1
Masterthesis
Thesis Supervisor: dr. Rafael Perez Ribas
Amsterdam Business School
Faculty of Economics and Business – Section Finance
Business Angel or Venture Capitalist?
The Investor’s Impact on Start-up’s Survival and Exit
Last Name, First Name:
Wiedekind, Johannes
Student ID:
11089067
Address:
Otzbergring 41
64846 Groß-Zimmern
Germany
E-Mail Address:
j.wiedekind@gmx.de
Programme:
MSc Business Economics, Finance Track
2
Table of Contents
I.
Statement of Originality
... 3
II.
Abstract ... 4
1 Introduction ... 5
2 Literature Review... 7
2.1 Private Equity ... 7
2.2 Business Angels ... 8
2.3 Venture Capitalists ... 9
2.4 Business Angels vs. Venture Capitalists ... 9
3 Hypotheses ... 11
4 Datasets and Variables ... 13
4.1 CrunchBase ... 13
4.2 Fit the Dataset and Variables ... 14
4.2.1 Main explanatory variables ... 15
4.2.2 Control Variables ... 16
4.3 Summary Statistics ... 17
5 Empirical Framework: Cox Proportional Hazard Model ... 21
5.1 Events in the Cox PHM ... 23
6 Empirical Results ... 25
6.1 Kaplan-Meier Survival Estimate ... 26
6.2 Survival Estimations for Entrepreneurial Companies in the PHM ... 27
6.3 Exit Estimations for Entrepreneurial Companies in the PHM ... 30
7 Discussion and Future Research ... 31
8 Conclusion ... 34
References ... 36
3
I.
Statement of Originality
This document is written by Student Johannes Wiedekind who declares to take full responsibility for the contents 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 contents.
4
II.
Abstract
This paper analyzes Business Angels and Venture Capitalists as investors for Start-ups and investigates which type of investor disproportionately backs more successful Start-ups in terms of exit and survival rate. Prior studies examined one single investor in comparison to all remaining ones and conclude that entrepreneurs are better off favoring Venture Capitalists, because of their high liquidity and their huge expertise in exit strategies. This paper uses CrunchBase’s extensive dataset together with a proportional hazard model to oppose Business Angels and Venture Capitalists. As the results will show, Start-ups backed by Business Angels have a 95 percent higher hazard to fail and a 14 percent lower probability to end up successful in terms of exit. These findings suit the related literature. It should be taken into account though that the results are significantly influenced by the fact that the funding amounts differ distinctly between both investor types. By controlling for provided liquidity, Business Angels appear to be the better option for entrepreneurs to accomplish a successful exit via IPO or acquisition. Consequently, Business Angels seem to have a non-monetary impact on entrepreneurial firms that is often superior to the one of Venture Capitalists.
5
1 Introduction
Politicians and Academics agree that company founders are a driving force for global growth as recently underpinned by the Free Democratic Party (FDP) in Germany. Soon after the Referendum in the UK about leaving the European Union, a poster with the following headline was spotted in the City of London:
“Dear start-ups, Keep calm and move to Berlin.”1
In most cases founders have to overcome financial constraints to start their businesses. The long debate on advantages and disadvantages of different external investors on entrepreneurial firms is not settled but rather intensified again recently. The number of financing opportunities for Start-ups and their transparency has diversified. Due to the increasing importance of internet platforms, information about different opportunities for external funding are easily accessible for entrepreneurs.
Business Angels (BAs henceforth) and Venture Capitalists (VCs henceforth) are a worthwhile alternative to cope with liquidity shortages. Both of them target companies that are in the same stage of development but differ in fundamental characteristics, such as governance and ownership structure. These characteristics are topics of many papers (e.g. Chahine et al., 2007; Chemmanur and Chen, 2014; Hellmann and Thiele, 2014), but these studies do not evaluate in detail the emerging differences in success prospects for the supported Start-ups. Unlike these papers the aim of this study is to find evidence or whether one of the two external investor types set off Start-ups to be more successful.
The analyzed success indicators are survival rate and exit rate in terms of an initial public offering (IPO) or an acquisition. Differing survival rates might be a crucial factor for entrepreneurs to consider when they evaluate the pros and cons of external financing. To shut down their Start-up before business really started to roll is one of the entrepreneur’s main fears. Furthermore, most of the prosperous founders already experienced a painful company shutdown before they got successful. To avoid failure in the first place and to find the most suitable investor the presented findings can be very beneficial for entrepreneurs. Additionally, investors might also profit from these findings to a great extent by learning about their own potential of improvement.
This thesis executes a proportional hazard model (PHM) to evaluate whether the supporting investor type is a determinant of Start-up’s failure and exit. The PHM is adopted from medical studies and has never been used before in this context. Applied on CrunchBase’s dataset on Start-ups it enables me to provide answers about different survival and exit chances. The CrunchBase platform supplies a reliable, extensive and systematic dataset about Start-ups for research purposes. The
1 Available under FDP’s homepage:
6
availability of such a dataset is anything but self-evident, because Start-ups do not want to reveal too many details about the profitability of its business to outsiders, other than potential investors.
The results indicate that successful Start-ups with regard to both measures are generally backed by VCs. VCs have a 95% higher chance to survive and a 14% higher chance to exit considering the simple model. However, after examining more closely the various determinants of success the outcome shifts. Business Angels turn out to back more companies that accomplished an exit if the model controls for the liquidity provided by the investor. Then BA backed Start-ups have an 84.6% higher probability to exit. These results show that the non-monetary support of BAs should not be underestimated and is of tremendous utility for Start-ups.
Both BAs and VCs are active during the earliest stage of a Start-up but have different targets and ways to address these targets. Related literature suggests that companies should favor VCs. Schmidt (2013) examines 252 Series A funding and concludes that BA backed companies perform worse compared to their VC backed counterparts. However, collaboratively supported ventures have the highest exit rates. The methodology he uses is a logit model, which has some similarities with the proportional hazard model used in this thesis. Chemmanur and Chen (2014) show theoretically why entrepreneurs should favor VCs over BAs, although they require higher rates of return. The basic PHM in this thesis suggests the best survival and exit rates for VC backed companies. For both measures, the success of VC supported Start-ups is also higher than of mixed supported Start-ups.
