The success of ET is defined and measured at both the individual and organisational levels through the resulting entrepreneurial competencies, entrepreneurial attitude and non-cognitive skills, business management skills, innovation, business start-up and growth. These ET outcomes are collectively determined as EHC, according to human capital theory (Becker 1962; Mincer 1958) and more specifically recent developments in EHC literature (Martin et al. 2013; Unger et al. 2011). Valerio et al. (2014:ix) report that EET programmes have proliferated. This ET trend has been developing since the mid-1980s because of the importance of promoting entrepreneurship as a propeller of economic growth and development (Ács, Desai & Klapper 2008:265; Botha et al. 2015:55; Moos 2015:26; Shane 2009:147). This study’s review of 17 South African ET programmes has shown that the majority (53%) have entrepreneurship content, whereafter some have management content (29%) and the remaining 18% have an even combination of entrepreneurship and management. This is consistent with the literature because a balance of entrepreneurship and management content is considered necessary for ET efficacy (Botha et al. 2015:63–64; Morris et al. 2013:352).

Furthermore, personal entrepreneurial skills training focused on entrepreneurial behaviours, and the entrepreneurial mindset, such as innovation, identifying and exploiting new opportunities and goal setting, results in significant improvements in the business performance of sales (17%), profits (30%) and innovation (offering new products and services) compared with the control group (Campos et al. 2017:1289–1290). A further study results again showed that personal initiative training yielded a statistically significant increase in sales and profits and the adoption of recommended business practices and innovation (Ubfal et al. 2019:18–19). Brooks, Donovan and Johnson (2018:197) found an average increase of 20% in weekly profits for mentees of an entrepreneur mentorship programme compared with ‘muted results’ for those receiving only business training programme. It may be concluded that ET has value, and that ET efficacy depends on developing entrepreneurial competencies, business management skills and personal entrepreneurial characteristics.

The study was conducted in South Africa, and the country’s economy has not performed well in recent years. This is evident by the country’s gross domestic product (GDP) growth (measured by production) of less than 2% per annum between 2015 and 2020, although it was higher in 2021 at 4.9% (World Bank, GDP growth [annual %] Data 2022). This low growth is compared with the preceding decade (2005–2015), in Figure 1. Furthermore, according to Herrington, Kew and Mwanga (2017:6–7), entrepreneurship in the country is dismal compared with sub-Saharan Africa and equivalent developing economies elsewhere.

Seminal research views the economic peak as the onset of economic recession, which may be understood in the business life cycle theory (Moore 1967:16; Mueller 1972:200–201). In this context, the trend line of Figure 1 is extremely worrisome because from 2006 to the time of this document’s publication, South Africa’s GDP growth trend has been downward. During the same time, South Africa experienced a persistent decrease in total early-stage entrepreneurship (TEA) from 2002 to 2016, and in 2016 was ranked 46th out of 65 economies for its TEA rate (Herrington et al. 2017:6–7). From the preceding context and rationale, effective ET is expected to lessen the dire situation of low business start-up rates and high numbers of existing enterprises closing in South Africa, which will contribute to increased economic output. The following subsection introduces the ET literature and problems currently experienced in the field.

The characteristics of ET must be well demarcated to achieve the study’s research objectives. Entrepreneurship training is intended for entrepreneurial performance outcomes and status in starting a business or achieving business growth, higher profits and wealth (Ács et al. 2018:1; Valerio et al. 2014:3). Therefore, ET for entrepreneurial performance has more to do with business growth, compared with day-to-day management (Botha et al. 2015:63–64; Morris et al. 2013:352). In comparison, EE programmes focus on building knowledge and skills about entrepreneurship according to Valerio et al. (2014:2), who build on Hynes’ (1996) seminal study for demarcating ET from EE. Entrepreneurship training generally lasts between 5 and 15 full days, but not necessarily continuous days (McKenzie & Woodruff 2013:54). In comparison, EE is longer, perhaps a university or school term or even semester- and year-long programmes. Often, ET consists of workshops, discussion groups, role modelling, case studies and practical workshop materials compared with the EE lecture style model. In addition, professionals often facilitate ET, many of whom are entrepreneurs as opposed to academics who present EE.

