Should I stay or should I go? Founder’s decision to leave an entrepreneurial venture during an industrial crisis

Abstract

We investigate if while experiencing an intensely negative industry-specific shock, skilled entrepreneurs may decide to leave the firm they founded, whereas founders who are less endowed with human capital may decide to continue their activity. Developing a stylised theoretical framework of the issue intended to derive the necessary and sufficient conditions for the emergence of this phenomenon, this study explores its occurrence in Italy during the ICT industry crisis from early 2000–2003 by analysing the individual stay/leave decision of a sample of 201 founders of 79 start-ups operating in the ICT services market.

Keywords:

• Entrepreneurial choice
• stay/leave decision
• high-tech entrepreneurship
• adverse selection
• founder human capital

JEL classification:

• L26
• L86
• M13

Acknowledgements

We gratefully acknowledge the support of FIRB 2003 (RBNE03ZLFW_002) and MIUR 2006 Italian national funds. We are indebted to participants to the 13th International Schumpeter Society Conference in Aalborg and the 30th Strategic Management Society Conference in Rome for helpful comments and suggestions. We are also grateful to the constructive remarks provided by the Editor and two anonymous reviewers. Responsibility for any remaining errors is solely of the authors.

Notes

¹ According to the Global Entrepreneurship Monitor (GEM) Observatory, approximately one-third of the entrepreneurs in modern advanced economies can be defined as ‘necessity entrepreneurs’; i.e. individuals who have become entrepreneurs because they have no other viable choice of occupation (Parker 2009). Evans and Leighton (1989) were among the first ones to document the numerical importance of this typology of entrepreneurs among young U.S. self-employed individuals.

² There also exist in the literature other studies that in the same vein as Wennberg et al. (2010) position themselves in-between an individual-and firm-level survival analysis. The recent work of DeTienne and Cardon (2012) represents an example. Here, the authors investigate the impact of founder human capital measured at individual level on the preferred prospective ways of exit for the firms these founders created. Their analysis highlights how different founders’ characteristics in terms of experience exert a differential impact on the modes of entrepreneurial exit.

³ For an analogous periodisation (and description) of the telecom boom and bust, see Fransman (2004) and Kam (2006). Note that the findings that will be presented later are rather insensitive to a more stringent definition of the burst period (i.e. from 2001 to 2003).

⁴ We also execute an additional analysis (see Section 6.2) on an enlarged sample of individuals including those whose firm failed or was acquired during the bust period early 2000–2003 and those individuals who were single founders and not part of a founding team. In this respect, note that Grilli (2011) performed a similar analysis on the same time span here analysed but based on firm-level data and highlighted how different levels and typologies of human capital collectively possessed by the founding teams could differently shape an ICT services firm’s exit dynamics.

⁵ Clearly, this assumption may be not valid in some industries. For example, in the case of the biotechnology sector the individuals with the highest human capital may be scientists with low entrepreneurial experience and managerial skills. However, the Italian ICT sector (which constitutes the test bed of the subsequent econometric analysis) has been characterised by massive entry during the boom period, which seems to be consistent with the existence of low entry barriers and the small incidence of star scientists on entry rates. Therefore, our assumption seems reasonable for this industry. Formally, this means that the distribution of two different types does not overlap.

⁶ Note also that the analysis is focused on an industry-specific crisis, i.e. a crisis that arises in a specific sector of the economy.

⁷ Formally, the first term of the profits function, v_e, increases with x_e. Moreover,  > v_e ∀e′ > e. Also, note that k_e(.) depends on the entrepreneurial ability of the agent, x_e. Let us assume that . This guarantees that the profits increase with human capital and/or entrepreneurial ability of the agent at any economic scenario. Furthermore, note that ρ_e may be also function of x_e.

⁸ If the best alternative option is to be an entrepreneur in an industry different from i, the value of this option depends both on the human capital and the entrepreneurial ability of the agent. That is, it should be: w_e(x_e, γ). This case can be analysed as the simpler situation described in the text: details are available on request.

⁹ We do not consider the case where an agent diversifies her/his activities (e.g. running a firm while part-time working as a salaried worker for another firm). Our data-set does not contain this information in a systematic way, and this is a limitation of our study. However, qualitative information we collected through interviews with firms’ owner-managers suggests the absence of numerically relevant hybrid activities among the Italian high-tech entrepreneurs surveyed. For an empirical analysis on hybrid entrepreneurship see Folta, Delmar, and Wennberg (2010). See also Li, Meng, and Zhang (2006) regarding the involvement of entrepreneurs in (political) activities other than simply running their firms.

