Can organizations influence the ROI of training?
A study on the moderating role of employee perceptions of
organizational investment, supervisor support and transfer of training.
Master Thesis – January 2017
University of Amsterdam | Amsterdam Business school MSc Management Studies | Leadership & Management
Supervisor: dr. Corine Boon Author: Haanstra, Manon Student number: 1088409
Statement of Originality
This document is written by student Manon Haanstra who declares to take full responsibility for the contents of this document.
I declare that the text and the work presented in this document is original and that no sources other tan those mentioned in the text and its references have been used in creating it.
The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.
Abstract
Organizations in the Netherlands spend over 3.3 billion euro’s on their yearly training budged. Nonetheless, the way training brings organizational value is still unclear. Mixed research results on the relationship between training and positive employee outcomes have directed HR scholars to focus on the individual level of analysis and they found for one employee’s perceptions to be an important influencer. This research has tried to amplify this line of thought and extended it to the level of the organization, the manager and the employee. In a multisource case study at a semi-governmental knowledge institution surveying 121 employees the author has tested how perceived investment in employee development (PIED), perceived supervisor support for training (PSST) and transfer of training (TOT) affects the training – employee job performance relationship. Also, the preliminary hypothesis was tested that a positive relation exists between both skill as knowledge training and job performance. Regression analyses performed showed a marginally significant relationship for PSST on the skill training – job performance relation and did not confirm the other hypotheses. Directions for future research are proposed.
CONTENT Introduction ... 5 LITERATURE REVIEW ... 8 SHRM ... 8 Training ... 8 Job performance ... 9
Benefits of training and development related to job performance ... 10
Employee perceptions as potential source of variance ... 11
PIED and outcomes of training ... 12
PSST and outcomes of training ... 13
(Perceived) TOT and outcomes of training ... 14
METHOD ... 16
Research procedure ... 16
Direct effect design ... 16
Moderating effect design ... 17
Sample ... 17
Measures ... 17
RESULTS ... 19
Common tests ... 19
Computing scale means ... 19
Correlation analysis ... 20 Regression analyses ... 21 DISCUSSION ... 24 Findings ... 24 Hypotheses 1a and 1b ... 24 Moderating effects ... 26 Hypotheses 2a and 2b ... 26
Hypotheses H3a and H3b... 26
Hypotheses 4a and 4b ... 28 Limitations ... 28 Theoretical implications ... 29 REFERENCES ... 32 APPENDIXES... 38 Survey questions ... 38
Introduction
When organizations started to look at their internal resources as sources of competitive advantage they also realized that the people they employee are strategic assets and important for positive organizational performance (Wright et al., 2001). This realization only became more truth in recent years. Many organizations nowadays strive to compete in global economy and adopt new technologies, making it increasingly important to differentiate on the basis of the skills, knowledge and motivation of the human capital (Aguinis & Kraiger, 2009). To influence these competences, HR practices are considered to have the most direct effect (Delery & Shaw, 2001). When human resources (HR) has the role of supporting the business strategy and achieving strategic goals like organizational performance, (Wright et al., 2001) we speak of strategic human resource management (SHRM).
Job performance is one of the outcomes that has a direct link to organizational performance (Delaney & Huselid, 1996). Job performance includes how employees behave in their job-specific tasks but also the way they shape the culture and environment of the organization (Ng & Feldman, 2008). Job performance can be further increased through HR practices of which employee training and development is one of the most important practices. With help of training, employees can develop themselves, they can perform their tasks more efficient and they are more motivated to use their competences (Dysvik & Kuvaas, 2008). Training is especially valuable in organizations that require firm-specific skills, which are not ready available in the labor market (Lepak & Snell, 1999). Combined with the increasing demands organizations put on their human capital and the awareness that human capital is a source of competitive advantage, training and development is - not surprisingly - an important subject for research.
Interestingly, studies on the relationship between HR practices and organizational performance give mixed results: while some show a large and significant positive relation between HR practices and performance, others show a small or non-significant relationship (Khilji & Wang, 2006). And if researchers have found mixed results for outcomes of HR practices, this implies that outcomes of training are also unclear. Dysvik and Kuvaas (2008) state that individual differences and different perceptions of training are important in understanding the relationship between training and employee outcomes. Employees often not react to the actual or objective environment – in this case the intended training – but rather to how they interpret things. Because of this mechanism, it is their perception of training that has
Thus far, relatively few studies have focused on individual experiences of HR practices. In recent years scholars have been calling for more research that examines employees perceptions of HR practices at the level of individual outcomes. In a firm, the same offer of HR practices does not automatically lead to uniform outcomes for all employees. Variance in the outcomes suggests moderators are at work causing these differences. Hence, in accordance with recent studies which have an alternate focus than ratings given by HRM professionals or line managers (e.g. Alfes et al., 2013), the present study intents to measure employees’ perceptions and with this tries to explain the variance in the effect training has on job performance at the individual level. It does so by investigating possible moderators in the area of employee perceptions. This is in line with Nishii (2008) who argues that in order for a HR system to link to performance in the desired ways, the perception of employees of the HR practices are important. This study builds on this theory by also proposing that employees’ perceptions of other moderators on this effect are important. It does so by investigate the relationship between the outcomes of training and the perceived supervisor support for training (PSST), perceived organizational investment in employee development (PIED) and perceived transfer of training (TOT). For example, it could be that some employees perceive their organization to invest at a moderate level in the development of their workforce, while other employees perceive the investment as very high, influencing the engagement of the employees and through this affects the way the employee is willing to put the learned in to use, which in its turn affects job performance.
To my knowledge, not many published study to date have considered PIED, PSST and TOT as moderating variables affecting the linkage between training and job performance. These individual components are examined because they exemplify three different levels of the processes causing variance in the effect training can have on job performance: the organization, the supervisor and the employee. At the highest level the organization and the extend to which the employee perceives the investment of the organization in the employee. Building on the social exchange theory (Ostroff & Bowen, 2000), the perceived investment of the organization in the development of the employee may result in employees that will reciprocate and because of that are more willing to put the learned into job tasks behavior. The second level is that of the supervisor and how the employee perceives that his supervisor supports him in developing new skills, abilities and knowledge through training. Alfes et al., (2013) have tested a model that suggests that the enactment of positive behavioral outcomes largely depends on perceived support and employees’ relationship with their line manager. At the lowest level is the employee
himself, to what extent does he feel that he is able to transfer the learned to his workplace and to what extend it has helped him carrying out his job tasks.