Do the different investment models complement or contradict each other? By analyzing the Early Stage of newly founded companies, it appears that the first funding round is often done by BAs, whereas VCs approach Start-ups in later funding rounds. Consequently, there are many companies firstly funded by BAs, followed by a funding from VCs. Hellmann et al. (2013, 2014) and Goldfarb et al. (2013) explore mixed funded companies intensively. They evaluate in which interrelation the VCs and BAs stand to each other. They pick up the work of Leavitt (2005) who shows that after liquid VCs enter a company the entrepreneurs leave BAs behind as “Burned Angels”. In contrast to Schmidt (2013) both Hellmann and Goldfarb examine that a collaborative support by VCs and BAs is not beneficial to Start-ups. Results of my paper show that a collaborative support is only beneficial in terms of survival for entrepreneurial firms if the initial investor is a BA but disadvantageous if the initial investor is a VC. These findings support the idea of VCs being “patient” investors during the first years of newly founded companies.
The remainder of this thesis is structured as follows: Section 2 introduces the two examined investor types and a theoretical background of the private equity sector in general. Subsequently, in Section 3 the development of the hypotheses is outlined. The CrunchBase dataset and important variables are the content of Section 4. The methodology used to answer the research question is a
7
proportional hazard model. An introduction and the fitting of the Cox model are given in Section 5. Section 6 finally applies the empirical framework and presents the results of the conducted regressions. The next section puts the results into perspective and discusses the findings with regard to the related literature. After a description of the study’s limitations and potential future research the final section concludes and sums up this thesis. All tables and references are available at the end of this work.
2 Literature Review
The literature review is divided into four sections, whereby the first section introduces the private equity environment in detail. Subsequently, and by quoting the relevant literature I explain what the reader has to expect when talking about Business Angels and Venture Capitalists. Finally, I point out from which perspectives other authors investigate the relation of the two investor types and which conclusions they draw in regard to my research aim.
2.1 Private Equity
Companies which are not listed on global stock markets and want to finance investments have to rely on external debt or private equity. The private equity segment contains investors like funds or wealthy individuals, who directly finance other companies to acquire equity ownership in return. Following this business model the private equity firms search for promising companies who can potentially increase their market value to divest its shares profitably in the upcoming years. Potential targets of private equity investors are often high tech companies because of the high growth opportunities. A performance comparison between private and public equity depends on the compared benchmarks and time ranges. Some studies come to the conclusion that private equity outperforms public equity and others vice versa. For example Robinson and Sensoy (2011) conclude that private equity outperforms public equity by 1.5% per year, whereas Phallipou and Gottschalg (2009) come up with a slight underperformance of private equity.
In their paper Robb and Robinson (2014) investigate capital structure decisions by Start-ups and reveal a “financing pyramid”. Entrepreneurs prefer outside debt over owner equity followed by debt from insiders. Fourth in the pyramid is outside equity. However, if a Start-up is backed by a private equity investor this investor plays a crucial role, with on average more than 50% of the gathered capital. The used dataset of Robb and Robinson is the Kauffman Firm Survey, which is a longitudinal survey of newly funded businesses in the United States and includes almost 5000 companies. They point out that Start-ups which are outside equity backed are mostly high-tech firms and have intellectual property in some form which falls in line with the CrunchBase dataset.
8
Private Equity investors are divided into three groups, depending on the stage of the firms they typically invest in. Firstly, Seed or Venture Capital Stage, where investors provide capital to innovative Start-ups for early expansion. The scale of these investments is relatively small with up to 10m (million) USD, but comes with a lot of failures and few great successes. Investments in the Growth and Development Capital Stage are for companies which require funds for future growth but already have an existing product or service. The typical financial instruments being used in this more expensive stage are convertible bonds. The stage with the highest turnover is Buyout, where the target is mostly financed by a high debt level. Those companies have a leading market position in its segment and stable cash flows. By implementing a business plan, the private equity fund targets to create value before exiting after 3-6 years.
Those are the three stages for private equity investors to acquire shares of a start-up with increasing deal size. This study mainly focuses on the Seed/Venture Capital stage and Development Capital stage with BAs and VCs as the typical investor.
2.2 Business Angels
Business Angels are a heterogeneous group of investors. This feature is what differentiates BAs from VCs the most. The original definition describes Business Angel as a high net-worth individual who holds shares of a Start-up. Nowadays BAs summarize a broader variety of investors. They can be high net-worth individuals but also just family and friends or syndicates of individual BAs. Some of them are active investors, some are passive and they do not necessarily have a past within the specific industry they invest in. An institutional difference between BAs and VCs is that BAs invest their own money. In doing so a BA supports a Start-up with strategic advice and often has a distributed network in the specific segment from which the entrepreneur profits to a great extend as well. BAs are the primary source of external capital when there is often only a promising idea for a business which faces financial constraints. Considering that the default rate of entrepreneurial companies at this stage is the highest, the BAs need some lucrative exits to compensate the defaulted investments. The BA’s support is characterized by a high commitment and mentoring. BAs mostly have only a few parallel investments and most of them are in one specific industry in which they have been active themselves. According to Politis (2008) this enables the investors to work close together with the Start-ups and be incorporated in strategic and future-oriented decisions to ensure success. Furthermore, Kerr et al. (2014) come to the conclusion that not the capital is the main support of a BA but rather their contributed know-how. Whether this contributed expertise outweighs the extraordinary high funding of VCs is also investigated in my paper. Madill et al. (2005) sum up that, since BAs invest their own capital, their investments do not have a time limit for an exit in contrast to VCs.
9
2.3 Venture Capitalists
A Venture Capitalist consists of General Partners (GP) that collect capital from Limited Partners (LP) to launch a fond. A fund is typically alive for about 10 years and during this time the fund’s General Partners invest the capital in different Start-ups. For their work the GPs receive a management fee between 1% and 3% of the committed capital as well as a variable fee of around 20% of the profits. Those profits are generated by successful exits of Start-ups they invested in. VCs invest in a greater variety of ups and at a later point in time compared to BAs. Due to the greater variety of Start-ups, the active involvement is relatively small. However, VCs also depend on some successful exits and try to govern a Start-up towards the direction of their interest as shown by Goldfarb et al. (2013) and Kaplan and Stroemberg (2003). Both papers worked out that VCs press for more control rights than BAs. Furthermore, the average invested amount is higher compared to a BA.
A crucial difference to BAs is that VCs have a dual identity in the investment process as both principal and agent. For their limited partners, they are agents and for the Start-up they are principals. This situation can lead to a conflict of interest as Chahine et al. (2012) showed. On the one hand, the GPs face a short-term pressure to generate profits by exits. On the other hand, a value maximizing strategy may be to hold the shares of the Start-up for a longer period. A study of Hellmann and Puri (2002) identifies a significantly shorter time to launch a product on the market by VC backed Start-ups.
