Why should banks provide entrepreneurship training seminars?

Abstract

This article examines an apparent contradiction at the heart of the provision of management training and advice for new and small firms. Assessments using self-report data show high levels of satisfaction, implying that the training/advice is effective and appreciated. In contrast, assessments using robust statistical methods point to modest, or even zero, impact upon firm performance. Accordingly, our core research question explores whether there is an identifiable performance benefit of management training or whether impact is limited to emotional attachment to the training provider – reflected in enhanced loyalty. We test this by examining the effects of a bank seminar provided for new enterprises and find it had no significant effects on either the survival or sales growth of participants. However, those new enterprises who participated in the seminar were significantly less likely to switch to a rival bank, implying the seminars may have induced a feeling of loyalty among clients. Finally, we discuss the implications for theory, for the bank and for the providers of training for new and small firms more widely.

Keywords

client retention entrepreneurial learning entrepreneurship education firm growth survival

Introduction

This article proposes that management training for new and small firms can have two outcomes – a performance effect and a loyalty effect. Training is theorised to enhance the knowledge and skills of the owner/manager (Burke and Hutchins, 2007); such skills are seen (Ireland et al., 2003) as key factors influencing the performance of a new small firm. However, this link is contested by Coad et al. (2013) who argue that new venture performance is close to a random walk and so unrelated to knowledge and skills. The loyalty effect is derived from theories arguing that the providers of training arouse positive emotions leading to changed mindsets (Cardon et al., 2005; Souitaris et al., 2007), which in turn leads recipients to display an emotional attachment to training providers. It is this which explains why the use of self-report data consistently points to a positive impact of training. The core research question posed within this article therefore, explores whether there is an identifiable performance effect from management training or whether impact is limited to emotional attachment to the training provider – reflected in enhanced loyalty.

To test these theories, first, we examine the impact of training provided for new and small firm customers by Barclays Bank. Based on our theories, we hypothesise that training could have commercial benefits because the trainee firm is able to apply the knowledge acquired to improve its performance – reflected in higher survival rates and higher growth rates than would otherwise have been the case. Barclays may also benefit if an improvement in performance lowers the risk of defaults and/or leads to a greater demand for a range of banking services. Our second test is of the loyalty effect. Here, we argue that the new/small firm that participates in training may feel an increased sense of satisfaction with Barclays, making them more loyal and so less likely to switch to an alternative bank. Training is therefore hypothesised to lead to a lower likelihood of switching.

To conduct an appropriate test of these theories requires a dataset that can deal with the crucial statistical issues of self-selection and comparisons with an appropriate control group. Thus, we apply the ‘powerful matching techniques’ called for by Rideout and Gray (2013: 347) in order to obtain estimates of the effect of knowledge enhancement on firm performance on the one hand and on loyalty-raising on the other. Specifically, we utilise Propensity-Score Matching (PSM) to address causality concerns (Caliendo and Künn, 2011; Pons Rotger et al., 2012). Our central result is that seminar participation has no significant impact on the survival or sales growth of the enterprise, confirming other statistical work showing that management training provides little or no performance enhancement among new and small firms (Bryan, 2006; Georgiadis and Pitelis, 2014; Norrman and Bager-Sjögren, 2010). However, we do find that those who participated in bank training were less likely to switch providers – offering support for the loyalty theory. The final novelty of the article is that we are able to report how the Barclays responded to these findings; it appears that despite the loyalty they appeared to generate, the bank is uncomfortable with its programmes being unable to influence new firm performance. Accordingly, it has chosen to shift the emphasis away from training and advice for very new, and sometimes pre-start, firms towards enterprises with a longer track record.

The remainder of the article is organised as follows. We begin with a literature review (section ‘Knowledge, resources and new firm performance’) to provide the reader with a background for our analysis. We then formulate hypotheses (section ‘Hypotheses’) relating to how the entrepreneurial training seminar might affect different dimensions of performance and loyalty. Section ‘Data’ presents the dataset, paying particular attention to the details of the entrepreneurial seminar in question. Section ‘Analysis’ contains the analysis, and section ‘Conclusion’ concludes.

Knowledge, resources and new firm performance

This section is in three parts. The first reviews the theoretical case and empirical evidence in support of a positive link between knowledge, resources and new firm performance. The second makes the alternative case that this relationship is likely to be insignificant. The final section reviews the evidence pointing towards knowledge enhancement, through training, arousing emotions and changing mindsets. The case is made that this leads to a loyalty bond between the provider of training and the new firm owner.

The case for a positive link between training/advice/counselling and the performance of new and small firms

There is an extensive literature, stretching back to Cooper (1993) linking knowledge and resources on the one hand with new firm performance on the other hand. More recently, Wiklund et al. (2009) review the myriad factors believed to influence the growth of new and small firms; these factors are divided into five groups: entrepreneurial orientation, the (external) environment, strategic fit, attitudes and, finally, resources. Wiklund et al. recognise the limitations of current knowledge; for example, the five groups clearly overlap with each other and are inter-dependent rather than independent. Then, even when the correlations between each group and with performance are identified, the direction of causation is rarely adequately explored. Despite excluding a role for chance, Wiklund et al. provide a valuable review of current knowledge and its limitations. They argue there is a long tradition of research that assumes that enterprises with access to more and better resources perform better Rocha et al. (2013). The more nuanced interpretations then distinguish between the ‘types’ of resources, how they are acquired and how they change in order to adjust to changing circumstances (Teece et al., 1997). Among the distinctions drawn between ‘types’ of resources is the separation between financial and non-financial resources. The former is the ability to access fixed and working capital – normally from own resources, fellow traders or from a financial institution such as a bank (Berger and Udell, 1998). Non-financial resources are conceptually rather more slippery, but normally include dimensions such as the human capital of the business owner(s) with the inference that these reflect their knowledge and skills.

Knowledge and skills within the enterprise are theorised to influence performance in several ways. The most direct is through the educational qualifications of the owner(s) – it being assumed that education either enhances these skills and qualifications reflect inherent intelligence (Van der Sluis et al., 2003). A second dimension of knowledge and skills is the expertise that owners bring from prior employment (Braguinsky et al., 2012) or business ownership (Coleman, 2007). Here it is assumed that, having faced broadly similar situations previously, the owner(s) have learnt either from mistakes or successes – or both. However, the owner’s current knowledge and skills include those acquired both before they began their current enterprise and while operating the current business – particularly through the development of networks (Vissa and Chacar, 2009). As such, knowledge is at the heart of non-financial resources. It can be acquired through the experiential route, as reflected in the examples cited above, but it may also be acquired through more formal modes of learning. Two of these are potentially relevant for the new/small business owner. The first is training, which can be considered as a group activity providing general knowledge. The second is consultancy which is the bespoke provision of information and guidance tailored to the specific requirements of an individual business.

We now make the case that, as with experiential learning, additions to the stock of knowledge of the new firm founder through training/advice/counselling will enhance firm performance. According to Chrisman and McMullan (2002, 2004), once the need for new knowledge is recognised by the venture founder, an experienced outsider directs and facilitates a learning process, often by combining tacit with explicit knowledge. Such entrepreneurs learn valuable skills and knowhow, enabling them to better face a multitude of challenges, apply their knowledge to resolve difficulties, better plan ahead to avoid potential problems and be better prepared for future challenges. Prior knowledge also helps the new firm founder to identify a higher number of, and more productive, entrepreneurial opportunities (Shepherd and DeTienne, 2005). Formal learning may also enhance social legitimacy and validation. Pons Rotger et al. (2012) make the distinction between knowledge enhancement and badging. The former is the skills the trainees gain which enable them to better position their products and enjoy higher sales. In contrast, badging is the formal qualification – the certificate obtained – which serves to reassure financiers or other traders of their legitimacy and minimum level of competence and honesty.

In short, knowledge is a core non-financial resource of the new venture owner. Owners with more knowledge are expected to have ventures that perform better. So, if this knowledge can be supplemented through training, it leads to better venture performance.

The case for no link between knowledge and the performance of new and small firms

Despite knowledge underpinning much of the entrepreneurship literature, the nature and presence of its link with performance remain an open and valid question:

Small firms literature generally reveals an expectation of a positive relationship between training and performance, while evidence for this is scanty. (Bryan, 2006)

Along similar lines, Jones et al. (2013) write,

A body of literature that broadly supports the argument that training positively influences business performance through enhanced productivity, quality, labour turnover, and financial results can therefore be seen to exist … It must also be highlighted, however, that this evidence is not unequivocal.

The case for the absence of a link between knowledge and new/small firm performance is based on two types of evidence. The first is theoretical and develops the concept of Gambler’s Ruin. The second is empirical and draws upon prior studies that have failed to find a link between the performance of new and small firms and several proxies for knowledge enhancement.