The aim of this study is to contribute to current understanding of the effects of training in two ways. First, the study tries to fill in a bit of the puzzle that surrounds the mixed results found on the effects training has on organizational outcomes. More specific, the study proposes that the employees’ perception of PSST, PIED and TOT influences the positive effect training has on job performance. In this study this is tested for both training focused on skills as well as training focused on knowledge. Besides adding to the understanding of how we can measure the effects of training, this study has practical implications. If there are as suggested multiple levels in the organization that can cause variance in the effect training has on job performance, organizations should be aware that strict paper policy regarding the content of training is not a strategy. This study will do so by answering the following research question:
What is the effect of training on employee job performance and how is this affected by the way employees perceive the supervisor support for training, the perceived investment in development by the organization and the perceived transfer of the training?
The study is conducted at a semi-governmental knowledge institution in the Netherlands. Its size is middle-big with approximately 1.800 employees of which a sample of 121 is used in this study. Employees work in teams of around 16 peers and work highly independent. A supervisor manages every team. The HR department states training as “very important”. Training has the aim enabling the employees to achieve the goals of the organization. The organization has acted on this by implementing an online training platform (the “Academy”) where an extensive range of training is offered from which the employees themselves can choose an appropriate selection.
This report is structured as follows. First, an overview of existing literature regarding the topics of interest for this study is given. Based on this, assumptions and hypotheses are presented in chapter 2. In the chapter following, the research structure, measures and sample are explained. The results of the research are presented in chapter 4 and in chapter 5 these results are discussed. This report ends with limitations and directions for future research.
LITERATURE REVIEW SHRM
The field of strategic human resource management (SHRM) has steadily grown in the last three decades. The first extensive research on SHRM was published in 1984 by Devanna, Fombrum and Tichy, in which they explored the link between business strategy and HR (wright et al., 2001). Through the years that followed, many studies investigated how SHRM contributes to the achievement of business goals and organizational effectiveness (e.g. Becker & Huselid, 2006; Bowen & Ostroff, 2004; Truss, 2001). With its focus on organizational performance rather than on individual performance, the field of SHRM differs from traditional HR management research (Becker & Huselid, 2006). This focus on organizational outcomes makes it complex to prove the added value of SHRM. For one it is unclear how, when and why the relation between SHRM and organizational outcomes is achieved (Boxall & Purcell, 2000). An organization is a complex whole with many variables that can differ. In order to understand the mechanisms that are in place, researchers started to study organizations at different levels, for example the level of the job group or the individual level (Wright & Nishii, 2007).
Also divergent from traditional HR is the debate in SHRM research about how the added value should be achieved: both the use of a ‘best fit’ approach and a ‘best practice’ approach are empirically endorsed. The best fit approach takes in account the relevant impact of contextual factors where best practice pre-describes certain HR practices that have universal success (Paauwe & Boselie, 2003).
Training
Employee training is one of the key practices of strategic HRM. It can be defined as “a planned effort by a company to facilitate the learning of employees” (Noe et al, 2008). Training enables employees to develop themselves and to perform their job tasks for their organization more effectively (Dysvik & Kuvaas, 2008) and has a positive impact on the performance of teams (Anguinis & Kraiger, 2009). Therefore, investment in training provides benefits for both the employee and the organization. Not surprisingly, training is a big expense for organizations. In the Netherlands, where this study is located, organizations together have spent 3.3 billion euro’s on training in 2010: more that 2% of the total costs of labor (CBS.nl). In the last two decades, training as part of SHRM is argued to become more and more important, due to the increasing employee demands the 21st-century brings. Organizations need to invest intensively in development, as employees are continuously required to update their knowledge, skills and work habits (Chen & Klimosky, 2007).
And, investment in training is more than upgrading individual abilities. Providing training opportunities also has a positive impact on the intrinsic motivation of the employee to use their competences at the workplace (Dysvik & Kuvaas, 2008). In addition, research reviewed by Shore et al. (2006) suggests that employees feel more obligated towards their organization when the level of organizational investment is high. This mechanism is part of the social exchange theory: the expectation that when one person does another a favor, this will be returned in the future (Wayne et al., 1997). Offering training opportunities to an employee is an example of a ‘favor’ done by the organization. The employee in return feels obliged to reciprocate the favor and will engage in behaviors that positively influence organizational performance. For example, the employee increases the internalization of organizational norms and values, puts more effort in meeting organizational goals and increases his productivity (Eisenberger et al., 1986).
Job performance
Job performance is how employees behave, and the expected organizational value of this behavior carried out over a period of time (Kraiger, 2003). A wide variety of criterion measures have been used to measure job performance, including salary, promotion, turnover and productivity indexes (Kraiger, 2003). In this study, job performance is measured on the basis of job appraisal.
Variance in employee behavior is expected to automatically lead to variance in the expected organizational value as well (Kraiger, 2003): if employee behavior improves (or gets worse), expected organizational value will increase (or decrease!). This cause-effect makes job performance an interesting subject for organizations to influence. Employee behavior is a complex concept and can be about almost anything a person does. Campbell (1990) defined no less than eight behavioral dimensions that are relevant to performance, which are: job-specific task proficiency; maintaining personal discipline; demonstrating effort; facilitating team and peer performance; non job specific task proficiency; communication task proficiency; supervision/ leadership; management/ administration. Although most of these behaviors are not core task activities per se, they do significantly influence organizational outputs by “shaping the organizational cultures and environments in which core task performance takes place” (Ng & Fieldman, 2008 p.392). The appraisal system at the organization in this study is broadly based on these behaviors, relevant dimensions are selected for every job title and supervisors are supposed to use these in their appraisals.
Benefits of training and development related to job performance
Training gives employees the opportunity to improve their knowledge, skills and behavior. Their improved competences in turn help to retrieve better job performance. For example, the employee is aware of legislation or has the skills to use the organizations’ systems. Colquitt et al. (2000) argue that from possible training outcomes, especially the acquisition and transfer of skill is predictive of individual job performance. This is very intuitively, but they also argue that training, focused not directly on job performance, indirectly does contribute to increased performance, for instance when training is focused on individual or team well being. Reasons for this can be found in the social exchange theory; employees who perceive their organization to be engaged in their development will feel a higher organizational commitment and are more motivated to reciprocate. This in its turn will leads to positive employee work behavior (Bartlett, 2001). A more straightforward finding is that of Guzzo et al. (1985). In their meta-analysis they found training to be the most powerful HR tool to enhance employee productivity.
