Analysis of the impacts of entrepreneurship training on growth performance of firms: Quasi-experimental evidence from Nigeria

Uchenna Efobi (College of Business and Social Sciences, Covenant University, Ota, Nigeria)
Emmanuel Orkoh (North-West University, Potchefstroom, South Africa)

Journal of Entrepreneurship in Emerging Economies

ISSN: 2053-4604

Publication date: 3 September 2018



Using quasi-experimental designs, the purpose of this paper is to study the effects of training entrepreneurs and such entrepreneurs going ahead to retrain its workers on the business high-growth performance.


This paper used a unique evaluation data from the National Business Plan Competition in Nigeria, organized by the Nigerian government in collaboration with the World Bank. The data was analyzed using the Propensity Score Matching technique and complemented with the Difference-in-Difference estimates.


The authors find from the estimation of this paper that those entrepreneurs who received standard evaluation training and goes ahead to retrain its workers experienced an expansion in the number of employees by two persons, an increase in innovation index by about 3 units. An increase in revenue is also observed, but this increase was not significant at the 1, 5 or 10 per cent levels.


This paper presents an interesting view point on how training within an entrepreneurial venture should be viewed as a ‘two sided coin’. This is such that training the entrepreneur is one side of the story, and the entrepreneur retraining its workers is another important side of the story.



Efobi, U. and Orkoh, E. (2018), "Analysis of the impacts of entrepreneurship training on growth performance of firms", Journal of Entrepreneurship in Emerging Economies, Vol. 10 No. 3, pp. 524-542.

Download as .RIS



Emerald Publishing Limited

Copyright © 2018, Emerald Publishing Limited

1. Introduction

Small businesses are vital in many developing economies for job creation, value chain addition and industrial growth. They can equally be strong sources of technological innovation for development (Coad and Tamvada, 2008; Michelitsch et al., 2011). Also, the traditional Industrial Organization literature suggests that new entrepreneurial ventures enhance market contestability, which is an important source of competition to spur growth within the economy (Tetteh and Essegbey, 2014). Despite the relevance of small businesses to economic and industrial growth, they face a number of challenges that affect their operational capacity and growth. Among such constraints are lack of technical knowledge, innovation, poor access to capital and market access. Some of these challenges translate into common characteristics that usually define their sizes and growth. For instance, the majority of these small businesses never expand beyond only the owner and a few employees (Nichter and Goldmark, 2009; McKenzie and Woodruff, 2015), with a possible size of 10 workers (Hsieh and Olken, 2014; McKenzie, 2017).

Noting the enormous contributions of small businesses to developing countries and the challenges that confront them, an important question, therefore, is whether policies that are directed at improving the capacity of both the owners (i.e. entrepreneurs) and employees of the small businesses will significantly help to overcome these constraints (at least to an extent) and result in high-growth outcomes. This inquiry is relevant considering that entrepreneurs in developing countries are seriously lagging behind in innovativeness and the extent of reforms that they bring into their business processes (Santarelli and Vivarelli, 2007). Most new small businesses in developing countries are founded as a last resort (Beck et al., 2005) and may not be based on a firm conviction that is tied to the expertise and know-how of the entrepreneurs in the particular sector of interest. For example, there are instances in developing countries where entrepreneurs with low competency engage in multiple businesses (in different sectors) to increase their income flow. As a result, only a few of newly established small businesses in developing countries succeed and are able to weather the harsh business environment that confronts their operations. About a third of newly formed businesses survive beyond two years, and about 90 per cent of those surviving will not grow at all or will be left with the same number of employees as when they started (Olafsen and Cook, 2016). In Nigeria, the statistics is not much different; the available evidence suggest that about 65 per cent of small businesses fail within three years of existence because of lack of technical experience and knowledge, among others (Central Bank of Nigeria-CBN, 2003; Obi, 2013). Therefore, providing empirical evidence on some of the factors that can help to improve the human capacity of both the owners of small businesses and their employees to achieve long-term efficiency and expansion will be relevant for industrial growth policy formulation.

This paper, therefore, contributes to the literature by advancing the findings of McKenzie (2017), who examined the impact of random assignment of grants on the entry, survival, profits and sales and employment growth of the beneficiary firms using the same data that we used for our analysis. Unlike McKenzie (2017) who focused on the impact of monetary incentives in propelling growth of the beneficiary firms, we emphasize the impact of a complete cycle of knowledge transmission on the growth of the firms in question. We achieve this objective by dividing our sample into two distinct groups (experimental or treatment group and comparison group) of firms and assess the impact of this complete cycle of knowledge transmission on their growth performance outcomes. The experimental group comprises firms whose owners received external training and initiated an in-house training system for their workers, while the comparison group comprises firms whose owners did not have an in-house training system for their employees but were part of the external training program. In the section on the analytical framework (Section 2), we explain in details the channels through which this complete cycle of knowledge transmission is essential for growth of firms.

Our research uses a comprehensive evaluation data from the National Business Plan Competition in Nigeria, organized by the Nigerian government in collaboration with the World Bank. The data contain a baseline survey for 2011 and a subsequent annual follow-up surveys for three years to enable adequate tracking of the entrepreneurs and their businesses. The main aspect of the survey that was of interest to this study are the information on the entrepreneurs’ participation in the business evaluation training in the base year, and information on the entrepreneurs’ action in retraining their workers over the years. Other important information from the survey includes those that measure our main outcome variables (business performance, innovation and the growth of the size of the employees of the business).

This study, therefore, is an ex-post inquiry from previous evaluation and uses a quasi-experimental design where the decision by an entrepreneur who was previously trained by the evaluation team to organize in-house training (or retrain) for its employees is based on the choice of the entrepreneur and not any specific experimental program. As a result, a propensity score matching technique is applied to net out the impact of such choice on the entrepreneur’s business. The counterfactual is estimated from the comparison group as earlier defined. The results from the analysis show that firms that are in the “treatment” group are significantly innovative and able to grow their employees than those in the comparison group. This result is observed only after three years of the entrepreneurs’ consistent implementation of in-house training programs for their workers.