Kirkpatrick’s seminal work argues that training design must match the trainees’ requirements and a needs analysis must be conducted based on the individual, task requirements and organisation (Kirkpatrick & Kirkpatrick 2009:3–4). Furthermore, the transfer of training from the learning context to its application must take place, which is defined by its design and delivery to give trainees the ability to apply their learning to their jobs (Baldwin & Ford 1988:86–88; Grohmann, Beller & Kauffeld 2014:85–86). Real-world examples and practice-oriented tasks during training increase their practical significance and have been shown to positively influence their transfer (Baldwin & Ford 1988:86–87). Training design must consider potential barriers to transfer by including reflection exercises on possible solutions to such barriers (Grohmann et al. 2014:99).

Training design must also consider individual trainee characteristics, training context and facilitation style (Alvarez et al. 2004:389; Kirkpatrick & Kirkpatrick 2009:3–4). For example, in the context of teaching mathematics, teachers modify task difficulty to the level of their students’ understanding, interest and experience (Son & Kim 2015:7–8). In the ET context, training design must meet the needs of the participant entrepreneurs. For example, training content must develop the competence required in the entrepreneurial context of starting and growing enterprises (Botha et al. 2015:63–64; Morris et al. 2013:352).

Furthermore, content and process decisions (such as the order of the curricula content) influence instruction quality and, therefore, influence learner outcomes. Entrepreneurship training must consider the pedagogy approach, which may affect learner outcomes (Täks et al. 2014:574). For example, engineering students studying entrepreneurship were overwhelmed by collaborative learning (team tasks). Confusion also resulted from the entrepreneurial task of being solution-driven (seeking answers of their own) instead of replicating materials that the teacher provided (Täks et al. 2014:574). Active versus passive learning techniques distinguish learning by doing, rather than just gaining knowledge by reading; this is an important element of entrepreneurial learning (Wang & Chugh 2014:3–4).

From the aforementioned, according to Alvarez et al. (2004:389), training efficacy is the study of organisational, individual and training programme variables that collectively influence training outcomes before, during and after training ensues. These variables may influence training outcomes negatively or positively; furthermore, ‘training effectiveness is a theoretical approach for understanding [training] outcomes’ (Alvarez et al. 2004:387). Tonhäuser and Büker (2016:129) similarly identified training effectiveness as a question of which determinants, as a process, are responsible for successful training, and the authors refer to ‘organisational, learning field and individual level’ determinants. General vocational training efficacy literature provided this study with well-established concepts, constructs and existing measurement scales of the variable (determinant), namely training design.

In addition, self-efficacy theory is suited for this study, as the theory argues that individuals are responsible for their own training outcomes (Bandura 1977). This may be applied to the ET setting, meaning that facilitators and participants of ET determine training outcomes. Bandura’s (1977) seminal work established that an individual’s belief in their ability influences events and outcomes in their life and career. Self-efficacy is defined as the belief in or expectation of one’s ability to successfully execute whatever behaviours are required to develop the specific outcomes one desires (Bandura 1977:193).

In more recent vocational training contexts, participant self-efficacy argues that trainees are responsible for their learning. Participant self-efficacy is relevant in the ET context because entrepreneurs with a higher need for achievement and who believe they can implement what they have learnt from ET will most likely perform better than those who do not (Holton, Bates & Ruona 2000:334–335; Salas et al. 2012:79). Entrepreneurial self-efficacy in the context of ET (Chen et al. 1998:296; Gielnik et al. 2017:336) provided the theoretical underpinning for the variable (determinant), ESE, at the individual level. Entrepreneurial self-efficacy is important because it reportedly influences ET outcomes, according to Gielnik et al. (2017:336) and Chen et al. (1998:296).

Human Capital Theory (HCT) is another important theoretical foundation for this study. The theory stems from seminal literature, which predicts an individual’s economic output will increase because of investment in their training (Becker 1962; Martin et al. 2013:212–213; Mincer 1958). More recently, HCT has been studied in the entrepreneurship context. Studies have found that high entrepreneurial task-related human capital contributes to entrepreneurial success through firm’s growth (Marvel, Davis & Sproul 2016:599; Unger et al. 2011:344), and the concept of EHC has been developed. Therefore, HCT is important in the context of this study to structure an understanding of the relationships between ET and EHC formation. From the discussion expounded earlier, the concepts, constructs, theories and existing measurement scales from earlier research have been used in this study. Entrepreneurship human capital, which the study postulates as the expected outcome of ET, is briefly introduced next.

