VC exit success : evidence from cost reducing technology shocks

(1)

Amsterdam Business School

MSc Finance: Corporate Finance Track

MSc Thesis

VC exit success:

Evidence from cost reducing technology shocks

Josca van Walsum

10210326

Thesis supervisor:

Dr. Jan Lemmen

(2)

1

Statement of originality

This document is written by Student Josca van Walsum who declares to take full responsibility for the content of this document. I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it. The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the content.

(3)

2

Acknowledgement

I would like to thank my supervisor Dr. Jan Lemmen for his guidance and feedback during the writing process. His encouraging words guided me in the right direction when needed, but still allowed for this research to be my own. Moreover I would like to thank my family, friends and fellow students, who have supported me in this process and shared ideas with me whenever I got stuck.

(4)

3

Abstract

As a result of technological shocks, the costs of starting new businesses have dropped drastically in online-and-software related markets. This has led to an increased “spray and pray” investment approach among venture capitalists, in which investors provide smaller levels of funding and governance to an increased number of startups. This thesis examines how the cost-reducing technological shocks affect VCs, by testing for changes among three indicators of exit success. Based on a sample of 14.953 US-based exits over a period of 2000 to 2012, the results show no evidence of a change in the likelihood to exit successfully (either through an IPO or trade sale). However, there is evidence of a shift from trade sales to IPOs. Furthermore, no significant results are found for the DuPont return on equity (ROE) hypothesis, implying no change in the expected return on equity of exiting ventures. Finally, the outcome of a Cox hazard analysis shows evidence of a decrease in time to exit, indicating the VC’s improved ability to exit portfolio firms at a faster rate after the shock. Although not all three indicators of success are proven to change as expected, the findings partly support the prediction that cost-reducing shocks have a positive impact on VC exit success.

(5)

4

3.2.1 Sample ... 26 3.1.2 Descriptive statistics ... 27 4. Empirical analysis ... 31 4.1 Model results ... 32 4.1.1 Hypothesis 1 ... 32 4.1.2 Hypothesis 2 ... 34 4.1.3 Hypothesis 3 ... 34 4.2 Robustness tests ... 38 4.2.1 Financial crisis ... 38 4.2.2 Alternative shocks ... 38 5. Conclusion ... 39 References ... 42 Appendix... 47

(6)

5

1. Introduction

Venture capitalists (VCs) are well-known for their investments in new, innovative businesses that aim to transform traditional markets (Wright & Robbie, 1998). Over the past 15 years the software and online industry has thrived, with increasing numbers of startups that use innovative technologies to offer new products or services. From Adyen offering electronic payment services, to Airbnb providing an online marketplace for short-term home renting, to Spotify offering an online music platform; all companies that would not have existed without today’s technology and that have revolutionized the informational ways of established firms.

In April 2014, McKinsey released a report on the growth of software and online-services and their potential to become billion-dollar giants. In the article, 3,000 software and online-services were analyzed over the period 1980-2012 of which 28% reached annual revenues of $100 million and 3% achieved annual sales of $1 billion. This indicates that only a small share of these firms has the potential of developing into world players and that VCs are presented with low probabilities of funding the next Uber or Airbnb. Additionally, technological shocks have significantly lowered the costs of starting new businesses, especially for online-related firms. Earlier work from Ewens et al. (2015) concluded that such technological shocks caused VCs to adapt their financing strategy fundamentally over the past decade. They report on an increased “spray and pray” investment approach in which investors provide little funding and limited governance to an increased number of startups which they are more likely to abandon. This leads to a disproportionate rise in innovations, where information on future prospects is revealed quickly and cheaply.

These developments raise questions about their effects on the performance and success of VCs; are the cost-reducing tech-shocks among booming online-and-software-related businesses beneficial or harmful to the success of VC-funding? Existing literature shows that VCs have unique skills to bring more expertise, experience and intensive monitoring services to their portfolio firms, compared to non VC-backed counterparts. It has been shown that this results in increased levels of performance and research & development (R&D) (Guo & Jiang, 2013), as well as reduced levels of underpricing for initial public offerings (IPOs), since capital markets recognize the quality of this monitoring as a positive signal (Barry et al., 1990). The increased “spray and pray” investment approach implies limited governance for each individual startup, diminishing these monitoring advantages and implying a disadvantageous effect for VCs on the one hand. However, beliefs about the future potential of ventures are revealed to the VC sooner and they are able to abandon unsuccessful firms quicker, since they have invested less in them and have more alternatives in their portfolio. Therefore, the increased “spray and pray” effect could on the other hand lead to higher chances of funding the next Uber or Airbnb and gaining excessive returns, whilst limiting the level of early-stage investment and risk per portfolio firm.

(7)

6 This thesis builds on the existing literature by analyzing the evolving field of VC investment in a time of technological progress, and studies whether cost-reducing technological shocks to certain industries, lead to a positive or negative effect on their exit success. The following research question is answered: How do technological shocks to the costs of starting new businesses affect the exit

success of Venture Capitalists? The outcome of this research matters for the following reasons: first,

it adds to a deeper understanding of technological changes to the cost of establishing new firms and its effects for financial intermediaries. Research on the mentioned technological shocks is limited, even though Ewens et al. (2015) show that there is significant impact on VCs and their investment strategies. Additionally, the research done by Ewens et al. (2015) is solely focused on the changing strategies of VCs, whereas this paper takes it further by looking at actual indicators of exit success. This allows for a concrete measurement of whether the general earnings model of VCs is more or less beneficial in current times of technological evolvement and whether the new funding strategies (increased “spray and pray” approach) pay off. Finally, this paper considers both exit timing and a full range of exit types (IPO, trade sale, secondary sale, write-off), whereas most existing literature mainly focusses on IPO exits. This allows for a complete analysis of “exit risk” for VC-backed firms (Giot & Schwienbacher, 2007).

The research question is answered by looking at a US sample of VC-backed firms over the period 2000-2012, using a difference-in-difference methodology to test for three hypotheses. Hereby the advent of Amazon Web Services (AWS) in 2006 is used as a technological shock; a cloud service by Amazon which dramatically lowered the initial cost of starting internet and web-based businesses (Ewens et al., 2015). The research question is tested using three measurements that together provide an overall indicator of exit success. The first hypothesis is that the AWS shock leads to an increase in exit success rate, whereby an exit is defined as successful when it is executed through an IPO or trade sale, as opposed to a secondary sale or write-off (Nahata, 2008). The second hypothesis states that the DuPont ROE at time of exit of affected firms increases after the shock. This is an indicator of return on equity and is measured as the product of a firm’s total asset turnover, return on sales (or operating profit margin) and equity multiplier (Loos, 2006; Turner et al., 2015). Finally, the third hypothesis predicts that the time to exit, defined as the number of years between initial investment and moment of exit (Nahata, 2008), decreases significantly.

The remaining part of this thesis is structured as follows. Section 2 provides a literature review on the general characteristics and effects of venture capital, the type of exit choice and its definitions of success, technological shocks and cloud computing. Section 3 explains the methodology of this research and describes the data used. Section 4 provides both the model results and additional robustness tests. The fifth and final section sets out the findings and provides recommendations and a discussion.