Gambler’s Ruin

The theory-based case is that following Gibrat’s Law, the performance of a new enterprise is best characterised as a random walk (Levinthal, 1991). The analogy provided by Coad et al. (2013) is that the new enterprise owner is faced with a similar situation to a gambler entering a casino and seeking to play a game of chance. In this Gambler’s Ruin model, performance at the table (win/lose) is a game of chance, but survival in business – as in the casino – depends upon access to a stock of chips or financial resources. So, those with more chips survive for longer, on average. While some scholars may feel uneasy with the concept of firm growth being essentially unpredictable, nevertheless a random process of firm growth seems to fit the data well (Denrell et al., 2015). Geroski explains that ‘The most elementary “fact” about corporate growth thrown up by econometric work on both large and small firms is that firm size follows a random walk’. (Geroski, 2000: 169). Random models of firm growth need not imply that entrepreneurs have no strategy or effort; however, ‘Chance models are, in fact, compatible with effortful managers who carry out deliberate actions’ (Denrell et al., 2015: 937). Gambler’s Ruin asserts that the probability of a win increases with duration at the table – so, as with any game of chance, the more tickets bought or dice rolled, the higher the chance of having a win, even though the probability of a win, for each ticket purchased, remains the same. On those grounds, even though they have no more entrepreneurial knowledge/talent than others, those with ‘deep pockets’ have a better chance of surviving (Coad et al., 2014).

Thus, knowledge plays no role in the pure Gambler’s Ruin model of the performance of new and small firms. However, most importantly, it does not imply the new firm founder either does not learn something or is somehow incapable of learning, and it certainly does not exclude entrepreneurs reporting to others that they have learnt. Instead, learning is excluded by definition since it is impossible to ‘learn to play the lottery’. This is primarily because the circumstances faced by the owner(s) of new and small firms fluctuate extensively and are so diverse that the opportunities for carrying learning from one context to another are minimal. This is in line with the evidence provided in the review by Thompson (2009) that it is learning through repetition which enhances performance, although even here diminishing returns set in quickly. In contrast, learning from wholly diverse circumstances is close to impossible.

Empirical evidence regarding knowledge and learning

One source of evidence on the absence of a strong link between knowledge enhancement in the form of training/consultancy/advice and performance is derived from evaluations of the impact of publicly funded training and consultancy programmes (Georgiadis and Pitelis, 2014; Norrman and Bager-Sjögren, 2010; Organisation for Economic Co-operation and Development (OECD), 2007; Pons Rotger et al., 2012; Wren and Storey, 2002). Several reasons for this weakness are proposed. Some may interpret this as evidence that the training is delivered by public sector organisations that lack a real understanding of the issues facing private enterprises at start-up. This, however, seems an unlikely explanation since the results for publicly funded and delivered training by ALMI in Sweden appear neither better nor worse than in other countries such as the United Kingdom where delivery has been through private sector consultants (Mole et al., 2008).

A second, and more challenging, explanation is that although entrepreneurial learning is often assumed a priori (Politis, 2005), the evidence is either very limited or non-existent, once that which is based on self-report data is excluded. For example, one frequently asserted link is that prior business-ownership experience provides knowledge which is then used to improve the performance of a subsequent venture. However, large-scale evidence from Germany (Gottschalk et al., 2013; Metzger, 2006), Denmark (Nielsen and Sarasvathy, 2011) and Portugal (Rocha et al., 2015) suggests that entrepreneurs who re-enter after failure perform worse than novices. Using US Panel Study of Entrepreneurial Dynamics (PSED) Data, Parker (2013) confirms his earlier findings (Parker, 2006) that entrepreneurs do not significantly adjust their priors in the light of experience. He concludes that any learning from prior business ownership is, at best, short-lived. Finally, prior work (Frankish et al., 2013) using the same dataset utilised in this study found that new business owners continue to engage in risky behaviour (unauthorised overdraft excesses), even when the serious consequences for the survival of their business were made clear to them.

However, even though business owners may not either have accumulated experiential expertise, either from formal education or from prior experience or in running their current business, they may benefit from expert advice from others. Specifically, those who seek out advice/training may be those who are more aware of their limitations and be more willing to learn from others than those who are more optimistic or confident about their skills. The case can still be made that training and advice can improve the knowledge-based resources available to the owners of some new and small businesses and hence, enterprise performance. Yet, we are only aware of a very small number of studies that examine this link between knowledge, resources and the performance outcomes for new and small firms using longitudinal data and sophisticated econometric techniques. The most notable recent example is by Georgiadis and Pitelis (2015) who examine a programme of workforce and management training launched by the British government to ‘foster growth, productivity and performance amongst SMEs in the accommodation and food services sector’. The novelty of their analysis is that, because training places were limited, these were allocated randomly. The core finding was ‘a weak or no effect of managerial and human resource management (HRM) training services on firm performance’.¹ Other work using this approach has been undertaken in less developed countries. For example, Karlan and Valdivia (2011) apply a randomised controlled trial (RCT) to data on Peruvian microfinance, and observe there is little or no evidence that entrepreneurship training sessions yield benefits in key dimensions (business revenue, profits, employment), although, interestingly in the context of our article, they improve client retention rates for the microfinance institution. Klinger and Schundeln (2011), however, obtain more positive results from three Central American countries, implying the debate remains open.

In short, if there is a substantial unexplained element to the performance of new ventures which can be characterised as randomness, then the opportunities for learning are small. It is this which constitutes the theoretical underpinning for large-scale statistical work that has consistently failed to find a clear link between training and SME performance generally and particularly, for new firms.

Enhancing emotional resources

The two theoretical approaches described above are examples of what Cardon et al. (2005) refer to as a hegemonic rational perspective. They argue that an alternative stance which offers novel insights into several aspects of entrepreneurial behaviour is the relationship approach. This provides insights into, for example, cognitive biases in risk assessment or persistence in entrepreneurship despite poor results. Most challengingly, they argue that it ‘sheds light on aspects of entrepreneurship that seem illogical from the hegemonic rational perspective’. Souitaris et al. (2007) widen this approach by stating, ‘Entrepreneurship can also be viewed from an emotional lens. The hero, apart from being intelligent and rational, has also to be passionate and emotional’ (p. 589). They make the case that emotional factors play a powerful role in the formation and development of new business ventures, with this theme being taken up by those such as Cardon et al. (2009, 2013) and Murnieks et al. (2014). It is argued that passion – a strong inclination towards certain activities the individual regards as important – plays an important role in behaviour across a wide variety of disciplines. Passion is argued to be an underlying cause of economic behaviour and, again, one likely to offer different predictions from those derived from the hegemonic rational perspective.

The role of affect on the entrepreneurial process generally appears to be a valid and promising area for research (Baron, 2008; Foo, 2011). Although previous research has investigated the role of emotions and affect on entrepreneurial behaviour during the opportunity identification stage (Grichnik et al., 2010; Welpe et al., 2012) and exit (Shepherd, 2003), nevertheless, we contribute towards filling a gap in the literature concerning how emotions influence the ‘middle’ of the entrepreneurial process (Cardon et al., 2012: 3). Our investigation also contributes to the gap in the literature concerning how entrepreneur emotions affect relations with other stakeholders (Cardon et al., 2012: 4), which in our case is the provider of banking services. Furthermore, although much evidence on the role of emotions in entrepreneurship comes from laboratory experiments (e.g. Foo, 2011; Grichnik et al., 2010; Welpe et al., 2012), we focus on a real-world setting.

Cardon et al. (2005) make the case that our understanding of the passions and emotions experienced by entrepreneurs are helped by analogies – most notably that between parent and child. The evidence for this is based on entrepreneurs frequently describing their business using terms that reflect this link – such as referring to it as ‘my baby’ (see also Handler and Kram, 1988). Thus, if entrepreneurs think of businesses as ‘their babies’, then we need to theorise about how new and small business owners view the provision of ‘support for their babies’. We make the case that business owners will view this support in a similar positive manner to that of a parent whose child is treated by a doctor. Reported reactions following the treatment are likely to be positive for three reasons. The first is that, as between the parent and the doctor, training for the owners of new and small firms arouses emotions and changes mindsets, with this being reflected in raised optimism and confidence or in providing the tenacity to overcome obstacles. Second, those participating in training may develop an emotional attachment to their mentors. Entrepreneurs who see that the seminar provider shows concern and support for their new businesses are likely to make favourable judgements or evaluations about the seminar provider (Baron, 2008), such as building up loyalty that leads to lower rates of bank account switching. The third factor pointing towards a positive response is that, since the decision to participate in training was made by the business owner(s), to respond negatively casts doubt upon their judgement.

In short, the case is that the provision of guidance and advice generates an emotional attachment to the providers of help. Where that training is provided by a bank, this emotional attachment could be captured through self-report data. However, the well-established limitations of self-report data in the entrepreneurship field (Bertrand and Mullainathan, 2001) point to the desirability of capturing this attachment through a directly observable metric that is, the observed decision relating to probability of switching to a rival bank.