In the context of this study, EHC comprises entrepreneurial competencies, entrepreneurial non-cognitive skills and business management skills. Entrepreneurial competencies are measured through the knowledge, skills and behavioural and attitudinal ability required for starting and growing a business (Kyndt & Baert 2015:4–7; Le Deist & Winterton 2005:39–40; Mungule & Van Vuuren 2016:5). Business management skills are measured through the management functions of leadership, planning, organising and controlling. In contrast, management skills typically relate to areas such as finances, sales and marketing involved in day-to-day management operations (Botha et al. 2015:63–64; Morris et al. 2013:352).

In addition to EHC in the form of cognitive skills, the concept of entrepreneurial non-cognitive skills refers specifically to ESE in the study. This is based on seminal research by Chen et al. (1998:296), who found that entrepreneurs have higher levels of self-efficacy in areas of innovation and risk-taking compared with non-entrepreneurs, such as managers. Newman et al. (2019:404–405) define ESE as the belief that one can successfully perform entrepreneurial tasks with outcomes such as business start-up and growth. Furthermore, ESE links with the theory of planned behaviour based on Ajzen’s (1991) work, namely that an individual’s belief in their ability increases their planning to start and grow a business (Krueger & Brazeal 1994:94). Therefore, ESE is positioned as an IVV in the study and is expected to mediate or moderate the impact of ET on EHC.

Aboobaker’s (2020:79) recent study finds a positive impact of ET on human capital development among university students but laments that few studies identify how ESE mediate such an impact (Aboobaker 2020:77). However, positive transfer climate is described as favourable policies and practices that ensure the transfer of training takes place (Baldwin & Ford 1988:86–88; Grohmann et al. 2014:85–86; Srimannarayana 2016:271) has been found to mediate the transfer of training (Schindler & Burkholder 2016:293). The literature on the mediating and moderating effect of ESE remains scarce and this creates the research gap that this study focuses on.

Self-efficacy theory posits that the belief in one’s ability will influence achieving the outcomes one desires (Bandura 1977:193). Thus, ET efficacy relies not only on design and delivery but also on who is receiving the training. Therefore, participant self-efficacy is relevant in the ET context because entrepreneurs with a higher level of ESE will be more positive so that they can, and will, learn from ET. On this basis, they are likely to outperform participants with poor ESE and then achieve higher levels of EHC after training (Holton et al. 2000:334–335; Salas et al. 2012:79). However, the opposite may be true when individuals with higher levels of self-efficacy underestimate the entrepreneurial task requirement and subsequently underperform because they lacked in preparation (Gielnik, Bledow & Stark 2020:496).

A majority (six) of the study’s hypotheses (see Section Methodology) are based on the expectation expounded earlier regarding ESE. Should the empirical study show support for the hypotheses, then the findings will contribute to the theory of self-efficacy in the context of ET. However, this is not unique as other studies have done so before (Miao, Qian & Ma 2017:89; Newman et al. 2019:404; Salas et al. 2012:84).

However, another aim of this research is to test ESE as the moderator and/or mediator of the effect that ESE has on the relationship between ET and EHC. This is a unique contribution of this study to the theory and practice of ET.

The study’s research hypotheses, listed as follows, were based on the research objectives stated earlier and were supported by the literature consulted and the study’s theoretical foundations:

The key constructs are operationalised as follows:

This study’s sample was an ET project in the Eastern Cape Province of South Africa. This project provided ET for potential entrepreneurs to develop their entrepreneurial competencies and business management skills. The sample was located through a collaboration with the ET project’s organisers, who allowed onsite data collection. In addition, consent forms were signed by all the trainees from whom data were collected. The researcher aimed for a study sample size of 440 potential entrepreneurs. However, it must be noticed that at the time of writing, the ongoing coronavirus disease 2019 (COVID-19) pandemic impacted the final study sample size. A total of 234 respondents were surveyed, using the intercept survey approach at the study site, which produced a net sample of 184 usable pre- and post-test surveys. This resulted in a related sample comprising 184 pre and 184 post participants, with a 78.63% usable survey response rate during the COVID-19 pandemic.