(8)

7

2. Literature review

2.1 Venture Capital

VCs typically invest in privately held, entrepreneurial companies. This distinguishes VC from mainstream corporate finance, which is mostly focused on the financing of investments and the undertaking of investments (Brealy & Myers, 1996). VCs are involved in the financing of new firms that are challenging the status quo or transforming the informational ways of established firms. These firms are often characterized by problems of informational asymmetry, since they are not quoted on stock markets at the time of investment. They may have the potential to achieve super-abnormal returns, but are in need of intensive monitoring due to the large levels of information asymmetry between insiders and outsiders (Wright & Robbie, 1998 and Gompers, 1995). VCs are able to take on significant investment risks through equity funding, since one extremely successful investment can offset a number of break-even or losing investments (Floriday & Kenney, 1988).

2.1.1 VC characteristics: monitoring and screening

Existing literature states that VCs possess the specific experience and skills to manage these kind of risky investments with large levels of informational symmetry, into innovative and successful companies. However, this can be both a result of active VC involvement (“monitoring”), or because VCs simply select the right firms to invest in (“screening”) (Bernstein et al., 2016). Several researchers have examined this causal relationship between monitoring and screening of VCs, and the impact on the firms they invest in; Bernstein et al. (2016) do an experiment in which they eliminate the “screening” advantage by randomly providing some companies with VC funds, whilst leaving others without. They conclude that when company selection is controlled for, VCs are still an important determinant of innovation and success. Croce et al. (2013) examine to what extent the improved performance of European VC-backed firms in high-tech industries is due to either “screening” or “value added” provided by VCs. They conclude that VC-backed firms do not exhibit a significantly different productivity growth than non-VC-backed counterparts before the first investment round (thus proving the absence of a screening effect). Conversely, after the first VC investment round, productivity growth is significantly higher for VC-backed firms, indicating a value-adding effect. Chemmanur et al. (2011) find evidence that VCs have both a screening and monitoring role in improving efficiency of their portfolio companies. They conclude that the efficiency levels of VC-backed firms before receiving venture funding are higher than for non-VC-VC-backed firms, evidence of a screening role, as well as larger growth in total factor productivity (TFP) after receiving venture funding for VC-backed firms, indicating a monitoring role. Finally, Korteweg & Sorensen (2010) examine the effects of a more experienced VC on its funded companies likelihood to go public. They develop an econometric model to distinguish the effect as a result of “influence”; VCs adding value

(9)

8 and bringing companies public at a higher rate, from “sorting”; VCs investing in better companies. They find both to be statistically and economically significant, where sorting turns out to be almost twice as important for explaining the differences in IPO rates across VC investors with different levels of experience.

2.1.2 Effects of VC on portfolio firms

Various research has been done on how the extensive monitoring, expertise, and experience of VCs affect the portfolio firms compared to non-VC-backed ventures. Guo & Jiang (2013) examine the contribution of VC investments to entrepreneurial firms in China, comparing the performance and R&D activities of VC-backed and non-VC-backed firms. They find that VC-backed firms perform better in terms of R&D investment, profitability, sales growth and labor productivity. The authors state that VCs first select firms that have higher levels of financial indicators such as profitability, sales growth, R&D investment, and then magnify these differences significantly after VC entrance (indicating both evidence of screening and monitoring).

Existing literature on the effects of VC-funding on the levels of IPO underpricing is extensive, but the outcomes are mixed. VC firms create limited partnerships to raise and invest capital. These “venture capital funds” typically have limited lifespans after which the money must be returned to the original providers. VCs can realize returns through an acquisition (Gompers, 1996), but most of the returns are gained from companies that eventually go public through an IPO (Sahlman, 1990). Gompers (1996) argues that if investors believe that successful VCs are funding ventures that eventually go public, then taking a portfolio firm public would be received as a signal that the VC is skilled at financing startup ventures. This is consistent with the grandstanding hypothesis, which predicts that younger VCs rush to take their portfolio firms public in order to establish a reputation and raise capital for new funds. Additionally, Barry et al (1990) research a set of VC-backed IPOs. They find that VCs are specialized in providing intensive monitoring services, take concentrated equity positions, take a position in the boards of their portfolio companies and maintain (part of) their investments beyond the IPO. Capital markets recognize the quality of this monitoring as a positive signal, resulting in lower IPO underpricing for VC-backed firms. Similarly, Meginson & Weiss (1991) find results indicating lower IPO underpricing and state that this is consistent with the certification role of VCs, in which they are known to recognize the true value of a company and thereby reducing underpricing. This leads to lowered total costs of capital and maximized net proceeds to the offering firm after the IPO. More recent work by Lee & Wahal (2004) contradicts these findings as they show evidence of larger IPO underpricing for VC funded IPOs. Using instruments to control for selection bias, their results show that VC-backed IPOs experience larger first-day returns, leading to increased future flows of capital into VC funds. The authors state that the

(10)

9 larger levels of underpricing increase costs to VCs, since wealth is being transferred from them to new shareholders. This effect can be explained by the grandstanding hypothesis of Gompers (1996); VCs are accepting the higher costs of underpricing since an IPO provides a signal of quality to potential investors.

Chemmanur et al. (2011) study the efficiency gains achieved by venture capitalists. They conclude that VC-backed firms have higher levels of overall efficiency than non-VC-backed firms. This is a result of both screening and monitoring, since the efficiency before VC investment is higher for VC-backed firms and the growth in efficiency subsequent to VC investment is larger for VC-backed firms as well.

Additionally, Puri & Zarutski (2012) examine the life cycle dynamics of VC-and non-VC-financed firms. Looking at both successful and failed VC-backed companies they find that the cumulative probability of failure is lower for VC-backed companies. They state that VCs have made large returns from a few successful exits, typically through IPOs, and argue that this could be due to one of two reasons. Firstly, VCs might push their portfolio firms to grow quickly and terminate firms with the least potential relatively rapidly. Secondly, judging whether a firm will be successful is difficult in the early stages of investment, therefore VCs equally nurture all portfolio companies over a certain period of time. The authors find that the lower cumulative failure rates of VC-backed companies are driven by a significantly lower likelihood of failure in the first years after receiving venture funding. These results are consistent with the second interpretation that VCs award all their investments a certain period of time to grow, but after this VCs are equally likely to shut their firms down as non-VC-financed firms.

Finally, VC-funded investments might have different risk-adjusted average returns than traded securities, as well as different betas and residual risk values. There are several reasons why investors might require a higher return for venture capital. First, private equity investments are characterized by larger levels of illiquidity for which investors want to be compensated. Second, private equity is generally held in large blocks which means that it might represent a significant part of the investor’s wealth making risk levels less spread. Thirdly, VC funds need to be compensated for the mentoring and monitoring role they typically provide to their portfolio firms (Cochrane, 2005). Chiampou & Kallett (1989) evaluate 55 privately held VC funds and find that the VC group offered high average annual returns of 17.5% as well as large standard deviations of 37.6%.

Korteweg & Sorensen (2010) address the importance of the dynamic selection problem in their research. This occurs if valuations of portfolio firms are only observed when they receive funding or are exited through either an IPO or acquisition. Both events are most likely to arise for well-performing ventures resulting into biased estimates of risk and return. The authors address this

(11)

10 problem by controlling for selection and find that it both significantly decreases the estimated returns and increases the riskiness, compared to previous studies.

