Fandos-Garrido, González-Soto, (2014) state that task variety and task complexity have an impact on employees interest in acquiring new knowledge and skills, they explain that this interest refers to employees' desire to address their job demands.
Further, the authors' findings indicate that job support, such as supervision and co-worker support, influenced employees' beliefs that successful job performance was going to be valued. Contrary to the findings, Zaniboni et al. (2011) found that job support was related to expectancy but not to valence.
Contingent rewards had a strongest but negative impact (β = 0.-214, P < 0.05) on valence (attractiveness and the perceived value of training results). This means that employee satisfaction with contingent rewards which are linked to current performance made training outcomes (valence) less attractive.
Regarding Model 3, the nature of work and supervision were the only two dimensions of JS that were found to have a significant and positive impact on TM (instrumentality). These results make sense, as employees who did not think they could apply what they had learned in practice were not motivated to take part in training. Moreover, it was found that co-workers and contingent rewards did predict valence and expectancy but not instrumentality. Thus, it could be claimed that in an employee's good relationship with co-workers, their support has an impact on expectancy, i.e. confidence in the ability to gain knowledge and skills, as well as the significance and importance of training outcomes, but not on instrumentality (rewards associated with learning).

RESEARCH CONTRIBUTION AND RECOMMENDATIONS
This study makes a number of theoretical and managerial contributions. From a theoretical perspective, it investigated the relationship between various JS dimensions (pay, fringe benefits, contingent rewards, promotion, supervision, co-workers, nature of work and communication) and TM. This enhances our understanding regarding the organizational as well as other contextual factors that influence TM; this in turn is perceived as fundamental to training effectiveness.
As nature of work and supervision were found to have a significant impact on all TM constructs, this study suggests that managers should take them into consideration to enhance employees' desire and enthusiasm to engage in training activities. For example, this study suggests that designing challenging, meaningful, autonomous work tasks enhances employees' satisfaction with the nature of their jobs, which in turn influences the level of their confidence in training's role and outcomes.
Supervisors are advised to build a bridge of good relationships between them and their subordinates, as well as between subordinates themselves. A cooperative and supportive climate enhances workers' confidence in their ability to learn and boosts their trust that training can lead to successful improvement in performance, which eventually influences the importance attributed to training outcomes.
Contingent reward was found to have a significant but negative impact on TM. This indicates a need to critical revision of rewards pol-icies, meaning that rewards policies should not be based on current performance only. A clear career development paths are advised to be created that determine the employees' T&D needs and link them with employees' progression and promotion.

LIMITATIONS AND FUTURE RESEARCH
This study was conducted in a specific context, Jordanian ministries, therefore, future research within other sectors, like private sector institutions in Jordan, would be a good contribution to examine other factors that influence TM in various sectors.
Furthermore, this study investigated only nine dimensions of JS based on Spector's model, which directs future research to investigate other aspects of JS.
Although the results show a positive assessment of JS as a whole, it is evident that employees have moderate satisfaction regarding pay practices, promotion, fringe benefits and contingent rewards, which indicates a need for more effort to study the Jordanian public sector compensation and benefit structure.
Additionally, this study investigates the impact of satisfaction on TM; however, the relationship between satisfaction, training motivation and desired training outcomes like training efficacy, employees' performance and organizational effectiveness was not investigated, leaving the door open for future research to investigate those relationships.
Finally, according to the time horizon, this study followed the cross-sectional path. Therefore, longitudinal research is needed to investigate employees' training motivation, for example, before and after conducting organizational reform programs.

CONCLUSION
This study concludes that four dimensions of JS (supervision, contingent rewards, co-workers and nature of work) predicted employees' motivation for training. However, the dimensions of JS explain a low but significant variance of the overall TM dimension. Furthermore, it was found that the nature of work scored the highest β value. This implies that employees who enjoyed and loved what they do, who felt their work was meaningful, had more enthusiasm to engage in training activities. Supervision was found to have a positive and significant β value, indicating that employees' satisfaction with their supervisors influenced employees' motivation for training. Employees' satisfaction with their co-workers had the least β value but a positive and significant influence on employee training motivation. Findings also show that contingent rewards had a negative and significant impact on TM, assuming that contingent rewards (e.g., appreciation and recognition for good work) led employees to focus on current accomplishments rather than long-term learning. The findings showed that four dimensions of JS (pay, promotion, fringe benefits and communication) did not have a significant impact on overall TM.
It was found that supervision predicted the three TM constructs, assuming that employees' satisfaction with their supervisors enhanced their confidence in their ability to learn and improve their skills. Furthermore, it was found that supervisors' support enhanced employees' beliefs that training could lead to an improvement in their performance, which overall influenced the importance attributed to training outcomes.
Furthermore, the nature of work predicted the three TM constructs, assuming that employees who were satisfied with and enjoyed their job were more confident in their abilities to learn through training, i.e. they believed that their efforts in training would enhance successful training performance. Moreover, they were more able to consider the essential role of training in improving their performance, which in turn enhanced the importance attributed to training outcomes.
Moreover, it was found that the nature of work and supervision were the elements that appeared to predict all constructs of TM, while the results indicate that employees' satisfaction with their co-workers and contingent rewards did predict valence and expectancy, but not instrumentality. Thus, it could be claimed that an employee's good relationship with co-workers, and co-workers' support have an impact on expectancy, i.e. workers' confidence in their ability to learn new knowledge and gain required skills, as well as the importance of training outcomes, but not on instrumentality (rewards associated with learning).