The rest of the paper is divided into five sections. Section 2 presents a brief overview of the literature and analytical framework, while Section 3 discusses the research method that includes information about the survey, the data and the estimation technique. Section 4 presents the econometric results, followed by Section 5, which concludes the paper with some directions for policy, and future research.

2. Overview of the literature and analytical framework

The literature on training and entrepreneurship growth and development can be classified into different domains. For instance, one aspect of the literature looks at the impact of entrepreneurship training on performance and growth indicators of firms (Njoroge and Gathungu, 2013; Tambwe, 2015). While some studies pay particular attention to retraining of firm employees and how this impacts the growth and survival of the firm, others examine the channel through which entrepreneurs transfer their innate attributes to their employees (Cardon, 2008; Li, Zhang, and Yang, 2017). Yet some others have emphasized the need for a shift from entrepreneurship training interventions that provide general business skills to a more focused training on subsistence and growth-oriented entrepreneurship (Titley and Anderson-Macdonald, 2015).

Despite this on-going discourse, the available literature provides pieces of evidence that are relevant for policy and further research. In their quasi-experimental study on the possibility of teaching entrepreneurial activity, Klinger and Schündeln (2011) found that receiving business training could significantly increase the probability of starting a business or expanding an already existing business. This finding corroborates evidence found in other studies that self-employment assistance programs (training) are viable policy tools to promote rapid reemployment of unemployed workers (Benus, 1995; Kosanovich et al., 2001; Dvoulety and Lukes, 2016). Blackburn (1990) studied training of workforce of small businesses in England and found a positive impact from such training on the performance outcomes of small businesses. In Romania, Rodríguez-Planas and Jacob (2010) found that a firm that engages in training and retraining contributes to improvement in its economic outcomes such as higher employment prospects.

Some empirical evidence from African countries such as Kenya and Tanzania suggests that entrepreneurship training is crucial for successful performance and growth of Micro and Small Enterprises (Njoroge and Gathungu, 2013; Tambwe, 2015). Mano et al. (2012) examined similar issue using a randomized experiment in Ghana and found that basic-level management training improves business practices and performance. Elert et al. (2015), and Karadag (2017) also found positive effects on entrepreneurial income and firm survival from participating in entrepreneurship education and training in high school.

In yet another study, Fafchamps et al. (2014) conducted an experiment of a small business plan competition in Ghana, where winners were selected to receive individual training. The authors, however, concluded that there was no significant impact of such training on growth of firms. Such findings can be related to the assertion by Titley and Anderson-Macdonald (2015) that there is the need for entrepreneurship training programs to separate dimensions of business expertise into different training courses focused specifically on aspects such as marketing, finance or operations. In their attempt to provide insight into the missing link between benefits such as confidence derived by management of firms that participate in management training scheme and the impact of the training on the performance of the firm in general, Westhead and Storey (1996) conclude that management training may not have a’priori determined impact on firms. The authors ascribed their conclusion to factors such as lack of impact, difficulty of attributing cause and effect, poor quality of training and very short periods of training.

So far, very limited literature exists on the extent to which retraining of employees by trained entrepreneurs impact firm performance and growth. In their analysis of the efficacy of self-employment training to unemployed and other individuals interested in self-employment, Michaelides and Benus (2012) found that such training was effective in encouraging unemployed participants to start their own business, leading to significant impacts in self-employment and overall employment immediately after program entry. They further found that the program enabled unemployed participants to remain self-employed and avoid unemployment even five years after program entry. However, the authors indicate that the program was ineffective in improving the labor market outcomes of participants who were not unemployed.

We argue in this paper that capacity development in an entrepreneurs’ business is a “two-sided coin”, such that training the entrepreneur and a further action by the entrepreneur to retrain his/her workers makes the complete cycle of knowledge transmission. Training programs that contribute to the advancement of the knowledge of both business owners and workers are more effective than those that are limited to only the owners of the business. Such one sided transmission of knowledge could lead to what we term “truncation of human capital development” in the business, which is the limited impact of human development effort in the business when such efforts are not transmitted to the development of other individuals in the business, and this will consequently affect productivity and growth. We therefore contribute to the literature by emphasizing the importance of trained entrepreneurs going further to retrain their workers.

The analytical framework that underpins the argument of this paper presents two channels through which the assumed positive relationship between business productivity and cash income that comes from a complete cycle of training can be achieved. First, transfer of knowledge from trained entrepreneurs to their workers will contribute to an improved stock of skills, efficiency and innovative capacity of the business. These will contribute to high productivity and improved income of the entrepreneur’s business in the long run. Second, depending on the content of the training program, the trained entrepreneurs are expected to improve their own efficiency and leadership skill. This will directly increase their own productivity and those of their workers granted that they are able to effectively influence their workers as a result of such skill. Some of these channels have been discussed in detail in some studies such as Mason et al. (2012) and Naude (2013).

Other available pieces of evidence that support the channels through which both sides of knowledge transmission can influence the growth of entrepreneur’s businesses include increase in the speed of business development (Fagerberg et al., 2009; Audretsch and Sanders, 2011; Szirmai et al., 2011), growth in self-efficacy, passion and other business operations (Shindina et al., 2015; Riel et al., 2015).

3. Research method

3.1 The matching process

The national business plan competition in Nigeria (YouWiN!)[1], from which the survey for this study was extracted, targets individual entrepreneurs who represent their varying businesses. These entrepreneurs were trained and randomly selected into the original experimental and non-experimental groups, based on the originally defined criteria for evaluation. From the original survey, our study further categorizes the individual entrepreneurs in the experimental group into two, where those entrepreneurs who participated in the initial evaluation training and then had consistently affirmed that they had operational in-house training for their workers for a period of three years (across the survey follow-up periods) were grouped as “treatment”. On the other hand, those entrepreneurs who participated in the original evaluation training, but reported that they did not have in-house training for their workers over the period of interest were classified as the comparison group. It is important to note that the groups of entrepreneurs who were surveyed by the national business plan competition were earlier selected randomly across the different states of Nigeria. Hence, our sample is representative of Nigeria.