2.4 Business Angels vs. Venture Capitalists
In 2011 OECD published a report which indicates that both BA and VC markets have approximately the same size.2 For a long time, Business Angels remained relatively uncovered by the academic literature. Though, this changed in the recent years also because of such reports. This thesis affiliates itself in the growing literature that investigates the interrelation between the two investor types and their impact on entrepreneurial firms.
The work probably closest to my paper is a book by Schmidt (2013). He studies an extensive dataset of Series A funding rounds to whether the investor choice has an impact on Start-up’s success. Series A financing is the first time a company obtains VCs support. His sample provides the same results if mixed support is ignored. However, the findings for Start-ups which are backed by a mixed group of investors are different in my thesis. He finds out that ventures have highest exit rates if financed collaboratively.
Chemmanur and Chen (2014) as well as Schwienbacher (2009) try to capture and consolidate the available information about this topic in a theoretical framework. The fact that they differ in crucial assumptions shows how much the opinions vary in this debate. Chemmanur and Chen (2014) think
2 For more information about the role of angel investors see:
10
that only VCs are capable of adding value to an entrepreneurial firm. Based on this assumption they come up with an explanation wherefore entrepreneurs initially obtain capital from BAs and then from VCs. Schwienbacher (2009) comes to another conclusion. His first assumption is that both investors may add value to a Start-up. Then he includes his knowledge about the higher liquidity of VCs that enables them to finance later funding rounds. BAs cannot participate in such subsequent rounds and that’s why BAs have higher value adding incentives than VCs. Since these two papers work with a theoretical model about the entrepreneur’s choice of capital structure they have to make such assumptions. Their papers give a valuable insight for the elaboration of my hypotheses.
Many other researchers explore the advantages and disadvantages of the investors in empirical papers and from different perspectives, whereby it is still a fact that the academic literature on VCs is much more extensive than for BAs. Da Rin et al. (2011) provide a detailed summary of this topic. As described above, BAs and VCs have different characteristics and as a result they occupy other positions in the Start-up.
Lerner (1994) and Ehrlich et al. (1994) examine one of those characteristics in their studies. The two types of investors follow different paths in terms of governance and monitoring. Nofsinger and Wang (2011) show that as opposed to BAs, which rely on trust and relational governance, VCs record their agreements in detailed contracts and ask for higher profits in return of a funding. The degree of investor’s specialization on particular investments plays a crucial role as demonstrated by Sorenson and Stuart (2001) and Politis (2008). Both take up different positions when it comes to the specialization of the investors. First-named base their conclusion on VCs that issue industry and/or country specific funds. This facilitates them to leverage their expertise for more effective due diligence and to focus resources essential for venture development. Consequently, VC networks are expected to be wider compared to BAs. What in turn might lead to the hypothesis that VCs have a stronger endorsement effect than BAs. On the other hand Politis (2008) points out that BAs typically have a smaller number of parallel investments than VCs what enables them to focus their attention on these companies. Furthermore, Kerr et al. (2014) take position in favor of BAs, with the argument that it is not uncommon that these high net worth individuals are some of the most sophisticated and active investors, with immense industry experience.
Typically BAs are the first investor in the earliest stage of a Start-up, followed by VC investments in a greater scope. This assumption is proven not only by my thesis but it is also empirically proven by Goldfarb et al. (2013). The average funding amounts in the first round are smaller compared to subsequent rounds what leads to BAs having weaker control rights than VCs. This finding is kind of paradox since the BAs demand less control rights in a stage when they take a higher risk.
11
As stated above, it is not uncommon that BAs and VCs support a Start-up collaboratively. This situation and the associated chances of success are investigated in this paper as well. Hellmann et al. (2013) opposed the contradicting hypotheses of the two investors being complements or substitutes. Analyzing a dataset of Start-ups from British Columbia, they find more evidence for the second hypothesis, which is in line with the results of this paper. In this thesis, Start-ups show a significantly better performance if they are backed by only one investor type for both success measures. In the following year Hellmann and Thiele (2014) published a paper about the relationship of BAs and VCs if a Start-up mixed both types of funding. The market environment in form of angel protection, market transparency and lower Start-up costs play a crucial role for the relationship of the investors but due to the higher liquidity of VCs they do not depend on BAs anymore and get in competition with each other. These results are based on a theoretical model that endogenously derives the size and competitive structure of the BA and VC market.
By comparing specific characteristics on basis of a dataset the above mentioned papers cover only a part of the overall puzzle of which type of investor is superior and should be preferred by entrepreneurs. However, there are two papers that try to answer a related question to mine, one for survival as success measure and one for exit as success measure.
Puri and Zarutskie (2012) compared the lifecycle dynamics of VC and non-VC financed firms, also to evaluate whether VCs disproportionately back firms with longer survival. They identify a lower failure rate for VC financed firms. A main difference to this paper is that their control group is not narrowed to BAs. Nevertheless, their study supports the findings of this paper that VCs provide enough resources to ventures to survive the observed time period. The high experience of VCs to realize IPO’s of their financed firms examine Ritter et al. (2011), whereby they point out that the GPs of a fund have a good network to underwriters as well as to star analysts. By the coverage of all-star analysts, the VCs aim for an extreme short-run run-up of the stock price after the IPO. The results of this thesis suggest that the high frequency of IPO’s and exits overall by VC backed firms is not induced by their management or network support but only driven by their huge funding amounts.
3 Hypotheses
The aim of this paper is to figure out which investor is more likely to stimulate success of an entrepreneurial company, whereby the focus is on high-tech companies. In this case survival time in years and exits via acquisition or IPO are success proxies. The two indicators of success are correlated and affected by similar variables. However, a successful exit is the investor’s ultimate goal and goes beyond simple survival until the end of the observation period of the used sample. The first hypothesis is as follows:
12
Hypothesis 1: The type of investor, by which a Start-up is backed, significantly affects the survival
and exit chance of a Start-up.
If there is any gap in success this have to be induced by the differences of the two investor types. A BA financed Start-up profits from the close guidance as well as the network of the Angel. Werth and Boeert (2013) show in their paper that the better the network of the BA in a specific industry the better the performance of the Start-up. However, according Hochberg et al. (2007) this effect of networks coincides also for VCs. Hence it is unlikely to be the crucial factor in favor of one investor assuming that both have the same possibilities to build a network.
Another difference is the invested amount and the financial situation of the investors in general. Although BAs are high net worth individuals, in general VCs invest higher amounts of money in Start-ups and have the required liquidity for further financial injection if it is appropriate. Leavitt (2005) studied the vulnerability of BAs when VCs as liquid investor negotiate contracts in their desire, leaving back the BAs as “Burned Angels”. The reason for the higher investments is also due to the later point in time when VCs finance a Start-up (Berk et al., 1999). At this later stage the Start-up requires higher funding for further growth but it is also a fact that many Start-ups never reach this stage because they already failed. Consequently, the pool of Start-ups when VCs invest is biased regarding success compared to the totality in the earliest stage when BAs invest.
