Effects of training needs assessment in enhancing employees’ skills and firm performance

Abstract

Purpose

Work-based learning is critical for enhancing employees’ skills and contributing to the firm’s performance. This paper aims to establish the effects of needs assessment on the relationship between training intensity as part of learning and how employees’ skills are reflected in firm performance.

Design/methodology/approach

The paper used the World Bank Tanzania Employees Skills Survey (TESS) dataset, which contains 424 firms. This paper estimated the moderated mediation model through partial least squares structural equation modelling (PLS-SEM) and employed the index of moderated mediation to determine if the model was correctly specified.

Findings

The results show that among three skills, i.e. technical, interpersonal/communication and work ethic skills, only the level of technical skills mediated the relationship between training and the firm’s performance. The index of moderated mediation suggests a threshold point for the firm’s training needs, above which the indirect effect of training on performance through technical skills starts to decrease. The negative correlation between the firm’s training needs and the indirect effect suggests that employees’ essential human capital qualities, viewed from the angle of their training needs, are among the key factors for executing effective training.

Research limitations/implications

This paper’s conceptual model is limited because it does not incorporate an education variable for the trained employees. In addition, it only conceptualized the perceived most important skills of interpersonal communication, technical skills and work ethic, despite there being other skills that could have been considered. Moreover, the data only measured the present skill level at three on the Likert scale, providing limited room for skill level variance.

Practical implications

Those who decide which training programme deserves priority given limited resources and the firm’s goals need to understand that training is an addition to what their employees already have and, thus, should make extra efforts to equip them with more knowledge relating to their assignments. Moreover, this understanding should extend to the employees themselves.

Originality/value

The paper introduced and showed the necessity of training needs assessment to increase the value of training in enhancing the firm’s performance. We propose a model for assessing training intensity through process analysis. The respective model depicts a threshold point for the firm’s training needs, below which the training will work.

Citation

Robert, N. and Mori, N. (2024), "Effects of training needs assessment in enhancing employees’ skills and firm performance", Journal of Work-Applied Management, Vol. ahead-of-print No. ahead-of-print. https://doi.org/10.1108/JWAM-05-2023-0046

Introduction

Various learning practices and policies substantially contribute to a firm’s sustainable competitiveness due to the unique features of human resources compared to other resources (Zabiegalski and Marquardt, 2022). Training and learning policies are among the policies that firms strategically employ to improve the productivity of their employees. Through these policies, employees gain access to training to help them perform their tasks well. The investment in employee training is primarily made to meet the firm’s needs and priorities at a particular time to maximize the expected output (Ballot et al., 2006; Becker, 1994; De Grip and Sauermann, 2012). Typically, no direct effect of training occurs on the firm’s output, but rather, the effect is transferred through predetermined responses while considering the existing circumstances (Chi et al., 2008). Yet evaluating the presence, strength and significance of the expected responses on the final output is needed to make informed conclusions on the total effect (Aragón-Sánchez et al., 2003; Preacher et al., 2007).

For firms to offer training for their employees, they need to assess their training needs. Training needs assessment helps determine whether training is the right solution for the employees. It is also an ongoing process of gathering data to determine what training needs exist and what skills are missing among employees so that training can be developed to help the firm accomplish its objectives. Training is human resources' most fundamental organizational activity to enhance competence and productivity. However, achieving employee training needs cannot be fully accomplished without a need analysis.

While most existing training evaluation models present the mediation process through which the training effect is transferred to the final output, few have considered the existing circumstances within the firms as determinants of the return to be obtained (Tharenou et al., 2007). While a handful of studies consider the moderation of the training effect transfer process, they do not exclusively reflect how training directly defines the status of the employees’ skills depending on their existing needs (Chi et al., 2008). In addition, there is a gap in the literature showing how training needs assessment assists firms in determining the type of training to be undertaken, the skills learned and the extent to which this assessment contributes to the firm’s performance (Mohd Hashim and Tasir, 2020). Addressing the gap in the literature and attempting to extend the information managers should consider when evaluating the value of training, this paper builds on Tharenou et al. (2007) conceptual model by considering the “when” factor in the evaluation process. This approach establishes not only the process involved in the transfer of the effect but also the circumstances under which training adds value, especially after a proper assessment of training needs. The question that this paper addresses is as follows: How does a training-needs assessment influence the relationship between the conducted training and the firm’s performance? We analyse training needs since this is known to be an approach that bridges the gap between performances from an expected state to the current state (Sharma, 2018).

