Survival of the funded: Econometric analysis of startup longevity and success.

• Author: Keogh, Daniel

INTRODUCTION

There is a long literature about why new businesses fail. Yet that literature appears inadequate in light of the fact that the failure rate of new businesses is 90% (Carrigan, 2020) in the United States, where those same new businesses account for as much as 50% of new job creation year-to-year across a host of industries (Fairlie et al., 2016). That combination of facts has led to the development of new capitalization strategies, including venture capitalists (VCs) and angel investors, exclusively for high-risk investment scenarios. Our research here improves existing models of startup survival and success, adding to the literature in several critical ways. We use multiple alternative models of success, interaction terms to permit nonlinear outcome response Functions, and nuanced estimation strategies, which all contribute to a more detailed understanding of startup success via financing choices.

Many investors in startups value analytical evidence on the quantitative traits of a firm which pre-determine success, but other investors look almost exclusively at the qualitative traits of the startup and the character traits of successful founders which predict survival (Kleinert et al., 2020; Wang et al., 2019). Still, others take a special interest in specific industries, with investment decisions informed largely by their familiarity and expertise in a certain field. These investment philosophies have likened startup investing to horse racing: one can bet on the jockey (the entrepreneur), the horse (the business), or the race (the industry/market), as pointed out by Kaplan et al. (2009). In practice, successful investors are informed by all three philosophies in tandem. Given that venture capital investment in startups exceeded $300 billion in 2020 despite the pandemic (Teare, 2021), it is expensive if we place the wrong bets. While the literature has thoughtfully explored the separate contributors of success, it has been very limited in measuring the interactions between those same factors.

This paper will model and test potential complementarities between financing strategies and the personal attributes of the entrepreneurs, using survival rates as a measure of success. In other words, it might be critical to recognize that some financing strategies work best in conjunction with other inputs/factors and less well on their own. We recognize that survival (i.e., "continued existence") masks a diversity of outcomes, so we also model revenue growth and employment levels to round out the picture. Finally, this is the first paper, of which we are aware, to correct for potential sample selection bias in performing startup-survival modeling. The resulting analysis is a test of the hypothesis that it is a combination of factors that matter for startup success, that no one financing strategy is superior on its own, but rather it is nuance and context (the presence of other factors) that are critical.

LITERATURE REVIEW

As young firms, startups lack long time-trends of metrics used to evaluate older businesses; the question of how to identify and measure startup success is an argument in the economic literature (Laitinen, 2019; Baluku et al., 2016). Research has diverged into predominantly two directions: a) through the presence of successful financing under the hypothesis that investment by a competitive source is a strong signal of success (e.g., Wang et al., 2019), and b) through metrics standard in evaluating older ventures. That latter path is exemplified by studies using firm survival (Hipp and Binz, 2020), sales growth (Bednar et al., 2018), turnover (Kim, 2020), or return on equity (Laitinen, 2019). Since objective success and financing outcomes are inextricably linked, investor financing may be determined, which then creates further success either objectively or via subsequent rounds of subjective investor decisions (Kleinert et al, 2021).

Typically, Cox proportional hazard functions are used to measure new-venture survival, although with varying levels of success (Cader & Leatherman, 2011; Delmar & Shane, 2006). Other survival-time regressions are also common depending on data availability (Bosma et al., 2004), and for non-binary indicators such as revenues or employment, more conventional maximum likelihood estimation of probits, logits, and tobits are traditional (Bosma et al., 2004; Delmar & Shane, 2006). However, there is inherent bias in these non-binary regressions as data panels are invariably unbalanced with missing values from failed firms; researchers have coped with this bias in a variety of ways (e.g., Boehmke et al., 2006; Cader & Leatherman, 2011).