The monitoring and advisory characteristics of VC funds make them capable of guiding their portfolio companies towards high returns. However, some researchers argue that this is a time-consuming role that creates tradeoffs between the magnitude of a VC’s activism and portfolio size: as the number of portfolio ventures increases, VC guidance will be spread out across more ventures and the firm’s prospects decrease. Jackson et al. (2012) examine the relation between VC activism and investment returns and test the hypothesis that an increase in the amount of portfolio firms, while closely assisting all of them, has a negative effect on investment returns. This is known as the profit destruction effect. Their results show that intense VC activism indeed leads to higher returns, however the profit destruction effect operates as well. They conclude that increases in portfolio size therefore imply risks of burdening limited VC resources and lowering returns.

2.2 Exit

The exit decision is a crucial decision in a firm’s life, both for the venture itself and the VC. For the company it will be the first time to access the public market and for the VC it is the first opportunity to liquidate some of its holdings. Understanding the aspects that determine the choice between an IPO or acquisition is therefore important for both entrepreneurs and venture capitalists (Bayar et al., 2010). In an IPO exit, a private venture typically sells part of its equity but the entrepreneurs often retain a significant stake of ownership and control. Whereas sellouts are transactions in which a public firm takes over the entire private corporation (Poulsen & Stegemoller, 2008).

2.2.1 Choice of exit type: IPO vs. acquisition

When deciding on the choice of exit of a portfolio firm, entrepreneurs and VCs may have different objectives, although existing literature provides mixed theories on these preferences. Some state that the founders of the startup prefer an IPO over a trade sale, as it allows them to retain a significant stake of ownership and control. VCs on the other hand may choose a trade sale over an IPO since it provides them with the opportunity to fully exit their ownership and realize their returns. Moreover, VCs face an additional risk during the lock-up period of an IPO, in which they are not allowed to sell all of their shares (Poulsen & Stegemoller, 2008). On the other hand, some state that an IPO is the most favorable exit type for VCs since it leads to the highest returns (Bayar & Chemmanur, 2011; Cochrane, 2005; Sahlman, 1990; Bock & Schmidt, 2015). They argue that the lockup period mitigates the information asymmetry problem between old and new investors by forcing old investors to remain invested for a period of time after the IPO, thereby providing a signal of quality to the public market (Bock & Schmidt, 2015; Gompers & Lerner, 2001).

(12)

11 Bayar et al. (2010) research what determines a private firm’s choice of exit between an IPO and acquisition. They refer to the “IPO valuation premium puzzle”: an increasing number of VCs and entrepreneurs choose to take the company public through an acquisition, even though IPOs tend to offer much higher payoffs. The authors mention competition as one of the crucial explanations for this phenomenon. A stand-alone firm will have to fight competition on its own after going public, whereas an acquirer is able to provide support and defend the firm against product market competition. Information asymmetry is mentioned as an additional reason. There is significantly less asymmetry between the acquirer and insiders compared to the asymmetry in the IPO market, leading to a relatively more stable acquisition value over time, compared to the more fluctuating IPO market value.

While these factors suggest that acquisitions are favorable over an IPO, disadvantages of acquisitions have been identified. Acquirers have considerable bargaining power over the exiting firm, which might allow them to extract value from firm insiders. Whereas investors in the IPO market might price the firm’s value more competitively, allowing insiders to retain the net present value of the firm’s projects. Additionally, entrepreneurs managing the private firm might lose their private benefits of control in an acquisition. In most cases the founding entrepreneurs are fired after the acquisition or choose to leave the firm themselves. Even when the founder remains active in the merged firm, he or she will have to carry out policies as instructed by the top management. In contrast, after an IPO the entrepreneurial founders have the possibility to continue as CEO of the company thereby maintaining (part of) their private benefits of control (Bayar et al., 2010).

Brau et al. (2003) define four key factors that affect the takeover versus IPO decision for private firms. First, industry-related factors such as the level of concentration within an industry, antitrust concerns and government scrutiny. High concentration industries might have less potential for consolidation, making takeovers less likely. Second, market-timing factors such as “hot issue” periods where periods of high returns result in booming numbers of IPOs (Lowry & Schwert, 2000), and investor sentiment in the IPO markets, in which investors are optimistic and willing to overpay for IPOs. Managers and underwriters are more likely to bring IPOs to the market during these periods (Lowry, 2003). Third, deal-specific factors such as firm size. For relatively smaller firms, an IPO can be costlier whereas their possibility of success as a small public firm might be limited. Finally, funding demand factors such as the need of funding for new investments and the costs of debt. Private firms in need of large funds are more likely to exit through an IPO than through a takeover. Additionally, following the pecking order theory, high costs of debt make acquisitions harder to fund and lead to a preference for IPO exiting (Myers & Majluf, 1984).

Additionally, Poulsen & Stegemoller (2008) examine which firm-specific characteristics lead to firms going public either through a trade sale or an IPO. Their results suggest that companies

(13)

12 exiting through an IPO are more likely to be high-growth firms with high valuation ratios and the need for funding alternatives to debt. Additionally, they conclude that IPO firms tend to face more capital constraints, have fewer intangible assets, are less likely to be in the development stage and are more likely to be VC-backed. Since VCs are skilled at choosing the right exit type for their portfolio firms.

Lian & Wang (2012) research the acquisition valuations of withdrawn IPOs to understand the impact of an IPO registration or withdrawal on the private firm’s valuation. They state on the one hand that an IPO withdrawal may be perceived as a failure and therefore be seen as riskier by the market. This will eventually lead to lower valuations when they want to retry entering the public market. On the other hand they argue, that before each withdrawal, a private firm has filed an IPO registration which may positively influence its acquisition valuation. First, an IPO registration can decrease the asymmetric information problem between investors and entrepreneurs through e.g. mandatory accounting disclosures. Second, an IPO registration reduces the valuation uncertainty since it has already been established during the book building process. Finally, private ventures have shown their ability to carry the costs of an IPO listing, a signal of their value and quality. The final effect of IPO withdrawal on the acquisition value therefore depends on which of the two effects is strongest. They conclude that the two effects offset each other for companies acquired after their IPO withdrawals, however when the “almost-public” firms are acquired before the IPO withdrawal they sell at a premium. This indicates a positive valuation effect of IPO registration and a negative valuation effect of IPO withdrawal.

2.2.2 Exit failure: secondary sale or write-off

As is discussed in section 2.2.1 above, entrepreneurs and VCs generally choose between an IPO or trade sale when exiting the portfolio. However, in some cases the choice is not up to them and heavy underperformance causes them to exit through a write-off. Hereby the VC fund decreases the reported value of a poorly performing firm down to zero, meaning that the portfolio venture is worthless and the invested money is treated as lost.

A fourth type of exit is through a secondary sale in which only the shares owned by the VC are sold to a third party, typically a strategic acquirer (Cumming & MacIntosh, 2003). Ibrahim (2012) argues that this is a “potentially game-changing” exit option that should be preferred over trade sales. The author states that secondary sales operate at individual investor level instead of startup level, which attends to each investor’s individual liquidity needs. Additionally, he mentions that secondary sales have the potential to limit agency costs between VCs and entrepreneurs. Startups can’t be forced into a premature IPO or acquisition to fulfill the investment needs of VCs.

(14)

13 Wright, Robbie & Albrighton (2000) contradict this statement and argue that secondary exits occur when better opportunities are lacking. They find that secondary sales often take place after a failed, attempted trade sale. Additionally, secondary sales may be chosen as exit when further financing is needed for product or market development strategies that the initial VC does not support. Finally, a secondary sale might be induced by poor performance that leads to replacement of the management team by outsiders. Following these notions, secondary sales can be seen as a last choice or failure compared to IPOs and acquisitions.