We use the matching technique to net out the effect of trained entrepreneurs setting up an in-house training program for their workers on the entrepreneur’s business outcome. There are some pre-conditions required for the matching technique to provide low bias and reliable evidence-based conclusion. They include:

  • the data for both the “treatment” and comparison groups should be collected using similar instruments;

  • both groups should have similar baseline characteristics so that, without the “intervention”, comparable outcomes can be expected of the two groups;

  • finally, the propensity score function should include similar explanatory variables for both groups (Heckman et al., 1997; Glazerman et al., 2003; Cook et al., 2008; Wanjala and Muradian, 2013). Considering the nature of the data collection process of the initial YouWiN! survey, the first pre-condition is already satisfied. The second and third preconditions are satisfied in our analysis, which will be subsequently discussed. Therefore, attributing the impact of entrepreneurs who participate in the evaluation training and have set-up in-house training program for their workers can be seen as the change in the outcome of interest, supposing it is measured as the difference in the outcome of entrepreneurs in the “treatment” group (Ti = 1) and those in the comparison group (Ti = 0), conditioned on the entrepreneurs’ status of having in-house training program for their workers (T).

Mathematically, the change in the outcome of interest is depicted as YiT=1 for the entrepreneurs in the “treatment” group and YiT=0 for those in the comparison group. The change in the outcome is, therefore, computed as:


Therefore, the average treatment effect will be:


To begin the discussion on the estimation strategy, it is important to point out that the entrepreneurs’ businesses is the unit of analysis for our study. We rely on the Propensity Score Matching (PSM) to identify comparable entrepreneurs from the “treatment” and comparison groups (Rosenbaum and Rubin, 1983). The PSM generates propensity scores, which it uses to match both groups of entrepreneurs based on their observed characteristics. The Logit model is used to estimate the propensity scores, where the action of trained entrepreneurs to set-up in-house training for their workers is seen as a binary outcome and regressed against the entrepreneur’s characteristics and those of the small business. Once the propensity score is derived from the logistic regression estimation, we then match the units in the “treatment” group with those in the comparison group based on their overlapping propensity scores that are within a common support area[3]. Our matching procedure was based on different algorithms[4]. Following Heckman et al. (1997), two conditions are considered in our analysis to validate the efficiency of the matching process. They include:

  • all important characteristics that explain the decision of the entrepreneur to retrain their employees are accounted for; and

  • the entrepreneurs in the “treatment” and comparison groups are similar based on the identified characteristics.

One important limitation of the PSM estimation is that it entirely depends on observable characteristics to accurately match entrepreneurs in both groups. However, there are some unobservable characteristics that can explain the entrepreneur’s action to set-up in-house training for their workers. We therefore complement the PSM with the Difference-in-Difference (DiD) technique (Gertler et al., 2011).

3.2 Applying the difference-in-difference approach

To further check the robustness of our PSM results, we apply the DiD estimation to adjust for other time-varying factors that may affect the outcome variables, as this approach eliminates further biases that are time dependent (Gertler et al., 2011). Essentially, applying the DiD approach controls for unobserved heterogeneities that may affect the outcome variables – apart from the decision of the trained entrepreneur to set-up in-house training program for his/her workers.

We used the first round of the follow-up (November 2012 and May 2013), which is the year immediately after the evaluation training and the third round of the follow-up (September 2014 and February 2015), given the need to allow for more time to observe the changes in the explained variable. The first round is classified as Year 1, while the third round is classified as Year 2. The “treatment” variable still remains the action by the trained entrepreneur to have in-house training program for his/her workers. Hence, the mathematical expression for the DiD estimation is:


The outcome variables “Y” include the business performance, innovation and the growth of the number of employees in the entrepreneur’s business. The usual observable characteristics as included in the PSM estimation are denoted as “Xi,t, while the error term is denoted as ‘εi,t.

3.3 Variables

The observable characteristics include:

  • Entrepreneur’s confidence level, which is transformed and measured as an ordered variable, where “1” represents not at all confident, “2” represents somewhat confident, “3” represents confident and “4” represents very confident for the following nine categories of inquiry – confident to come up with an idea for a new business product, estimate accurately the cost of a new business venture, estimate customer demand for a new product or service, sell a product or service to a customer, identify good employees, inspire, encourage and motivate employee, search for reliable suppliers, persuade lenders for finance and correctly value a business for sale;

  • Entrepreneur’s gender, which is a dichotomous variable, where “1” for male gender and “2” for female;

  • Number of business owned by the entrepreneur, which is a count variable capturing the number of entrepreneur’s businesses; and

  • The quality of the entrepreneurs’ involvement in the business, which is measured as the number of hours that the entrepreneur devotes to the particular business and the number of hours that the entrepreneur spends on other businesses in a typical month.

The entrepreneur’s business observable characteristics included in the analysis are:

  • the business size, which is measured as the number of customers that the business has;

  • ownership status of the business, which is measured as a dichotomous variable where “1” represents sole-proprietorship and “0” otherwise;

  • access to credit, a dichotomous variable which takes on the value “1” a for positive affirmation that the business has access to formal credit, and “0” otherwise;

  • total asset of the firm, which is another monetary measure of the business size, an estimates of the total assets owned by the firm; and

  • external environment, measured as a dichotomous variable, where “1” represents a positive affirmation that the firm has been confronted with paying bribe and “0” otherwise. The selected characteristics are informed by literature on the factors that influence internal policy decisions in small businesses (Chell, 1985; Stewart et al., 1998; McMahon, 2001; Bridge et al., 2003; Ayuso and Navarrete‐Báez, 2017).

3.4 Outcome variables

The outcome variables are measured as follows: business performance is computed as the total sales of the firm, which is measured based on the values in the Local Currency Unit. This variable was logged in the estimation models to increase its predictability. The innovation variable is computed as an index from a weighted response to the following questions:

  • whether the small business has introduced a new product;

  • improved an existing product or service;

  • introduced new business process;

  • implemented new design or packaging;

  • introduced new marketing channel;

  • new method of pricing, new approach to advertising;

  • new database and supply chain;

  • new way of organizing work;

  • new quality control standards;

  • engaged in outsourcing;

  • licensed a new technology; and

  • obtained a new quality accreditation. Each of these indicators was given equal weights of 1/12, such that the aggregate value of innovation ranges from 0 (low innovation) to 1 (high innovation). The last outcome variable is the job creation capacity of the business, which is measured as the number of new jobs (new employment) that the business creates in the current year.