Training can be focused on the three determinants of job performance: declarative knowledge, procedural knowledge and skill, and motivation (Campbell, 1990). Declarative knowledge is factual knowledge and knowledge about principles and procedures. Procedural knowledge and skill is the combination of knowing what to do and being able to actually do it. It includes skills such as cognitive, psychomotor, physical and self-management skills. Motivation is the choices made about how much and how long effort is exerted and can be a side effect of both knowledge as skill training. This focus on knowledge and skill can be found at the studied organization as well: training is designed to either develop skills or to develop knowledge.
Concluding, training focused on knowledge as well as on skills will have a positive effect on job performance, through the development of knowledge, skills and behavior and because of reasons found in the social exchange theory. Accordingly, I expect:
H1a: There is a positive relationship between training focused on developing
knowledge and employee job performance
H1b: There is a positive relationship between training focused on developing
Employee perceptions as potential source of variance
The main focus of SHRM research is on relating the variance in HR practices across organizations with variance in performance (Wright and Nishii, 2007). In other words, taking the HR strategy of two organizations and compare the respective organizational performance. In a more recent stream of SHRM research, researchers have identified the unit of analysis in organizations as a variable that has multiple levels; an important proposition for understanding variance. In this stream not only variance between organizations is examined but also the variance that exists within a single firm. Wright and Nishii (2007) for example have proposed a model, which gives consideration to job groups and the individual level as sources of variation. The model presents the processes that are in place in order for HR practices to have effect on organizational performance. For every one of the processes the level of analysis is specified as a potential source of variance. One of these processes is the perception of the employees of the HR practices, describing how the objectively HR practices are subjectively interpreted and how this affects the main relationship between a practice and the outcomes.
It is not only HR practices that are subjectively interpreted by employees. People in general perceive reality in a different way and attribute their own explanations to the events they encounter (Fiske & Taylor, 1991). Because of this I reason that employees not only subjectively interpret HR practices but also other factors surrounding these practices. And in order to understand the relation between training and employee outcomes - both objective measures - we have to take in account other factors open for interpretation as well. This research builds on this idea and on the theory of Wright and Nishii (2007) proposing that employees’ perceptions of the supervisor support, organizational investment and transfer of training are also important sources of variance. Their perceptions influence the general effect training has on job performance and causes the relationship to vary between employees. These three constructs were chosen because they exemplify three different levels of the environment surrounding training and the effect training has on job performance: At the highest level the organization and the perceived investment of the organization in the employee. Following the level of the supervisor and the way the employee perceives his supervisor supports him in gaining new competences through training. At the lowest level is the employee-training-job relation; to what account does the employee perceive that the learned helps him with his job tasks.
PIED and outcomes of training
Sustained competitive advantage - a long-term advantage that enables outcompeting the competitors - can be created through investing in employee development (Lee and Buvold, 2010). Organizations invest in employee development when they supply training opportunities and enable their employees to solve new job tasks they are given in the future (Lee and Bruvold, 2003). Besides the obvious benefits of perceived investment in employee development (PIED) - positive performance behaviors through the development of competences - a positive assessment by employees of their organizations’ commitment may create more. It could create a dynamic relationship in where employees have a greater sense of engagement towards their organization. Employees believe that their contribution is valued (Keep, 1989; Lee & Bruvold, 2003) and that the organization cares about their employability (Lee and Bruvold, 2003), and because of this they feel they should reciprocate at least at the level they feel their organization is investing in them (Lee and Bruvold, 2003). This will show itself for example in employees who use their (new learned) competences in their job tasks (Ostroff and Bowen, 2000). Because of this, it is expected that the return on investment on training will be higher with higher levels of PIED.
Interesting is that not the hours spend in training count the most for positive employee outcomes. It is the amount of training the employee feels his organization is providing (Bartlett, 2001). So, according to this theory, in the case two employees spend the same amount of time on training, the employee who believes his organization supplies the most training will demonstrate the most positive employee outcomes! This implies that perception of investment in employee development might be of more importance than the actual attending to training. This is similar to findings of Alfes et al. (2013); they have tested a model that suggested that positive employee behavioral outcomes largely depend on the perceiving of the organizational support.
These findings on PIED give an interesting line of thought: can differences in the training – job performance relationship be partly explained by how employees perceive the investment in employee development of their organization? Thus, would a higher level of PIED also mean that, because employees are more willing to put the learned in use, there is a greater positive effect of training on job performance? This study will try to prove this. By testing this, it also answers to the calling of Lee and Bruvold (2003) for more research on PIED and the effect it has on employee behavior. Accordingly, I expect:
H2a: The positive effect of training focused on knowledge on employee job
performance is moderated by the level of PIED, in such a way that when PIED is high, the positive effect is stronger.
H2b: The positive effect of training focused on skills on employee job
performance is moderated by the PIED, in such a way that when PIED is high, the positive effect is stronger.
PSST and outcomes of training
The effects of developmental activities are reinforced by perceived supervisor support for training (PSST) (Baldwin and Magjuka, 1997) and, as Kuvaas and Dysvik (2010, p.142) state “it may represent a necessary condition for effective implementation of HR practices”.
PSST is a derivative of perceived supervisor support (PSS). PSS refers to employees’ views concerning the degree to which their immediate supervisor values their contributions and cares about their well-being (Kuvaas & Dysvik, 2010). Specific, if supervisors are supportive by way of involving themselves in their employees’ development, holding positive expectations, and providing daily support for such initiatives, for instance by facilitating time and space for the appliance of training content, such behavior should increase the effects of developmental activities (Baldwin and Magjuka, 1997). Supervisor support is a diverse construct and could include activities like encouragement, goal-setting activities and modeling of behaviors (Baldwin and Ford 1988). Applying this theory to training, supervisors can jointly set training goals prior to the training. After training, supervisors can ensure that employees have the opportunity to use the learned competences. Furthermore, the supervisor can offer extrinsic rewards for employees that utilize their new competences, for example in praise or more interesting assignments. Reinforcement processes can also work in reverse; a supervisor that ignores or even attacks the usage of new skills can cause the learned behavior to evaporate (Baldwin and Ford, 1988).
I therefore expect that PSST has a positive influence on the relation between training and job performance. This is in line with results from the study of Olivero (1997) who found that training on its own increased productivity with 22,4 percent while training in combination with coaching (e.g. goal setting, feedback) increased productivity with no less than 88,0 percent. Accordingly, I expect:
H3a: The positive effect of training focused on knowledge on employee job
performance is moderated by the level of PSST, in such a way that when PSST is high, the positive effect is stronger.