2.2.3 Determinants of exit success

Various papers have examined the different factors that affect the exit success of a portfolio firm. Nahata (2008) proposes a new measure of VC firm reputation and finds that companies backed by more reputable VCs are more likely to exit successfully (exit through an IPO or acquisition), access public markets faster and have higher asset productivity at IPOs. Hereby two measures of reputation are used. First, the cumulative market capitalization of VC-backed IPOs. Since IPOs typically bring the highest returns to venture investors they can be seen as the most successful way to exit (Gompers & Lerner, 1999). In addition to this, IPOs create a “buzz” and increased market visibility for the selling VC. Therefore a VC can be considered as highly reputable when a fair share of its portfolio firms exit through an IPO. Second, the VCs share of aggregate investment in the industry. A high share of investment indicates a large share of fund commitments by limited partners (LPs) who select funds based on reputation. Additionally, entrepreneurial ventures tend to choose more reputable VCs over higher valuation offers (Hsu, 2004), indicating that more reputable VCs have a larger set of investment opportunities to choose from which leads to a higher share of aggregate VC investment.

Strese et al. (2018) argue that the measure of exit success does not solely depend on financial or categorical success indicators. Instead they develop a new measure of exit success based on the individual entrepreneur’s perception defined as: perceived exit performance (PEP). This is measured along the following four dimensions: personal financial benefits, personal reputation, employee benefits, and firm mission persistence. The authors conducted three studies with independent samples and find that entrepreneurs who carefully plan their exits, perform better on the personal reputation and firm mission persistence indicators, but improved planning does not affect the dimensions of financial and employee benefits. Additionally, they conclude that an entrepreneur’s previous industry experience positively influences all PEP dimensions, except personal reputation.

Streletzki & Schulte (2013) analyzed 64 startups funded by German VCs and examined the selection criteria that lead to “high-flyer exits”; exits that returned more than five times the money invested by the VC in the first-round. The results show the following dimensions as predictors of

(15)

14 high-flyer exits: targeting the business-to-customer market, being located in a metropolitan cluster and close to the lead investor, raising VC funding prior to the proof of concept level and having strategic partners raising the first VC investment round.

When looking at personal characteristics of an entrepreneur, most existing literature focusses on its impact on successful creation of business. However, Lee & Lee (2015) argue that entrepreneurs do not only create ventures but also yield the created value from the startup at the moment of exit. Therefore, they examine how personal characteristics of an entrepreneur affect successful entrepreneurial exit (SEE) They find that entrepreneur human capital, such as education and industry experience, does not affect SEE. In contrast, entrepreneur labor, such as the number of working hours put in by the founder, does significantly affect SEE. In this study an exit is defined as successful when the venture firm is sold to another business or merged with another business, and unsuccessful when its operations are permanently stopped.

Milosevic (2018) continues on this topic of human and social capital, focusing on France which has, in contrast to the US market, a poor performing VC industry that is characterized by tax subsidies and government funding. The author shows that VC performance and fundraising are disconnected in poor performing markets, such as France. She argues that this is due to the social networks, such as government intervention through public funding and tax-subsidized investments, that keep facilitating funding for low performing VC ventures. Additionally she examines how the entrepreneur’s experience in R&D and experience in investment banking affect VC performance. The results show that both factors predict better trade sales than IPOs, as opposed to entrepreneurial experience, which tends to lead to IPOs. This indicates that public networks of VC firms and their managers do not lead to successful exits and can contribute to distortions in the VC market by funding underperforming startups.

Finally, Lee et al. (2014) investigate the relation between exit success and institutional and cultural distances. Hereby the cultural distance is defined as the differences in cognitive patterns and values across borders, these can lead to hindering of information sharing, reduced trust and increased transaction costs. Institutional distance is described as differences in regulations and assumptions that govern business relations. These can increase the risk of sanctions, legitimacy loss and again additional transaction costs. The authors find that both cultural and institutional differences negatively impact exit success in the form of an IPO or trade sale.

(16)

15

2.2.4 Time to exit

When weighing possible exit options, VCs not only assess how they can cash out but also how long they need to stay involved before their investment pays off. Piot & Schwienbacher (2007) therefore state that the exit decision has two crucial elements; the type of exit and the exit timing. Existing literature shows that VCs use contractual agreements that guarantee them certain intervention rights, one of which allows them the right to force an exit (Gompers, 1997; Kaplan & Stromberg, 2003). Additionally, it is well-known that the lifespan of VC funds is limited (typically 10 years) and eventually the invested money needs to be returned to the original providers (Gompers, 1996; Gerasymenko & Arthurs, 2013).

Hsu (2013) argues that the time to exit, or incubation period length, influences a VC’s future fundraising. He states that in case of an IPO, the effects are two-sided: on the one hand, a short incubation period may lead to lower realized returns to the capital providers of VC funds (limited partners). This may prevent them from investing again in the VC in the future. On the other hand, shorter exit time indicates a VC’s ability to quickly exit portfolio firms. An ability that is perceived as a strength by LPs, as it allows VCs to make and exit more investments and therefore generate larger returns at the fund level.

Gompers (1996) describes the grandstanding hypothesis in which a quicker exit rate may be caused by younger VCs rushing to take their portfolio firms public to establish their reputation among LPs. Since VCs have the right to force an exit, this may cause them to take their investments public earlier than would maximize their returns (Gompers, 1996). Therefore the grandstanding hypothesis expects IPOs backed by young VCs to be inferior to IPOs that are backed by more established VCs, since they go public prematurely. Wang, Wang & Lu (2003) state that rushed IPOs are expected to show lower post-IPO results compared to the pre-IPO level, as they tend to be associated with window dressing by the public market.

Giot & Schwienbacher (2006) examine the dynamics of the exits of 6000 VC-backed firms. Regarding exit timing, they find that VC-backed ventures first are more likely to exit through an IPO. Additionally, they find evidence that portfolio firms have fewer and fewer opportunities of an IPO exit, as time progresses. In other words, firms can get selected for an IPO relatively fast, however if they do not go public quick enough, their chances of doing so drop drastically. For trade sales they find that the peak is reached much later and the possibilities of exiting after the peak decrease much slower than for an IPO. Which confirms the notion that a trade sale is a more common form of exit than an IPO (Giot & Schwienbacher, 2006).

(17)

16

2.3 Technological shocks

The availability of theory on the effect of technological shocks is limited. Florida & Kenney (1988) were one of the first to address the relation between venture capital and innovation processes among start-up ventures. A technology cycle can be described as an S-curve, passing through three stages: emergence (phase of rapid growth), consolidation (phase of steady expansion) and maturity (phase of growth decline). The authors state that VC is most relevant in the first stage of innovation and great breakthroughs; a phase that is characterized by uncertainty, experimentation, low barriers of entry and diseconomies of scale. They conclude that venture capital has transformed the innovation process in the US, as they provide funds and assist in the foundation of new high technology businesses. Additionally, they argue that VCs have active networks of e.g. entrepreneurial companies, financial institutions and universities which, together with the information flow at their disposal, lower many of the risks related to starting a new enterprise.