Hypotheses

The case has been made for and against the impact of some form of knowledge enhancement on the performance of new and small firms. To test this, we identify two measures of firm performance, survival and sales growth, and formulate the following hypotheses:

Hypothesis 1a: Businesses where Owners received Bank training experience faster sales growth than otherwise comparable businesses.

Hypothesis 1b: Businesses where Owners received Bank training have higher survival rates than otherwise comparable businesses.

We also made the case that, even if performance is not enhanced by participating in a training programme, it is possible that this generates a sense of loyalty towards the provider of training. For a commercial organisation such as a bank, the case for providing free/subsidised training has three elements. The first, reflecting H1a and H1b above, is that the training enhances the performance of the enterprises leading to lower defaults for the bank. The second is that training enhances enterprise growth leading to a greater demand for a range of financial services provided by the bank.

The third case is that, even if neither of the above occurs, the enterprise appreciated the training provided. It is taken to be an indicator of concern on the part of the bank for its customer; it may give the business owners greater insights into business issues and so improve their enjoyment of the business. In short, even if there are no ‘bottom line’ effects, training may improve the bond between the business owner and the bank, leading to greater loyalty reflected in a longer relationship between them. This emotional attachment is manifest in a sense of greater loyalty to the bank and measured as a lower probability of switching to an alternative provider of banking services. We therefore, hypothesise the following:

Hypothesis 2: Businesses where Owners received Bank training will be less likely to switch to a rival bank.

Data

We analyse the customer records from Barclays Bank. It provides the primary current (checking) account facility for just over 20% of all businesses in England and Wales with sales of less than £1 million; their active customer base in this market is in excess of 500,000 firms. Firms in our dataset are present in all sectors (except for financial services). We focus on a cohort of new firms that began to trade for the first time between April and June 2004. Only those firms with trading activity over the period April–June 2004 are included, and so dormant businesses are excluded. The key advantage of these data is that every financial transaction is monitored with both the bank and the business owner(s) having a strong incentive to ensure that the data are accurate and timely. Once operational, the bank is able to accurately gather information on the customer’s activity by overseeing its account. We have over 6000 observations at start-up, with all financial transactions collated on a monthly and then quarterly basis. Although there is considerable sample attrition, we can track the performance of firms until they exit. It therefore overcomes the flaws noted by Yang and Aldrich (2012) that characterise virtually all data on new enterprises.

Several other features of commercial bank account data require emphasis. The first is that the creation of a business bank account is NOT conditional on the use of any other banking service such as a deposit account, an overdraft facility or a term loan. Accordingly, it is NOT a sample of bank borrowers, but instead a sample of all new business bank customers. However, of course, we are able to identify whether a customer has access to, and makes use of, loans or overdrafts as well as the scale of this finance. A second characteristic is that, unlike much of continental Europe (Ongena and Smith, 2000), multiple banking is rare among new firms in the United Kingdom. Hence, the transactions passing through the account are likely to capture in full the activities of the business. Third, the bank is able, by account monitoring, to quickly assess when a personal account is being used for business purposes. It therefore is able to accurately identify when trading starts – even if it is not informed of this by the customer – and provide appropriate account facilities. As part of this process, new business owners were asked to complete a voluntary questionnaire relating to their prior employment and educational attainment, together with some personal details such as age and gender, as well as information on the sources of advice or support approached prior to start-up.

Start-Right Seminar

A key variable of interest relates to participation of a new business owner in the free Start-Right Seminars (SRSs) offered by Barclays. A regular programme of seminars (more than 70 a year) was provided across England and Wales, run by local Enterprise Agencies – organisations providing wider business support services – on behalf of Barclays. These seminars lasted between two and three hours – which is comparable with Solomon et al. (2013) who take three hours as the cut-off between high and low experience of face-to-face counselling, and Chrisman and colleagues, whose samples focus on small businesses with minimum five hours of counselling (e.g. Chrisman and Katrishen, 1994; Chrisman and McMullan, 2004). To the extent that these training seminars are simply too short to impart valuable advice or have an effect on performance and behaviour, we would expect no significant support for any of our hypotheses.

The seminars were open to potential and recent start-ups, that is, those thinking about going into business and those who had just done so. There was no requirement to attend the seminars as a condition of opening an account with Barclays. Indeed, it is likely that some participants eventually used another banking provider or decided not to start-up at all. The emphasis of the material presented was on concise practical advice from experienced business advisers covering a number of critical areas required to take a prospective owner from potential to established trading, including developing a business idea, drawing up a business plan, sourcing and managing business finances, marketing, tax and other legal obligations on businesses and finding the right banking solutions to suit the requirements of the firm. Given the voluntary nature of the seminars, participants may or may not have been representative of the broader set of individuals going into business at this time, in terms either of their immediately observable characteristics or their resources, aims and ambitions. We take account of this self-selection when choosing our econometric strategy.

Performance variables

Our treatment variable SRS participation is evaluated in terms of turnover growth, survival and bank account switching. Our lag structure is such that SRS participation (which occurs around the time of entry) has its effects on survival, growth or switching, in the years after entry, up to and including the sixth year after entry). Our measure of size (and growth) is a measure of firm sales (to be precise, we look at credit turnover which corresponds to the value of payments into the business current account). Annual turnover growth is calculated by taking log-differences (Coad, 2009; Tornqvist et al., 1985). We can also identify which firms switch bank to use an account at a different bank. Furthermore, as noted earlier, in the United Kingdom, the account at a single bank is likely to capture the full trading activities of the new venture.

Control variables

Our database boasts a rich set of control variables that we use to ensure that the treatment group (SRS participants) is as comparable as possible to the control group (SRS non-participants). We begin with characteristics of the founder. A gender dummy is included to control for gender-related factors. Age and age-squared are included as controls, because of the expected non-linear relationship between founder age and venture performance. Education may also affect firm performance as well as opportunity recognition and entrepreneurial learning, and so we include four dummy variables reflecting the degree of education attained (none, General Certificate of Secondary Education (GCSE), A-level, Degree or higher). We also control for the founder’s business experience (either personal business experience or parental business experience) because the latter may affect learning and performance outcomes. Controls are also available for firm-level characteristics that can be expected to affect entrepreneurial behaviour and performance – the number of owners of the venture, the legal form of the venture (dummies for company, partnership, or sole trader), as well as 12 industry dummies. We also control for firm size by using the log of annual turnover, and lagged growth (measured in terms of log-differences taken at an annual frequency). Further information on these control variables is found in Table 7 in Appendix 1.

Table 1 contains summary statistics for the full sample, as well as for subsamples of SRS participants compared with non-SRS participants, and Switchers compared with non-Switchers. This suggests SRS participants are somewhat younger, better educated, more likely to be female and less likely to have prior business experience (either personal or parental business experience) than non-participants. The ventures of SRS participants have fewer owners, and with regards to legal form, they are more likely to be sole traders.

Table 1.

Summary statistics for year 1: full sample, SRS, non-SRS, switchers and non-switchers.

Variable	Full sample		Non-SRS		SRS		t-test p-values	Non-switchers		Switchers		t-test p-values
	(about 6671 obs.)		(about 6622 obs.)		(about 49 obs.)		t-test p-values	(about 6247 obs.)		(about 424 obs.)		t-test p-values
	Mean	Std Dev.	Mean	Std Dev.	Mean	Std Dev.		Mean	Std Dev.	Mean	Std Dev.
Turnover	116,868	514,242	117,225	515,874	64,587	127,181	0.0661	116,724	529,336	118,709	251,397	0.0090
log(turnover)	10.55	1.50	10.55	1.50	10.01	1.49	0.0000	10.55	1.49	10.49	1.61	0.0000
Age	39.04	10.23	39.06	10.24	36.32	9.47	0.0000	39.02	10.25	39.46	10.03	0.0336
Education dummy
None	0.23	0.42	0.23	0.42	0.04	0.20	0.0000	0.23	0.42	0.19	0.39	0.0000
GCSE	0.33	0.47	0.33	0.47	0.35	0.48	0.5633	0.33	0.47	0.33	0.47	0.6762
A-level	0.17	0.38	0.17	0.38	0.24	0.43	0.0007	0.17	0.38	0.19	0.39	0.0138
Degree or higher	0.27	0.44	0.27	0.44	0.37	0.49	0.0002	0.27	0.44	0.29	0.45	0.0763
Business experience
Self	0.72	0.45	0.72	0.45	0.53	0.50	0.0000	0.72	0.45	0.70	0.46	0.0091
Family	0.63	0.48	0.63	0.48	0.57	0.50	0.0263	0.63	0.48	0.63	0.48	0.8658
No. of owners	0.16	0.36	0.16	0.37	0.06	0.24	0.0000	0.16	0.36	0.16	0.37	0.4825
Male owner	0.81	0.39	0.81	0.39	0.63	0.49	0.0000	0.81	0.39	0.76	0.43	0.0000
Legal form dummy
Company	0.37	0.48	0.38	0.48	0.00	0.00	0.0000	0.37	0.48	0.40	0.49	0.0160
Partnership	0.13	0.34	0.13	0.34	0.22	0.42	0.0000	0.13	0.34	0.16	0.36	0.0004
Sole trader	0.49	0.50	0.49	0.50	0.57	0.50	0.0073	0.50	0.50	0.45	0.50	0.0000
Industry dummy
Agriculture	0.01	0.10	0.01	0.10	0.00	0.00	0.0830	0.01	0.10	0.02	0.15	0.0000
Manufacturing	0.05	0.22	0.05	0.22	0.06	0.24	0.4523	0.05	0.22	0.04	0.20	0.0317
Construction	0.15	0.35	0.15	0.35	0.08	0.28	0.0016	0.15	0.35	0.13	0.34	0.0136
Motor trades	0.04	0.18	0.04	0.18	0.06	0.24	0.0152	0.04	0.19	0.03	0.17	0.1967
Wholesale	0.02	0.15	0.02	0.15	0.02	0.14	0.7840	0.02	0.15	0.02	0.14	0.1712
Retail	0.12	0.32	0.12	0.32	0.18	0.39	0.0007	0.12	0.32	0.11	0.32	0.3141
Hotels and catering	0.09	0.29	0.09	0.29	0.06	0.24	0.0524	0.09	0.29	0.10	0.30	0.3801
Transport	0.04	0.19	0.04	0.19	0.06	0.24	0.0328	0.04	0.19	0.04	0.19	0.9759
Property services	0.03	0.18	0.03	0.18	0.04	0.20	0.5706	0.03	0.18	0.04	0.20	0.1304
Business services	0.26	0.44	0.26	0.44	0.24	0.43	0.5553	0.26	0.44	0.25	0.44	0.5355
Health, educ., social	0.02	0.15	0.02	0.15	0.04	0.20	0.0583	0.02	0.16	0.01	0.12	0.0008
Other services	0.16	0.37	0.16	0.37	0.14	0.35	0.3259	0.16	0.37	0.20	0.40	0.0000