A concept of behavioral finance is overconfidence, which describes the human nature of inflated self-esteem and is proven by Odean et al. (2011) and Malmendier and Tate (2005). If you were successful in the past it is more likely that your level of overconfidence is high. Both, the high net worth BAs and the GPs of Venture Capital funds, can refer to impressive careers. Nevertheless, in a VC firm with several partners and employees there are control mechanism implemented, which also monitor actions driven by overconfidence. An autonomous BA might lack such supervision and is more likely to go for Start-ups which are risky investments under neutral observation.
Most of the time BAs have only a small number of active investments and are in a lively exchange with the Start-ups they invested in (Robb and Robinson, 2014). The know-how in the relevant industry is often extraordinary and the Angel can refer to expertise of a long successful career. On the contrary, a VC has many parallel investments, but it is difficult to evaluate how many investments are solely supervised by one GP. The closer collaboration between investor and Start-up exists with BAs.
Most of the differences signal that in case of doubt a VC backed Start-up will be more successful. This leads to the following second hypothesis:
Hypothesis 1a: Get funding is always a positive signal for the Start-up that its business idea has high growth potential. However, the supportive factors induced by Venture Capitalists are higher
13
than for Business Angels and hence the survival rates of Start-ups are higher for VC backed and highest for mixed backed firms.
Additional importance has the different time horizon and investment target if the rate of acquisitions and IPOs is considered. Bruton et al. (2010) show that investor type does matter for the performance of IPO firms. Furthermore, by comparing British and French IPOs they find out that also the institutional environment of the country of origin in connection with the investor lead to contrasting performance outcomes. The time horizon for VC investments is limited to the life of the fund. A BA on the other hand doesn’t have to take into account any deadlines. Under pressure to exit its investment position, a VC is more eager to push and prepare a up to an IPO or offer the Start-up as acquisition target. Since the complete business strategy of a VC takes aim to resell in a limited period, the third hypothesis reads as follows:
Hypothesis 1b: The chance of a successful Exit is highest for Start-ups financed by VCs. The strategy of VCs is designed to precipitate Exits and VCs have the capacity as well as the experience to realize an acquisition or an IPO.
4 Datasets and Variables
To analyze and measure success of Start-ups, an extensive dataset is needed which indicates investors, change of ownership structure as well as information and date of failures. Such a dataset is provided by CrunchBase and had to be extended and merged with other sources for additional variables and checking purposes. Beside some restrictions about age and filtering of Start-ups that are neither backed by an Angel nor a VC this study relies on random sampling in contrast to Whalen (1991) who uses a choice based sampling approach in his paper about bank failure, also analyzed by a proportional hazard approach.
4.1 CrunchBase
CrunchBase was founded in 2007 by Mike Arrington who is the founder and former editor in chief of TechCrunch.3 Starting as a crowd source database to track Start-ups only covered on TechCrunch, it nowadays is the most extensive dataset about ventures worldwide. CrunchBase describes itself as “the leading platform to discover innovative companies and the people behind them”.4 The platform is located in the Silicon Valley and was originally founded for entrepreneurial companies out of the high-tech industry and still covers this market almost completely. Active users of CrunchBase are not only entrepreneurial firms, but also investors as BAs and VCs, Organizations and founders. Through
3
The website of TechCrunch with information and latest news about the news platform is available under:
https://techcrunch.com/ 4
14
this channel all active members can find potential business partners, position oneself on the market and follow the recent trends of the industry. CrunchBase also includes data about established firms which do not fulfill the characteristics of a Start-up anymore. The extensive dataset does not only include data about the companies and its products. Moreover, the platform also contains information about the people and investors behind the Start-up, location, current status, raised amount per funding round and much more. Professionals can edit the available, public information and subsequently this pass through an approval process before going online. Nevertheless, this form of editing by outsiders might be unfavorable for the affected company in certain circumstances.
Two variables are not available via the CrunchBase platform. Firstly, outgoing investors are not tracked and consequently one has to assume that the investors of the first round are still active investors at later stages. Secondly, the exact acquired ownership of investors is not traceable through the different exports.
CrunchBase gives unrestricted access for research purposes, which is used for this thesis. Data exports are available in form of Excel and CSV reports whereby in this case I exploit the CrunchBase REST API because the platform recommends it on their homepage as the most complete source of data. I merge the different required API requests (funding rounds, funded companies and investors) to get the final dataset which is used to run the Cox regressions.
4.2 Fit the Dataset and Variables
This section describes the data analyzed in the thesis. The CrunchBase API request for the funding rounds provides a longitudinal dataset with detailed information about the funding history of 75,172 companies with overall 130,902 rounds of funding. This particular request was executed on the 30th of April 2016 and due to the daily updates of CrunchBase there will be deviations if a download is conducted afterwards. Crucial for this study is the variable funding round type which is given for each observation as well as date of funding and the raised amount.5 The Start-ups have to declare here which type of investor they want to address with a funding round. For simplification Seed and Angel are consolidated to Business Angel and then all types different to Business Angel and Venture Capital are summarized as Other. To measure success I merge the request for companies, which besides
founding date and total raised amount, also contains the variable Status, with the primary dataset. The variable Status indicates if a company is active, failed or has been acquired. For failed or acquired companies an additional variable with the date of the event is available. Furthermore, the CrunchBase export about IPOs and acquisitions is merged as well as the ISO 3166 code-list to assign the country codes to continents.6
5
Noted variables in the text are written in italics and as they are issued in the CrunchBase exports
6 The ISO 3166 was introduced 1974 and is a 3-digit coding list for countries. For more information see
15
To run the Cox model I have to calculate age variables based on the specific entry dates. These different age variables are current age, closing age, age at funding round, whereby I drop negative ages since it is assumable that those are entry errors. Furthermore, ventures that did not disclose necessary variables as funding date or status are dropped, because I cannot measure success for those companies in my model. There are 41,897 firms left in the sample with 88,292 corresponding funding rounds.
Additionally there are dozens of different industry categories which I summarize by logical mapping to five core categories as can be seen in Table 1. With more than 40% of the whole sample the category with the most records is Information Technology what is not surprising, thinking of the origin of CrunchBase in the Silicon Valley. As a proxy of companies leverage the fraction of the
raised amount in debt over the total raised amount is created. To get a more complete sample I calculate approximations for the closing age. In many cases the status of the firm indicates “closed” but the closing date variable is missing. For those Start-ups the closing age is estimated based on the other closed companies. Over all closed companies the average time span is calculated between the last funding round and the closing date. Then I add this average to the age when the last funding round happened but only if the closing date is missing. The same approach is deployed for the exit event, if IPO date or acquisition date is missing.