Using the TESS, the PLS-SEM is employed to predict the value of training for firms´ returns. The moderated parallel mediation model, guided by Hayes (2018) from his statistical arguments and suggestions, is then formulated (Hayes, 2009, 2015, 2017, 2018; Hayes and Scharkow, 2013). Moreover, other authors, like Bauer et al. (2006) and Namazi and Namazi (2016), support Hayes’s model.

The value of training is linked to human resource outcomes, as observed in our case through three employee skills perceived as necessary by the top management. However, it also depends on the fundamental quality of the employees themselves (Colquitt et al., 2002). Furthermore, the management determines if the existing training needs result from a lack of other human capital sources, such as low levels of education, less experience on the job or negative impacts from family background (Becker, 1994). Training needs might be high, but if the capacity of the trained employees is low, the knowledge imparted will be less valuable (Blundell et al., 1999).

This paper is organized as follows: In section two, the paper presents a theoretical and empirical literature review, acknowledging previous studies' contributions that provided inputs in formulating the conceptual model and hypotheses to be tested. Section three describes the data used, identifies the variables, specifies the model and presents the analysis methods. The findings and discussions are presented in sections four and the final section offers the conclusions and implications.

Literature review

The theoretical literature on training effects suggests several models that explain the systematic way in which training might lead to expected firm performance, that is, the mediation process (Kirkpatrick and Kirkpatrick, 2006; Yamnill and Mclean, 2001). This current study is guided by the expectancy theory, also known as the process theory (Vroom, 1964). This theory argues that a person who expects a desired consequence will be encouraged to make a decision that will give rise to that result. In this case, managers need to know their employees' expectations, divergent skills and abilities and assign tasks based on their capabilities and competencies.

The conceptual model by Tharenou et al. (2007) links to this theory by suggesting steps for assessing and conducting training. Part of their aim was to establish clear causal linkages in the process and address the intervening variables by displaying the mediation role, as seen in Figure 1. However, their model considered the “how” part of training and omitted the “when” part. As explained by Preacher et al. (2007), where necessary, the analysis becomes richer when the “how of the when” or “when of the how” is taken into account during model formulation. This is even more important in training evaluation models since the effect is not direct but is determined mainly by different circumstances.

Several different empirical works, like that of Becker and Huseild (1998), have emphasized the use of mediation models to realize the value of training. A decade later, Zinovieff and Rotem’s (2008) study on training evaluation methods argued that most training evaluations were not efficiently done and therefore came to incorrect conclusions. They emphasized that evaluating a particular implemented programme aims to provide empirical inputs for policymakers and decision-makers, so the information should be rich. Nevertheless, few empirical business studies have been written based on the moderated mediation models. The difference between this paper and the existing studies is how we define training needs, employee skills as part of training intensity and the methodology used to test the model.

Training needs assessment collects information about an expressed or implied firm’s needs that could be met by training its employees (Barbazette, 2006). Mohd Hashim and Tasir (2020) pointed out that training needs assessments are widely used as an evaluation tool by firms to determine if training is the best solution to their problems. Building on this, this paper argues that the training needs assessment moderates the relationship between the actual implementation of the training and the enhanced skills.

Several studies have analysed the value of training in different contexts but have not used process analysis (Hayes, 2017). For instance, Aragón-Sánchez et al. (2003) assessed training value in terms of employees’ involvement, human resource indicators, quality and labour productivity. The analysis needed to be extended to firm output. Another group of authors did not use human resource outcomes when examining the value of training. For example, Chi et al. (2008) conducted a process analysis to determine if foreign direct investment (FDI) training was effective for existing needs. In their moderated mediation model, one more mediator was expected after training to determine if the conducted FDI training correlated with the employees' required skills post-training before observing the final effect on the firms’ performance. The lack of human resource outcomes in the model should not be considered lightly since the performance of a firm, as observed from sales, can be determined by several factors apart from training (Aragón-Sánchez et al., 2003; Vega-Jurado et al., 2008). The limitation observed in Chi et al.’s (2008) model was also observed in the work of Ng and Siu (2004), who argued that the value of training can be objectively determined by observing a firm’s productivity through its sales. Similarly, they defined the value of training through a direct effect, missing the process of transferring the effect.