Kortum and Lerner (1998) examine the influence of venture capital on patented inventions in the US. They argue that patent applications among VC-backed firms are indicative of innovation for two reasons. First, VC-backed companies might opt to patent inventions quicker than non-VC-backed firms because they fear that their ideas might be exploited by the venture investors. Second, investors might have difficulties distinguishing the quality of patent holdings, therefore ventures may apply for patents to signal the value of their technologies to potential investors. The authors find that over the period 1983-1992, venture capital accounted for 8% of the industrial innovations. The results of their research show that the amount of VC activity in an industry significantly spurs the amount of patenting.

Instead of analyzing the effect of venture capital on innovation, the reverse has also been examined; how technological shocks affect VC activity. Hsu (2013) examines the effect of industry-specific technological changes on IPO timing of VC-backed firms and question what consequences such timings might have. According to the VC fundraising hypothesis, VCs are motivated to exit their portfolio firms earlier during positive technological changes in the industry, in order to raise future funds. This hypothesis arises from 2 factors: First, positive technology shocks raise the expected levels of productivity and provide attractive investment opportunities (Chemmanur & Fulghieri, 1999), which creates an incentive for VCs to attract additional funding. Second, the effect of the length of the incubation period on VC’s future fundraising, which can be both positive and negative. On the one hand the performance of a portfolio firm after post-IPO is directly linked to the returns of VCs’ providers of capital (LPs); shorter incubation periods lead to decreased post-IPO performance. This may withhold LPs from investing in the VC after short incubation periods. On the other hand, a short incubation period indicates a VC’s ability to be flexible and quickly exit their portfolios firms, allowing them to make and exit more investments, executing more IPOs and therefore generating

(18)

17 larger total returns for limited partners. This may encourage limited partners to invest in VCs after a period of short incubation. The authors’ findings are consistent with the VC fundraising hypothesis. They conclude that VCs are more likely to both shorten incubation periods and take their portfolio firms public during high levels of technological change in the industry. In addition to that, VCs that execute IPOs with a short incubation period are able to raise higher levels of funding afterwards.

2.4 Cloud computing

The technological shock that is used in this research is the introduction of the Amazon Web Services (AWS) in 2006. This service provides a wide range of cloud solutions such as; on-demand delivery of compute power, database storage, applications and other IT sources. All products are offered via the internet and on a pay-as-you-go basis (aws.amazon.com). The services provided by AWS were initially developed for Amazon’s internal infrastructure before made available to a worldwide public. Ewens et al. (2015) state that the introduction of cloud computing was a defining moment at which the costs of starting certain businesses dropped significantly. These cost reductions are possible due to a number of reasons. First, users only pay when they consume and for how much they consume. Instead of making a large upfront investment in e.g. datacenters and servers, entrepreneurs can now have these costs growi along with the success of their business. Second, the AWS platform provides large economies of scale, since hundreds of thousands of users are aggregated in the cloud spreading the costs and eventually leading to lower prices. Third, entrepreneurs no longer have to guess their infrastructure capacity needs beforehand. Over-or-underestimating these needs might lead to extra costs afterwards. Fourth, as the service is provided in a cloud computing environment, speed and agility within the organization are substantially increased. Fifth, business owners no longer have to worry about running and maintaining their data centers, and can therefore fully focus on their own products and customers. Finally, AWS provide entrepreneurs with the ability to go global in just a few clicks, as the application can easily be deployed worldwide (aws.amazon.com).

Several web-articles, a.o. by Clarke (2012) and Vogels (2011), argue that the timing of opening Amazon’s cloud computing services up to developers in 2006, was not anticipated by entrepreneurs and investors. The introduction of AWS is seen by many practitioners as a groundbreaking moment that effectively lowered the costs of starting web-based and internet businesses. Before 2006, firms had to invest heavily upfront for a small chance that their venture would be a success and there were no abilities to freely rent hardware that could scale with demand. The advent allowed entrepreneurs to scale up their business as demand grew, making it possible to learn about the success and viability of a venture before investing heavily (Ewens et al., 2015). The technology is used by a large share of startups, a.o. firms like Netflix, Pinterest and OMGPOP. In

(19)

18 addition to that, some major multinationals such as Samsung and Unilever are using the cloud computing services (Clark, 2012).

Ewens et al. (2015) study how technological shocks to the cost of starting new businesses have affected the investment strategies of venture capitalists. They conclude that these cost-reducing technological shocks caused venture capitalists (VCs) to adapt their financing strategy fundamentally over the past decade. They report on an increased “spray and pray” investment approach in which investors provide little funding and limited governance to an increased number of startups which they are more likely to abandon. This leads to a disproportionate rise in innovations where information on future prospects is revealed quickly and cheaply.

3. Empirical design

3.1 Methodology & hypotheses 3.1.1 Hypotheses

This research is focused on examining whether technological shocks that reduce the startup costs of certain businesses have an effect on the success level at which VCs exit their portfolio firms. Following existing literature, three measures are selected that together will provide an extensive indication of exit success. The first one is the exit success rate, in which an exit is defined as successful when it occurs through either an IPO or trade sale. The second measure is a firm’s DuPont ROE, an indicator of the investor’s return on equity based on profitability, asset efficiency and financial leverage. Finally, the speed at which a firm is exited is used, this is measured by the number of years between the initial VC-funding and the moment of exit. Following existing theories, this thesis will test for the following hypotheses:

H1: The AWS shock leads to an increase in exit success rate

As is described by Ewens et al. (2015), the introduction of Amazon Web Services in 2006 has led to an increased “spray and pray” approach by venture capitalists in US markets. Hereby VCs invest in an increased number of firms affected by the shock, providing each startup with a smaller amount of funding and governance than before. This is explained by the groundbreaking characteristics of Amazon Web Services that significantly reduce the costs of starting web-based and internet businesses, providing VCs with an opportunity to get informed about the success and potential of a startup without heavily investing upfront.

Following the existing literature, this could on the one hand lead to beneficial effects in the exit success rate of VCs. Beliefs about the future potential of ventures are revealed to the VC sooner

(20)

19 and they are able to abandon unsuccessful firms quicker, since they have less invested in them and more alternatives in their portfolio. At the same time, the lowered cost to starting new businesses allows a new group of entrepreneurs to receive early stage financing, where they would not have received any funding before. These are mostly startups with a lower probability of success but excessive returns once they turn out successful (Ewens et al., 2015). Therefore, the increased ‘spray and pray’ effect could lead to higher chances of funding the next Uber or Airbnb and gaining excessive returns, whilst limiting the level of early-stage investment and risk per portfolio firm.

On the other hand, the increased number of portfolio firms leads to each startup receiving less governance in the early stage, precisely the moment at which mentorship and governance are needed the most (Ewens et al., 2015). Venture capitalists have the unique experience and skills to manage high risk, high reward startups, partly because of their intensive monitoring role (Bernstein et al., 2016). The increased “spray and pray” effect as a result of the AWS shock could therefore lead to a lower exit success rate, since the level of mentorship and governance, one of VCs’ main strengths, decreases for early stage investment firms. Jackson et al. (2012) examine the “profit destruction effect” in which an increase in the amount of portfolio firms has a negative impact on VC investment returns. They argue that increases in VC portfolio size generate risks of weighing down limited VC resources. However, when looking at the overall effect they conclude that the positive effects outweigh the profit destruction effects, and that VC activism predicts higher returns even when the profit destruction effect is in play.

Since the existing literature provides reasons to expect a stronger rise than fall in exit success rates as a result of the shock, the hypothesis is stated positively, one-sided. The introduction of AWS is therefore expected to cause an increase in the exit success rate of affected firms.