H3b: The positive effect of training focused on skills on employee job
performance is moderated by the level of PSST, in such a way that when PSST is high, the positive effect is stronger.
(Perceived) TOT and outcomes of training
Transfer of training (TOT) is assumed to have an important role in how training can positively affect job performance (Olivero, 1997). Because of this, there has been a major effort from researchers to “understand the antecedents and consequences of the transfer of training process” (Velada et al., 2007, p.283).
One of the components studied is the average percentage that is actually transferred from the training environment to the workplace. For example, Wexley and Latham (2002) suggest that while directly after training forty percent of the content is transferred, after one year only a modest fifteen percent of the learned is still maintained. Research has tried to explain why this happens. At least two explanations for unsatisfactory transfer of training may be found in the design of the training. First, it is important to provide the right antecedents for the transfer of training, for example it needs instructions how to use it in the work environment and opportunities to practice the learned in the workplace, in order for the learning not only to take place but also be transferred (Yamnill and McLean, 2001). Second, the design of the training needs to be appropriate for the work design. The competences of the employees have to be aligned with the work design to have optimum change on transfer of training (Xiao, 1996). Holton (1996) illustrates in his transfer of training model that besides transfer design, motivation to transfer and transfer climate are important influencers of transfer of training. Motivation to transfer revers (Yamnill and Mclean 2001). This can vary per employee and, according to Tannenbaum et al. (1991), is influenced by the extent to which training meets the trainees’ expectations. Motivation can fluctuate and decline over time. The perception of the employee of his work place also influences the level of transfer, which we call the Transfer Climate. Yamnill and McLean (2001) explain how it influences how much of the learned an employee actually is able to use in the work place.
over a period of time” (p. 56). About this definition there is a strong consensus, as well as about the fact that acquisition of competences through training is of little value if the transfer doesn’t take place (e.g. Yamnill and McLean, 2001; Goh, 1998). Training offers employees new knowledge but it is only worth something to the organization if the employee changes his behavior afterwards: either because he is putting this knowledge to use or through a positive change of aptitude.
The above is proven in the1996 study of Xiao, whose findings were that in order to improve employees’ capacity, training is of essential necessity and transfer of training is important in facilitating the usage of new capacities. Concluding, training needs transfer in order to result in positive organizational outcomes and the amount of transfer is a gliding scale. Because transfer of training influences the actual usefulness of offered training, I assume it to have a positive effect on job performance. Accordingly I expect:
H4a: The positive effect of training focused on knowledge on employee job
performance is moderated by the level of TOT, in such a way that when TOT is high, the positive effect is stronger.
H4b: The positive effect of training focused on skills on employee job
performance is moderated by the level of TOT, in such a way that when TOT is high, the positive effect is stronger.
METHOD Research procedure
The study was conducted at semi-governmental knowledge institution in the Netherlands and involved all employees that made use of training offered by the “Academy” (online training platform) in the year 2015. In this case study, a multi-source design was used. Using multiple sources for data collection was decided to be most suited: training is measured very accurately through data stored by the online platform and job performance through the annual performance cycle, while perceptions of the almost 700 employees will be measured best with a survey. A survey is the most common approach to assess the relationship between HR practices and employee attitudes (Guest, 1999)
The study is semi cross-sectional. Most part of the data is collected at one point of time. Job performance is measured over two successive years and because of that measures the change over these years, making the study a little less cross-sectional. For validity, all the different divisions and teams were surveyed. This resulted in a broad mix of employees that participated in the study.
Direct effect design
The direct effect design tries to answer the question which effect training has on job performance. Data about employee job performance was collected from the HR department and is the measurement at the end of year 2014 and 2015, making it possible to calculate the change in performance in the year 2015. Data about received training will be collected from the online training platform and is a sum of the followed training in the year 2015. This is visualized as follow:
Figure 2: model of the direct effect design
Performance appraisal
2015
ultimo 2015
Moderating effect design
The moderating effect design tries to answer the questions what the effect is of the way employees perceive supervisor support, the investment in employee development and transfer of training on the relationship between training and job performance. To test the moderating effects, data was collected through the use of a web-based survey. The content of the survey and invitation email can be found in the appendixes.
All participants received a notification email containing a link. In the invitation the purpose of the research was explained as well as an assurance that the answers would be completely anonymous and kept strictly confidential. The researcher ensured anonymity by using coded versions of the employee personnel numbers for connecting data from the different sources. The items in the survey are all self-reported and the questions are based on prior used questions and scales.
In total, 586 invitations for the survey were send out. Of these, 178 surveys were (partly) filled, of which 121 were fully completed: a response rate of 20%.
Finally, demographic information, such as age and tenure at the individual level - used as control variables - was provided by the HR department.
Sample
The sample for this study consists of 121 employees of the researched organization. A total of 586 employees were invited to fill in the survey; all of the employees who at least once made use of training supplied by the Academy in the year 2015. The list of the employees was obtained from the Academy database. Following on the invitations, 121 usable responses of the survey were received. 43 (36%) of the respondents are female and 76 (64%) are male; of them 7 (6%) are managers and 112 (94%) are non-managers. The average age was 44 and the average tenure 10 years.
Measures
Job performance, the dependent variable, is expressed in the employees’ performance appraisal and extracted from data administered by the HR department. The appraisal is based on the employees’ functional profile and the set of seven to ten competences that are linked to this profile. These competences are related to the dimensions Campbell (1990) has defined, for example analytical skills as job-specific task proficiency, initiating as demonstrating effort and coaching on leadership as facilitating team and peer performance. The final appraisal is spread
over a 7-point scale: excellent (in grade: 10), very good (9), good (8), more than sufficient (7), sufficient (6), does not fully satisfy (5) and insufficient (4). The aim for management is to differentiate in such way that the whole scale is used. In addition, due to many factors appraisals can vary from year to year. After all, it appraises the performance of an employee during a specific year. For this research, the performance over the years 2014 and 2015 will be used to calculate the change in performance over the year 2015.
Training, the independent variable, is measured in the amount of PE (‘Permanente Educatie’) points the employee has earned in one year. One PE point stands for one hour of training. Points can be earned for either training focused on knowledge or on skill. A training can be finished in a couple hours or can last for multiple days. The online training platform offers an extensive range of training that employees can choose from. Employees need to discuss with their manager which trainings they will attend with the rule of thumb that they at least have to spend 24 hours on knowledge training and 16 hours on skill training. Employees are free, in consultation, to spend more hours on training. An academic year starts in March and ends in February, due to performance agreements and in this research is measured in this way. The online training platform has a build-in function to extract reports, which will be used to measure the hours and focus of the training followed by the employees.