CrunchBase was founded back in 2007 and since then it developed to the platform with probably the highest market coverage for ventures. However, in the dataset are also many observations about funding rounds that happened before 2007 which are recorded retrospectively. For testing the equality of survivor function the creation of a round age group variable is done. The tests do not reject the null hypotheses that the survivor functions are equal for the age groups older than 10 years in contrast to the groups younger than 10 years. Hence, I exclude entrepreneurial firms which are funded earlier than 2007. By extension this leaves a sample of 30,044 companies and 57,424 funding rounds. Finally, leftover firms without information about continent and industry are dropped what leads to a reduction to 52,484 rounds done by 28,212 companies.
4.2.1 Main explanatory variables
The central explanatory variables are the investor types, whereby it is differentiated between Business
Angel, Venture Capitalist and BA+VC. The applied procedure is as follows: If a company is exclusively financed by a Business Angel or a Venture Capitalist throughout all funding rounds they are in this specific ranking. If a company was only financed by one of those two investors and by investors which fall into the category Other, they are still classified under Business Angel or Venture
Capitalist. There is a third case, which is headlined as BAs+VCs. If a Start-up is financed by both types of investors it is classified as such, whereby it is irrelevant whether the funding by VCs is higher
16
than BAs or vice versa. For all categories a binary variable is created taking 1 if the company is financed by this investor and 0 otherwise. This leads to three variables of interest, which are tested in different combination to answer the research question if Start-ups are disproportionately backed by a specific investor. The regression tables report the hazard ratios for the type of investor and not the coefficients. The interpretation of the hazards is as follows: The baseline value is always 1 and to this the given ratios has to be compared. A figure greater than 1 indicates for survival as success measure that the chance for failure is higher for Start-ups that got the “Treatment” of this specific investor. Accordingly, a value between 0 and 0.99 can be interpreted as an advantage for the entrepreneurial company in terms of survival.7 Those interpretations are always done in comparison to the rest of the sample, which functions as the control group. That is why it is of special importance to consider all types of investors which are analyzed in a particular regression because the inclusion of a second investor may change the control group and hence the hazard ratios. In contrary to coefficients in a PHM the hazard ratios can be interpreted in absolute terms. For example a hazard ratio of 1.1 for an investor type means that the companies in his portfolio have a 10% higher chance to fail per year compared to those companies which are in the control group with “Treatment”=0.8
4.2.2 Control Variables
The regression outputs might indicate that one investor type is superior to the other, although the reason is rather that this type of investor is mainly active in a more successful industry for example. To avoid such biased conclusions it is necessary to include controls in the model.
To check for influential variables I include several controls in the model which are the Round
Count, Industry, Continent, Years, ln Total Funding, Company Size and Leverage.
For the controls Round Count, Industry and Continent I created per category a binary variable. Since the PHM get along without an intercept, one dummy variable to omit has to be chosen. These are 1st Round, Information Technology and US and Canada. As a consequence the interpretation of these covariates is in comparison to the omitted category. The assumption for Industry and Continent is that they stay the same over time. I have to do this, because those two variables are only available for the first Round of a Start-up. However, since it is very unlikely that a Start-up completely abandons the region or industry where it is found, the assumption should be valid in most cases.
Furthermore, year fixed effects are included in several regressions to control macroeconomic effects over the observed period. Those could be a significant change in interest rate or a bubble in the real estate market. A bubble in the real estate market would lead to easier accessible loans due to
7
For a better understanding of the shape of hazard rates see the paper of Aalen and Gjessing (2001) 8
The time units in the sample are years and consequently the interpretation of hazard ratios in the PHM has to be made in terms of years. Months, weeks or even days as time unit would be possible but the results would lead to the same general conclusion although the hazards would be different
17
higher potential collaterals, what might lead to a situation when only entrepreneurs with risky projects ask for support by BAs.9
Company Size is based on the number of employees of the Start-up. The number of employees is seldom available in the sample. Hence it is rarely included in the model, because the variable alters the sample.
A proxy of Leverage is also included in form of a dummy variable to evaluate whether Start-ups which got debt are significantly better than others.
Additionally to binary control variables, there are continuous covariates included in the model. The interpretation of those variables is the same as for the main explanatory variables. For example a ℎ < 1hazard for ln Total Funding in the survival regressions shows an increasing likelihood to survive the more funding a company accumulates. This covariate is of special interest, since it controls for the huge difference in funding amounts between the different investors.
4.3 Summary Statistics
The analyzed dataset for this thesis contains information about 52,484 funding rounds done by 28,212 companies. A summary of the sample from different perspectives is given in this section. The standard deviation is high across all tables what brings out that a funding round is not limited by predefined levels but vary from no provided resources at all to comparatively extreme high amounts.
Figure 1 shows an overview of funding rounds per year. The increasing number of funding rounds reflects the importance of CrunchBase since 2007. When there were only 494 registered rounds in 2007 it scale up to 12,292 rounds in 2015 what represents a twentyfold increase. The drop in 2016 affiliates to only five months observed up to May. As mentioned above the standard deviations of the funded amounts is very high what is also reflected in the mean values plotted in Figure 1. Those are driven to a great extent by large outliers. Those outliers can be several billions of dollars. For example, a funding round in April 2016 about 2.3 billion USD financed by VCs what is partly explaining the high value in 2016.
The median raised amount shows an opposed trend over time to the number of rounds. It decreases after 2007 until 2013 from over 1.2m (million) USD to 0.8m USD. In this period also the number of failures increases significantly. However, the reason for this is most likely not only a development caused by lower liquidity of the PE market and the market overall. Instead the reason is behind the higher popularity of CrunchBase for Start-ups as well as for investors what leads to higher
9 This effect of real estate shocks on corporate investments is studied by Chaney, Sraer and Thesmar (2012)
18
inflows of smaller companies and also investors specialized for smaller investments. Hence, the later years in this sample are more representative due to the greater market coverage.
Figure 1:
Raised Funding per Year
Notes: This Figure shows the average raised funding amount from 2007 until May 2016 on the right y-axis and the number of funding rounds on the left y-axis.