H2: The AWS shock leads to a higher DuPont ROE

The DuPont analysis is a well-known decomposition of return on equity (ROE) and is an indicator of a firm’s performance. It is a product of asset turnover (AT), which indicates the efficiency rate at which assets are converted to sales, return on sales (ROS) or profit margin which is defined as operating income divided by sales, and equity multiplier (EM) which is calculated by dividing a firm’s total assets by total equity (Loos, 2006; Turner et al., 2015). In this research the DuPont rate of firms is observed at the moment of exit.

Based on Ewens et al. (2015), the introduction of AWS leads to an increased “spray and pray” effect in which information on a startup’s future prospects is exposed fast and at low costs. VCs are providing more startups with a first round of funding but at the same time they are more likely to abandon those ventures if their prospects turn out unpromising. Guo & Jiang (2013) find that

(21)

VC-20 backed firms perform better in terms of profitability. They argue that VCs first select firms with higher profitability and then significantly magnify these differences after including them in their portfolio (evidence from both monitoring and screening). As is suggested under hypothesis 1 as well, the larger “spray and pray” effect increases the chance for VCs to find investments that are going to provide them with super-abnormal returns and leaves them with a portfolio that has a higher ratio of successful to average or failing firms (Ewens et al., 2015). Therefore, it can be argued that the overall DuPont ROE is expected to be higher for firms affected by the AWS shock.

On the contrary, one could argue that the cost reducing AWS shock leads to a decrease in DuPont ROE since (part of) the monitoring role of the VCs is limited by the increased portfolio. Similar to the reasoning under hypothesis 1, the “profit destruction effect” as described by Jackson et al. (2012), may (partly) weigh down VC resources such as governance and mentoring. However, this effect is again not expected to exceed the benefits since monitoring is only part of VC’s skills in managing their portfolio firms. Additionally, Jackson et al. (2012) show that the benefits of spreading chances among more startups at a lower cost outweigh the profit destruction effects.

Therefore, the hypothesis predicts an increase in the DuPont ROE of VC-backed firms that are affected by the introduction of AWS.

H3: The AWS shock leads to a significant decrease in time to exit

According to existing literature, a firm’s time to exit is an important indication of VC success (Black and Gilson, 1998; Wang and Wang, 2012). Hereby the time to exit is defined as the number of years between an IPO/trade sale exit and initial VC funding, following Nahata (2008).

Ewens et al. (2015) find an increased “spray and pray” effect as a result of the AWS shock. They show that information about the future potential of portfolio firms is revealed relatively fast and VCs are more likely to abandon them in an early stage if they rate the potential as insufficient. This leads to an expected decrease in time to exit, since an increased number of firms is exited at a quicker rate.

Although less likely, one might argue an expected increase in time to exit for treated firms, after the introduction of AWS. Espenlaub et al. (2015) argue that the active involvement of VCs through their expertise, effort and networks, will pay off most in the initial phases of investment via high levels of marginal value added. At the same time, increased effort means increased marginal costs for VCs. In later stages, the marginal value added is expected to decline together with the effort put in by VCs. Therefore, VCs will choose to plan their exit when the marginal value added drops below the marginal costs of active involvement. In the increased “spray and pray” effect found by Ewens et al. (2015), VCs provide their portfolio companies with lower levels of monitoring and

(22)

21 guidance in the early investment stages. If VCs determine the ideal moment of exit by balancing the costs and benefits of maintaining their investment, the “spray and pray” effect may lengthen the duration to exit, since initial effort is lower and more time is needed to “make up” for lost value added in the crucial, early stages. However, this effect is expected to be relatively small since the active involvement of VCs is still present, only to a smaller extent, and the timing of exits is based on more factors than solely the right balance of marginal costs and value added, such as e.g. market-timing (Lowry & Schwert, 2000).

A decrease in time to exit can be interpreted as both less and more successful for VCs, where existing literature provides more evidence of the latter. Hsu (2013) states that on the one hand, it can be seen as beneficial for VCs since it demonstrates their ability to quickly exit portfolio firms. This is recognized as a desirable skill by (potential) LPs because it allows them to make and exit more investments, thereby generating larger returns at fund level. Furthermore, Ragozzino & Blevins (2016) state that a short investment duration is a signal of appropriate internal governance by the startup to potential investors. They argue that VCs use staged capital financing and extension of fund duration as a control mechanism for ventures that do not (yet) have sufficient internal checks and balances and that face incentive misalignment issues. Additionally, the innovativeness of a startup’s concept tends to decrease over time (Ragozzino & Blevins, 2016), whilst the chance of competitors increases (Carrow, Heron & Saxton, 2004), creating incentives for VCs to exit their portfolio investments in a timely manner. In contrast, a longer incubation period may lead to larger realized returns to the capital providers of VC funds.It allows venture capitalists more time to add value to the portfolio venture (Espenlaub et al., 2015). VCs are known for holding the specific skills and experience to manage starting businesses into successful firms. A large part of these skills are related to active involvement through guidance and monitoring of the portfolio firm(Bernstein et al., 2016; Croce et al., 2013; Chemmanur et al., 2011). An increased time to exit means more opportunities for VCs to share their experience and apply their guiding skills to the development and growth of the startup, although these effects are expected to be outweighed by the above mentioned benefits of a shorter duration to exit.

Since the existing theory provides more extensive argumentation to expect a decrease rather than an increase in the venture’s time to exit as a result of the shock, the final hypothesis is one-sided. The expected decrease in time to exit brings both positive and negative consequences for the VCs, but since the existing literature describes more arguments in favor of the positive ones, a decrease is regarded as a greater success for VCs.

(23)

22

3.1.2 Methodology

The methodology used in this thesis is a difference-in-difference framework to test for differences in the exit success of VC-backed firms defined as treated, before and after the introduction of AWS. The following regression equation will be used to test for the first hypothesis (Ewens et al., 2015):

𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑗𝑗𝑗𝑗𝑗𝑗= 𝛽𝛽1𝑇𝑇𝑇𝑇𝑆𝑆𝑇𝑇𝑇𝑇𝑆𝑆𝑇𝑇𝑗𝑗∗ 𝑃𝑃𝑃𝑃𝑆𝑆𝑇𝑇𝑗𝑗+ 𝛽𝛽2𝑋𝑋𝑗𝑗+ γ𝑗𝑗+ ρ𝑗𝑗

Where β1 is the main coefficient of interest showing the interaction effect of treated * post on the dependent variable. Xi represents control variables for entrepreneurial firm characteristics including:

major industry group and the separate “treated” variable. Hereby major industry group is either one of six Company Venture Economics Primary Industry Classes, as defined by ThomsonOne: “Communication and Media”, “Medical/Health/Life Science”, “Computer Related”, “Semiconductor/Other Electronics” or “Biotechnology”. The definition of treated industries is made on a more detailed level using 17 industry sub-groups. This allows for within-industry variation among treated versus non-treated firms, since VCs usually have a wide industry range for their investments. The definition of treatment is further explained in section 3.1.3. The separate “post” variable is dropped due to multicollinearity issues caused by the inclusion of time FE. To control for time effects, the variable γt is added which corresponds to year FE. This incorporates the varying

macroeconomic impact across different exit years, such as interest levels, “hot issue” periods and the overall state of the economy. Finally, VC-firm fixed effects are included through ρj, to allow for

within-VC dynamics when estimating the main independent variable β1.