Regarding the mediation test, most empirical in business and management studies have used the traditional Baron and Kenny (1986) test, which was reconsidered in several studies (Hayes, 2009, Hayes, 2017; Zhao et al., 2010). The guidelines provided by Hayes (2017, 2009, 2015, 2018) were used primarily in communication, psychology and clinical studies and less in business and economics studies. In this study, we also address the methodological limitations of other empirical studies regarding establishing the value of training needs and skills built by accommodating the recent mediation test guides.

Conceptual model and hypothesis formulation

The conceptual model of this paper extends the mediation model of Tharenou et al. (2007) by including firm training needs as a moderating factor. This paper also expands the human resource outcomes to target more specific skills perceived as necessary for the firm’s output. Figure 2 portrays the conceptual model of this paper.

Few studies have evaluated the effect of training needs assessments on the performance of firms. A study by Sharma (2018) assessed the types of training required for employees in the real estate industry. The results showed that employees needed more technical skills training than soft skills. Kaewkunha and Sukying (2021) had similar results. From the literature, we argue that when individual skill gaps are correctly determined, the training implementation is expected to become effective and address the firm’s training needs (Daniels, 2003; Freel, 1999). In addition, previous studies focused on the direct effects of training needs assessments. At the same time, we hypothesize its moderating role in the relationship between the actual training implementation and the change in the skills required. Therefore:

H1.

A firm’s training needs to positively moderate the relationship between training and employee skills change.

The meaningful improvement of the needed employee skills after training is a priority for the firm’s decision-makers to ensure a significant return on their investment (Ballot et al., 2006; Yiu and Saner, 2005). Sadaf (2014) also noted that a skilled workforce through training could significantly enhance the employees’ performance and effectiveness. Therefore:

H2.

There is a positive relationship between the employees’ skill status and firm’s performance.

According to Kraiger et al. (2004), training should be linked to the existing firms needs to become effective. These authors argued that a needs assessment should determine how the training strategy will be implemented to obtain the forecasted return to the firm through human capital. The empirical literature also supports the argument that training leads to expected returns for the firm when addressing existing needs (De Grip and Sauermann, 2012). This hypothesis intends to determine if the model is conceptualized correctly. Hence:

H3.

The total indirect effects between training and the firm’s firm performance is moderated by the training needs

While Hypothesis 1 and 2 were tested and concluded in the first stage of the moderated multiple mediation model (analysed under the PLS-SEM (Hayes, 2017, 2015), hypothesis 3 was tested through the index of moderated mediation (Hayes, 2015).

Methodology

Data

This paper employs secondary data from the 2015 TESS conducted by the Enterprise Analysis Unit and the Education Global Practice of the World Bank Group. As detailed in the TESS database, stratified random sampling was used to obtain and design the sampled firms based on eight pre-identified economic activities (subsectors) following ISIC [1] code revision 3.1. The surveyed firms were categorized according to size (small, medium and large) [2]. The respondents were either the top managers or a single manager with significant experience with the firm to ensure the appropriateness and validity of the data. The questionnaire focused on the firm’s skill levels and development through training. The total observations from the database included 424 firms.

To establish reliability, Cronbach’s alpha test was used to establish the strength of the consistency, particularly for the employees’ skills status data, measured through Likert point scale questions and training needs data. The key was ensuring that the respondents consistently measured the same concepts when observing the total score variance. The test results for employees’ skills status were 0.74 and for training needs, they were 0.71, which are acceptable levels.