Figure 2 shows the same output as Figure 1 but categorized for the different types of investors. Over the last ten years the number of BA and VC backed funding rounds increased to more than 10,000 per year. Especially the number of provided financing by BAs shows a boost. Starting with a lower count in the years from 2007 until 2010, the turning point is reached at 2011 and since then the highest fraction of rounds represents BAs.10 The mean raised amount in the 25,523 (~49%) Angel rounds are 0.76m USD. The development over the last years underlines the increasing importance of BAs, also caused by the growing importance of BA syndicates. However, the average raised amount per year remains relatively stable.
With ~44% VCs finance a major part of funding rounds and steadily increase since 2007. The median amount gathered by Start-ups in this category is 4m USD, with a peak in 2015. This supports the findings of Kaplan and Stroemberg (2003) that VCs provide ventures with much more liquidity
10 This is the result of the adjusted dataset for this thesis. The overall CrunchBase dataset might show a slightly
different picture for later download due to the daily updates of the data
0 2 0 0 0 4 0 0 0 6 0 0 0 8 0 0 0 R a is e d A m o u n t p e r R o u n d i n 1 0 0 0 U S D 0 5 0 0 0 1 0 0 0 0 1 5 0 0 0 # R o u n d 2006 2008 2010 2012 2014 2016 Year
19
than BAs. By comparing the median, one can also ensure that this result is not only driven by large outliers within the VC segment.11
Figure 2:
Funding Characteristics per Investor
Notes: This Figure shows the average raised funding amount split for investor types from 2007 until May 2016 on the right y-axis and the number of funding rounds on the left y-axis.
The category Other summarizes ten different investor/funding types as for example debt, post-IPO equity or undisclosed investors. Those investors provide mostly equity but also debt and are active in 3,744 funding rounds. With ~7% it represents the smallest class of the sample and is rarely present in 2007 and 2008, what is due to the exclusion criteria described in section 4.2. Even though the median funding amount by those investors is with 0.77m USD much higher compared to BAs median and lower to VCs.
The average amount provided by a VC is ten times higher than by a BA. Hereon back to two main assumptions. Firstly, the higher amount indicates that VCs indeed finance ventures in mature stages, assuming that higher funding is required at later stages. Secondly, if the companies financed by VCs have more money at their disposal it is very likely that they can survive a longer time period although they are in a non-profitable state.
11
Median for Business Angel in the observed time period is 0.38m USD
0 5 0 0 0 1 0 0 0 0 1 5 0 0 0 R a is e d A m o u n t p e r R o u n d i n 1 0 0 0 U S D 0 2 0 0 0 4 0 0 0 6 0 0 0 # R o u n d 2006 2008 2010 2012 2014 2016 Year # BA # VC # Other
20
Table 1 gives an overview of the 28,212 companies in the sample by analyzing it from four different perspectives. Firstly, the current status of the companies, followed by the represented industries and continents. Finally the treatment groups which are classified in three categories to conduct the survival and successful exit analyses. The company with the highest total funding is Uber that collected in 15 rounds over 10 billion USD. For this table the USD figures do not represent values in a specific round but the total funding by a company up to the 30th of April 2016.
The majority of the 28,212 samples’ companies (~89%) are still operating and the rest of ~10% successfully exited (~6%) or closed its business (~4%). The small quantity of closed businesses might seem very low and biased at the first sight but since the greatest part of companies just joined CrunchBase in the last five years it gets a realistic figure. Unsurprisingly, the companies which undergo a public offering or are acquired receive the highest funding and closed companies the smallest amounts. The most recent IPO represented in the sample is one by an Indian firm, specialized on micro financing, called Ujjivan Financial Services from the 27th of April 2016.
There are almost 50 different industries that can be distinguished by CrunchBase. However, for the purpose of this thesis I accumulate them with reasonable criteria to five different industries. As mentioned before CrunchBase initially started as a platform for US based high-tech companies. This matter of fact is reflected in the high share of Start-ups active in the Information Technology (~44%) whereas Financial Services (~5%), as a category, plays a minor role. Nevertheless Science and Health
Care (~11%) industry raise the highest amounts (median).
The highest share of companies in the sample is from the US and Canada (~62%), followed by
Europe (~20%). Native in Latin America and Caribbean is every 50th company. However, Start-ups of other parts of the world (Africa, Asia and Australia) collect the highest amounts of total funding with 20.4m USD on average, what is mainly driven by the liquid Asian private equity market.
The last section of Table 1 presents the classification of the companies based on their investors as explained in Section 4.2.1. The data shows that there is a higher quantity of BA backed and a lower quantity of VC backed firms and it is relatively rare that a Start-up is financed by both types of investors. Furthermore, it is visible again that a VC provides much higher funding than a BA. The high discrepancy between mean and median values clarifies again that some huge outliers shift the mean to high figures.
Table 2 illustrates the distribution and key figures within the sample for the chronology of funding rounds instead of years as in Figure 1 and 2. This cross-table presents several characteristics of the founding rounds, namely Investor, Status and Industry. The intention here is to get an impression about the dynamics of funding by different investors over the lifetime of a Start-up. Besides, all observations of funding rounds greater than 4 are summarized for clarity.
21
With a count of 28,212 most observations are first rounds and the number decreases the higher the funding round count is. A reversed progression is taken by the mean figures. Starting with an average of “only” 2.5m USD in the first round it goes up until 10.8m USD in the fourth round and higher.
The Investor category of Table 2 supports the findings of Goldfarb et al. (2013). Firstly, the pattern of the frequency shows that in first rounds mainly BAs are involved but this ratio changes in subsequent rounds in direction of VCs. Moreover, the mean invested amount of VCs is ten times higher than that of BAs.
In terms of the category Status the table has to be read as follows: For example, companies who are indicated as closed in the columns for the first round actually closed their business after an external investor firstly submitted his support and so forth for the subsequent rounds. Unfortunately, data for companies which did not even do a first funding round so far are not available. An interesting finding here is that the number of failures does not extremely decrease over rounds. This shows that despite several investors did a valuation and decided to invest in a particular company there is still prone to failure. On the contrary, the trend for acquisitions and IPOs is as expected. The more funding rounds a Start-up passed through the more likely it exits. On the one hand, this is caused by a higher number of investors who support the company with liquidity and expertise, on the other hand during several funding rounds the visibility on the market increases. This holds not only for the happening (Count) of an IPO, but also for the raised amount prior to public offerings. For example a public offering after the first round realized before 13.7m USD and one after the second round significantly more with 17.1m USD. This part of the table gives an important insight which distribution the later analyzed events have within the sample.
The Information Technology segment executes the main share of funding rounds. Striking is the development for companies of the Science and Health Care industry. Starting as the industry with the highest average funding the other sectors move closer in subsequent rounds until they surpass science and health care in the +4th round. The reason for this is most likely the related expenses to develop a service or product. Whereas entrepreneurial firms with scientific background require huge R&D investments in the early stage, other Start-ups require less investment in the beginning. For further expansion these other Start-ups have to enter new markets what is accompanied by high costs for new offices, marketing et cetera.