The t-test p-values compare non-SRS and SRS means, and non-Switchers and Switchers means. Number of observations is slightly lower than the figure appearing at the top of the columns for some variables due to a small number of missing observations (exact numbers available from the authors upon request).

Analysis

We investigate our hypotheses using regression and matching techniques. Matching is a quasi-experimental technique that is preferable to multivariate regressions in terms of obtaining causal estimates from observational data because more care is taken to establish an appropriate control group (Nichols, 2007).

SRS: survival and growth

Regressions

Tables 2 and 3 contain regression results where the dependent variable is survival or growth (respectively), and find similar results for the general absence of any effect of SRS participation on performance.

Table 2.

Logit duration regressions for the determinants of survival.

	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)
	Full	Year 3	Year 4	Year 5	Year 6	Matched sample (SRS propensity)
						Year 3	Year 4	Year 5	Year 6
Log turnover (lagged)	0.0521** (0.0225)	0.318*** (0.0321)	0.314*** (0.0363)	0.331*** (0.0452)	0.478*** (0.0454)	0.345*** (0.0372)	0.324*** (0.0413)	0.313*** (0.0507)	0.500*** (0.0532)
Turnover growth (lagged)	0.285*** (0.0349)
Age	0.0611*** (0.0170)	0.0356 (0.0252)	0.0362 (0.0334)	0.0828** (0.0372)	0.0434 (0.0458)	0.0592* (0.0312)	0.0657 (0.0405)	0.0845* (0.0470)	−0.0349 (0.0609)
Age-squared	−0.000586*** (0.000202)	−0.000298 (0.000297)	−0.000315 (0.000398)	−0.000721 (0.000446)	−0.000493 (0.000531)	−0.000568 (0.000380)	−0.000689 (0.000492)	−0.000795 (0.000578)	0.000398 (0.000729)
Education dummies
GCSE	−0.00403 (0.0718)	−0.0243 (0.110)	−0.115 (0.142)	0.221 (0.157)	0.0546 (0.178)	0.378** (0.191)	−0.190 (0.268)	0.234 (0.275)	−0.249 (0.386)
A-level	−0.0676 (0.0840)	−0.162 (0.127)	−0.113 (0.166)	−0.0185 (0.182)	0.0363 (0.219)	0.181 (0.204)	−0.264 (0.286)	−0.0877 (0.293)	−0.284 (0.409)
Degree or higher	−0.0358 (0.0790)	−0.0758 (0.124)	−0.349** (0.152)	0.000959 (0.171)	−0.0904 (0.197)	0.243 (0.201)	−0.523* (0.279)	−0.00976 (0.286)	−0.452 (0.392)
No business experience	−0.0265 (0.0809)	−0.0981 (0.123)	0.0676 (0.160)	0.424** (0.188)	0.172 (0.200)	−0.0430 (0.141)	0.128 (0.186)	0.439** (0.203)	0.342 (0.246)
Sources of advice
EABL	0.0710 (0.0919)	0.147 (0.140)	−0.0248 (0.165)	0.295 (0.207)	0.365 (0.245)	0.0671 (0.154)	−0.0615 (0.181)	0.283 (0.237)	0.383 (0.282)
Accountant	0.0411 (0.0569)	0.00303 (0.0882)	0.108 (0.110)	−0.0748 (0.124)	0.0860 (0.141)	−0.0653 (0.101)	0.116 (0.129)	0.0616 (0.146)	−0.0181 (0.167)
Solicitor	−0.109 (0.122)	−0.246 (0.181)	−0.195 (0.228)	0.0587 (0.285)	−0.0217 (0.292)	−0.195 (0.203)	−0.187 (0.263)	−0.105 (0.311)	0.0301 (0.338)
College	0.197 (0.138)	0.532** (0.236)	0.151 (0.243)	−0.164 (0.293)	0.581 (0.373)	0.738*** (0.266)	0.208 (0.258)	−0.129 (0.320)	0.883* (0.453)
SRS	−0.0855 (0.355)	−0.166 (0.507)	0.795 (0.867)	−0.426 (0.679)	−1.465** (0.745)	−0.255 (0.502)	0.712 (0.836)	−0.459 (0.693)	−1.550** (0.719)
PYBT	−0.0228 (0.262)	0.936* (0.553)	−0.940*** (0.362)	0.827 (0.742)	−0.721 (0.543)	1.041* (0.623)	−1.108*** (0.417)	1.152 (0.996)	−1.496** (0.668)
Family	0.00346 (0.0599)	0.0584 (0.0921)	−0.106 (0.112)	0.267* (0.137)	−0.240 (0.147)	−0.00867 (0.105)	−0.116 (0.128)	0.272* (0.157)	−0.382** (0.173)
Other	−0.110 (0.109)	0.0623 (0.161)	−0.327* (0.184)	0.0667 (0.266)	0.0288 (0.297)	−0.00558 (0.181)	−0.490** (0.200)	0.0566 (0.315)	0.340 (0.378)
Volatility (lagged)	−0.472*** (0.0315)	0.0927 (0.0589)	−0.0417 (0.0642)	−0.0680 (0.0742)	0.0755 (0.0832)	0.0980 (0.0689)	0.0363 (0.0768)	−0.0550 (0.0887)	0.0817 (0.0977)
Overdraft excess (lagged)	−0.118*** (0.0401)	−0.250*** (0.0658)	−0.128* (0.0769)	−0.255*** (0.0860)	−0.247** (0.103)	−0.168** (0.0782)	−0.112 (0.0896)	−0.239** (0.102)	−0.240* (0.124)
(Continued)
OD XS duration (lagged)	−0.815*** (0.0934)	−0.658** (0.310)	−0.483** (0.242)	0.434* (0.237)	−0.198 (0.260)	−0.563 (0.358)	−0.430 (0.284)	0.538* (0.289)	−0.256 (0.286)
Authorised OD use (lagged)	0.145*** (0.0487)	−0.0204 (0.0987)	−0.0938 (0.0890)	0.147 (0.108)	−0.213* (0.109)	−0.0814 (0.116)	−0.0971 (0.104)	0.235* (0.130)	−0.204 (0.134)
Extent of auth. OD use (lagged)	−0.315*** (0.0999)	0.295 (0.296)	−0.183 (0.195)	−0.207 (0.235)	−0.217 (0.223)	−0.0572 (0.347)	−0.290 (0.228)	−0.390 (0.275)	−0.247 (0.282)
No. owners	0.107 (0.0798)	−0.0288 (0.127)	0.103 (0.158)	0.320* (0.183)	−0.376** (0.181)	0.189 (0.165)	0.109 (0.188)	0.254 (0.225)	−0.367 (0.227)
Male owner(s)	0.166** (0.0713)	0.122 (0.108)	−0.110 (0.138)	0.226 (0.155)	0.130 (0.193)	−0.0876 (0.122)	−0.153 (0.150)	0.154 (0.173)	0.0670 (0.212)
Legal form dummies
Partnership	−0.476*** (0.0858)	−0.402*** (0.132)	−0.510*** (0.163)	−0.113 (0.197)	0.132 (0.235)	−0.344** (0.152)	−0.452** (0.189)	−0.256 (0.226)	0.233 (0.281)
Sole trader	−0.115 (0.0704)	−0.0457 (0.107)	−0.00810 (0.136)	0.0953 (0.148)	0.578*** (0.175)	−0.0313 (0.126)	−0.0149 (0.157)	0.0655 (0.177)	0.674*** (0.206)
Industry dummies	yes	yes	yes	yes	yes	yes	yes	yes	yes
Region dummies	yes	yes	yes	yes	yes	yes	yes	yes	yes
Constant	0.631 (0.493)	−2.600*** (0.742)	−1.983** (0.913)	−4.144*** (1.073)	−4.187*** (1.213)	−3.491*** (0.838)	−2.265** (1.082)	−4.354*** (1.280)	−2.787* (1.461)
Observations	11,636	3,956	3,073	2,510	2,097	2,873	2,215	1,784	1,483
Pseudo R-squared	0.1244	0.1183	0.1340	0.1234	0.1671	0.1209	0.1341	0.1193	0.1819