Variables

The primary dependent variable is firm performance, also referred to as productivity. This variable defined the previous year’s sales (Zwick, 2006). The independent variable is training, measured as the number of employees who received training in the past two years (Zwick, 2012). The moderating variable is the firm’s training needs, measured from a composite of five questions related to improving production levels, implementing new technology and effective sales and marketing. The mediating variable is the status of employee skills, illustrated by three skill variables: interpersonal communication, technical and work ethics. The control variables are measured as a binary variable equal to one if the firm has 20 employees or less and zero otherwise. The sector measured as a binary variable is equal to one if the firm is in the manufacturing sector and zero otherwise. Table 1 summarizes the variables.

Model specification

The model assumes that training impacts employees’ skills and matches the firm’s existing needs. In that case, the study used a moderated parallel multiple mediator model whereby the path from training intensity to the firm’s performance/productivity was mediated by the present status of the three skill levels, as seen in Figure 3. Training needs are included a moderator on the training path for the three employees’ skills. This establishes the level of the respective employees' skills to meet the existing training needs. Hayes (2017) also adopted a similar model.

This model is called a first-stage moderated parallel multiple mediation model as the moderator only covers path A of the model (Hayes, 2015). In estimating the model, two equations are involved to obtain the effect of X (training) on Y (firm performance) through the mediator variables. While the first equation represents the effect of X on M (skills), the second examines the effect of M on Y when controlling for X.

(1)Mi=KMi+a1iX+a2iW+a3iXW+eMi

(2)Y=KY+c′X+b1M1 +b2M2 +b3M3 +σSizeDummy+φSectorDummy+eY

where Y is firm performance, K is constant and i = 1, 2 and 3 are the number of mediator variables. X is training, W is the firm’s training needs, M₁ is interpersonal and communication, M₂ is technical and M₃ is work ethic. It is important to note that our interest in this paper is focused on the moderated indirect effect (as per hypothesis 3).

From equation (1), the effect of X on M is (a_1i + a_3iW), and its moderating effect is equation (2), where the effect of M on Y is b_i. Different scholars, including Hayes (2018) and Preacher et al. (2007), explain how to deduce the indirect effect of X on Y by finding a product of the conditioned effect (X on M, moderated by W) and the indirect effect of M on Y. The obtained indirect effect is expressed as the linear function of the moderator variable, as seen in equations (3), (4), and (5). Since the mediation variable is measured using three skills, it implies three specific indirect effects of X on Y to be tested. The first one is through M1 (interpersonal and communication), the other is through M2 (technical) and the last is through M3 (work ethic). The aim is to understand their presence, strength and significance.

(3)λM1=(a11+a31W)b1=a11b1+a31b1W

(4)λM2=(a12+a32W)b2=a12b2+a32b2W

(5)λM3=(a12+a33W)b3=a13b3+a33b3W

Index of moderated mediation

From equations (3) to (5), we can examine the index of moderated mediation (λ) for each specific indirect effect, M₁, M₂ and M₃, respectively (Hayes, 2015). The index, which also measures the fitness of the moderated mediation model in this case, intends to quantify the strength of the training needs (W) to increase or decrease the indirect effect of X on Y (i.e. training on firm performance) through each specific skill. The logic behind this choice is to measure the relationship between the moderator and the indirect effect to assess whether the index differs from zero. In that regard, the respective index through M₁ is a₃₁b₁, M₂ is a₃₂b₂ and M₃ is a₃₃b₃. These indices are not meant to test the causality of the mediation but rather to interpret the results obtained from the estimation, establishing the model’s fitness. They are intended to observe how the statistically determined mediated effect varies by the moderator variable in the model (Edwards and Lambert, 2007; Preacher et al., 2007). Not having a moderated direct effect in equation (2) does not affect the extracted index of moderated mediation (Hayes, 2015). However, caution is taken when interpreting the index, considering the possible limitations of the data employed, e.g. normality. In that case, a confidence interval (CI) bootstrapped technique using resampled data is used during the causality estimation.