The proposition that a firm's financing technique can explain success is a popular thought (Baum & Silverman, 2004; Huyghebaert et al., 2007; Ahmed &Cozzarin, 2009; Yankov et al., 2014). However, if that decision is itself endogenous, a function of characteristics of the market or founder, then the story and model must become more complex. Several studies model the first endogenous stage, investment criteria, via the effects of the entrepreneur, industry, and the firm's strategy on venture success (Kleinert et al., 2020; Wang et al., 2019; Van Gelderen, 2004). Previous literature has found strong links between the entrepreneur and the firm's financing, so this begs questions as to when both are accounted for, if either of these have effects on new venture success (Sanyal & Mann, 2010; Baum & Silverman, 2004).

In a two-stage model, financial intermediaries not only select which firms get financing but influence survival and other success outcomes. Baum and Silverman (2004) describe how venture capital, for example, identifies potential and offers validation as well as the coaching and resources that a startup needs to survive: not just funding but portfolio company alliances, or advisors. However, the effects were entangled, since more funding correlated with founder characteristics, more alliances, more intellectual capital and more human or network capital, making it impossible to determine the true "causes" of success (Baum & Silverman, 2004) This opens debate about the differences between financial, human capital and social capital, and how each affect new-venture success (Bosma et al., 2004; Zankov et al., 2014; Larson, 1992). In a sense, founder traits come first, and determine the type and amount of funding that a startup may receive. In their research, Sanyal and Mann (2010) analyze how an entrepreneur's assets, communication of relevant information, and personal characteristics predict what type of financing they pursue or attain. They find that more educated entrepreneurs are more likely to pursue debt-financing while serial entrepreneurs are just as likely to self-fund, pursue external debt, or external equity due to mitigated information opacity.

Given the choice of funding, Bosma et al. (2004) quantified the effects of financing strategy, controlling for talent. They conclude that human capital (such as education or startup experience) and social capital (such as a geographic location or ties to industry professionals) play a decisive role in predicting survival, profit, and employment. Cooper et al. (1994) conclude that general human capital such as education level and demographics play a stronger role in success than managerial know-how such as past entrepreneurial experience and advisors do. However, there is active debate on their relative effects (Baum & Silverman, 2004; Bosma et al., 2004; Zankov et al., 2014). Delmar and Shane (2006) added the additional insight that the distinction between general and managerial-specific human capital varies with the age of the startup. In essence, there is a strong correlation or complementarity between factors predictive of success, with little clarity on which comes first (Bapna, 2019).

Founder identity attributes were found to be statistically significant by Banir (2014) in his paper evaluating determinants of gender differentials in the entrepreneurial space. Models that closely resemble this study include such controls whenever the entrepreneur is evaluated (Bosma et al., 2004; Sandberg & Hofer, 1987). Clearly, factors beyond financing and founder attributes must also be considered. For example, Conti et al. (2013) found that patents, especially in certain industries, are significantly and largely predictive of new venture performance. Other studies use intellectual property variables to control for novelty of a product and innovative capacity of the firm (Baum & Silverman, 2004; Sanyal & Mann, 2010). Not controlling For industry or sector may skew the results (Yankov et al., 2014). With many investors looking exclusively at specific industries, it is important to account For the fact that this selection bias in the investment process may not be explained by the venture financing variables (Sanyal & Mann, 2010; Cooper et al., 1994). Hence, we will be careful to include panel effects for each economic sector in the analysis that follows.

Building on this literature then, this paper proposes an empirical test of the hypothesis that financing strategy effectiveness is significantly dependent upon (and complementary with) founder characteristics. We hypothesize a structure for that exploration in the subsequent section, a model that reflects the complexity while attaining clear results.

METHODOLOGY

Suppose that survival is modeled as a binary outcome using the Cox proportional hazard function so that

[Please download the PDF to view the mathematical expression]

where coefficients are estimated for survival (s) on each input's separate subtypes (/), which might affect survival probabilities independently. Specifically, financing is divided into six categories: angels, equity companies, venture capital, debt, government funding, and Friends/Family/Fools (FFF) sourcing. Collaboration, or competitive advantage, is recorded as a series of four binary indicators for the...