Perceived Investment of Employee Development (PIED), the first moderating variable, was measured using a six-item scale developed by Kuvaas & Dysvik (2009). An example item is “My organization invests heavily in employee training and development”. In the current study, PIED reported a Cronbach alpha coefficient of .877.
Perceived Supervisor Support for Training (PSST), the second moderating variable, was measured using a six-item scale developed by Santos & Stuart (2003). An example item is “My supervisor regularly discusses my training needs with me”. In the current study, PSST reported a Cronbach alpha coefficient of .894.
Transfer of Training (TOT), the third moderating variable, was measured using an adapted six-item scale developed by Xiao (1996). An example is “I can accomplish my job tasks more efficient than before training”. Two questions were adapted for the type of organization the survey was conducted in: in the Xiao study the research took place in a production environment where this study took place in a knowledge environment. The word ‘faster’ in two sentences was replaced by ‘more efficient’. In the current study, TOT reported a Cronbach alpha coefficient of .902.
For all three moderating variables, a five point Likert-scale was used, anchored from 1= strongly disagree to 5 = strongly agree. The questions were provided in Dutch, which first have been translated by the Translation back-translation procedure.
RESULTS Common tests
To test the hypotheses IBM SPSS Statistics combined with Hayes’ (2012) PROCESS procedure is used. Preliminary tests were conducted to check the data consistency. Frequency tables were run to check on missing data. This check is not performed for the survey questions because the survey did not accept non-responses. There was no missing data found in the dataset so it was left in its original form. Furthermore, the data was checked on errors: for both illegitimate codes and for illogical relationships. Two respondents were assigned a ‘0’ on their appraisal, a number that is not part of the appraisal range. These were most likely hires at the end of the year and therefore not appraised. These respondents were deleted from the dataset to ensure the quality of the statistical analysis.
Next, frequency distributions were used to check the distribution of the variables. When a distribution is normal, the majority of scores will lie around the distributions’ center and the further away from the center, the fewer the scores (Field, 2013). For each variable, a histogram was composed. None of the graph showed a deviation from normality. All variables did showed a slightly leptokurtic distribution, of which training showed a value of +.049 and job performance of +.406; for all variables there are more scores than average located in the center.
Finally, reliability checks for the constructs PIED, PSST and TOT were run. To test if the underlying questions actually measured what they are supposed to do, Cronbach’s alpha was carried out. All three constructs’ Cronbach’s alpha returned greater than 0.8, which indicates a high level of internal consistency. In none of the constructs removing a question would substantially increase the Cronbach’s alpha. Therefore, all questions were used in the analysis.
Computing scale means
Next to the directly measured variables, a new variable was calculated that displays the change in appraisal between the year 2014 and 2015. The 2014 appraisal is subtracted from the 2015
appraisal. If the appraisal went from 6 to 8, the change would be +2. The new variable is called ApChange.
New variables were also calculated to enable the measurement of the interaction effects. All questions underlying a construct were grouped and mean centered to form the new variables PIEDc, PSSTc and TOTc.
Correlation analysis
A bivariate correlation analysis was run to access the relation between the independent and dependent variables. All variables are numerical and to examine correlation between numerical data, Pearson’s product moment correlation coefficient, ‘r,’ is best used (Saunders & Lewis, 2012). The value for this test can differ between 1 and -1, where 1 represents a perfect positive relation and -1 a perfect negative relation. Also given is the p-value, representing the significance of the correlation. A relationship >.5 is considered strong, a relationship >.2 moderate. If the probability of the statistical occurrence is small (0.05 or lower), the relationship is called statistical significant (Saunders & Lewis, 2012).
Unfortunately, between none of the variables a statistical significant relationship was found. Hours of training in skill (r = .008, p = .618) as well as hours of training in knowledge (r = -.053, p = .568) showed no correlation with ApChange. PIEDc (r = -.064, p = .490), PSSTc (r = -.06, p = .519) and TOTc (r = -.002, p = .982), the moderating variables, also show no correlation with ApChange.
Control variables age, gender, tenure and manager/employee showed no correlation with any of the variables, with the one exception of manager/employee (r =-.257 p =<.01) that showed a moderate correlation with PIED, implying that managers intend to score the Investment in Employee Development by their organization higher than employees do. However, for the aim of this research, this is not relevant. This results makes it possible to do further analysis without controlling for these variables.
Table 1: Means, standard deviations, scale reliabilities and scale inter-correlations
Regression analyses
Even though there were no direct correlations found between the different variables, it is still interesting to see if there is a moderating effect of PIED, PSST and TOT on the training – job performance relation (hypotheses 2a to 4b). A moderating effect is in place when a variable “affects the strength and/or direction of the relation between a predictor and an outcome: enhancing, reducing, or changing the influence of the predictor” (Fairchild & McKinnon, 2008, p.89). To find out if any of these effects is going on in the data, regression analyses were performed. On forehand, z-scores were composed for all the variables: this score indicates the number of standard deviations the original score is away from the mean. With z-scores, all scores are standardized and better to compare.
Previous conducted correlation analysis showed no correlation for any of the control variables and therefore no control variables were added to the models.
Hypotheses 1a and 1b were tested using a linear regression analysis. The hypotheses state that if the hours spent on training – skill training for hypothesis 1a and knowledge training for hypothesis 1b - goes up, job performance will also go up. For both tests the coefficients of the Beta were not significant (skill: β =.008, p = .932; knowledge: β = -.053, p = .568). Thus, a change in the hours spent on training, either on skill as on knowledge training, do not increase job performance.