Mediation analysis

There are several methods of analysing mediation models, the most famous being the causal step approach suggested by Baron and Kenny (1986). In addition to the three steps in their approach, determining a significant relationship between independent and dependent variables before proceeding with the mediation test is emphasized (Baron and Kenny, 1986). Several scholars, however, have reconsidered the arguments presented by Baron and Kenny (Hayes, 2009; Zhao et al., 2010). Hayes (2009), for example, argued that the mediation effects should still be tested even when there was no direct relationship between independent and dependent variables. His argument is predicated on the possibility of no significant direct effect being observed between X and Y; in contrast, multiple indirect effects in between might determine the relationship between the two variables. Sometimes, other indirect effects might be positive and others negative, which blunts their effect. Still, the mediation role of a certain moderator is worthy of being tested as a specific indirect effect. This study aligns with Hayes’s (2015) argument and continues to test for specific indirect effects, regardless of possible direct effects on X and Y.

Estimation technique

This study employs PLS-SEM estimation to obtain the coefficients for the first and second equations. The PLS-SEM analysis reduces the bias of using the observed variables and represents the correct method for prediction purposes. It presents the already bootstrapped results when testing for significance (Hayes, 2018; Preacher et al., 2007). According to Hayes (2009), because the sampling distribution of the indirect effect is not always normal, the estimated parameters are conditioned for 95% bootstrapped confidence intervals in order to obtain the empirical representation of the sampled distribution of the indirect effect.

Zhao et al. (2010) proposed five possible conclusions from the mediation test, which we also adopted. These include complementary mediation, competitive mediation, indirect-only mediation, direct-only non-mediation and no-effect non-mediation. While the signs of an indirect effect and a direct effect are the same, the signs of the two effects differ in competitive mediation. In indirect-only mediation, only indirect effects are significant; in direct-only non-mediation, the indirect effects are not significant. Nothing is significant in the case of no-effect non-mediation. In this case, the direct effect is X on Y when controlling for M and the identified firm size and sector covariates.

Findings and discussions

Descriptive statistics

General descriptions were generated to understand the advantages realized by a firm’s investment in training. Table 2 compares the existing firm’s training needs and employees’ skill status ratings before and after the training period. The results show that, despite similar needs, not all firms trained their employees and those who did so had an advantage since approximately 5% more firms responded that their employees’ skills were above those required. In terms of work ethics, more trained firms responded negatively. This response can be partly explained by the likelihood of a decline in employees’ work discipline due to increased market value from the knowledge acquired.

Figure 4 compares the productivity of the two groups and the results show that the firms that have trained employees reported more output than those that did not. These statistics still need to be interpreted with caution due to the possibility of a causal effect. There are chances that firms with more sales revenue have more capacity to train their employees compared with others.

Despite the possibility of causality, the fact that there was an increment in firms’ output after the training, as shown in Figure 5, cannot be ignored. While the sales increase for the group that trained their employees was, on average, USD 423,000, the untrained firm group only had an average increase of USD 76,500.

PLS-SEM results

As presented in Table 3, the SEM estimation results include the path coefficients and the 95% bootstrapped CI to reflect the significance of the effects. The path coefficients include direct effects, from the independent to the dependent variable and indirect effects, which go through the mediating variables (Dhar, 2015). The results show that the direct effect c’ that presents the relationship between training and firm performance is not significant. Nevertheless, the nonsignificant direct effect provides more interest to continue studying the conditioned indirect effects of the employees’ skills, which is one of the potential factors in explaining the effectiveness of the training conducted (Hayes, 2009, 2017, 2018). This is typical of human resource expected outcomes, as postulated in Tharenou et al.'s (2007) conceptual model.

Usually, multiple mediators explain the relationship between the predictor and the outcome variables. The final revealed direct effect results from several indirect effects, with different signs of effect possibly cancelling each other (Preacher et al., 2007). Aligning with the goal of this paper, we only tested the mediation role of the three identified employees’ skills. Nonetheless, other indirect effects influence the direct effect witnessed, which is not in the scope of this paper.

Results for hypothesis 1: “Firm’s training needs positively moderate the relationship between training and identified employees’ skills status” are shown in Table 3. Here, a significant moderated relationship between training and the employee’s skill status is only reflected in interpersonal and technical skills. However, from the interaction variable path coefficient, when the firm’s training needs increase, there is a decreasing effect from training to technical skills (a₃m₂ = −0.1, p < 0.05). The result suggests that there is a threshold point for training to work, above which training alone might not work, but other human capital sources should be observed.