5 Empirical Framework: Cox Proportional Hazard Model
The used methodology to test the three stated hypotheses is a Cox proportional hazard model (PHM). Sir David Cox (1972) developed this model. After publishing it 1972, it was usually applied for studies of medicine and biology as for example by Crowley and Hu (1977) and Binet et al. (1981).
22
The Cox model is one possibility to analyze survival time data beside multiple discriminant analysis (MDA), logit and probit models. The main idea of this model is the proportional hazard assumption which assumes a constant impact of a certain variable on the risk of failure. In general, the Cox model analyzes the hazard for a predefined event and provides an estimate of the treatment effect on this event. Consequently, it cannot only be used for research about survival time (survival as measure of success) but also for other time-to-event data as the probability of successful exit. In contrast to MDA, logit and probit models it is not necessary to make any assumptions for the distribution.
Survival data can usually be described by its density function ( ), its survival function ( ) = ( ≥ ) , its hazard function ℎ( ) = ( )( ) and its failure function ( ) = 1 − ( ). With ( ) as the chance of a Start-up to survive t. The distribution of time to failure could be expressed in terms of ( ) or with a hazard function ℎ( ) (Henebry, 1996). However, the hazard function is more commonly used because of the recommendation by Cox and Oakes (1984). Whalen (1991) applied the proportional hazard model for bank failure and his notation is adopted in the following equations:
ℎ( ) = lim→ = ( < < + | > )=− ′( )( ) (1)
This function ℎ( ), also called Mill’s ratio, can be interpreted as the instantaneous risk to fail after a Start-up survived until time t and T is the estimated time to failure. On the right hand side is the Survival function ( ) with a reverse interpretation. Time t is the independent variable and represents the time to failure in the hazard function.
Based on the hazard function the Cox model is as follows:
ℎ( |#$) = ℎ ( )exp (()#)+ ⋯ + ($#$) = ℎ ( )exp ((′í#,$) (2)
or equivalently as a linear equation:
ln ℎ( |#$) = ln ℎ ( ) +()#)+ ⋯ + ($#$ (3)
Where #$ denote a collection of characteristic variables for entrepreneurial company i and ( is the regression coefficient of how much the variable affects the hazard. Applying the studies of Lane et al. (1986) about bank failure on business failure of new firms, the following statements can be made. The baseline hazard function is ℎ ( ) and can be seen as the average hazard function for a centralized Start-up in the sample. This baseline is the hazard for a Start-up if all predictors are equal to 0 and functions as a substitute for the intercept in the PHM. Hence, the explanatory variables only multiply the hazard of the Start-up by a constant factor. The test for the proportional hazard assumption for the
23
sample of this thesis shows that this is only true for younger Start-ups. The main reason for this is the insufficient number of funding rounds which are observed before the launch of CrunchBase in 2007. Therefore the sample is restricted to companies founded after 2007. Section 4.2.2 presents several explanatory control variables. However, the main variables of interest are survival of a Start-up and successful exit.
The following equation expresses the likelihood .$ that company i fails at time t compared to an average company: .$ =∑ 0ℎ( $|#$) 1( $)ℎ2 $3#14 5 16) = exp (#, $() ∑516)01( $)exp (#,1()= exp (#, $() ∑1∈89exp (#,1() (4)
With :$ as the risk just before time $. Following Honjo (2000) the partial likelihood function L in the multiplicative hazards model under consideration of the type of investor has this form:
. = ; <∑ 0exp (#,$() 1( $) 5 16) exp (#,$()= >9 = 5 $6) ; <∑ exp (#exp (#,$(),1() 1∈89 =>9 5 $6) (5)
Another advantage of the proportional hazard model is that it can work with right-censored data. The used dataset in this thesis is right-censored, what is that many of the Start-ups are still active up till the end of the observed period and didn’t experience the event of failure or successful exit. Left-censored companies in this case would be companies with unknown funding round dates. Those Start-ups are excluded from the sample, although it introduces bias into the analyses, as described in a note by Amemiya (1999). An adjustment for left-censored data is a very complex issue and goes beyond the scope of this thesis, although there are approaches to correct for it for example by D’Addio and Rosholm (2002).
5.1 Events in the Cox PHM
Success can have several appearances and depends on the context. However, in this thesis success is tracked in the occurrence of two different events.
Start-ups are newly founded businesses which have to prove that they can be profitable in a niche by offering an innovative service or product. Unless they will prove this they are sooner or later dropped out of the market. The event of closing is the first indicator of success, which is set in the PHM model. All funding rounds per Start-up are available in the dataset and thus usable for the model. For each observed funding round the binary variable Survival Status indicates either 1 if the company is still active or 0 if the company closed its business. Hence, Survival Status is always 1 until the last observation per Start-up, when it can change to 0 in dependence of the current status. For
24
companies which are still active until the end of the observed period the event did not occur and hence their data is censored. An acquired or IPO status of a Start-up also takes 1 for the Survival status, because the company did not experience the event of failure. In this thesis, I do not consider multiple events per company in form of more than one failure or exit.
This is different for the second event which is set in the PHM. An exit is an attractive opportunity to cash out their position in the Start-up and for most investors it is the final goal of an investment to cash out with one big transaction. Hence, the exit is an event which is observed as a success for both involved parties. Exit in form of a public offering or an acquisition is the second success measure. Hereby it is of importance that all funding rounds occurring after the event are not considered. This means that a funding round for post-IPO equity for example is not included in the model. The experience of the event exit is always the last observation for a Start-up and the variable Exit Status takes the value 1 for exited firms and 0 otherwise. Figures 3 and 4 give a brief overview about the count of events over the observed period. Overall there are 1,345 failures with a peak in year 2011 as presented in Figure 3. Most of failures happen to ventures which are backed by BAs and least for those backed by both types of investors.
Figure 3:
Closing Companies per Year
0 1 0 0 2 0 0 3 0 0 # F a ilu re s 2006 2008 2010 2012 2014 2016 Year
# Failures overall # VC Failures
# BA Failures BA+VC Failures
Notes: This Figure shows the event Closing or Failure over time from 2007 until May 2016. The addition of BAs, VCs and BA+VC Failures results in the overall Failure figure.
25
A similar picture shows Figure 4, but for the event exit. The number of IPOs remains relatively constant over time in the small double-digit range whereas the number of acquisitions is at its peak in 2011.