Dependent variable: ApChange
β t p
Independent variables: Hours spend on Knowledge training .0265 .3306 .7416
Hours spend on Skill training -.0528 -.5725 .5680
Table 2: results from the regression analyses
Variables M SD 1 2 3 4 5 6 7 8 9 10 1. Gender .36 .482 -2. Age 44.08 9.964 0,147 -3. Tenure 10.10 8.20 -0,114 ,630** -4. Mgr/ empl .94 .236 -0,109 -0,052 -0,115 -5. Knowledge training 15.24 13.524 -0,111 -0,05 0,065 -0,062 -6. Skill training 8.71 9.755 -0,04 -0,121 -0,07 0,128 -0,046 -7. PIED 3,8641 .59816 0,093 -0,027 -0,088 -,257** -0,048 -0,108 (.877) 8. PSST 3,3059 .76527 -0,059 0,124 -0,044 0,016 -0,046 -0,103 ,427** (.894) 9. TOT 3,3855 .55296 -0,062 0,045 -0,007 -0,052 0,061 -0,086 ,355** ,298** (.902) 10. Appraisal change -.10 .915 -0,013 -0,007 0,071 0,168 -0,053 0,008 -0,064 -0,06 0,002 -**Correlation is significant at the 0.01 level (2-tailed)
Hypotheses 2a to 4b were tested using the PROCESS procedure. This procedure combines many functions of statistical methods for testing moderating effects in one or multiple models. To test hypothesizes 2a to 4b, for each individual hypothesis PROCESS model 1 was used: a simple moderation model where the relationship between predictor (X) and outcome (Y) is influenced by a moderating variable (Z). This moderating effect can be described by the following equation:
𝑌 = 𝑎 + 𝑏𝑋 + 𝑐𝑍 + 𝑑𝑋𝑍
Hypotheses 2a to 4b all have the proposition that how employees perceive something (either PIED, PSST or TOT) influences the relation between training and job performance. Regression analyses were made on Hours spend on Knowledge training and Hours spend on Skill training. The results can be found in table 3.
Hypotheses 2a and 2b propose an interaction effect between PIED and the training-job performance relationship. Two tests were performed. First, Hours spend on Knowledge training was used as independent variable (IV) (regression coefficient for the interaction effect XZ = .1039, t (119) = 1.1769, p=.2417) and second, Hours spend on Skill training was used as independent variable (regression coefficient for the interaction effect XZ= .0265, t (119) = .3306, p=.7416). For both tests, the coefficients for the interaction effect (XZ) were not statistically different from zero. Thus, a change in the level of PIED does not affect the relation training – job performance.
Hypotheses 3a and 3b propose an interaction effect between PSST and the training-job performance relationship. The same two tests were performed, first the Hours spend on Knowledge training was used as IV (regression coefficient for the interaction effect XZ= .1249, t (119) = 1.2476, p=.2147), for this test the coefficient of the interaction effect is not significantly different from zero. Thus, a change in the level of PSST does not affect the relation between knowledge training and job performance. Second, Hours spend on Skill training was used as IV (regression coefficient for the interaction effect XZ = .1770, t (119) = 1.7867, p=.0766). For this test, the coefficients for the interaction effect were marginally statistically different from zero. Therefore, a change in the level of PSST has a marginal positive effect on the relationship between Skill training and job performance. For better insights in this interaction effect, the coefficients of the model were visualized (graph 1). The plotted
interaction shows an interesting intersecting effect. When an employee spends a low amount of hours on skill training, the increase in his job performance is higher when he perceives low support from his supervisor than when he perceives high support. In the latter situation the change in job performance even becomes negative! This is exactly opposite from when an employee spends a high amount of hours on skill training: in this case when the employee perceives low support from his supervisor, the change in job performance is negative while for an employee who perceives he is highly supported by his supervisor shows an increase in job performance. Thus, even though no evident significant interaction effect was found, the interpretation of this marginally significant interaction is that a low amount of hours spend on skill training is best accompanied by low PSST when trying to increase job performance, while this increase in the situation of a high amount of hours spend on skill training is best gained with a high amount of PSST.
Last, hypotheses 4a and 4b were tested. An interaction effect between TOT and the training-job performance relationship is proposed. Hours spend on Knowledge training was used as IV in the first test (regression coefficient for the interaction effect XZ= -.0741, t (119) = -.7588, p=.4495) and Hours spend on Skill training as IV was tested second (regression coefficient for the interaction effect XZ= -.1096, t (119) = -1.2763, p=.2044). As well as for the first tests, the coefficients for the interaction effect were not statistically different from zero. Thus, a change in the level of TOT therefore does not affect the relation training – job performance.
Interaction Model coeff t p
PIED x Hours spend on Knowledge training .1039 1.1769 .2417 PSST x Hours spend on Knowledge training .1249 1.2476 .2147 TOT x Hours spend on Knowledge training -.0741 -7588 .4495
PIED x Hours spend on Skill training .0265 .3306 .7416 PSST x Hours spend on Skill training .1770 1.7867 .0766 TOT x Hours spend on Skill training -.1096 -1.2763 .2044
Graph 1: The interaction effect of PSST on the skill training – job performance relationship
DISCUSSION Findings
The primary aim of this research was to answer the overall question: ‘What is the effect of training on employee job performance and how is this affected by the way employees perceive their supervisor support for training, the perceived investment in development by the organization and the perceived transfer of the training?’ Answering this question, with the use of theory and completed with a case study, could contribute to the current literature and give further direction in the ways organizations structure the environment surrounding training, in order to positively affect job performance. The overall question was decomposed in four sub-questions and its related hypotheses. Except for a marginal effect for PSST on the skill training – job performance relationship, none of the other hypotheses were confirmed and thus had to be rejected. In the following part of the discussion, these main findings regarding the hypotheses will be critically reviewed. Potential explanations for the results will be offered and literature is cited to give further insights.
Hypotheses 1a and 1b
With regard to hypotheses 1a and 1b, it was expected that an increase in training, both knowledge based as skill based, would lead to an increase in job performance. This expectation was based on previous findings in current literature. For example Bartel (1992) found a positive
-0,25 -0,2 -0,15 -0,1 -0,05 0 0,05 0,1 0,15 0,2
Low hours skill training High hours skill training
ApCh
an
ge
Low PSST High PSST
correlation between training and growth in employee salary, which she explains by the increase in performance value for the organization. However, these hypotheses were not confirmed.
With the use of literature, I will try to give possible explanations for not finding a relation between training and job performance in this study. For one, even though in general training has a positive effect on job performance, this can vary per type of organization. Different types of organizations are engaged in different activities and need different knowledge strategies (Laursen & Mahnken, 2001). The creation, integration and utilization of knowledge can be stimulated by HR practices in a number of ways (Lado & Wilson, 1994) and can involve different types of learning (Pavitt, 1984). For example, increased delegation of responsibility may be a better practice in knowledge intensive organizations than offering formal training (Jensen and Meckling, 1992). As the organization in this study is a knowledge intensive organization, the development in job performance could very well be based on the desired outcomes of training, but achieved in different ways. Employees might have developed themselves for example with job-rotations or learning on the job.
A second explanation can be found in the ambiguities of performance appraisal. Ouchi (1980) suggests that neither employee behavior nor their outputs can be assessed with precision due to misunderstood cause-effect relations. For example, when employees operate in teams, it’s hard to determine exactly who has contributed what to the result. A second complexity is the possibility of appraisal bias (DeNishi & Smith, 2014), for example when a manager gives all his followers somewhat the same rating. These arguments suggest that using a different indicator than appraisals might be better suited to measure job performance. Possibly employees that followed more training than average, in fact did do their job more efficient or effective than their peers, but this has not been appraised as such.
A third explanation might be in the measurement of training. In this research the variable training was measured by the hours spent on training. It can differ between employees, for example due to cognitive or motivational reasons, how much they actually learn at training. For example Colquitt et al. (2000) have researched how individual traits influence training capability. This could cause a great deal of variance between employees, muddling the results of the effectiveness of training in this research.
A last explanation can be found in the difference between the moment training took place and the measurement of job performance. Training can be followed throughout the year and does not have to be spread equally. Because of this, there is a lot of variance between the last day of training and the measurement, which can significantly moderate the size of the
Moderating effects
The absence of a relation between training and job performance in this study directly affects the proposed effects of PIED, PSST and TOT on this relationship. If training supposedly does not improve job performance, than the importance of the question if the overall effect of training is enhanced either by how the employee feels about the investment of the organization in his development, the support he receives from his supervisor or the transferability of the training to his work place is less obvious. Still, there was the possibility to find some interacting effects, meaning that an effect of training on job performance would only appear in certain circumstances. This was indeed found, only marginally, for the effect of PSST on the skill training - job performance relationship. For the other hypotheses, no interaction effect was found. Different explanations can be offered for these outcomes. These explanations are described in the next paragraphs.
Hypotheses 2a and 2b
It was expected that the employee perception of the investment in employee development of their organization would be partly responsible for differences in the training – job performance relationship. This proposition was reflected in hypotheses 2a and 2b, stating that an increase in PIED would positively affect the effect training has on job performance. This proposition is based on combined findings from current literature. Bartlett (2001) for example found in his study that PIED has a larger effect on employee outcomes than the actual hours spend on training. Unfortunately, in this study these hypotheses were not confirmed.
A possible explanation could be found in the data. There is a clear similarity in the respondents’ judgments of the organizational investment in development. The mean (3.86 on a 5 point scale) is high and the standard deviation low (SD= .59). Even when adjusted the data for this, it can still imply that there is just not a real difference in how employees perceive organizational investment. In that case, logically no interaction effect can be present. Possibly, the clear communication strategy of the organization is responsible for this uniformity.
Hypotheses H3a and H3b
It was expected to find a part of the differences in the training – job performance relationship explained at the level of how the employee perceives the support from his supervisor. This proposition was reflected in hypotheses 3a and 3b, stating that an increase in PSST, would lead
to an increase in the positive effect training has on job performance. This assumption is based on previous findings. For example, Olivero (1997) found management support accountable for no less than ¾th of the increase in productivity followed after training. In the current study only a marginally effect (p<.08) was found for PSST, and only there where training was focused on development of skill. For knowledge training no similar result was found.
A closer look at the marginally significant interaction effect for PSST on the change in job performance after skill training displays an interesting story (graph 1). High perceptions of supervisor support only increases the skill training – job performance relationship when the hours spend at skill training is high as well. If the hours spend on skill training is low, high PSST has a negative effect on job performance. I will offer potential explanations for this interesting effect. The fact that the relationship is only marginally significant, however, must not be forgotten.
An explanation could be that when a supervisor is not perceived as supportive for training, for example when he does not jointly set goals and does not discus training needs, this might imply lack of recognition from the supervisor for skill training. If one of his employees spends a lot of time on training his skills, he is not appreciative of this time spending or not receptive for the results. This may also be dependent on the employees’ job tasks. For every task other KSAs are necessary to have for employees to perform the job (Noe et al., 2012), and the importance of skills capacities can vary between jobs. On the other hand, supervisors who are perceived as highly supportive for training, setting joint goals and discus training needs, may appreciate the effort and effects of skill training and will reward this corresponding. Either way, this is a complex outcome, showing it is not as simple as instructing supervisors to be more supportive to increase the effect of skill training. Although marginally, this result adds to our understanding of the complexity of HR practices, in this case training. This is in line with Nishii et al. (2008) who also found that the same set of HR practices do not exhibit the same effects within a single organization. Other moderating variables might be at work and therefore more research at the individual level is necessary for better understanding of these complexities. A phenomenon that might be an explanation for the absence of findings on knowledge training can be that social support, covering also supervisor support, may be of different importance for certain job types. Fitzgerald and Kehrhahn (2003) state that employees in an autonomous job context rely less on supervisor support in their decision to transfer training. This is confirmed by Ng and Sorensen (2008), who state that social support is particularly beneficial for employees that are constant under internal and external pressure, for example due
may not benefit in the same way. The employees of the organization in this study have very autonomous jobs and therefore might not have a very strong need for supervisor support.
Hypotheses 4a and 4b
The last two hypotheses state that an increase in transfer of training would lead to an increase in job performance. Meaning that the employees’ perception of the fit and usefulness of training would be partly responsible for the differences in the training – job performance relationship. These assumptions were based on findings from current literature. Xiao for example found is his study in 1996 that transfer of training is of importance in facilitating the usage of new capacities. Unfortunately, the results from this study did not confirm these hypotheses.
A possible explanation can be found in the difference between motivation to transfer and intention to transfer. The question is if the respondents in this study were able to distinguish their willingness to transfer from actual using the learned on the job. Al-Eisa et al. (2009) explain the difference between motivation and intention to transfer as follow: “Whereas motivation to transfer refers to a desire to initiate transfer, transfer intention refers to a commitment to initiate transfer”. So in the researched organization, the desire might be present and not so much the commitment.
A second explanation can be found in the maintenance of transfer. Wexley and Latham (2002) found that from the learned immediately after training forty percent is transferred while a year later only fifteen percent of transfer is left. Measuring TOT directly after training therefore gives a different result than measured later in time. Combining this with the knowledge that employees are free to spread training in the way that suits them, early in the year, at the end or an equal distribution, causes the measure of TOT to be ambiguous.
Limitations
The contributions that can be drawn from the current research should be viewed with its limitations in mind.