There are three main motivations for the choice of our outcome variables. They include: first, our measure considers the different dimensions of the entrepreneurs business that show its capacity for a continuing existence. Second, the ability of an entrepreneur to be profitable in business and to be able to grow and hire more workers is a fundamental indicator of sustainable business development and industrialization (Schoar, 2010). Third, some of these measures are favored in recent empirical literature that considers high-growth business potentials (Mason et al., 2012; McKenzie, 2017).

3.5 Description of survey data and descriptive statistics

The original survey (Nigeria Youth Entrepreneurship Survey) contains a baseline survey in 2011. Three annual follow-up surveys were gathered, with the first round in November 2012 and May 2013, second follow-up in October 2013 and February 2014 and the third follow-up in September 2014 and February 2015. The evaluation training was conducted in the baseline year, and consequently, the entrepreneurs self-reports their actions to have an in-house training program for their employees across the follow-up surveys. We used the cross-sectional data from the third follow-up survey in September 2014 and February 2015 for the matching procedure[2].

The survey contains individual, household and extensive firm (business) level data. For the firm level data, there are very detailed information about the inputs and outputs, human resource and other additional information that are relevant for our analysis. From the survey, we focused on only entrepreneurs who own an operational business and report whether their businesses operate training programs for their employees or otherwise. Hence, the final sample for our study is made up of only 133 entrepreneurs in the “treatment” group and 1,468 entrepreneurs in the comparison group.

The descriptive statistics of the characteristics of the entrepreneurs and their businesses computed from the third wave of the survey, which is our main data set for the matching, are reported in Table I. It is evident from the table that most of the entrepreneurs in the survey were male, representing over 80 per cent for both the entire sample and the sub-groups. The entrepreneurs – the entire and the sub-groups – own only one business. Comparing the entrepreneurs in the “treatment” and those in the comparison group, there is a significant difference in the number of businesses owned by the entrepreneurs across the two groups. Likewise, the number of hours that entrepreneurs invest in their businesses, and their confidence level significantly differs across the two groups. The entrepreneurs in the “treatment” group put in more hours in running their business, and they are more confident than their counterpart in the comparison group.

With regards to the entrepreneurs’ business characteristics that are reported in Table I, it is evident that the differences in the number of customers of the businesses of entrepreneurs in both groups are not significant. This is also applicable to the number of hours that the entrepreneur spent on other businesses apart from the primary business. However, significant difference was observed for the ownership status, access to credit, corruption problem that confront the business and the total assets of the businesses. About 65 per cent of the entrepreneurs in the “treatment” group operate a sole-proprietorship type of business, unlike the comparison group (47 per cent), while 29 per cent of the businesses of the entrepreneurs in the “treatment” group have access to credit, compared to only 17 per cent in the comparison group. Also, more of the businesses in the “treatment” report corruption as a major issue that they are confronted with, compared to those in the comparison group. Finally, the average size of the businesses in the “treatment” group (based on total assets) is about three times larger than those in the comparison group, and this difference is significant at 1 per cent level.

The kernel density plot is also used to further present the outcome variables across the two groups of entrepreneurs’ businesses. The kernel density plots in Figure 1 reveals that in all the plots (a-c), the density for the entrepreneurs in the “treatment” group perfectly overlaps with those of the comparison group. This suggest that there is a rightward bias for innovation, revenue and job creation outcome of the businesses of entrepreneurs in the “treatment” group, relative to those in the comparison group. The implication is that the businesses of the entrepreneurs who were trained and went further to organize an in-house training program for their workers have a high-growth potential than those in the comparison group. The graph also supports the earlier observation in the descriptive statistics of the three outcome variables in Table I that there is a significant difference in the outcome variable for those businesses in the “treatment” group compared to the comparison group.

4. Econometric results

We begin the econometric analysis by presenting the results from the logistic regression and the balancing tests from the matching process. Table II shows that gender, access to credit, exposure to institutional crisis (like corruption), size of the firm and the number of hours that the entrepreneur put in other businesses apart from his current business were significantly associated with the choice of trained entrepreneurs having to set up an in-house training program for workers. The correlation between the choices of training the employees, the size of the entrepreneur’s businesses and the number of hours that the entrepreneur spends on other businesses apart from his current business follow logical expectation. For instance, entrepreneurs with large business size, but with insufficient time input in their current business, rely more often on training their workforce to enhance efficiency and to reduce the cost of monitoring. The significant coefficient for the gender variable may be linked to the fact that more male entrepreneurs are likely to engage in training their workers, especially when considering the social setup in Nigeria where men desire to gain industry competitive advantage. Yet more entrepreneurs with access to credit tend to be more aligned with training their workers. But with increased business exposure to corruption, entrepreneurs tend to reduce their implementation of in-house training for their employees.

The next step in the PSM estimation is to present the results of the balancing quality checks, which are reported in Figure 2 and Table III. Figure 2 shows the propensity score distribution of the two groups of entrepreneurs. It is evident from the figure that the entrepreneurs in the “treatment” group have equivalent matches from those in the comparison group. There is an adequate overlap between the two groups of entrepreneurs to justify the use of PSM. The region of common support is wide enough to generate adequate match for the PSM estimation. The comparison of the differences between the two groups, in terms of the overall covariance distribution (mean and median absolute bias) and the model fit (pseudo R2 and LR-test) before and after the matching, are presented in Table III. The results for the NNM, KM and RM in Table III suggest that the pre-matching differences in the observable characteristics of the entrepreneurs (across the two groups) are significantly reduced after the matching. For instance, the mean absolute biases are significantly reduced for the three matching algorithm, and the p-values of the LR test are no longer significant for post-matching.

4.1 Growth performance differences: Matching and regressions

The matching estimates are presented in Table IV. In addition, the Ordinary Least Square (OLS) estimation techniques with the three matching algorithms (NNM, KM and RM) were estimated and included in Table IV for robust checks.