The dependent variable Successjit, is a dummy variable equal to 1 if the portfolio firm exits

through an IPO or trade sale and equal to 0 if the exit was a secondary sale or write off. Following previous literature, this indicates whether a VC exit is successful or not (Nahata, 2008; Thomas et al., 2011).

A second fixed effect regression is done including the same independent variables as in regression 1. In this case, the dependent variable is the DuPont rate of VC-backed firms at the moment of going public:

𝐷𝐷𝑃𝑃𝐷𝐷𝐷𝐷𝐷𝐷𝑗𝑗𝑗𝑗𝑗𝑗 = 𝛽𝛽1𝑇𝑇𝑇𝑇𝑆𝑆𝑇𝑇𝑇𝑇𝑆𝑆𝑇𝑇𝑗𝑗∗ 𝑃𝑃𝑃𝑃𝑆𝑆𝑇𝑇𝑗𝑗+ 𝛽𝛽2𝑋𝑋𝑗𝑗+ γ𝑗𝑗

The DuPont rate, or the DuPont equation, provides an analysis of what drives a firm’s profitability by combining measurements of its profitability, efficiency and capital structure into one ratio. It is a product of a firm’s profit margin or return on sales (ROS), total asset turnover (TATO) and equity multiplier (EM). (Loos, 2006; Turner et al., 2015). Hereby the profit margin captures by how much

(24)

23 revenue exceeds costs, total asset turnover indicates the efficiency rate at which assets are converted to sales, and the equity multiplier measures the share of a firm’s assets that is financed by equity (Turner et al., 2015).

𝐷𝐷𝑃𝑃𝐷𝐷𝐷𝐷𝐷𝐷 = 𝑃𝑃𝑃𝑃 ∗ 𝑇𝑇𝑇𝑇𝑇𝑇𝐷𝐷 ∗ 𝐷𝐷𝑃𝑃

𝐷𝐷𝐷𝐷𝑆𝑆 = 𝐷𝐷𝑂𝑂𝑆𝑆𝑇𝑇𝑇𝑇𝑇𝑇𝑂𝑂𝑂𝑂𝑂𝑂 𝑂𝑂𝑂𝑂𝑆𝑆𝑃𝑃𝑖𝑖𝑆𝑆_{𝑁𝑁𝑆𝑆𝑇𝑇 𝑆𝑆𝑇𝑇𝑠𝑠𝑆𝑆𝑆𝑆}

𝑇𝑇𝑇𝑇𝑇𝑇𝐷𝐷 =_{𝑇𝑇𝐴𝐴𝑆𝑆𝑇𝑇𝑇𝑇𝑂𝑂𝑆𝑆 𝑇𝑇𝑃𝑃𝑇𝑇𝑇𝑇𝑠𝑠 𝑇𝑇𝑆𝑆𝑆𝑆𝑆𝑆𝑇𝑇𝑆𝑆}𝑁𝑁𝑆𝑆𝑇𝑇 𝑆𝑆𝑇𝑇𝑠𝑠𝑆𝑆𝑆𝑆

𝐷𝐷𝑃𝑃 =𝑇𝑇𝑃𝑃𝑇𝑇𝑇𝑇𝑠𝑠 𝑇𝑇𝑆𝑆𝑆𝑆𝑆𝑆𝑇𝑇𝑆𝑆_{𝑇𝑇𝑃𝑃𝑇𝑇𝑇𝑇𝑠𝑠 𝑆𝑆𝑒𝑒𝑆𝑆𝑂𝑂𝑇𝑇𝑒𝑒}

In this research the DuPont rate of firms is observed at the moment of exit. Since the above described financials are only available for public firms, this analysis will only be done for firms that exited through an IPO.

Finally, a third regression is run, where the dependent variable is the speed at which VC-backed firms exit successfully:

𝑆𝑆𝑂𝑂𝑆𝑆𝑆𝑆𝑇𝑇𝑗𝑗𝑗𝑗𝑗𝑗 = 𝛽𝛽1𝑇𝑇𝑇𝑇𝑆𝑆𝑇𝑇𝑇𝑇𝑆𝑆𝑇𝑇𝑗𝑗∗ 𝑃𝑃𝑃𝑃𝑆𝑆𝑇𝑇𝑗𝑗+ 𝛽𝛽2𝑋𝑋𝑗𝑗+ γ𝑗𝑗

The dependent variable is measured by the log of years between initial funding and the time of exit. Similar to regressions 1 and 2, the main independent variable β1 represents the effect of the interaction between “treated” observations post-shock. Then, Xi represents a number of control

variables. Again, the 6 major industries mentioned under regression 1 are included to allow for within-industry impact of treatment, which is defined at a more granular level of 17 sub-industries. Xi

also includes the separate interaction term “treated”. In addition to that, the following binary variables are included to indicate the life-cycle stage of the startup at the moment of exit; “Buyout/Acquisition”, “Later Stage”, “Expansion”, “Early Stage” and “Startup/Seed”. Finally, three remaining variables are added to control for: the investment round number, equity value and syndicate size at the moment of exit. ρj are year FE that represent the year of exit. This incorporates

the varying macroeconomic impact across time, such as interest levels, “hot issue” periods and the overall state of the economy. The separate “post” variable is dropped as it is multicollinear with the included year FE.

(25)

24 The measure of speed to exit raises some concerns since companies in existence beyond 2016 may still exit successfully, creating a sample period that is right-censored at the year 2016. To solve for this censoring problem, a Cox hazard model is used. This model, also known as the Cox regression model, is a semiparametric model in which the hazard function is not dependent on a specific distribution of the survival time and therefore allows for time-varying exit market conditions (Crichton, 2002). It enables for a comparison of the survival rate between the treated and control group. In this research survival is defined as a firm exiting successfully through either an IPO or acquisition. The dependent variable is the logarithm of time to exit, which is taken from the date of each VC’s first funding in the portfolio company (Nahata, 2008).

3.1.3 Defining treatment & control

Since not all firms were equally affected by the advent of cloud computing, the definition of treatment is based on their online presence. However, solely looking at this raises some endogeneity concerns, as there might be other characteristics specific to online firms that affect the levels of VC exit success. To solve this, definition of treatment is based on industry-level rather than firm-level. Hereby industries are considered as “treated” based on the likelihood of benefiting from the AWS arrival. This industry-segment level of exposure to the treatment is measured by the share of firms within the industry that contain words such as “e-commerce”, “web”, “online” and “hosting” in their activity description pre-period (Ewens et al., 2015). The control group is then defined as the remaining industries that do not have significant likelihood of benefiting from the shock. This in order to meet the common trend criteria of the diff-in-diff method, in which the control group needs to be similar to the treatment group, except in the exposure to the shock.

Table I shows the number of sample firms per industry and the percentage of companies that contain at least one of the above mentioned key words in their activity description. In this table, the distinction between treated and control observations is made on a 10% word share threshold. Panel A demonstrates relatively higher word shares compared to the non-treated industries, where percentages are close to zero. For all industries a comparison is made between pre-period and total-period observations. Panel A shows an overall increase in the word share whereas in panel B the percentages remain roughly similar when post-period observations are included. This implies a significant growth in the number of investments in firms that are most likely to be affected by the shock, which is in line with the “spray and pray” theory stated by Ewens et al. (2015). In this theory, the authors argue that due to the cost reduction effects of starting new, online-related businesses, VCs tend to increase the number of online-related startups in their portfolio.