The study used empirical research by employing a quasi-experimental research design that differs from true experimental research when it does not have a control group (Cohen, Manion & Morrison 2007:282–283), as occurred with this research. The study followed the one-group pre-post-test experimental design. According to Cohen et al. (2007:282), this may be presented as O₁ → X →O₂, where O₁ is the pre-test baseline measurement (self-reporting survey), X is the treatment (ET) and O₂ is the post-test measurement (self-reporting survey). Experimental research is the most authoritative type of research design to show that casual relationships exist between variables (Cohen et al. 2007:265; Cooper & Schindler 2014:192–193). Experimental research has at least two cross-sectional measurements, meaning that time is required to elapse between the measurements (Blumberg, Cooper & Schindler 2011:197). This was a suitable design for the study to test the impact of ET on the experimental group before and after they participated in the ET programme.

Notably, a study’s unit of analysis and the research respondents (whom data are obtained from) do not need to be the same (Blumberg et al. 2011:166). This applies to this study, as the research respondents are the ET participants in the experimental group, but the unit of analysis is the ET programme itself. This decision of the unit of analysis is based on the argument that the ET cannot be successful unless it meets the needs of those being trained (Kirkpatrick & Kirkpatrick 2009:3–4).

The causal research hypotheses of this study will be accepted or not based on probability, which is determined by statistical significance (Christensen, Johnson & Turner 2014:180; Urdan 2011:62). Statistically significant relationships indicate that a real relationship exists between the IV and DV, and the relationship cannot be because of random factors. In this study, the IV was training design, and ESE was the IVV, while the DV was EHC, which comprised entrepreneurial competencies and business management skills. Before the inferential statistics, the sample descriptives are first presented.

Table 1 refers that the study sample comprised more females (49%) than males (37%), although 14% opted not to answer this question. Most participants were youth with an average sample age of 27 years, with 72.3% of the sample aged between 18 and 35 years. Furthermore, the majority (66%) completed at least Grade 12, while 6% achieved a diploma or university degree, but 16% did not answer this question. The majority (74%) were unemployed. This ET project was therefore founded on youth entrepreneurship to alleviate unemployment and poverty.

Table 2 illustrates the inferential statistical procedures and purposes, as well as data and sample types used. It provides a brief explanations of the various inferential tests for the study.

As illustrated in Table 2, in this study, the experimental group sample remained the same during the pre- and post-test and was, therefore, a related sample (Cohen et al. 2007:586–587). However, there were males and females and other demographic factors used to separate the sample into independent groups. The individual inferential test procedures that were considered for the study are now expounded upon.

The direct effects illustrated in Table 3 were represented by Hypotheses 1–5. It must be emphasised that simple linear regression first tested these relationships, individually and independent of any interaction, to illustrate how the three main constructs act directly on one another during training. The linear regression tests were necessary to be conducted first, to answer the first study objective, and thereby lay the basis to proceed to answer the second study objective.

Firstly, Table 3 indicates that the related-sample Wilcoxon signed-ranked test shows that entrepreneurial competencies’ median increased by 0.679, a statistical (p < 0.001) and practically significant improvement (d = 0.52). Together with the regression result (r = 0.471; R² = 0.221; F-statistic = 51.184), it can be inferred that this ET programme was well designed and catered to the participants’ needs. This ET’s success largely resulted from training design, with the unstandardised coefficient β = 0.574 as a predictor of entrepreneurial competencies. Thus, it is concluded that every one-point increase in the standard deviation of training design increased entrepreneurial competencies by approximately 58%.

Secondly, Table 3 further indicates that the related-sample Wilcoxon signed-ranked test shows business management skills’ median increased by 0.727, which was statistically (p < 0.001) and practically significant (d = 0.53). The regression (r = 0.400; R² = 0.160; F-statistic = 34.017) showed that the training design was successful in developing participants’ business management skills (unstandardised coefficient β = 0.563). Therefore, it is concluded that every one-point increase in the standard deviation of training design increased business management skills by approximately 56%.