The only significant indirect effect on technical skills also implies that the training provided was more inclined towards technical knowledge and less towards other skills. In this case, our first hypothesis (hypothesis 1) is partially supported. We suggest that employees’ essential qualities, defined by their education level, experience and other sources, must be considered when evaluating the value of training.

A positive relationship between employees’ skills status and firm’s performance is hypothesis 2. Table 3 shows that, among the identified skills, only technical skills significantly positively influenced the firm’s performance (b₂ = 0.14, p < 0.01). These results imply a significant indirect effect of employees’ technical skills on the firm’s performance (95% CI = 0.04 to 0.24). For other skills, there was no significant indirect effect from training on the respective employees’ skill status, but their status does not influence the firm’s performance. This suggests that, although the two skills, interpersonal/communication and work ethic skills are important for the firm, their influence on performance is neither significant nor material. In this regard, hypothesis two (2) is partially supported only by technical skills. The results for hypotheses one and two suggest that technical skills fully mediate the relationship between training and the firm’s performance.

Results of the index of moderated mediation

This index establishes the relationship between training needs as the moderating variable and the indirect effects established through technical skills. This relationship is employed to test the third hypothesis, which is whether the mediation process as a whole is moderated by the firm’s training needs. In this case, no new statistical estimations were undertaken. However, an interpretation of the existing ones was conducted, and the conditional indirect effect of training intensity on the firm’s performance through technical skills was first established as a linear function (Figure 6) (Hayes, 2018; Preacher et al., 2007). The function is built from the conditioned indirect effect of training on technical skills (a₁₂ + a₃₂ W) and the effect of technical skills on the firm’s performance (b₂). The index is presented here as

λM2=(a12+a32W)b2 =a12 b2+a32 b2 W=(0.16 * 0.14)+(−0.1*0.14)W=0.0224 −0.014W

where 0.0224 is the intercept and −0.014 is the slope.

This equation determines the effect of training needs on the relationship between training and the firm’s performance through a technical skills mediator (a₃₂b₂). From the index, the indirect effect of training intensity on the firm’s performance through technical skills decreased by 0.014 with a unit increase in training needs. On the other hand, this value implies that a unit decrease in training needs results in a 0.014 increase in the respective indirect effects. These results are reinforced by the information obtained from the interaction plot in Figure 6, which presents the reaction of technical skills status in response to the increase in training intensity at different training need levels. Nevertheless, the fact that the index is not “zero” implies that the established indirect effect in this model is linearly related to training needs, as mentioned in hypothesis three and that the model was correctly specified (Hayes, 2015). We then accept hypothesis three (3) and suggest that the training needs to moderate the relationship between training intensity and the firm’s performance through technical skills.

Figure 6 further shows a significant positive relationship between training intensity and technical skills outcomes that occurs when there are low training needs. On the other hand, when training needs are above average, the relationship between training and technical skills is almost zero until the threshold point is reached. Then, the reaction becomes negative (Table 3 – interaction results).

Discussion

Empirically, when the firm’s training needs are high, the training is expected to have a more significant impact on employees’ skills if the training conducted responds to the need in place (Daniels, 2003; Macheke, 2012). Our results were contrary to this regarding technical skills. The theoretical literature explained that training in response to the existing skills demanded by the firm is to be determined through needs assessment (Kraiger et al., 2004). However, the results suggested otherwise. First, only technical skills mediated the training intensity effect on the firm’s performance. Skills are positively related to training intensity (when training needs are low) and the firm’s performance. Although firms identified three critical skills for their operations, the results suggested that most of the training taken was on technical skills. This result is not surprising, as the most important skills in the firm do not necessarily mean the most needed skills by the employees at that particular period. It always depends on the operational needs of the firm concerning the prioritised goals.

Secondly, when training needs were higher, the employees’ technical skills from the training were significantly low. The index even suggests that with an increase in training needs, the indirect effect of training on the firm’s performance through technical skills decreases with an increase in training needs. According to Darvas and Palmer (2014), several factors can explain this unmatched supply and demand, including the wrong type of training provided (topic) and the training quality. Still, more importantly, the essential qualities of the employees partly contribute to the firm’s training needs. In our findings, with low training needs, the training complemented existing individual capacity, significantly improving employees’ technical skills. The threshold point, which defines if training will work, is built from individual abilities that can differ from one person to another due to natural potential and conditioning factors like education and experience. However, on average, a firm needs to observe training levels when planning for them to ensure maximum positive effects.