(26)

25

Table I – Treated vs Non-treated industry segments primary industry sub-based

This table reports the allocation of treated and non-treated industries, extensive industry descriptions are provided in the appendix table X. An industry is defined as treated, when it is likely to be affected by the introduction of AWS. This is measured by the fraction of firms that have at least one of the following words in their activity description: “online”, “hosting”, “web” or “e-commerce”. The resulting industry-level exposure is then used to distinguish treated and control industries at a 10% cutoff. Panel A shows the observations for treated industries and panel B shows the observations for the control industries. Hereby N presents the firm count within each industry segment and word share displays the industry-level exposure as mentioned above. 2000-2005 refers to VC-exits before the technological shock, whereas 2000-2012 refers to the entire sample.

Industry Panel A: Treated industries

2000-2005 2000-2012

N Word share N Word share

Business Serv. 182 13% 407 14% Communications 387 14% 571 16% Computer Hardware 154 14% 285 18% Computer Software 1,118 26% 2,113 31% Consumer Related 425 12% 919 15% Financial Services 119 12% 290 10% Internet Specific 758 56% 1,678 61%

Industry Panel B: Non-Treated industries

2000-2005 2000-2012

N Word share N Word share

Agr/Forestr/Fish 15 0% 35 0% Biotechnology 238 1% 489 2% Construction 64 3% 141 1% Industrial/Energy 424 2% 1,026 3% Manufact. 125 5% 283 7% Medical/Health 464 2% 963 3% Other 55 0% 123 4% Semiconductor/Electr. 357 4% 496 5% Transportation 145 8% 335 7% Utilities 29 0% 61 7%

Following the described definition of treatment, the following industries are classified as being most exposed to the AWS shock, using a 10% pre-period word share threshold: “Business Services”, “Computer Hardware”, “Computer Software”, “Consumer Related”, “Communications”, “Financial Services” and “Internet Specific”. Most affected industries provide a service rather than a product or are computer or internet related and therefore tend to benefit more from the introduction of Amazon Web Services (Ewens et al., 2015). This leaves the remaining industries as non-treated: “Agriculture/Forestry/Fishing”, “Biotechnology”, “Construction”, “Industrial/Energy”, “Medical/Health”, “Other”, “Semiconductor/Electrics”, “Manufacturing”, “Transportation” and Utilities”. These are in general more traditional industries providing core products or services, often with large upfront investments and starting costs that are less sensitive to technological shocks such as the AWS advent. Additionally, the final regression will be performed a second time using an

(27)

26 increased threshold of 25% word share to define treatment. Hereby solely “Internet Specific” and “Computer Software” are defined as treated industries. The increased threshold is expected to lead to a larger coefficient, since the industries are more likely to be affected by the shock. A more detailed description of each industry is given in the appendix, table X.

3.2 Data & descriptive statistics 3.2.1 Sample

The data used for this research is based on US, VC-backed firms. Information on their exits, such as, exit type, total funding to date and years to exit, is gathered from the ThomsonOne database. The industry categories used to separate the treated and control groups are also retrieved from ThomsonOne and are based on primary industry subsector, further described in the appendix table X. Additionally, Capital IQ provides extensive data on public firms and is used to obtain financials and information on the sample firms. Since the sample only consists of VC-backed firms, most data are not publicly available. Therefore information and input for the DuPont analysis is only researched for firms that exited through an IPO.

To allow for a proper comparison between exits of initial funding before the shock to exits of investments after the shock, this study uses a primary sample of exits that occurred in the period 2000-2012, both 6 years before and after the introduction of AWS. This is because the period should be long enough to include enough treatment variables; the sample-average number of years until exit is approximately 5 and Nahata (2008) & Gompers and Lerner (2000) argue that firms should be allowed a minimum number of 4 years for a successful exit. Yet, the sample period is not too extensive in order to limit the effects of other shocks after 2006.

The pre-period is defined as all exits that occur through the years 2000-2005, which automatically implies that the initial funding of these startups also takes place before the shock (however, they may also take place pre-2000). Exits are defined as post-period solely when both the exit and the investment take place in the period 2007-2012, in order to observe the effect on exit success of the AWS shock and the response of VCs in their investment strategies.

Consistent with Nahata (2008) when testing for the effect on time to exit in regression 3, the same sample is used that consists of venture investments and exits during the period 2000-2012, as well as an additional period to observe exit types until the beginning of 2017 . This allows firms funded in the final sample year 2012, the minimum of 4 years for a successful exit (Nahata, 2008 & Gompers and Lerner, 2000). After the beginning of 2017, firms that have not yet exited successfully are defined as failed exits. This creates a sample that is right-censored at the beginning of the year 2017, since these investments still may exit successfully after this point. The Cox Hazard model is used to account for this right-censoring when examining the effects on time to exit.

(28)

27

3.1.2 Descriptive statistics

Table I shows the descriptive statistics of the firm sample used in this thesis. The mean, median, standard deviation, and minimum and maximum values are displayed for all relevant variables.

Table II – Descriptive statistics

This table displays the summary statistics of the sample, which consists of 14.953observations and is measured over a period of 13 years, from 2000-2012. Years to exit is the difference between the year of the first VC investment and the year of exit. Firm age is defined as the age of a venture at the moment of VC-exit. Total funding to date displays the amount of funding the startup has received at the moment of exit, in million USD. M&A deal value states the total deal value in case of a M&A exit, in million USD. Asset turnover, indicates the efficiency rate by which a firm converts its asset investments into sales. Return on sales is measured by: net income/sales. Equity multiplier measures the ratio assets/equity. DuPont ROE is the product of asset turnover, return on sales and equity multiplier. IPO exit is a binary variable that equals 1 if a VC-exit occurs through an IPO and 0 otherwise. Trade sale exit is a binary variable equal to 1 in case of an exit through trade sale and 0 otherwise. Write off exit is a binary variable that equals 1 if a portfolio firm seizes to exist, and 0 otherwise. Exit success is a binary variable that equals 1 if a VC exits a portfolio firm either through an IPO or trade sale, and 0 otherwise. Firms affected is a binary variable equal to one if the firm includes at least one of the following words in their business description: “online”, “e-commerce”, “web” and “hosting”. Finally, Firms post period equals a value of 1 when a firm has both its initial investment and exit in the post-shock period 2006-1012.

Variable N mean Median S.D. Min Max

Years to exit 14,134 5.56 4.80 3.82 .10 48.7

Firm age 13,572 15.91 9.00 20.10 -10 261

Total funding to date (mil) 11,380 69.66 24.18 423.12 0 37,605.00

M&A deal value (mil) 4,472 363.92 110.00 1,238.28 .05 32,105.38

Total asset turnover 770 .82 .70 .82 0 7.40

Return on sales 787 -5.42 -.22 34.59 -125 2.31

Equity multiplier 768 1.68 1.25 4.27 -8.25 16.5

DuPont ROE 766 -.18 -.05 3.22 -3.96 3.73

IPO exit 14,953 .09 0 .29 0 1

Trade sale exit 14,953 .63 1 .48 0 1

Secondary sale exit 14,953 .13 0 .33 0 1

Write off exit 14,953 .05 0 .23 0 1

Exit success 14,953 .73 1 .44 0 1

Firms affected 14,953 .22 0 .41 0 1

Firms post period 14,953 .31 0 .46 0 1

The table shows that the sample average time to exit is around 5.5 years and that firms are on average around 13.5years old at the time of exit. Looking at the finances, the average sample firm