The OLS results show that there is a significant increase in the innovation, revenue and job creation outcome of entrepreneur’s business as a result of taking further actions to set-up a training program for its workers. The OLS estimates for the three outcome variables are within the same range as those of the matching algorithms. For the NNM matching algorithm, for instance, the innovation outcome of the entrepreneur’s business from training its workers significantly increased by 3 to 4 point, compared to what it would have been assuming its workers were not trained. These results are consistent across the different matching algorithms. For the job creation outcome, we found that there is a significant increase in the number of jobs created within the entrepreneur’s businesses assuming the entrepreneurs support consistent in-house training of its workers. This increase was about two new jobs created annually. For the revenue outcome, we do not find a consistent significant increase in the revenue size for the entrepreneur’s businesses that support in-house training of its workers. Although the result was positive – suggesting a positive impact – we cannot verify this impact considering that it was not significant at either 1, 5 or 10 per cent levels of significance.

To check the robustness of these findings, we perform the Rosenbaum bounds test reported in Table V. The test shows the probability values from the Wilcoxon’s signed rank test, which presents the highest critical values that; the average treatment effect on the treated remains significantly different from zero. From the table, we see that the probability value is quite close to the estimated values in Table IV. The estimates in Table V indicates that the results in Table IV are valid assuming there is no hidden bias because of unobserved confounder. Thus, even the presence of unobserved differences in the covariates would not change our result. This is especially for the innovation and job creation model. However, for the revenue model, we need to exert some level of caution in inference, considering that the model is highly sensitive to unobserved differences. More so, the data does not allow us make inferences about the content of the training and the extent to which it influences the results.

4.2 Further analysis using the difference-in-difference approach

The results of the DiD estimations are presented in Table VI, and it is evident that the estimates corroborate the results of the PSM in Table IV. As earlier observed in Table IV, the value of the outcome variables of entrepreneurs who initiated an in-house training for their workers were significantly higher than those who did not initiate such training in their businesses. The innovation and job creation outcome variables remain positive and significant, but the revenue variable remains insignificant, despite that it was positive as in Table IV. The size of the impact is within the same range as that of Table IV for the outcome variable – job creation. However, for the innovation variable, the DiD result shows a slightly higher increase compared to the result in Table IV. This increase may be as a result of the sensitivity of this variable to some unobserved factors that are conditioned on time.

The findings in Tables IV and VI corroborate those in previous studies that training and other human capital development activity that are organized within entrepreneurs’ businesses have a significant positive impact on business outcomes Our results are in line with those of Millennium Challenge Corporation (2012), who used both quasi and pure experimental design to evaluate the impact of training and development on the outcome of agricultural entrepreneurs (farmers) in Honduras. Specifically, our findings support those of Duy et al. (2014), who find that the impact of investment in human capital on performance of the small and medium enterprises (SMEs) in Vietnam results in a significant positive impact on short-term performance, but not revenue and profit of the SMEs. More so, fast growing entrepreneur businesses are such that seek to meet their skill requirements through substantial training of their employees as well as searching for other human capital development activities that can improve their business outcomes (Mason et al., 2012).

5. Conclusion

Training programs that are directed at entrepreneurs are seen as an important catalyst for business growth and development. Likewise, entrepreneurs’ implementation of policies that support in-house retraining for their employees is also supposed to be important for high-growth business performance. Nonetheless, there is generally a lack of strong evidence to explain the later relationship. In this study, we provide empirical evidence to explain this relationship using a unique dataset from the Nigeria Youth Entrepreneurship Survey, which is part of the Youth Enterprise with Innovation in Nigeria (YouWiN!) Impact Evaluation survey. Using the propensity score matching technique, we examine the difference in revenue, innovation and employee growth of firms where the entrepreneurs were trained and they went ahead to set-up in-house training for their workers, compared to those who were trained but did not have an in-house training for their workers. We also conduct some sensitivity checks using the Rosenbaum bounds test and the DiD estimation technique to support our main finding.

From our analysis, we find that entrepreneurs who were earlier trained by the evaluation team and who went further to organize in-house trainings for their employees outperform their counterparts in the comparison group (i.e. those who never had any training for their workers). The experimental group were found to be more innovative and able to grow their businesses in terms of size of employee. We do not find a significant effect of retraining of employees by trained entrepreneurs on the revenue of the entrepreneur’s business. The positive effect of having an in-house training for employees on innovation and the growth in the number of employees are explained based on a logical expectation that employees who get trained will naturally be more innovative. For the growth in employee size, we infer that engaging in in-house trainings will boost the productivity and profitability of the small business, which will further broaden the capacity of such business to employ more individuals (Haltiwanger et al., 2015).

The results imply that policies that encourage just the training of entrepreneurs may be limited in the scope of impact if steps are not taken to ensure that the trained entrepreneurs go further to retrain their workers in their businesses. McKenzie (2017) did not emphasize this aspect, probably because the authors focused on the impact of the financial grant that was given to some participants of the initial evaluation training on the high-growth outcome of their businesses. Hence, our study suggest that it is important for entrepreneurs who receive such training to retrain their workers to have a better business outcome.

Just like many other scientific studies, ours is not void of some caveats that should be noted when interpreting our results. First, we used a survey data on entrepreneurs who self-reported that they organized in-house training program for their employees after receiving the initial evaluation training. Information pertaining to the content of such in-house training was not clearly stated in the survey. While we emphasize the importance of our finding, we are of the opinion that data from a well-structured instrument for data collection that include information on the specific subject matters covered in the original training and subsequent in-house trainings, will be very important for future analysis. These will also enable future studies to be in a position to provide detailed information on how the in-house training program could affect business growth and performance.

The second caveat that should be observed is that our data does not reflect whether the in-house training organized by the entrepreneurs was spurred by the earlier training that they received from an outside source. While we follow logical reasoning and strong assumptions that the in-house training organized by the entrepreneur will have contents of the earlier training that these entrepreneurs received from an outside source, it will be important for future studies to have a clear understanding of whether the in-house training contains contents of the previous training that the entrepreneur was exposed to. Noting these caveats, our findings should be seen as a first guide on how training entrepreneurs and those steps taken to retrain the workers of such entrepreneurs affect business growth and performance.