Treatment (SRS participation) occurs around the time of start-up, and we investigate its effects on survival into years 3–6, where years are either pooled together (results for the full sample in column (1) labelled ‘Full’) or years are taken individually (Columns (2)–(9)). Robust standard errors in parentheses.

***

p < 0.01; **p < 0.05; *p < 0.1.

Table 3.

OLS regressions of the determinants of turnover growth.

	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)
	Full	Year 3	Year 4	Year 5	Year 6	Matched sample (SRS propensity)
						Year 3	Year 4	Year 5	Year 6
Log turnover (lagged)	−0.115*** (0.0115)	−0.110*** (0.0164)	−0.0982*** (0.0200)	−0.174*** (0.0372)	−0.0862*** (0.0190)	−0.121*** (0.0203)	−0.0970*** (0.0248)	−0.206*** (0.0459)	−0.0786*** (0.0229)
Turnover growth (lagged)	−0.0313 (0.0255)
Age	0.00592 (0.00569)	0.00777 (0.0103)	0.0106 (0.0112)	−0.000285 (0.0118)	0.00992 (0.0121)	0.00884 (0.0136)	−0.00154 (0.0153)	0.00461 (0.0172)	0.00342 (0.0167)
Age-squared	−8.83e–05 (6.60e–05)	−0.000111 (0.000121)	−0.000117 (0.000128)	−3.50e–05 (0.000138)	−0.000136 (0.000141)	−0.000150 (0.000166)	3.48e–05 (0.000184)	−9.40e–05 (0.000210)	−1.97e–05 (0.000200)
Education dummies
GCSE	−0.00355 (0.0225)	−0.0680* (0.0400)	0.00256 (0.0431)	0.0622 (0.0499)	0.0223 (0.0471)	−0.106 (0.0732)	0.0809 (0.0878)	0.0224 (0.113)	0.101 (0.123)
A-level	−0.000606 (0.0279)	−0.0303 (0.0467)	0.0307 (0.0516)	0.108* (0.0643)	−0.129** (0.0625)	−0.0580 (0.0782)	0.0953 (0.0947)	0.0619 (0.123)	−0.0306 (0.129)
Degree or higher	−0.0111 (0.0268)	−0.109** (0.0468)	−0.0749 (0.0534)	0.181*** (0.0577)	0.00867 (0.0572)	−0.136* (0.0797)	−0.0109 (0.0969)	0.133 (0.119)	0.104 (0.130)
No business experience	−0.0263 (0.0272)	−0.0379 (0.0490)	0.0757 (0.0519)	−0.105 (0.0642)	−0.0455 (0.0496)	−0.0401 (0.0580)	0.0848 (0.0611)	−0.101 (0.0780)	−0.0359 (0.0581)
Sources of advice
EABL	0.0281 (0.0322)	0.0531 (0.0537)	−0.0136 (0.0734)	0.0452 (0.0667)	0.00940 (0.0626)	0.0634 (0.0608)	0.0207 (0.0873)	0.0495 (0.0817)	0.0631 (0.0637)
Accountant	−0.0101 (0.0193)	0.0302 (0.0339)	−0.0678* (0.0369)	0.0391 (0.0426)	−0.0490 (0.0414)	0.0380 (0.0420)	−0.0666 (0.0470)	0.0463 (0.0531)	−0.0538 (0.0506)
Solicitor	0.00852 (0.0458)	−0.0863 (0.0828)	0.171** (0.0811)	−0.144 (0.103)	0.128 (0.101)	−0.115 (0.0976)	0.204** (0.0994)	−0.0990 (0.126)	0.172* (0.0989)
College	0.0840** (0.0361)	0.107 (0.0699)	0.0720 (0.0635)	0.0767 (0.0791)	0.0783 (0.0770)	0.120 (0.0784)	0.116* (0.0620)	0.0972 (0.0868)	0.0732 (0.0789)
SRS	−0.0789 (0.139)	−0.00422 (0.234)	−0.210 (0.345)	−0.259** (0.132)	0.306* (0.179)	0.00923 (0.237)	−0.225 (0.345)	−0.310** (0.139)	0.299 (0.183)
PYBT	−0.266** (0.125)	−0.241 (0.197)	−0.189 (0.247)	−0.666** (0.274)	0.243 (0.165)	−0.485** (0.202)	−0.212 (0.363)	−0.823*** (0.294)	0.180 (0.236)
Family	−0.00954 (0.0199)	0.0525 (0.0356)	0.0107 (0.0370)	−0.0856** (0.0430)	−0.0629 (0.0457)	0.0386 (0.0433)	−0.00812 (0.0449)	−0.0900* (0.0526)	−0.0416 (0.0569)
Other	−0.0224 (0.0392)	−0.0601 (0.0695)	−0.0778 (0.0707)	0.142* (0.0832)	−0.0882 (0.0958)	−0.0821 (0.0850)	−0.0973 (0.0904)	0.100 (0.107)	−0.167 (0.120)
Volatility (lagged)	−0.165*** (0.0190)	−0.161*** (0.0330)	−0.123*** (0.0373)	−0.216*** (0.0399)	−0.151*** (0.0371)	−0.173*** (0.0396)	−0.123*** (0.0456)	−0.209*** (0.0497)	−0.152*** (0.0423)
Overdraft excess (lagged)	−0.0137 (0.0158)	−0.0429 (0.0263)	−0.0140 (0.0328)	0.00448 (0.0330)	−0.000759 (0.0362)	−0.0359 (0.0325)	−0.00605 (0.0405)	0.00562 (0.0421)	−0.000619 (0.0406)
(Continued)
OD XS duration (lagged)	−0.190*** (0.0714)	−0.129 (0.114)	−0.136 (0.119)	−0.101 (0.156)	−0.372** (0.179)	−0.287** (0.132)	−0.0795 (0.140)	−0.0312 (0.201)	−0.394** (0.171)
Authorised OD use (lagged)	0.0383*** (0.0143)	0.0303 (0.0255)	0.0369 (0.0240)	0.0619* (0.0359)	0.0307 (0.0315)	0.0530 (0.0324)	0.0284 (0.0289)	0.0842* (0.0466)	0.0189 (0.0395)
Extent of auth. OD use (lagged)	−0.117*** (0.0310)	−0.0813 (0.0587)	−0.110* (0.0569)	−0.161** (0.0695)	−0.0955 (0.0669)	−0.120 (0.0738)	−0.145* (0.0740)	−0.234** (0.0920)	−0.0472 (0.0802)
No. owners	0.0134 (0.0269)	0.0787 (0.0501)	−0.00982 (0.0504)	−0.0380 (0.0601)	0.00809 (0.0525)	0.0697 (0.0691)	−0.0397 (0.0689)	0.0278 (0.0784)	−0.0307 (0.0613)
Male owner(s)	0.00634 (0.0249)	−0.00667 (0.0444)	0.0344 (0.0511)	0.0353 (0.0540)	−0.0277 (0.0511)	−0.0148 (0.0505)	0.00843 (0.0541)	0.0347 (0.0630)	−0.0136 (0.0565)
Legal form dummies
Partnership	−0.0977*** (0.0294)	−0.0579 (0.0509)	−0.0883* (0.0503)	−0.205*** (0.0690)	−0.0806 (0.0701)	−0.0419 (0.0624)	−0.0682 (0.0606)	−0.196** (0.0814)	−0.0583 (0.0872)
Sole trader	−0.144*** (0.0253)	−0.130*** (0.0416)	−0.192*** (0.0474)	−0.188*** (0.0626)	−0.0583 (0.0555)	−0.0939* (0.0520)	−0.227*** (0.0596)	−0.221*** (0.0781)	−0.0382 (0.0667)
Industry dummies	yes	yes	yes	yes	yes	yes	yes	yes	yes
Region dummies	yes	yes	yes	yes	yes	yes	yes	yes	yes
Constant	1.258*** (0.194)	1.199*** (0.318)	0.886** (0.354)	2.152*** (0.499)	0.802** (0.380)	1.361*** (0.405)	1.068** (0.531)	2.396*** (0.687)	0.869* (0.483)
Observations	9,443	3,073	2,510	2,097	1,763	2,215	1,784	1,483	1,245
R-squared	0.045	0.053	0.041	0.099	0.065	0.059	0.044	0.116	0.067

Treatment (SRS participation) occurs around the time of start-up, and we investigate its effects on growth (t), where t denotes the year. Years are either pooled together (results for the full sample in column (1) labelled ‘Full’) or years are taken individually (Columns (2)–(9)). Robust standard errors in parentheses.