Thirdly, Table 3 shows that the related-sample Wilcoxon signed-ranked test shows ESE’s median increased by 0.375, which again was statistically (p < 0.001) and practically significant (d = 0.41). However, the regression showed a weak positive impact (r = 0.269; R^{2 =} 0.072; F-statistic = 14.129), indicating that training design was less successful in developing ESE than developing entrepreneurial competencies and business management skills. With an unstandardised coefficient β = 0.313, it is concluded that every one-point increase in the standard deviation of training design resulted in ESE increasing by approximately 31% – only half of the prediction magnitude of training design on entrepreneurial competencies and business management skills. This result alludes to the independent personal nature of ESE, which is rooted in the theory of self-efficacy (Bandura 1977) and within the non-cognitive domain (Salas et al. 2012:84).

Fourthly, Table 3 highlights that the regression showed a strong and statistically significant positive impact (r = 0.613; R² = 0.375; p < 0.001; F-statistic = 103.875) of ESE on entrepreneurship competencies. With an unstandardised coefficient β = 0.609 and a medium-sized practical significance (d = 0.52), it was surprising to find that the impact of ESE on entrepreneurial competencies was much larger than that of training design on entrepreneurial competencies (r = 0.471; R² = 0.221). This is a unique insight as testing the two constructs within a single study for their independent impact on EHC is unprecedented. It is concluded that ESE was essential for improving EHC during ET in the context of this study – more so than training design as a one-point increase in the standard deviation of ESE increased entrepreneurial competencies by approximately 61%.

Lastly, Table 3 highlights that the regression showed a moderate and statistically significant positive impact (r = 0.552; R² = 0.305; p < 0.001; F-statistic = 78.008) of ESE on business management skills. With an unstandardised coefficient β = 0.675 and medium practical significance (d = 0.53), it must again be observed that the impact of ESE on business management skills was much larger than that of training design on business management skills (r = 0.400; R² = 0.160). This was a unique insight, as mentioned earlier. This finding again supports the aforementioned importance of ESE for improving EHC during ET. Therefore, it is concluded that with a one-point increase in the standard deviation of ESE, business management skills increased by approximately 68%. Based on Table 3, and the earlier discussion. Thus, Hypotheses 1–5 were accepted.

Tonhäuser and Büker (2016) emphasise that ET variables act independently and as a system to contribute to the success or failure of ET outcomes. This study pursued this line of enquiry and found independent relationships between ET and EHC variables. Following this, the study further investigated interactive moderating and/or mediating effects through Hypotheses 6–9. See Table 4, for more in-depth information.

Firstly, Table 4 illustrates that the moderation interaction expected in the first interaction effects was insignificant, with zero change seen in the R² value (p = 0.700). Therefore, it is concluded that ESE did not moderate the relationship between training design and entrepreneurial competencies. It must be noticed that the moderation model – when compared with the simple regression model – shows that the former statistically significant impact (r = 0.574; p = 0.000) of training design on entrepreneurship competencies became insignificant (β = 0.28; p = 0.130). At the same time, ESE became the statistically significant predictor (β = 0.58; p = 0.000). This hints at a mediation effect which is consistent with the rules of mediation effects (Hayes & Preacher 2014; Kenny 2021).

The first mediation test (H₈) results in Table 4 show that the 95% bias bootstrap confidence interval was entirely above zero: 0.398–0.772. This result confirmed that ESE was a mediator of the relationship between training design and entrepreneurial competencies. This also shows how, in the model, ESE reduced the previous effect training design had on entrepreneurial competencies (Table 4). Entrepreneurial self-efficacy (path b) became the statistically significant predictor (β = 0.585; p < 0.001) of entrepreneurial competencies. Based on Table 4 and the previous discussion, H6 is unsupported and H8 was accepted.

This finding partially explains why ET efficacy was found to be unreliable in prior studies (Martin et al. 2013; Premand et al. 2016; Rauch & Hulsink 2015), where some studies found positive results (Botha et al. 2015; Valerio et al. 2014; Van Vuuren & Botha 2010) and other studies did not (Nabi et al. 2017; Rauch & Hulsink 2015). Prior studies that have not measured participants’ ESE levels cannot account for it as an explanation of ET ineffectiveness. The opposite is also true, and the uncertainty remains constant; in other words, even when studies find successful ET outcomes, the comprehensive understanding of why the ET was successful remains incomplete.