Kernel density plots (High-growth outcomes across groups)

Figure 1.

Kernel density plots (High-growth outcomes across groups)

Propensity score distribution

Figure 2.

Propensity score distribution

Descriptive statistics

Total ‘Treatment’ Comparison
Variable Mean SD N Mean SD N Mean SD N tx2
Entrepreneurs Characteristics
Gender (1 = male, 2 = female) 0.85 0.36 1,581 0.89 0.32 133 0.84 0.37 1,448 −1.41
Businesses owned (#) 1.29 0.62 1,581 1.38 0.67 133 1.28 0.62 1,448 −1.87*
Hours put into business (#) 44.26 22.63 1,581 48.95 26.05 133 43.83 22.25 1,448 −2.50**
Hours put into other businesses (#) 12.24 15.84 838 10.68 12.98 133 12.37 16.07 1,448 0.83
Confidence level (1 = low; 4 = high) 1.99 1.57 1,601 3.24 0.31 133 1.93 1.59 1,468 −9.49***
Business Characteristics
Ownership status (1, Sole proprietor; 0, otherwise) 0.48 0.50 1,601 0.65 0.48 133 0.47 0.49 1,468 −3.94***
Customers of business (# of customers) 90.65 315.72 1,601 102.32 235.72 133 89.95 320.03 1,468 −0.23
Access to credit (1 = yes if access to credit and 2 = no) 0.18 0.39 1,581 0.29 0.46 133 0.17 0.38 1,448 −3.55***
External environment (1 = yes if business is confronted with corrupt demands – e.g. bribe) 0.10 0.29 1,581 0.16 0.37 133 0.09 0.29 1,448 −2.46**
Total asset of the firm (‘000, 000) 6.44 1.65 1,601 16.60 27.30 70 5.88 15.50 1,468 −7.37***
Outcome Variables
Innovation (0, low and 12, high) 4.663 3.385 1,601 7.278 2.689 133 4.426 3.342 1,468 −9.563***
Total monthly sales (value in LCU, ‘000, 000) 0.904 2.335 1,496 1.265 2.446 133 0.870 2.323 1,363 −1.872*
Job creation (# of new employment) 7.863 9.035 1,542 10.323 9.498 125 7.640 8.963 1,417 −3.290***

LCU means local currency unit, which is the Nigerian Naira. The total asset of the firm and the Hrs. spent working in other businesses are presented in their real values and not in their logged form as in other tables in Section 4. This is to display the actual values before presenting their logged coefficient for the estimation results

Logistic regression

Dependent variable: Implementation of retraining workers Coefficient Robust standard error
Gender (1 = male, 2 = female) −0.844* 0.445
Businesses owned (#) −0.277* 0.146
Hours put into business (#) 0.671* 0.342
Confidence level (1 = low; 4 = high) −1.093 1.099
Ownership status −0.484 0.632
Customers of business (# of customers) −0.008 0.001
Access to credit (1 = yes if access to credit and 2 = no) −1.307** 0.659
Corruption (1 = yes if business is confronted with corrupt demands – e.g. bribe) −1.693** 0.677
Total asset of the firm (log)* 0.473** 0.241
Hrs. spent working in other businesses (log)* 0.273* 0.142
Constant −1.935 3.213
Pseudo R2 0.200
Wald Chi2 86.66 (0.000)
N 1180

The value in parenthesis is the probability value of Wald test. The superscript *, ** and *** imply significant levels at 10, 5 and 1 per cent, respectively

Matching quality

Matching algorithms Models Sample Total sample pseudo R2 LR chi-square p > Chi-square Mean bias Median bias
Five nearest neighborMatching (NNM) Innovation Unmatched 0.196 16.82 0.078 30.8 26.5
Matched 0.090 3.480 0.968 16.7 16.5
Revenue Unmatched 0.189 16.28 0.061 31.7 29.0
Matched 0.107 4.01 0.856 18.6 17.1
Job creation Unmatched 0.194 16.35 0.038 35.7 37.5
Matched 0.043 1.66 0.990 14.2 11.2
Kernel matching (KM) Innovation Unmatched 0.196 14.73 0.099 29.5 25.5
Matched 0.107 3.65 0.933 18.0 11.2
Revenue Unmatched 0.177 14.73 0.099 29.5 25.5
Matched 0.094 3.65 0.933 18.0 11.2
Job creation Unmatched 0.194 16.35 0.038 35.7 37.5
Matched 0.053 3.61 0.891 17.9 14.7
Innovation Unmatched 0.196 16.82 0.078 30.8 26.5
Matched 0.083 3.21 0.976 16.1 17.9
Radius matching Revenue Unmatched 0.177 14.73 0.099 29.5 25.5
Matched 0.067 2.59 0.978 14.9 16.3
Job creation Unmatched 0.194 16.35 0.038 35.7 37.5
Matched 0.086 3.34 0.911 16.7 12.2

The N for generating this table is 1,180 observations. The pstest for each of the individual observable characteristics that also shows the matching quality are available upon request. They were not included in this report for space

Estimated average treatment effect

Innovation 2.852*** 2.957*** 3.779*** 3.292***
(0.000) (0.008) (0.000) (0.000)
Revenue 341519.7* 231357.1 222967.7 53968.5
(0.094) (0.557) (0.859) (0.959)
Job creation 1.953*** 2.300* 2.294** 2.397*
(0.000) (0.070) (0.032) (0.086)
N 1601 1180 1070 1110

Probability values are in parenthesis. The superscripts *, ** and *** imply significant levels at 10, 5 and 1 per cent, respectively