Figure 4:
Successful Exited Companies per Year
6 Empirical Results
The described Cox PHM model is applied to the adjusted CrunchBase dataset and the results are presented in Tables 4-7. The central interpretations are those about the different investor types, which are Business Angel, Venture Capitalist and BA+VC. However, also the hazard ratios of the controls give some hints about the dynamics of the different explanatory variables and their influence on survival and exit. The indication about year fixed effects takes place only in form of Yes (included in the model) and No (excluded from the model). Furthermore, the Wald chi-square statistic is included to show the goodness of fit of the PHM in this specific form.
Additionally, a log-rank test for the three different investor types is performed in advance and shows for both success measures a good fit, as can be seen in Table 3. The first log-rank test for the survival measure shows strong imbalance between the different types and that the survival rates are not equal. A less clear picture shows the second part of the table, which includes the test for the exit
0 1 0 0 2 0 0 3 0 0 4 0 0 # E x it s 2006 2008 2010 2012 2014 2016 Year
# Exits overall # IPO # Acquired
Notes: This Figure shows the event Exit over time from 2007 until May 2016. The addition of IPO’s and Acquisitions results in the overall Exit figure.
26
measure. Only significant on the five percent level the observations of exit does not deviate that much from its expected values.
6.1 Kaplan-Meier Survival Estimate
The Kaplan-Meier survival estimate provides a first suggestion about the influence of the main explanatory variables, which are the different types of investors, on the probability of survival and exit of the supported firms. This estimate measures the fraction of the individuals who experiences an event to the sample population and was invented by Kaplan and Meier (1958).12 Tracked over time the survival estimate can be plotted to a curve as one of the simplest ways to deal with a censored dataset. Figure 5 shows the probability of survival on the y-axis plotted over time on the x-axis. Naturally at time t=0, when the Start-ups get their first funding by either a BA or VC, 100% of the companies are still active in their business. For all the different groups it is conspicuous that the survival rates have only a small negative slope. This is caused by the sample and the small number of closed firms overall. Nevertheless, the trend indicates a better chance of survival for those who are backed by VCs or both investors over companies which are financed by BAs. After 5 years of existence the path of Start-ups financed by both types of investors takes another more positive path than VC’s. This finding falls in line with the second stated hypothesis.
Figure 5:
Estimated Survival Function
Notes: This Figure shows failure patterns over time for companies backed by the different investors. The function is measured by the Kaplan-Meier Survival Estimate.
12
For more information about the model behind this survival estimate see Kaplan and Meier (1958)
0 .8 0 0 .8 5 0 .9 0 0 .9 5 1 .0 0 s u rv iv a l ra te 0 2 4 6 8 10 time in years
27
Figure 6 presents the second measure of success used in this paper in terms of Kaplan-Meier. This is the probability of an exit either via getting acquired or going public, again split by the type of investor. As can be seen from the exhibit up to year five it does not make a significant difference which type of investor is involved. However, this statement just holds for the first five years. By far the lowest exits are undertaken by BA backed Start-ups and the highest by those with VC support. The results indicate the validity of the third hypothesis. VCs actively guide their supported companies to IPOs and acquisitions and have the experience to realize them.
Figure 6:
Cumulative Estimated Exit Function
Notes: This Figure shows exit patterns over time for companies backed by the different investors. The function is measured by the Kaplan-Meier Survival Estimate and plotted cumulatively.
6.2 Survival Estimations for Entrepreneurial Companies in the PHM
All regressions follow the proportional hazard model by Cox as described before, whereby HR in the topline indicates that the presented coefficients are the hazard ratios. Table 4 and 5 are about survival as success measure and are performed in 14 different regressions overall. Table 6 and 7 apply the same procedure for the measure exit. The general interpretation of the hazard ratios is as follows: A HR<1 stands for a lower chance of failure if survival is the event (Table 4 and 5) in the PHM and for a lower chance of successful exit if exit is the event (Table 6 and 7), always in comparison to the rest of the sample. The interpretation for HR>1 is vice versa.
Table 4 presents all variables with its related hazard ratio. The Wald chi-square statistic shows an overall good fit of the used model.
0 .0 0 0 .1 0 0 .2 0 0 .3 0 e x it r a te 0 2 4 6 8 10 time in years
28
Regressions (1), (3) and (5) perform a basic model with only one investor type and the round control. As a consequence of the binary form of the control variable the first round is dropped throughout all regressions. The hazard ratios for rounds are in these regressions of similar magnitude and significance. With values smaller than 1 they show that the more funding rounds a company obtains the less likely they fail. In the case of a Start-up which is financed by BAs and VCs this effect is extraordinary large. The chance to fail after the 4th round (HR: ~0.2) is 80% lower than after the 1st round. The results for the investor types partly support the first impression from the Kaplan-Meier survival estimates that in comparison to the other investors the positive impact of BAs seems to be limited. Compared to the other two investor classes companies which obtain funding from BAs have a 94.7% higher chance to fail, whereas VCs successfully manage to keep their investments operating. In fact based on the hazard ratio for BA+VC in (5), VCs alone seem to perform better than in collaboration with BAs, what argues against the estimates in Figure 5 (Kaplan-Meier). Anyhow, since the HR is not even significant on the ten percent level this outcome does not meet the minimum criteria of significance. The basic regressions support the second hypothesis that in general the investor indeed make a difference for the chance of survival and that in particular VCs are superior to BAs.
Regressions (2), (4) and (6) extend the simple model with covariates for Industry, Continent and
Year fixed effects. The HRs for BAs and VCs stay significant at the one percent level and marginal expurgate from the baseline value 1. Moreover, the impact have the same trend as before but move closer to one if the controls are included, accompanied in (2) with a reduction in significance. In these regressions, the differential situations in industries and continents are evaluated for the first time. As mentioned before the baseline for industries is the dropped Information Technology sector and for continents North America represented by US and Canada. The industries Science and Health Care and Other Industries are significant in contrary to the other two which are Financial Service and Sales
and Marketing. We saw before in the summary statistics that especially in the first funding rounds Start-ups in the science and health care sector obtain the most funding. With hazard ratios to fail lower than 0.5 they compensate their investors for the huge need of funding. Other industries have 25%-30% better chance of survival compared to tech companies.
The continent figures implicate that Start-ups in Europe, Africa, Australia and Asia have more positive future prospects, whereby Latin Americas HR is close to 1 but insignificant. An explanation which is not far to seek is that CrunchBase is located in the US and has more renown there. Also smaller and more risky US ventures tap this source. European and Asian Start-ups in contrary will just be active on CrunchBase if they know its advantages and are keen to search bigger investors from overseas.