This study provided empirical support for ESE as a root cause of ET efficacy such that higher ESE influences ET outcomes positively. Furthermore, the study proved that ESE was an individual-level quality that interacted (mediated) with an organisation-level quality (training design). Therefore, differing levels of ESE at the start of ET influence participants’ learning and progress. In the programme studied, participants registered a high average baseline score (above 75%) for ESE in the pre-test measure, showing that high ESE pre-programme influenced the ET efficacy, by interacting with the training design. This finding is unique overall and illustrates that the study and practice of ET must consider it a system, whereby individual- and organisational-level qualities act together in influencing ET efficacy.

Secondly, Table 4 illustrates that the interaction in the second moderation (H₇) test was also insignificant, with zero value evident between the lower confidence interval −0.89 and upper confidence interval 0.52. Furthermore, no interaction and zero change were seen in the R² change (p = 0.61); therefore, H₇ was unsupported. Similarly, this moderation model found that the former statistically significant impact (r = 0.563; p = 0.000) of training design on business management skills was largely reduced and became insignificant (β = 0.19; p = 0.44). Thus, ESE became the statistically significant predictor (β = 0.66; p = 0.00); again, this finding is consistent with the rules of mediation (Hayes & Preacher 2014; Kenny 2021).

Finally, H₉ was confirmed by the last row of Table 4 showing the 95% bias bootstrap confidence interval entirely above zero: 0.43–0.89. Therefore, ESE was again a mediator of the relationship between training design and business management skills. With ESE, path b became the statistically significant predictor (β = 0.66; p < 0.001) of entrepreneurial competencies. This finding is unique because former studies have not undertaken such testing of interactive effects between ET variables. Based on Table 4 and the given discussion, H7 is unsupported and H9 was accepted.

As cited, Nabi et al. (2017:278) report that too few ET studies employ true experimental research designs with pre- and post-test measures. The study has contributed a quasi-experiment with a pre- and post-test measure and employed multiple regression statistical methods to study the interactive effects of ET variables. The use of novel regression statistical procedures, in the context of ET research, furthered the understanding of how ET variables act as a system to increase ET efficacy. Thus, future studies are recommended to continue designing studies that employ multiple regression statistics and pursue such powerful statistical analysis. Additionally, the study has made a pioneering contribution.

This study has made an empirical contribution to how interactive effects of ET qualities at the organisational, learning and individual levels act as a system to bring about training success. This contribution enhances the understanding of ET efficacy. This has been argued as necessary to contribute to vocational training theory overall and specifically to why so many training programmes have poor results (Alvarez et al. 2004; Tonhäuser & Büker 2016; Valerio et al. 2014). This adds to the existing vocational training and more specifically ET literature. Future studies are recommended to consider such interactions and regard ET as a system.

This study has provided empirical support for ET as a tool to develop entrepreneurs and entrepreneurship to create jobs and grow an economy. The researcher recommends that policymakers continue to shape entrepreneurship policy frameworks that include ET. It is hoped that the insights developed in this study will assist policymakers in gaining more confidence in the value of ET, the direction for ET and the required resources and funding for ET success.

The study emphasised that a needs analysis is the foundation and ‘gold’ standard of ET design. Such a needs analysis may include interviewing participants before training, as performed by Van Vuuren and Botha (2010). Therefore, it is recommended that future ET design ensures a fit between ET content and delivery, participant entrepreneurs and their industry context, which will ensure that ET is not overly generic but caters to participants’ specific needs. The design of ET requires expertise and focuses across several qualities. The custodians of this design process must include qualified practitioners who have the required academic qualifications, experience starting and growing a business, and facilitator efficacy in ability, confidence and motivation of the facilitator’s performance.

Finally, the study contributes clear empirical support that personal attributes, specifically ESE, have a huge impact on ET efficacy. This extends the implication of Bandura’s (1977) seminal work on self-efficacy theory into the realm of ET and EHC outcomes. This contribution builds on the work of Salas et al. (2012:84), Miao et al. (2017:89) and Newman et al. (2019:404) and cannot be claimed as a unique contribution. However, a pioneering contribution is made through the finding that ESE mediates the impact of ET on EHC outcomes. This bluntly indicates that no matter how well designed and delivered ET is, without high levels of participant ESE, ET is likely to be ineffective. Further research is called for in support of developing this intriguing direction for the study of ET, ESE and EHC outcomes in the context of self-efficacy theory.