Rosenbaum bounds test

Outcomes Gamma (Γ) U.Bound p-value L.Bound p-value U.Hodges-Lehmann L.Hodges-Lehmann
Innovation 1 0.001 0.001 3.417 3.417
1.25 0.002 0.000 3.223 3.587
1.5 0.004 0.000 3.106 3.691
1.75 0.008 0.000 3.048 3.835
2 0.012 0.000 2.976 3.913
2.25 0.017 0.000 2.861 3.918
2.5 0.022 0.000 2.776 4.047
2.75 0.028 0.000 2.678 4.082
Monthly Sales 1 0.976 0.976 774,793 774,793
1.25 0.991 0.948 839,197 711,542
1.5 0.996 0.911 874,043 653,457
1.75 0.999 0.869 896,764 576,930
2 0.999 0.824 957,966 477,997
2.25 1.000 0.779 989,886 273,002
2.5 1.000 0.734 1,100,000 171,521
2.75 1.000 0.690 1,200,000 140,552
Job Creation 1 0.070 0.070 2.025 2.025
1.25 0.013 0.032 1.309 2.240
1.5 0.020 0.015 0.934 2.383
1.75 0.027 0.007 0.701 2.668
2 0.034 0.003 0.486 2.883
2.25 0.041 0.001 0.241 3.094
2.5 0.047 0.001 0.033 3.202
2.75 0.053 0.000 −0.054 3.480

Double difference estimations

Multivariate linear regression
CoefficientProbability value
Estimated impact on innovation 5.002*** (0.000)
Estimated impact on revenue 170000 (0.682)
Estimated impact on job creation 2.390*** (0.003)

N for the regression models for innovation, revenue and job creation are 1,601, 1,496 and 1,542, respectively. The superscripts *** imply significant levels at 1 per cent


Audretsch, D.B. and Sanders, M. (2011), “Technological innovation, entrepreneurship, and development”, in Szirmai, A., Naude, W. and Goedhuys, M. (Eds), Entrepreneurship, Innovation, and Economic Development, Oxford University Press, Oxford.

Ayuso, S. and Navarrete‐Báez, F.E. (2017), “How does entrepreneurial and international orientation influence SMEs’ commitment to sustainable development? empirical evidence from Spain and Mexico”, Corporate Social Responsibility and Environmental Management, Vol. 25 No. 1, pp. 80-94.

Beck, T., Demirguc-Kunt, A. and Levine, R. (2005), “SMEs, growth, and poverty: Cross-Country evidence”, Journal of Economic Growth, Vol. 10 No. 3, pp. 199-229.

Benus, J.M. (1995), “Self-employment programs: a new reemployment tool”, Entrepreneurship Theory and Practice, Vol. 19 No. 2, pp. 73-85.

Blackburn, R. (1990), “Job quality in small businesses: electrical and electronic engineering firms in dorest”, environment and Planning A, Vol. 22 No. 7, pp. 875-892.

Bridge, S., O’Neill, K. and Cromie, S. (2003), Understanding Enterprise, Entrepreneurship and Small Business, 2nd ed., Palgrave Macmillan, New York, NY. ISBN 0-333- 98465-X.

Caliendo, M. and Kopeinig, S. (2008), “Some practical guidance for the implementation of propensity score matching”, Journal of Economic Surveys, Vol. 22 No. 1, pp. 31-72. doi: 10.1111/j.1467-6419.2007.00527.x.

Cardon, M.S. (2008), “Is passion contagious? the transference of entrepreneurial passion to employees”, Human Resource Management Review, Vol. 18 No. 2, pp. 77-86.

Central Bank of Nigeria-CBN (2003), Small and Medium Industries Equity Investments Scheme (SMIEIS), CBN Training Centre, Lagos.

Chell, E. (1985), “The entrepreneurial personality: a few ghosts laid to rest?”, International Small Business Journal: Researching Entrepreneurship, Vol. 3 No. 3, pp. 43-54.

Coad, A. and Tamvada, J.P. (2008), “The growth and decline of small firms in developing countries”, The Papers on Economics and Evolution, Working Paper 0808, available at:

Cook, T.D., Shadish, W.R. and Wong, V.C. (2008), “Three conditions under which experiments and observational studies produce comparable causal estimates: new findings from within-study comparisons”, Journal of Policy Analysis and Management, Vol. 27 No. 4, pp. 724-750.

Duy, N.K., Thanh, P.T., Oanh, N., Tam, N.D. and Vu, T.T. (2014), “Impact evaluation of training on firms’ performance – the case of small and medium enterprises in vietnam”, networking meeting of MERN, Phnom Penh, Cambodia 8-9 September 2014.

Dvoulety, O. and Lukes, M. (2016), “Review of empirical studies on self-employment out of unemployment: do self-employment policies make a positive impact?”, International Review of Entrepreneurship, Vol. 14 No. 3, pp. 361-376.

Elert, N., Andersson, F.W. and Wennberg, K. (2015), “The impact of entrepreneurship education in high school on long-term entrepreneurship performance”, Journal of Economic Behaviour and Organisation, Vol. 111, pp. 209-223.

Fafchamps, M., McKenzie, D., Quinn, S. and Woodruff, C. (2014), “Microenterprise growth and the flypaper effect: evidence from a randomized experiment in Ghana”, Journal of Development Economics, Vol. 106, pp. 211-226.

Fagerberg, J., Srholec, M. and Verspagen, B. (2009), “Innovation and economic development”, in, Hall, B., and, Rosenberg, N. (Eds), Handbook of the Economics of Innovation, Vol. 2, pp. 733-1256.

Gertler, P.J., Martinez, S., Premand, P., Rawlings, L.B. and Vermeersch, C.M. (2011), Impact Evaluation in Practice, The World Bank, Washington, DC.

Glazerman, S., Levy, D.M. and Myers, D. (2003), “Non-experimental versus experimental estimates of earnings impacts”, Annals of the American Academy, Vol. 589 No. 1, pp. 63-93.

Haltiwanger, J., Hyatt, H. and McEntarfer, E. (2015), “Do workers move up the firm productivity job ladder”, Paper presented at the 2015 SOLE conference, available at:

Heckman, J., Ichimura, H. and Todd, P.E. (1997), “Matching as an econometric evaluation estimator: Evidence from evaluating a job training programme”, Review of Economic Studies, Vol. 64 No. 4, pp. 605-654.

Hsieh, C. and Olken, B. (2014), “The missing missing Middle”, Journal of Economic Perspectives, Vol. 28 No. 3, pp. 89-108.