***

p < 0.01; **p < 0.05; *p < 0.1.

Table 2 contains probit regressions of the determinants of survival. Starting with column (1), we see that for the full sample (cross-sections pooled together), survival is not significantly associated with SRS participation. However, there are a number of factors that are significantly associated with survival, such as lagged size, lagged growth, age (non-linear effect), gender, legal form, as well as the bank account variables (volatility, overdraft behaviour). Columns (2)–(5) in Table 2 repeat the probit regressions on individual years and obtain similar results, SRS participation is generally insignificant although it is negatively associated with survival in the sixth year. Columns (6)–(9) of Table 2 (and also Table 3) contain results whereby efforts are made to ensure that the treatment group (SRS participation) closely resembles the control group (non-participants). This is done as it has been argued that evaluations of entrepreneurship training and education should pay more attention to issues of self-selection and the need for an appropriate control group (Duval-Couetil, 2013). For the matched sample, we begin by calculating the propensity score for a venture’s founder to participate in SRS, and drop all those observations that correspond to SRS non-participants that do not resemble SRS participants, in order to get more accurate estimates of the causal effect of SRS participation on survival. More specifically, we calculate the propensity score of being an SRS participant (using a probit regression with SRS participation as a binary dependent variable), and create an appropriate control group from the non-SRS firms by removing those non-SRS firms that do not lie on the area of common support with respect to the SRS propensity score (Table 6 in Appendix 1 shows details of the propensity-score regression, and Figure 1 in Appendix 1 shows the distribution of the SRS propensity for switchers and non-switchers). Columns (6)–(9) present results that are similar to those obtained previously with regard to the determinants of survival. Of particular interest is the general absence of an effect of SRS participation on survival (although again there is a negative significant effect in year 6).

Table 3 investigates the determinants of turnover growth. Results for the full sample (column (1)) and for individual years (columns (2)–(5), see also columns (6)–(9) for the matched sample) observe that faster growing ventures differ according to a number of factors such as lagged size, bank account volatility and overdraft behaviour, legal form. Our main finding from Table 3, however, is that SRS participation has no clear effect on growth. In most cases, SRS participation has no significant effect, and in two of the three cases where we do find a significant effect, this effect is negative.

Matching estimates

PSM

An ideal test of the effects of SRS participation on performance would be a RCT, whereby candidates are randomly placed into treatment and control groups (i.e. SRS participation or non-participation), and are forced to comply with their allocated treatment, and then their subsequent performance is compared across treatment/control groups. However, RCTs in economic contexts are expensive and complex to set up, and may even be unethical (Imbens and Wooldridge, 2009), and so we will take an alternative approach. Instead of RCT-style experimental data, our analysis exploits observational data, where we can observe the choices made by individuals, as well as some of their characteristics and performance outcomes, but we cannot be sure what made them choose their treatment status (i.e. whether or not to attend the SRS). Hence, we cannot rule out self-selection bias (Heckman et al., 1996). SRS participants may differ from non-participants in systematic ways (i.e. in terms of observed characteristics, and also probably in terms of unobserved characteristics), that may also be related to their subsequent performance, which may confound the causal chain from treatment to performance outcomes.

We are interested in comparing the performance outcomes Y_i for entrepreneur i after receiving the treatment (i.e. Y_i(1)) with the performance outcome for the same entrepreneur had they not received the treatment, that is, Y_i(0), to obtain a treatment effect τ_i:

τ_{i} = Y_{i} (1) - Y_{i} (0)

However, we cannot observe both Y_i(1) and Y_i(0) for the same individual; instead, one of these states will be observed while the other will be the ‘counterfactual’. This problem can be addressed by calculating the (population-level) Average Treatment Effect (ATE):

τ_{ATE} = E (τ) = E [Y (1) - Y (0)]

However, a drawback of this estimate is that individuals are heterogeneous, and instead, some individuals may self-select into the treatment group. Instead, one may prefer the Average Treatment Effect on the Treated (ATT), which focuses on those individuals who actually did receive the treatment (i.e. SRS participants):

τ_{ATT} = E (τ | D = 1) = E [Y (1) | D = 1] - E [Y (0) | D = 1]

where the treatment indicator D equals 1 if the individual participated in SRS, and 0 otherwise. Unbiased estimates of E[Y(1)|D = 1] can be calculated by observing the outcomes of SRS participants. Unbiased estimates of the counterfactual, E[Y(0)|D = 1], are not possible because of a missing information problem (Pons Rotger et al., 2012) because we cannot observe individuals who chose to participate in SRS but ultimately did not participate. Instead, we observe individuals who didn’t choose SRS participation and didn’t participate in SRS, which is problematic since E[Y(0)|D = 1] ≠ E[Y(0)|D = 0].

PSM allows identification of τ_ATT using two main assumptions (Caliendo and Kopeinig, 2008). The first is ‘selection on observables’ which maintains that performance outcomes and SRS participation are independent for individuals with the same characteristics X. Under this assumption, we have E[Y(0)|D = 1, X = x] = E[Y(0)|D = 0, X = x] which provides us with a counterfactual. Put differently, we assume that all characteristics X that may affect SRS participation and also performance outcomes simultaneously are observed in our data. Unobserved characteristics are assumed to have no influence on SRS participation and performance outcomes, which may not be entirely realistic (and cannot even be verified). PSM’s second main assumption is the ‘overlap’ assumption, which maintains that SRS participants will ‘overlap’ in characteristics space with non-participants, and are hence comparable according to observed characteristics. This second assumption is verified in the following analysis by way of verifying overlap in propensity scores. For more on PSM, see, for example, Caliendo and Kopeinig (2008).

Matching results

We begin by looking at the effect of SRS on firm performance, measured in terms of survival and turnover growth, using PSM (Table 4). In most cases, SRS participation has no benefits regarding survival or growth – the estimates are usually negative but not statistically significant. For the survival estimates for the full sample, however, the effect of SRS performance on growth is significantly negative (t-stat = −1.96). Table 5 contains multidimensional nearest-neighbour matching estimates (following Abadie et al., 2004) providing further support for our PSM results. Taking together our regression estimates and our matching estimates for the SRS treatment effect, we find no evidence that SRS has any effect on survival or growth, which offers no support to our Hypotheses 1a and 1b.

Table 4.

PSM estimates for the Average Treatment effect on the Treated (ATT) of SRS participation on survival and growth.

	Coefficient	SE	t-statistic	Total obs. on support
Survival
Full sample	−0.091	0.047	−1.96	16,241
Year 2	−0.092	0.080	−1.15	5,038
Year 3	−0.096	0.099	−0.98	3,339
Year 4	0.100	0.087	1.16	1,907
Year 5	−0.074	0.126	−0.59	1,554
Year 6	−0.301	0.177	−1.70	982
Growth
Full sample	−0.180	0.126	−1.43	11,431
Year 2	−0.240	0.242	−0.99	3,338
Year 3	−0.023	0.251	−0.09	1,905
Year 4	−0.244	0.375	−0.65	1,554
Year 5	−0.148	0.188	−0.79	982
Year 6	0.210	0.268	0.78	44

Treatment (SRS participation) occurs around the time of start-up, and we investigate its effects on survival (t) or growth (t), where t denotes the year since entry. Years are either pooled together (results for the full sample) or years are taken individually.

Matching covariates are lagged log turnover, age, age-squared, education dummies, prior business experience dummy, lagged values of trading variables (account volatility; authorised overdraft excess dummy and amount; unauthorised overdraft excess dummy and amount); number of owners, gender of owner(s) and dummies for legal form, industry and region.

Table 5.

Multidimensional nearest-neighbour matching (following Abadie et al., 2004) estimates for the Sample Average Treatment Effect (SATE) of SRS participation on survival and growth.