This conclusion resembles that of Blundell et al. (1999), suggesting that the basic quality of the trainee determines how effective the training provided will be. Although their paper was related to training and education qualities, in this study, we align the training needs displayed by the firms as a reflection of their employees’ competencies. From that perspective, when a firm reveals more training needs, it implies that its employees are less competent and thus, less is expected when they receive training. Similar results were presented by Chi et al. (2008) that when the training needs were high, the implemented training led to a decrease in organizational performance, which can be defined as the trainings not addressing the existing needs. Generally, training will be declared effective when it significantly addresses the skill needs and so affects firm performance positively (Freel, 1999). However, the level of skill demand determines how effective the training will be if it manages to address a significant portion of the need. Some scholars present different arguments that having a stock of training for an employee is a strategy that can be used to address the existing demand and impact firm performance positively (Ballot et al., 2006). Nonetheless, skill assessment and consideration of the capability of the employees are still important attributes for ensuring the value of training, even when a stock of knowledge and skills is created (Darvas and Palmer, 2014).

Conclusion

This paper employed the moderated parallel multiple mediation model to examine the effect of the training the employees attended, moderated by the firm’s training needs, on their skills and overall performance. The basic assumption was that when skills are needed, the planning for investing in the training was done strategically, and the influence of training on employee skills would be witnessed. Then, the firm’s performance would respond positively to the employees’ skill status.

Among the identified skills, interpersonal/communication skills, technical skills and work ethic, only technical skills fully mediated the training effect on firm performance. The respective skills bootstrapped indirect effects. From the SEM results, the interaction variable of training and training needs revealed a significant 10% average negative relationship between technical skills and the firm’s training needs. However, the respective skills exhibited a significant positive relationship to the firm’s performance. Moreover, from the index of moderated mediation, it is concluded that the model is correctly specified, which means that indirect effects are linearly related to training needs. The index suggests that a unit decrease in training needs causes an increase in the indirect effects of training on the firm’s performance through technical skills. The results can be seen in the interaction plot, which shows that when training needs are low, technical skills respond more positively to training intensity.

The results have a number of implications. First, the effect of training through a process analysis, understanding how and when the effect is realized, must be established. It is one thing to determine whether the training affected learning and the firm’s performance, yet it is another to communicate how the effect was transferred and under what circumstances. The delayed effect of training was not given special attention in this paper due to the nature of the data used, which was cross-sectional in nature (Konings and Vanormelingen, 2010; Wooldridge, 2002). While other effects, like technical skills, are immediate due to their nature, soft skills might take longer to affect firm performance as they depend on employees' attitudes (Thang et al., 2010; Tharenou et al., 2007). In this case, a panel dataset could be more appropriate. Nonetheless, the sustainability of training and knowledge provision, as suggested by Birou et al. (2019), is important to accommodate the delayed response as well as continuity and strengthening of the transferred skills for sustainable outcomes.

The second implication is linked to the role of training needs in training implementation. Although the results showed that training was more effective when training needs were low, that did not mean that the firms did not have training needs. Instead, it simply means two things: first, there is a threshold point for training needs below which the training will work. That means when training needs are too severe, the employees are desperate for skills and they can question their essential human capital even to digest what will be obtained from training if attended. The fundamental capacity of employees plays a more significant role in determining whether training will work, even if the proper topics are selected. However, the conceptual model of this paper did not incorporate an education variable for the trained employees and link it to the training needs observed, as it would have made the model particularly complex. Still, that could have been an exciting addition to the analysis to determine the nature of the employees trained and relate their quality to the training needs. Hence, as an area for further research, there is value in determining if there is a correlation between training needs and employee capacity, using the education variable as a proxy, thereby concluding why the training did not have positive effects on skills when training needs were so high. In addition, this paper conceptualized only the perceived most important skills as mediators; other researchers can consider other skills to observe the mediation role in training effects.