¹⁰ We use the term ‘number’ for expositional simplicity, but since there is a continuum of individuals we mean ‘measure’.

¹¹ As the framework is purely micro-founded, an increase of the number of entrepreneurs may mean that new firms are founded, or that some previously non-entrepreneur agents participate with own shares in previously founded firms. Similarly, a reduction of the number of entrepreneurs may imply that some founders leave the firms they have founded (but the firms still exist), or that some firms close (when all founders leave, or the firm has a single founder and the founder exits). The empirical analysis in the next sections is primarily based on co-founded firms that were still alive after the exit of one of the founders (see later in Section 4 for details and a theoretical justification of this approach), but evidence is also provided by enlarging the focus on exited firms and single-founded start-ups (see Section 6.2).

¹² Formally, this requires that , with e ≠ e′. That is, if the sensitivity of the profits of the marginal type-e agent to the economic situation is higher than the sensitivity of the best alternative option, the same must hold for any other type. It is worth noting that this is not necessary for adverse selection to occur, but it allows deriving an easily interpretable necessary and sufficient condition for adverse selection (see later).

¹³ However, it should be noted that in the case of entry, there would be other factors affecting the decision of an agent. Indeed, as noted by Holland and Garrett (2015), a persistence decision (i.e. whether to stay or leave) is different from a start-up decision (i.e. whether to enter or stay out), as the former entails less uncertainty than the latter. This aspect is not considered in the present model.

¹⁴ Note that the model we proposed is based on the individual decision of an agent. This allows considering the relative likelihood of exit among individuals of higher versus lower human capital both within a firm and across firms. This is consistent with our sample (see the subsequent sections), where we have both homogeneous firms (where all entrepreneurs are endowed with the same level of human capital) and heterogeneous firms (where the entrepreneurs have different levels of human capital).

¹⁵ This institution registers all business activities on the basis of fiscal codes and provides (upon payment request) eventual exit information on firms and entrepreneurs over time.

¹⁶ We also excluded from the analysis those few founders who exited the start-up before the end of the boom period, because our aim is to analyse the stay/leave founder’s decision during an industrial crisis. If one may suspect that some founders could have anticipated the arrival of the crisis before its beginning, note that the inclusion in our empirical analyses of the four founders exited in 1999, leaves our findings totally unaltered.

¹⁷ On the need of controlling in empirical entrepreneurial studies with appropriate sample choices on the inherent great amount of unobserved heterogeneity at a upper level (in our case at firm level) with respect to the unit of analysis and core of the investigation (individual-level decision-making in our context) see among others, Wennberg et al. (2010, 366, 367).

¹⁸ Employment is frequently used as a proxy for firm size in firm survival studies (see e.g. Mata, Portugal, and Guimarães 1995). Different measures such as total assets (e.g. Agarwal and Audretsch 2001) or physical output (e.g. Thompson 2005) are less common. Note that the use of the total amount of capital at foundation as an alternative measure of firm size yields very similar results (available upon request from the authors) to those presented in the next paragraph.

¹⁹ Hellmann and Puri (2002) document that VC investors favour the recruitment of external managers, thus contributing to their managerial ‘professionalization’, Bottazzi, Da Rin, and Hellmann (2008) show that European VC firms helped portfolio companies in recruiting outside directors and senior managers in 40.8 and 48.4% of the deals they analyse, respectively. See also Schäfer, Werwatz, and Zimmermann (2004).

²⁰ The covariate provided by the National Association of Italian Companies (Centro Studi Confindustria) refers to 1992 and it is calculated as the average of a series of indexes: per capita value added, share of manufacturing out of total value added, employment index, per capita bank deposits, automobile–population ratio, and consumption of electric power per head.

²¹ Controlling for the role of founders after their exit from the venture would also be an important feature of the econometric analysis. Unfortunately, the current data-set does not include systematically this information. We acknowledge this as a limitation of our study.

²² For a similar application (albeit in a different context) of a panel probit model with random effects using firm (instead of time) dimension for modelling individual decisions, see Colombo and Delmastro (2004). It is worth noticing that a fixed effects model cannot be estimated since most of the independent variables are firm-specific and do not vary across founders.

²³ Note that different choices as regards the survival distribution, like a lognormal or log-logistic distribution, lead to almost identical results to those presented here. We excluded from the survival model’s specification the variable Age so to avoid any risk of a tautological correlation with the duration variable. However, it is worth noting that that its introduction leaves our findings again unchanged.

²⁴ Some theoretical justifications in terms of overall efficiency of the firm-number overshooting in entry and the subsequent exit can be found for example in Jovanovic and MacDonald (1994) and Ericson and Pakes (1995).