	Coefficient	SE	z-stat	p-value	Observations
Survival
Full sample	−0.070	0.056	−1.26	0.209	16,874
Year 2	−0.049	0.107	−0.46	0.647	5238
Year 3	0.024	0.118	0.20	0.839	3956
Year 4	0.075	0.117	0.64	0.521	3073
Year 5	−0.079	0.143	−0.56	0.578	2510
Year 6	−0.194	0.168	−1.16	0.248	2097
Growth
Full sample	−0.035	0.132	−0.26	0.793	13,399
Year 2	0.077	0.225	0.34	0.731	3956
Year 3	0.121	0.296	0.41	0.684	3073
Year 4	0.023	0.339	0.07	0.946	2510
Year 5	−0.244	0.295	−0.83	0.407	2097
Year 6	0.300	0.292	1.03	0.304	1763

SRS and customer behaviour

We now examine Hypothesis H2 – whether an individual switches their bank account to a rival bank. Of the 6198 businesses that did not participate in SRS, 424 switched bank account during our period of observation (i.e. their first 6 years). This corresponds to 6.84% of the sample, which we take as our baseline frequency of switching. There were 49 businesses that participated in the SRS. If the frequency of switchers were the same as for non-SRS firms, we would expect 49 × 0.0684 = 3.352 switchers. In fact, among SRS firms, there were zero switchers. The unconditional probability of this happening at random is (1 − 0.0684)⁴⁹ = 0.031048. Hence, the proportion of switching among SRS participants is less than expected, and the difference is statistically significant at the 5% level.

However, there may be systematic differences between SRS and non-SRS participants that may be driving this different switching behaviour. Omitted variable bias may make a comparison of unconditional expectations unreliable. Thus, we seek to control for other influences. This cannot be done in a regression framework (e.g. probit with probability of switching as the dependent variable) because SRS participation perfectly predicts non-switching. Instead, we calculate the propensity score of being an SRS participant (using a probit regression with SRS participation as a binary dependent variable), and create an appropriate control group from the non-SRS firms by removing those non-SRS firms that do not lie on the area of common support with respect to the SRS propensity score (Table 6 in Appendix 1 shows details of the propensity-score regression, and Figure 1 in Appendix 1 shows the distribution of the SRS propensity for switchers and non-switchers). Removing the members of the control group of non-SRS that are poor matches to SRS participants from among the non-SRS firms leaves us with 4278 comparable non-SRS firms, of which 310 (7.25%) were switchers. We then apply this frequency of 0.0725 to the 48 firms in the treatment group (SRS firms), leading us to expect 3.478 of these SRS firms to switch. In fact, there were 0 switchers among SRS firms. The probability of this occurring is (1 − 0.0725)⁴⁸ = 0.0270. We therefore conclude that SRS participants were significantly less likely to switch bank accounts (results significant at the 5% level). Therefore, our results provide support to Hypothesis 2.

Taken together, we find that the ventures of SRS participants do not have different survival or growth outcomes (no support for H1a or H1b), although they do have a lower probability of switching to a rival bank (support for H2).

Conclusion

This article has addressed a core contradiction at the heart of much theoretical and empirical research on new ventures. It is that new venture owners who have participated in training report its impact positively (OECD, 2007, 2013), yet several rigorous statistical analyses using objective measures of performance find little impact. Our results suggest that it is possible to be satisfied with a training scheme even if there are no significant performance benefits. Relatedly, firm performance outcomes may be much more random, and hard to influence, than satisfaction levels. The case made here is that the heavy use of case material, relying on self-report data, influences the respondents to view the impact of training in a positive light because the training enhances a sense of loyalty from the trainee. This loyalty is reflected in both a positive self-report response but also in our case of bank-provided training, a reduced likelihood of switching to another bank. Although we have no direct observations on emotions, we nevertheless interpret this as suggestive of an emotional attachment to a mentor.

The findings can be considered as a reflection of the two very different approaches to offering insights into entrepreneurship and particularly new ventures. On one hand, the econometric approach using actual performance data fails to show an impact of training on either survival or growth of new ventures. On the other hand, in contrast to this hegemonic rational perspective, we find, even without resorting to self-report data, evidence that suggests an influence at an emotional level. Training appears to enhance the bond between training provider and trainee, leading to the development of loyalty, which results in a lower observed likelihood of the venture transferring its accounts to another bank. It therefore, is an example of an apparently illogical decision; reporting positively on training provision when there is no performance impact but this is better understood once emotional factors are included. Appendix 2 discusses how Barclays Bank has reacted to these findings to strategically reorient its entrepreneurship seminars.

We recognise of course that even this dataset has its limitations. First, we cannot entirely rule out the possibility of endogeneity between SRS participation and probability of switching bank, although we have no strong a priori reason to believe that those who self-select into SRS participation have a lower probability of switching (conditional on other observed characteristics). The ideal test, from an econometrician’s perspective, is that used by Georgiadis and Pitelis (2014) where individuals are randomly allocated to participate or not in the training seminar, and then forced to comply with the random allocation outcome. Given that our analysis was undertaken as such, it is possible that some bias may remain in our estimates – for example, if individuals who self-select into SRS participation are more likely to benefit (in which case we would see an upward bias) or if those ‘pessimists’ who see no benefits might actually benefit more than anticipated (in which case we would see a negative bias). A second limitation is that we only observe a relatively small number of SRS participants. Although we observe that zero out of 49 participants switched, a larger sample size might have shown stronger results (e.g. if zero out of a thousand participants switched).

Further work might also compare the ‘treatment’ and ‘control’ groups with a ‘placebo’ group, which receives no substantive knowledge or skills, but gets the emotional benefits (confidence, optimism) and certification effects from attending such a course. However, the unifying requirement of further work has to be to use appropriate methods and data (Greene, 2009; Meager et al., 2003) to adequately test this core tenet of entrepreneurship theory – that of a link between knowledge enhancement and new venture performance.

Footnotes

Appendix 1

Appendix 2 Acknowledgements

We are grateful to Pelin Demirel, Elena Kulchina, Cher Li, Kristian Nielsen and seminar participants at Nottingham University Business School, as well as Sue Marlow (Editor) and our three anonymous reviewers for many helpful comments. J.S.F. and R.G.R. write only in a personal capacity and do not necessarily reflect the views of Barclays Bank. Any remaining errors are ours alone.

Funding

Alex Coad gratefully acknowledges financial support from the Economic and Social Research Council (ESRC), Technology Strategy Board (TSB), Department for Business, Innovation and Skills (BIS) and NESTA on grants ES/H008705/1 and ES/J008427/1 as part of the IRC distributed projects initiative, as well as from the AHRC as part of the FUSE project.

Notes

Author biographies

Alex Coad is an Economic Analyst at the European Commission (Institute for Prospective Technological Studies, Seville, Spain). His research interests focus on the economics of innovation and entrepreneurship. He is an Associate Editor at Research Policy.

Julian S Frankish is Head of Business Economics and Research for Barclays Corporate Banking.

Richard G Roberts is Hon. Professor of Small Business Research, University of Birmingham. Prior to retirement in 2014, he was SME Market Analysis Director at Barclays Bank.

David J Storey is in the School of Business, Management and Economics at the University of Sussex. His research interest is in new and small firms.

References

Abadie

Drukker

Herr

. (2004) Implementing matching estimators for average treatment effects in Stata. The Stata Journal 4(3): 290–311.

Baron

(2008) The role of affect in the entrepreneurship process. Academy of Management Review 33(2): 328–340.

Berger

Udell

(1998) The economics of small business finance: The roles of private equity and debt markets in the financial growth cycle. Journal of Banking and Finance 22(6–8): 613–673.

Bertrand

Mullainathan

(2001) Do people mean what they say? Implications for subjective survey data. The American Economic Review 91(2) (Papers and Proceedings of the Hundred Thirteenth Annual Meeting of the American Economic Association (May 2001), pp.67–72).

Braguinsky

Klepper

Ohyama

(2012) High-tech entrepreneurship. Journal of Law and Economics 55(4): 869–900.

Bryan

(2006) Training and performance in small firms. International Small Business Journal 24(6): 635–660.

Burke

Hutchins

(2007) Training transfer: An integrative review. Human Resource Development Review 6(3): 263–296.

Caliendo

Kopeinig

(2008) Some practical guidance for the implementation of propensity score matching. Journal of Economic Surveys 22(1): 31–72.

Caliendo

Künn

(2011) Start-up subsidies for the unemployed: Long-term evidence and effect heterogeneity. Journal of Public Economics 95(3–4): 311–331.

10.

Cardon

Foo

M-D

Shepherd

. (2012) Exploring the heart: Entrepreneurial emotion is a hot topic. Entrepreneurship Theory and Practice 36: 1–12.

11.

Cardon

Gregoire

Stevens

. (2013) Measuring entrepreneurial passion: Conceptual foundations and scale validation. Journal of Business Venturing 28: 373–396.

12.

Cardon

Wincent

Singh

. (2009) The nature and experience of entrepreneurial passion. Academy of Management Review 34(3): 511–532.

13.

Cardon

Zietsma

Saparito

Matherne

. (2005) A tale of passion: New insights into entrepreneurship from a parenthood metaphor. Journal of Business Venturing 20(1): 23–45.

14.

Chrisman

Katrishen

(1994) The economic impact of small business development center counseling activities in the United States: 1990-1991. Journal of Business Venturing 9: 271–280.