Overall, various human capital stakeholders in firms play significant roles in ensuring that training adds value to the firm. These include the planners who perform needs assessments and which skills development programmes should be prioritized. Given the limited resources and firm goals, decision-makers who decide which training programmes have priority and the employees themselves should understand that training supplements what they already have. So they should make extra efforts to equip themselves with more knowledge. As work changes, firm-provided training may become more relevant for sound economic and social outcomes (Martins, 2021).

Figures

Figure 1

Theoretical model

Figure 2

Conceptual process model for training effectiveness – role of training needs

Figure 3

Statistical model

Figure 4

Firm performance for the trained and untrained groups

Figure 5

Firm output increment within the period of the training (in USD)

Figure 6

Interaction plot

Table 1

Variable measurements

Variable	Symbol	Measure
Dependent
Firm performance	Y	Deflated last fiscal year sales values
Independent
Training	X	(%) of employees who received training for the past two years by the time the survey was conducted
Moderating
Firm training needs	W	Composite score from the dummy five responses of questions: any difficulties on - maintaining production levels according to demand - maintaining consistency quality - implementing new technologies or improving production processes - developing new products and - effective sales and marketing
Mediating
Employees skills	M	The rate was in three-Likert scale: below required (1), as required (2) and above required (3) - Three perceived most important skills (interpersonal communication skills, technical skills and work ethics) were used as parallel mediators
Control
Firm size	SizeDummy	Dummy: ≤ 20 employees (1), >20 employees (0)
Firm sector	SectorDummy	Dummy: Manufacturing (1), Non-manufacturing (0)

Source(s): Created by authors

Table 2

Percentile table for skill level and training needs: trained and untrained groups

	Trained = Yes				Trained = No
Percentiles	M1 (Interpersonal)	M2(Technical)	M3(Work ethic)	Training needs	Interpersonal	Technical	Work ethic	Training needs
1%	1	1	1	0	1	1	1	0
5%	1	1	1	0	1	1	2	0
10%	2	1	1	0	2	1	2	0
25%	2	2	2	0	2	2	2	0
50%	2	2	2	0	2	2	2	0
75%	2	2	2	2	2	2	2	2
90%	3	3	3	4	2	2	2	4
95%	3	3	3	5	2	3	3	5
99%	3	3	3	5	3	3	3	5

Note(s): On the skills level: 1 = below required, 2 = as required, 3 = above required and Training needs is a composite score, 0–5

Source(s): Created by authors

Table 3

Mediating effect – statistical model

		Outcome
		M1: interpersonal skills	M2: technical skills	M3: work ethics skills		Y: firm performance
X: Training	a1	0.16***	0.16***	0.02	c'	0.02
W: Train needs	a2	−0.07	0.02	0.02
XW: Training X Train needs	a3	−0.06	−0.1**	−0.05
M1 (Interpersonal skills)					b1	−0.03
M2 (Technical skills)					b2	0.14***
M3: (Work ethics skills)					b3	−0.03
Sizedummy					Σ	−0.24***
Sectordummy					Φ	−0.08
		95% bootstrap CIa
X: Training		0.07, 0.25	0.05, 0.26	−0.1, 0.15		−0.08, 0.13
W: Train needs		−0.16, 0.02	−0.07, 0.11	−0.07, 0.11
XW: Training X Train needs		−0.12, 0.00	−0.21,-0.01	−0.17, 0.07
M1 (Interpersonal skills)						−0.13, 0.06
M2 (Technical skills)						0.04, 0.24
M3: (Workethics skills)						−0.13, 0.07
Size dummy						−0.36,-0.15
Sector dummy						−0.19, 0.01

Note(s): Path coefficients (***p < 0.01; **0.01 < p < 0.05) and ^aPercentile bootstrap CI based on 10,000 bootstrap samples

Source(s): Created by authors

Notes

1. Food processing (ISIC15), textile and garments (ISIC 17 and 18), fabricated metal products (ISIC 28), furniture (ISIC 36), construction (ISIC 45), hotel and restaurant (ISIC 55), transport (ISIC 60 and 61) and information technology (ISIC 72).

2. Small (less than 20 employees), Medium (20–99 employees) and Large (more than 100 employees)