A first limitation is in the largely cross-sectional design of the study. An assumption of the model is that the independent variable causes the effect on the outcome variable and that the three moderators affect this relationship. Unfortunately, this relationship cannot be proven in a cross-sectional design and only be supported with the use of theory (Kong et al., 2012). This research has put some effort in avoiding this by examining the change in appraisal between ’14 and ’15, making this study slightly longitudinal. However, it does not overcome causal
inference problems because of the risk on a delayed effect of training. It is questionable if training in the year ’14 immediately affects appraisals given the end of that same year. Other studies that test similar relations could consider an experimental or longitudinal design. When constructed in that way, causality inference can be made on the relationships – when they are found.
A second limitation is the possibility of common method variance. The survey questions are based on the evaluation of the employees, something according to Fiorito (2002, p.217) can lead to an “artificial correlation across questions due to mood or other contaminants”. This study did use other sources as well, objective training data and structured appraisals, which are less subjected to mood. Using different sources for the variables reduces the potential influence of common source and common method biases (Podsakoff et al., 2003). It might be useful to repeat this study using a 7-point Likert scale instead of a 5-point Likert scale. Possibly this will better capture the delicate differences between employees and in this way overcomes the risk of answering in a desirable way.
A third limitation can be found in the use of one firm in the study. The use of one firm has both drawbacks and advantages (Den Hartog et al., 2012). The generalizability to other firms and sectors may suffer due to this focus. In the meantime, the variable training is constant because all the researched employees are offered the same training opportunities, not influenced by manager implementation. Because of this, the measured training only varies in hours spent at training and the focus of the training which both are measured and part of the research model. This gives more confidence about the results than would be in a multiple organizational study. A last limitation can be found as well in the area of generalizability: due to the fact that the sample only includes Dutch employees, it is difficult to draw conclusions for employees of other countries. For example in non-western countries, the influence of a manager or the perception of investment in employee development could be very different, caused by cultural differences; Hofstede (1980) for example explains a difference in power distance in the different cultures. Repeating this study in cultural deviating countries would be valuable.
Theoretical implications
The more studies on HR practices, employee perceptions and different levels of outcomes are performed, the clearer the complexity of the HR – organizational performance relationship becomes (e.g. Nishii et al., 2008; Wright and Snell, 2001; Kuvaas and Dysvik, 2010). In this research I proposed a positive effect for training on employee job performance. Contrary to the
PSST a marginally significant moderating effect was found and solely where it involved skill training. For all the other propositions no moderating effects where found either. Hopefully, in the same way the mixed results on the effect of HR practices invited scientists to research underlying aspects of training to gain better understanding of the phenomena, this research is an invitation for studies on employee perception of different aspects surrounding training.
A recent development in SHRM research, which is a direct response to this complexity, is the recognition of the unit of analysis as important. When it comes to investigating the training – job performance relationship not only the implemented training as a whole should be considered. Moderators can be identified and many of them are influenced by the perception of the employees. Therefor, measuring moderators at the level of the organization, only measuring the way they are strategically designed and implemented, is not sufficient to understand differences at the level of the employee. This study has tried to add to this understanding by examining perceptions on organizational investment in development, supervisor support and transfer of training that have not been researched before. The marginal significant effect on the relationship between skill training and job performance found for PSST confirms that differences between employees can be a result of moderators. In certain situations, higher PSST leads to better job performance. It would be valuable if in future research this effect would be examined more closely to understand the mechanisms that are in place. Furthermore, this finding supports the idea that the unit level of analysis is of importance in understanding the complex effect HR practices have. Researchers still have many complexities to sort out and for this, more and extensive research is necessary.
Practical implications
Many organizations try to influence employee outcomes in a positive way by facilitating training opportunities. And because training is such a costly HR investment, increasing its rate of return is very profitable to organizations. An interesting question for organizations is how they can assure that the investment results into the highest possible positive outcomes. Many early researchers have examined in which way the design of training can accomplish that goal (e.g. Kirkpatrick, 1967). Later studies also included other levels of the training process, for example the implementation and how employees perceive training (e.g. Khilji & Wang, 2006; Dysvik & Kuvaas, 2010). This research tried to build further on the concept that employee perceptions on certain aspects affects the relationship between training and employee performance. It did so by looking into how employees perceive the investment of the
extent they perceive applicability of the learned to their work environment. These aspects are all opportunities for organizations to influence and to that end, have tools to stimulate the positive effect training has. An outcome of this research is a marginally significant effect for perceived supervisor support on skill training effectiveness. For HR practitioners, this suggests that assessing the level of PSST can provide valuable information about why skill training has or has not led to increased job performance. More in general, organizations should be aware that implementing training is about more than just the content of the training. Other variables can also be influenced to achieve optimal results.
In addition, theory on training effectiveness suggests that organizations should also be aware that the kind of organization, for example knowledge intensive vs. production intensive, is relevant. For example, increased delegation of responsibility may be a better practice in knowledge intensive organizations than offering formal training (Jensen and Meckling, 1992).
REFERENCES
Aguinis, H., & Kraiger, K. (2009). Benefits of training and development for individuals and teams, organizations, and society. Annual review of psychology, 60, 451-474.
Al-Eisa, A. S., Furayyan, M. A., & Alhemoud, A. M. (2009). An empirical examination of the effects of self-efficacy, supervisor support and motivation to learn on transfer intention. Management Decision, 47(8), 1221-1244.
Alfes, K., Shantz, A. D., Truss, C., & Soane, E. C. (2013). The link between perceived human resource management practices, engagement and employee behaviour: a moderated mediation model. The international journal of human resource management, 24(2), 330-351.
Baldwin, T. T., & Ford, J. K. (1988). Transfer of training: A review and directions for future research. Personnel psychology, 41(1), 63-105.
Baldwin, T. T., & Magjuka, R. J. (1997). Training as an organizational episode: Pretraining influences on trainee motivation. Improving training effectiveness in work organizations, 99-127.
Bartel, A. P. (1992). Training, wage growth and job performance: evidence from a company database (No. w4027). National Bureau of Economic Research.
Bartlett, K. R. (2001). The relationship between training and organizational commitment: A study in the health care field. Human resource development quarterly, 12(4), 335-352.
Becker, B. E., & Huselid, M. A. (2006). Strategic human resources management: where do we go from here?. Journal of management, 32(6), 898-925.
Boon, C., Belschak, F. D., Den Hartog, D. N., & Pijnenburg, M. (2014). Perceived Human Resource Management Practices. Journal of Personnel Psychology.