15.

Chrisman

McMullan

(2002) Some additional comments on the sources and measurements of the benefits of small business assistance programs. Journal of Small Business Management 40: 43–50.

16.

Chrisman

McMullan

(2004) Outsider assistance as a knowledge resource for new venture survival. Journal of Small Business Management 42: 229–244.

17.

Coad

(2009) The Growth of Firms: A Survey of Theories and Empirical Evidence. Cheltenham and Northampton, MA: Edward Elgar.

18.

Coad

Frankish

Nightingale

. (2014) Business experience and start-up size: Buying more lottery tickets next time around? Small Business Economics 43: 529–547.

19.

Coad

Frankish

Roberts

. (2013) Growth paths and survival chances: An application of Gambler’s Ruin theory. Journal of Business Venturing 28: 615–632.

20.

Coleman

(2007) The role of human and financial capital in the profitability and growth of women-owned small firms. Journal of Small Business Management 45(3): 303–319.

21.

Cooper

(1993) Challenges in predicting new firm performance. Journal of Business Venturing 8(3): 241–253.

22.

Denrell

Fang

Liu

(2015) Chance Explanations in the Management Sciences. Organization Science, 26(3): 923–940.

23.

Duval-Couetil

(2013) Assessing the impact of entrepreneurship education programs: Challenges and approaches. Journal of Small Business Management 51(3): 394–409.

24.

Foo

M-D

(2011) Emotions and entrepreneurial opportunity evaluation. Entrepreneurship Theory and Practice 35: 375–393.

25.

Frankish

Roberts

Coad

. (2013) Do entrepreneurs really learn? Or do they just tell us that they do? Industrial and Corporate Change 22(1): 73–106.

26.

Georgiadis

Pitelis

(2014) The impact of employees’ and managers’ training on the performance of small- and medium-sized enterprises: Evidence from a randomized natural experiment in the UK service sector. British Journal of Industrial Relations. Epub ahead of print 4 August. DOI: 10.1111/bjir.12094.

27.

Geroski

(2000) The growth of firms in theory and practice. In: Foss

Mahnke

(eds) Competence, Governance, and Entrepreneurship. Oxford: Oxford University Press, pp.168–186.

28.

Gottschalk

Greene

Höwer

. (2013) If You Don’t Succeed, Should You Try Again? The Role of Entrepreneurial Experience in Venture Survival. Mannheim: ZEW.

29.

Greene

(2009) Assessing the impact of policy interventions: The influence of evaluation methodology. Environment and Planning C: Government and Policy 27(2): 216–229.

30.

Grichnik

Smeja

Welpe

(2010) The importance of being emotional: How do emotions affect entrepreneurial opportunity evaluation and exploitation? Journal of Economic Behavior & Organization 76: 15–29.

31.

Handler

Kram

(1988) Succession in family firms: The problem of resistance. Family Business Review 1(4): 361–381.

32.

Heckman

Ichimura

Smith

Todd

(1996) Sources of selection bias in evaluating social programs: An interpretation of conventional measures and evidence on the effectiveness of matching as a program evaluation method. Proceedings of the National Academy of Sciences 93: 13416–13420.

33.

Imbens

Wooldridge

(2009) Recent developments in the econometrics of program evaluation. Journal of Economic Literature 47(1): 5–86.

34.

Ireland

Hitt

Sirmon

(2003) A model of strategic entrepreneurship: The construct and its dimensions. Journal of Management 29(6): 963–989.

35.

Jones

Beynon

Pickernell

. (2013) Evaluating the impact of different training methods on SME business performance. Environment and Planning C: Government and Policy 31(1): 56–81.

36.

Karlan

Valdivia

(2011) Teaching entrepreneurship: Impact of business training on microfinance clients and institutions. Review of Economics and Statistics 93(2): 510–527.

37.

Klinger

Schundeln

(2011) Can entrepreneurial activity be taught? Quasi-experimental evidence from central America. World Development 39(9): 1592–1610.

38.

Levinthal

(1991) Random walks and organizational mortality. Administrative Science Quarterly 36(3): 397–420.

39.

Meager

Bates

Cowling

(2003) An evaluation of business start-up support for young people. National Institute Economic Review 186(4): 59–72.

40.

Metzger

(2006) ‘Once bitten twice shy?’ The performance of entrepreneurial re-starts. Discussion Paper 06-083. Mannheim: ZEW.

41.

Mole

Hart

Roper

. (2008) Differential gains from Business Link support and advice: A treatment effects approach. Environment and Planning C: Government and Policy 26(2): 315–334.

42.

Murnieks

Mosakowski

Cardon

(2014) Pathways of passion: Identity centrality, passion and behavior among entrepreneurs. Journal of Management 40(6): 1583–1606.

43.

Nichols

(2007) Causal inference with observational data. The Stata Journal 7(4): 507–541.

44.

Nielsen

Sarasvathy

(2011) Passive and active learning from entrepreneurship. An empirical study of re-entry and survival. DRUID Working Paper no. 11-12. Aalborg: DRUID.

45.

Norrman

Bager-Sjögren

(2010) Entrepreneurship policy to support new innovative ventures: Is it effective? International Small Business Journal 28(6): 602–619.

46.

Ongena

Smith

(2000) What determines the number of bank relationships? Cross-country evidence. Journal of Financial Intermediation 9(1): 26–56.

47.

Organisation for Economic Co-operation and Development (OECD) (2007) OECD Framework for the Evaluation of SME and Entrepreneurship Policies and Programmes. Paris: OECD Publishing.

48.

Organisation for Economic Co-operation and Development (OECD) (2013) Skills Development and Training in SMEs, Local Economic and Employment Development (LEED). Paris: OECD Publishing.

49.

Parker

(2006) Learning about the unknown: How fast do entrepreneurs adjust their beliefs? Journal of Business Venturing 21(1): 1–26.

50.

Parker

(2013) Do serial entrepreneurs run successively better-performing businesses? Journal of Business Venturing 28(5): 652–666.

51.

Politis

(2005) The process of entrepreneurial learning: A conceptual framework. Entrepreneurship Theory and Practice 29(4): 399–424.

52.

Pons Rotger

Gørtz

Storey

(2012) Assessing the effectiveness of guided preparation for new venture creation and performance: Theory and practice. Journal of Business Venturing 27(4): 506–521.

53.

Rideout

Gray

(2013) Does entrepreneurship education really work? A review and methodological critique of the empirical literature on the effects of university-based entrepreneurship education. Journal of Small Business Management 51(3): 329–351.

54.

Rocha

Carneiro

Varum

(2015) Serial Entrepreneurship, learning by doing and self-selection. International Journal of Industrial Organization. 40: 91–106

55.

Shepherd

(2003) Learning from business failure: Propositions of grief recovery for the self-employed. Academy of Management Review 28(2): 318–328.

56.

Shepherd

DeTienne

(2005) Prior knowledge, potential financial reward, and opportunity identification. Entrepreneurship Theory and Practice 29: 91–112.

57.

Solomon

Bryant

May

. (2013) Survival of the fittest: Technical assistance, survival and growth of small businesses and implications for public policy. Technovation 33(8–9): 292–301.

58.

Souitaris

Zerbinati

Al-Laham

(2007) Do entrepreneurship programmes raise entrepreneurial intention of science and engineering students? The effect of learning, inspiration and resources. Journal of Business Venturing 22: 566–591.

59.

Teece

Pisano

Shuen

(1997) Dynamic capabilities and strategic management. Strategic Management Journal 18: 509–533.

60.

Thompson

(2009) Learning by doing. In: Hall

Rosenberg

(eds) Handbook of Economics of Technological Change. Amsterdam: Elsevier/North-Holland, pp. 429–476.

61.

Tornqvist

Vartia

(1985) How should relative changes be measured? The American Statistician 39(1): 43–46.

62.

Van der Sluis

van Praag

Vijverberg

(2003) Entrepreneurship Selection and Performance: A Meta Analysis of the Impact of Education in Industrialized Countries. Washington, DC: World Bank.

63.

Vissa

Chacar

(2009) Leveraging ties: The contingent value of entrepreneurial teams’ external advice networks on Indian software venture performance. Strategic Management Journal 30: 1179–1191.

64.

Welpe

Sporrle

Grichnik

. (2012) Emotions and opportunities: The interplay of opportunity evaluation, fear, joy, and anger as antecedent of entrepreneurial exploitation. Entrepreneurship Theory and Practice 36: 69–96.

65.

Wiklund

Patzelt

Shepherd

(2009) Building an integrative model of small business growth. Small Business Economics 32: 351–374.

66.

Wren

Storey

(2002) Evaluating the effect of soft business support upon small firm performance. Oxford Economic Papers 54: 334–365.

67.

Yang

Aldrich

(2012) Out of sight but not out of mind: Why failure to account for left-truncation biases research on failure rates. Journal of Business Venturing 27(4): 477–492.