Karadag, H. (2017), “The impact of industry, firm age and education level on financial management performance in small and medium-sized enterprises (SMEs): evidence from Turkey”, Journal of Entrepreneurship in Emerging Economies, Vol. 9 No. 3, pp. 300-314, available at:

Klinger, B. and Schündeln, M. (2011), “Can entrepreneurial activity be taught? Quasi-Experimental evidence from Central america”, World Development, Vol. 39 No. 9, pp. 1592-1610.

Kosanovich, W.T., Fleck, H., Yost, B., Armon, W. and Siliezar, S. (2001), Final Report: Comprehensive Assessment of Self-Employment Assistance Programs, DTI Associates, Arlington, VA.

Li, J., Zhang, J. and Yang, Z. (2017), “Associations between a leader’s work passion and an employee’s work passion: a moderated mediation model”, Frontiers in Psychology, Vol. 8 doi: 10.3389/fpsyg.2017.01447.

Mano, Y., Iddrisu, A., Yoshino, Y. and Sonobe, T. (2012), “How can micro and small enterprises in Sub-Saharan Africa become more productive? the impacts of experimental basic managerial training”, World Development, Vol. 40 No. 3, pp. 458-468.

Mason, G., Robinson, C. and Bondibene, C.R. (2012), “Fast-growing firms, product strategies and skills development”, Nesta, Working Paper No. 12/12.

McKenzie, D. (2017), “Identifying and spurring High-Growth entrepreneurship: Experimental evidence from a business plan competition”, American Economic Review, Vol. 107 No. 8, pp. 2278-2307.

McKenzie, D. and Woodruff, C. (2015), “Are bad business practices holding back small firms in developing countries? CEPR’s policy paper”, available at:

McMahon, R.G.P. (2001), “Growth and performance of manufacturing SMEs: the influence of financial management characteristics”, International Small Business Journal, Vol. 19 No. 3, pp. 10-28.

Michaelides, M. and Benus, J. (2012), “Are self-employment training programs effective? evidence from project GATE”, Labour Economics, Vol. 19 No. 5, pp. 695-705.

Michelitsch, R., Saliola, F. and Bernt, J.V. (2011), “Assessing private sector contributions to job creation: IFC open source study”, International Finance Corporation, available at:

Millennium Challenge Corporation (2012), MCC’s First Impact Evaluations: Farmer Training in Five Countries, MCC, Washington, DC.

Naude, W. (2013), “Entrepreneurship and economic development: theory, evidence, and policy”, Institute for the Study of Labour – IZA, Discussion Paper Series.

Nichter, S. and Goldmark, L. (2009), “Small firm growth in developing countries”, World Development, Vol. 37 No. 9, pp. 1453-1464.

Njoroge, C.W. and Gathungu, J.M. (2013), “The effect of entrepreneurial education and training on development of small and medium size enterprises in githunguri District-Kenya”, International Journal of Education and Research, Vol. 1 No. 8, pp. 1-22.

Obi, R. (2013), “65% new businesses die within first 3 years in Nigeria – NASME”, Business News, available at:

Olafsen, E. and Cook, P.A. (2016), “Growth entrepreneurship in developing countries: a preliminary literature review”, Working Paper, The World Bank Group, Washington, DC.

Riel, A., Tichkiewitch, S. and Paris, H. (2015), “Preparing researchers for entrepreneurship based on systematic innovation training”, Procedia Engineering, Vol. 131, pp. 940-944.

Rodríguez-Planas, N. and Jacob, B. (2010), “Evaluating active labor market programs in Romania”, Empirical Economics, Vol. 38 No. 1, pp. 65-84.

Rosenbaum, P.R. and Rubin, D.B. (1983), “The central role of the propensity score in observational studies for causal effects”, Biometrika, Vol. 70 No. 1, pp. 41-55.

Santarelli, E. and Vivarelli, M. (2007), “Entrepreneurship and the process of firm’s entry, survival and growth”, Industrial and Corporate Change, Vol. 16 No. 3, pp. 455-488.

Schoar, A. (2010), “The divide between subsistence and transformational entrepreneurship”, in Joshua, L. and Scott, S. (Eds), NBER Innovation Policy and the Economy, NBER, Boston, MA, Vol. 10, pp. 57-81.

Shindina, T., Lysenko, Y. and Orlova, N. (2015), “Entrepreneurs training in Innovation-Oriented society”, Procedia-Social and Behavioural Sciences, Vol. 214 No. 5, pp. 1098-1108.

Stewart, W.H., Jr, Watson, W.E., Carland, J.C. and Carland, J.W. (1998), “A proclivity for entrepreneurship: a comparison of entrepreneurs, small business owners, and corporate managers”, Journal of Business Venturing, Vol. 14 No. 2, pp. 189-214.

Szirmai, A., Naude, W. and Goedhuys, M. (2011), “Entrepreneurship, innovation, and economic development: An overview”, in Szirmai, A., Naude, W. and Goedhuys, M., (Eds), Entrepreneurship, Innovation, and Economic Development, Oxford University Press, Oxford.

Tambwe, M. (2015), “The impact of entrepreneurship training on micro and small enterprises’ (MSES) performance in tanzania”, Business Education Journal, Vol. 1 No. 1.

Tetteh, E.K. and Essegbey, G.O. (2014), “Firm level innovation: the case of ghanaian firms”, European Journal of Business and Innovation Research, Vol. 2 No. 2, pp. 1-18.

Titley, J. and Anderson-Macdonald, S. (2015), “When entrepreneurship training matters most”, International Growth Centre Blog.

Wanjala, B.M. and Muradian, R. (2013), “Can big push interventions take Small-Scale farmers out of poverty? Insights from the sauri millennium village in Kenya”, World Development, Vol. 45, pp. 147-160.

Westhead, P. and Storey, D. (1996), “Management training and small firm performance: why is the link so weak?”, International Small Business Journal, Vol. 14 No. 4, pp. 13-24.


The authors are grateful to the World Bank and David McKenzie for making their data available. The authors are also grateful to the participants at the 2017 Conference of the Economic Society of South Africa, where the first draft of the paper was presented. The authors appreciate all comments from the participants. Finally, the authors acknowledge the comments of the reviewers, which were helpful in improving the final version of this paper. As usual, all other errors and opinions are those of the authors.

Corresponding author

Uchenna Efobi can be contacted at: