The impact of venture capital funding on innovation
during the period 2004 – 2011
Amsterdam Business School
Name Marcha van der Boon
Number 10218114
BSc in Economics and Business Specialization Finance and Organization Field Finance
Supervisor Ilko Naaborg Completion 1 July 2014
Abstract
This thesis analyses the impact of venture capital funding on innovation across 20 different countries during the period 2004-‐2011. A regression is used to examine the impact of venture capital funding on innovation, where innovation is measured by the yearly patent counts issued per country. The results show a significant, positive relation between venture capital and innovation and a significant, negative effect of the current financial crisis on the number of patented innovations. This means that venture capital funding has an impact on patented innovations during the current financial crisis. These results appear quite robust. However, there are concerns about a causality problem. Although most researchers suggest that venture capital stimulates innovation, some researchers argue that innovations induce venture capital investments. Future research can investigate the impact of the entire crisis period on venture capital as a financing source for innovation, to look at the total impact of the crisis. Or if there is a new parameter to measure innovation, the study can be repeated to check whether the results of this research are robust.
Keywords: Venture capital, investment, innovation, R&D expenditures, start-‐up firms, financial crisis
Table of Contents
1. Introduction ... 2
2. Literature review ... 4
3. Hypothesis, Methodology and Data ... 8
3.1. Hypothesis and Methodology ... 8
3.2. Data and descriptive statistics ... 11
4. Empirical results ... 14
4.1. Empirical Results ... 14
4.2 Robustness check ... 19
5. Conclusion and discussion ... 21
References ... 23 Appendix A ... 25 Appendix B ... 26 Appendix C ... 27
1. Introduction
In the economy there are several drivers of economic growth and value creation, of which one important determinant is innovation (Schwienbacher, 2008). A main channel via which innovation is financed is via venture capital funds (Kortum and Lerner, 2000). Venture capitalists invest in risky start-‐up firms which have an uncertain future. They are an important financial intermediary for these young and small firms, because without the venture capital these firms perceive difficulties in gaining the necessary financial assets to develop and innovate (Gompers and Lerner, 2001). Since these start-‐up firms are a
contributor to innovation and since they are financed with venture capital, is venture capital an important contributor to innovation (Schwienbacher, 2008). Venture capital funding contributed to the success of many successful new firms, such as Microsoft, Google, Dell, Intel Computer and Apple. All these firms
received venture capital in their first stage of development (Da Rin et al., 2006). Kortum and Lerner (2000) demonstrate that venture capital really has a positive impact on innovation.
On 15 September 2008 the collapse of Lehman Brothers caused the current global financial crisis and also had an impact on the venture capital industry (Block et al., 2010). Block et al. (2010) show that the current financial crisis has a negative impact on venture capital funding. There is a dramatic decline in venture capital activities. They also show that the effect differs across industries and countries.
The stimulating effect of venture capital funding on innovation might be disturbed due to the current financial crisis. This impact will be stronger the more the venture capital industry is affected by the crisis. However, this has never been studied before. This thesis investigates the following research
question: ‘Does venture capital funding have an impact on the number of patented
innovations during the current financial crisis?’. Although most studies focus on
different industries within a country, this research focuses on 20 different countries. This research is also different compared to previous studies with respect to the time period, because it takes the current financial crisis into account. This is interesting because the crisis affects many countries. And as already mentioned, venture capital is a critical driver for innovation, whereas
innovation is important for economic growth.
This thesis makes use of a panel data analysis to test whether venture capital funding has a significant impact on the number of patented innovations during the current financial crisis. A method for analysing panel data is the ‘fixed effects regression’ (Griliches and Hausman, 1986). The data are collected from different databases. Innovations are measured by the total patent counts which are collected from the OECD Patent Database (OECD Patent Database, 2014). The data on venture capital activities are from the Thomson One database (Thomson One, 2014).
This thesis is structured as follows. After the initial introduction section the second section discusses the literature review, it contains the theoretical background and the empirical evidence. The third section presents the
methodology and the data sources, which are required for the regression and for testing the hypothesis. Then, the fourth section presents the empirical results and some robustness checks. Finally, in section five the conclusion is presented.
2. Literature review
This section includes the main theories in the existing literature and the empirical evidence found in line with these theories.
First of all, the theories about venture capital are discussed. Venture capital is a central source of finance for start-‐ups in innovative industries and is defined as equity or equity-‐linked investments in young, privately held
companies (Nanda and Rhodes-‐Kropf, 2013). The investor is a financial intermediary and is active as a director, an advisor or a manager of the firm (Kortum and Lerner, 2000). The investor has a close involvement with the firm, he has a network of contacts and also gives the firm the right expertise and knowledge about markets. Therefore, venture capital is used to support
economic growth and the innovative activities of firms (Da Rin et al., 2006). Zider (1998) estimates that more than 80% of the money invested by venture capital investors goes into building the infrastructure required to grow the business (in expense investments and the balance sheet). He states that venture capital is a financing form between sources of funds for innovation and traditional, lower-‐ cost sources of capital. The main alternatives for venture capital financing are business angels (also known as private individuals), banks, government and self-‐ financing (Hellman and Puri, 2000). However, the reason why the venture capital industry exists is by the structure and rules of capital markets (Zider, 1998). Because start-‐ups are a highly risky target, banks will only finance a new business when there are hard assets that secure the debt. The problem is that many start-‐ups have few hard assets (Zider, 1998). Hard assets are physical or tangible assets, which carry intrinsic value. Therefore, innovative start-‐ups suffer from credit constraints and this may be overcome by venture capital funds (Da Rin et al., 2006).
According to Zider (1998) the venture capital industry consists of four main players: young entrepreneurs who need funding, investors who want high returns, investment bankers who need firms to sell and the venture capitalists who make money for themselves by making a market for the other three.
Figure 1: The venture capital cycle
Source: Zider, 1998
The venture capital cycle is important for understanding the venture capital industry (Gompers and Lerner, 2001). In the first stage of the cycle a capitalist raises a venture fund. Second, the fund proceeds through investing, monitoring and adding value to the firms. Then the venture capital firm has to make
successful deals and returns capital to the investors. Finally it renews itself by raising additional funds (Gompers and Lerner, 2001). Block and Sandner (2009) explain that venture capital is particularly important in the early periods of a firm’s life. At this point in time the firm begins to exploit innovative activities. Venture capitalists provide capital to typically small and young firms, with high levels of uncertainty and high information asymmetry (Schwienbacher, 2008). Since these start-‐up firms are a contributor to innovation and since they are financed with venture capital, is venture capital an important contributor to innovation (Schwienbacher, 2008). Also the OECD (1996) argues that venture capital is crucial for innovation. Particularly for start-‐up firms it is hard to undertake high-‐risk innovative activities (OECD, 1996). Venture capitalists are willing and able to provide capital to these firms. They show that this is
confirmed by empirical evidence of technological revolutions, which have been led by venture capital-‐backed firms. It is also important to mention that
innovation is a critical driver of economic growth and value creation (Rosenberg, 2004). Schumpeter (1939) defines innovation as ‘making something new’. He separates innovation from invention and says that innovation is learning something new with respect to theoretical or practical knowledge, while invention does not necessarily induce innovation.
However, although venture capital is a main source for the financing of innovation, Gompers and Lerner (2001) explain four challenges with respect to an investor’s willingness to invest in innovative activities. First of all, the high uncertainty about the future. There is not only uncertainty in a firm’s
development possibilities, but also in the industry and in the whole market (Gompers and Lerner, 2001). Second they cite information asymmetry. This is a situation where the various players in the venture capital cycle have differences in knowledge with respect to their information inside and outside the firm. The third challenge is about the soft assets, such as patents. These assets cannot serve as collateral in the case a firm goes bankrupt, because they provide too little value (Gompers and Lerner, 2001). And the fourth challenge is the volatility of market conditions. Venture capitalists have to choose the best moment to invest in order to minimise the total risk, therefore market timing is important for investors (Gompers and Lerner, 2001). According to Gompers et al. (2008) venture capitalist are actually good at timing market conditions, they have to invest at the right time.
Another challenge that became visible was the financial crisis (Block and Sandner, 2009). On 15 September 2008 the bankruptcy of Lehman Brothers was announced, which caused the great recession. Many financial institutions were affected and lost value. The only way to save these institutions from bankruptcy was by government funds (Block and Sandner, 2009).
In the second part of this section, the empirical evidence with regard to the venture capital industry is discussed. It is argued that venture capital financing spurs innovation (Hellman and Puri, 2000). Hellman and Puri (2000) provide evidence that venture capital financing can have an impact on the development path of a start-‐up firm. Their sample consists of 173 start-‐up firms that are located in California’s Silicon Valley chosen independently of financing, so that they have both venture and non-‐venture capital-‐backed firms. Using a duration model with time-‐varying covariates they show that there is a significant reduction in the time to bring a product to the market. Kortum and Lerner
(2000) also find a positive relation between venture capital funding and innovation. They investigate this relation in reduced-‐form regressions across twenty industries in the U.S. manufacturing sector over a three-‐decade period,
controlling for R&D expenditures. They find that venture capital funding is accompanied with an increase in innovation. Venture capital accounted for 14% of innovations (Kortum and Lerner, 2000). Nanda and Rhodes-‐Kropf (2013) show that increased venture capital ensures that investments shift to more innovative start-‐ups by lowering the cost of experimentation and allowing them to make riskier, more novel investments. They use a sample of 12.285 US-‐based start-‐ups that received their first venture capital funds between 1985-‐2004, but follow these firms until the end of 2010. Using multivariate analysis they find their results.
The collapse of Lehman Brothers caused the current global financial crisis and also had an effect on the venture capital industry (Block et al., 2010). Block and Sandner (2009) analyse the effect of the current financial crisis on venture capital investments in US Internet firms. Using regression analysis they find that the financial crisis is accompanied with a 20% decrease in the venture capital disbursements. They show that firms in later stages of the venture capital cycle are more negatively affected. Also Block et al. (2010) show that the current financial crisis has a negative impact on venture capital funding. They argue that both firms in early-‐ and later stages of the venture capital cycle are negatively affected. However, they say that the effect differs across industries and countries.
Theory and empirical evidence suggest that venture capital stimulates innovation (Kortum and Lerner, 2000) and that venture capital funding
decreases due to the current financial crisis (Block et al., 2010). The research in this thesis relates to the existing literature because the stimulating effect of venture capital funding on innovation might be disturbed due to the current financial crisis. The research by Kortum and Lerner (2000) focuses only on the United States in a period before the crisis. However, the financial crisis affects many countries. Because Block et al. (2010) show that this effect differs across countries; the research in this thesis focuses on the relation between venture capital funding and innovation during the financial crisis, but takes several countries into account. Based on the previous theories, it is expected that
venture capital funding has a smaller impact on patented innovations during the current financial crisis.
Hirukawa and Ueda (2011) argue that because of causality the relation between venture capital and innovation should be interpreted carefully. There are two possible hypotheses, the ‘venture capital-‐first hypothesis’ and the ‘innovation-‐ first hypothesis’ (Hirukawa and Ueda, 2011). The venture capital-‐first
hypothesis means that venture capital investments stimulate innovation. However, the innovation-‐first hypothesis shows a reversed causality that
innovations induce venture capital investments. The demand for venture capital increases through the entry of new technology (Hirukawa and Ueda, 2011). Although Hirukawa and Ueda (2011) find evidence supporting the innovation-‐ first hypothesis, Trajtenberg (1990) find supportive evidence for the
attractiveness of patents as an indicator for innovation. His research supports the venture capital-‐first hypothesis. Though, Lerner (2002) states that both venture capital funding and innovations could be positively related to the arrival of technological opportunities. This means that on the one hand venture capital could spur innovation, but on the other hand there is a possibility that
innovation increases because venture capital reacted to a technological shock which lead to more innovation (Lerner, 2002). Kortum and Lerner (2000) address these causality concerns and show that venture funding has a strong positive impact on innovation. These results support the venture capital-‐first hypothesis. Therefore, as seen from the literature the most empirical evidence is found for the venture capital-‐first hypothesis.
3. Hypothesis, Methodology and Data 3.1. Hypothesis and Methodology
This section discusses the hypothesis and the model.
A regression will be used to test whether venture capital funding has a significant impact on the number of patented innovations during the current financial crisis. The following model is used to analyse this:
𝐼𝑛𝑛𝑜𝑣𝑎𝑡𝑖𝑜𝑛!,! = 𝛽!+ 𝛽!∗ 𝑉𝑒𝑛𝑡𝑢𝑟𝑒 𝐶𝑎𝑝𝑖𝑡𝑎𝑙 𝐷𝑒𝑎𝑙𝑠!,!+ 𝛽!∗ 𝑉𝑒𝑛𝑡𝑢𝑟𝑒 𝐶𝑎𝑝𝑖𝑡𝑎𝑙 𝐸𝑥𝑝𝑒𝑛𝑠𝑒𝑠!,! + 𝛽!∗ 𝑅&𝐷 𝑒𝑥𝑝𝑒𝑛𝑑𝑖𝑡𝑢𝑟𝑒𝑠!,!+ 𝛽!∗ 𝑆𝑡𝑎𝑟𝑡𝑢𝑝 𝑓𝑖𝑟𝑚𝑠!,!+ 𝛽!∗ 𝐶𝑟𝑖𝑠𝑖𝑠!+ 𝑢!,!
where 𝑢!,! = 𝛼! + 𝜀!,!, i is the country and t is the time period. The dependent
variable in the regression is innovation. Innovation is measured by the yearly patent counts issued per country (Hagedoorn and Cloodt, 2003). Two different variables will be used to measure venture capital funding. The first independent variable is a country’s yearly total venture capital expenditures. The second independent variable is a country’s yearly total number of completed venture capital deals (Kortum and Lerner, 2000).
Kortum and Lerner (2000) state that venture capital funding and patenting could be related to the arrival of technological opportunities.
Therefore, the variable ‘R&D expenditures’ is used in the model to control for the technological opportunities. However, there could be another relation with respect to innovation. Baumol (2002) predicts that there is a positive relation between the number of entrepreneurs and the amount of patents applied within a country and Almeida and Kogut (1997) state that start-‐up firms discover new technological areas by innovating in less busy areas. Therefore, a country with a higher number of start-‐up firms could have more patented innovations. The variable ‘start-‐up firms’ controls for this.
Quantitative data of 20 countries are collected, which are based on the size of venture capital activity. The period that will be analysed is 2004 – 2011. This period includes the current financial crisis during the period 2008 – 2011. The dummy variable ‘crisis’ is added to the model to determine the impact of the financial crisis. The dummy variable is equal to one if the time period ‘t’ is during the financial crisis (2008 – 2011) and zero otherwise.
Table 1: Selected countries
Australia Ireland Norway Singapore Sweden
China Israel Poland South Africa Turkey
France Italy Portugal South Korea United Kingdom
Germany Japan Russia Spain United States
The data is collected for 20 different countries observed at 8 different time periods and is called longitudinal data or panel data (Stock and Watson, 2012, p. 390). A method for analysing panel data is the ‘fixed effects regression’ (Griliches and Hausman, 1986). There are four assumptions for the fixed effects regression
(Stock and Watson, 2012, pp. 404-‐405). The first assumption is that ui,t has
conditional mean zero. The second assumption which Stock and Watson (2012, p. 404) make, is that the variables for one entity are distributed identically to, but independently of, the variables for another entity. The third assumption they mention is that large outliers are unlikely and the fourth assumption states that there is no perfect multicollinearity. Fixed effects regression is a method for controlling for omitted variables in panel data, since the omitted variables vary across the different countries but do not change over time (Stock and Watson, 2012, p. 396). The model decomposes the error term, ui,t, into a unit-‐specific and
time-‐invariant component, αi, and an observation-‐specific error, εi,t (Stock and
Watson, 2012, pp. 396-‐370). Stock and Watson (2012, p. 396) state that the fixed effects regression has for each country a different intercept, which absorb the influences of all omitted variables that differ from one country to the next but are constant over time. These are all the variables, which determine innovation in the ith country, but do not change over time. An example of such a variable is the
several policies used in each country (Da Rin et al., 2006). Therefore, with the fixed effects model the effects of the independent variables on the dependent variable can be estimated using the changes in the variables during the selected period (Griliches and Hausman, 1986). With an OLS regression there will be omitted variable bias because of the correlation between the unobservable factors and the variables in the regression (Stock and Watson, 2012, p. 221).
However, if some omitted variables are constant over time but vary across countries while others are constant across states but vary over time, then the alternative for the fixed effects model can be used. This is the ‘random effects model’ (Stock and Watson, 2012, p. 402).
To test whether the fixed effects model is more efficient than the random effects model, it is appropriate to run a Hausman test. The Hausman test tests a more efficient model against a less efficient, but consistent model, to ensure that the more efficient model also has consistent results (Stock and Watson, 2012, pp. 402-‐403). The null hypothesis is that the coefficients estimated by the efficient random effects estimator are the same as the coefficients estimated by the
consistent fixed effects estimator. When the p-‐value is significant the fixed effects model is used.
The hypothesis of this research is that venture capital funding has a smaller impact on the number of patented innovations during the current financial crisis (in the period 2008 – 2011). The number of patented innovations decreases. Kortum and Lerner (2000) show that increases in venture capital are associated with higher innovation. Their estimates suggest that venture capital may have accounted for 14% of innovations. However, Block and Sandner (2009) analyse the effect of the current financial crisis on venture capital and show that there is a decrease of 20% in venture capital due to the financial crisis. Therefore, when there is a decrease in venture capital, the number of innovations will also decrease, whereby venture capital funding contributes less to innovation in times of crisis. It is expected that the coefficient on the dummy variable ‘crisis’ is negative and the coefficients on the venture capital measures will be lower during the financial crisis than before the financial crisis, but these coefficients remain positive. The hypothesis tests if the impact of venture capital funding on the number of patented innovations is smaller during the crisis than before the crisis, which is indicated by a negative sign of the crisis dummy.
3.2. Data and descriptive statistics
This section lists the data sources.
First of all, the dependent variable will be discussed. The dependent variable is innovation and is measured by the total yearly patent counts issued per country (Hagedoorn and Cloodt, 2003). The data on patent counts is collected from the OECD (Organization for Economic Co-‐Operation and Development) Patent Database (OECD Patent Database, 2014). This database supplies patent indicators that are appropriate for statistical analysis and covers data on patent applications to the US Patent and Trademark Office (USPTO). The data comes primarily from the latest version of the EPO’s Worldwide Patent Statistical Database (PATSTAT).
Patents are defined as a method to protect an innovators idea (OECD, 2010). Therefore, the numbers of patent applications is used as an indicator of the amount of new ideas being produced and patents are therefore an indicator for innovation (Kortum and Lerner, 2000). Trajtenberg’s (1990) findings
indicate that patents are a good indicator of innovation. He declares that patents are the only manifestation of innovation activities covering every field of
innovation in several countries and over long time periods. It is a measure for innovation performance. Griliches (1998) states that a patent is issued by an authorized public institution. It is a document that grants the right to exclude anyone else from the production or use of a specific new device, apparatus, or process for a stated number of years. The aim of the patent system is to encourage innovation and technological progress (Griliches, 1998).
Secondly, the two independent variables ‘venture capital deals’ and ‘venture capital expenses’ are discussed. These two variables are used as a measurement for venture capital (Kortum and Lerner, 2000). The data on these variables are collected from the Thomson One database (Thomson One, 2014). Thomson One supplies data on a broad range of financial content including venture capital information.
The variable ‘venture capital deals’ is defined as the total number of completed venture capital deals per country per year. These deals are financed with venture capital. The variable ‘venture capital expenses’ is the total venture capital expenditures per country per year. These expenditures are expressed in millions.
Thirdly, the control variable ‘R&D expenditures’ is discussed. The data on this variable is collected from the OECD Database Research and Development
Statistics (RDS) (OECD Database R&D Statistics, 2014). This database provides a range of recent data on the resources devoted to R&D and is based on the data reported to OECD and Eurostat.
R&D expenditures affect, besides venture capital, innovation activities (Kortum and Lerner, 2000). Brown et al. (2009) state that financing of R&D is a critical input to innovation and economic growth. Therefore the control variable ‘R&D expenditures’ is used to control for this. The R&D expenditures are in millions of national currency, also per country and per year.
Finally, the other control variable ‘number of start-‐up firms’ is discussed. This is the total number of start-‐up firms per year per country. These data is collected
from the database Orbis (Orbis, 2014). Orbis supplies company information across the globe.
The total number of start-‐up firms contains both small-‐ and large start-‐up firms in each country. Although large start-‐up firms have more patents in well-‐ established areas, small start-‐up firms have more patents in smaller, less known areas. And because all these start-‐up firms affect innovation activities, the variable ‘number of start-‐up firms’ controls for this effect (OECD, 2010).
In table 2 are the descriptive statistics and in table 3 are the cross-‐correlations of the several variables. Appendix A contains the detailed descriptive statistics, where ‘between’ and ‘within’ indicate the descriptive statistics for ‘between countries’ and ‘within periods’ respectively. ‘Overall’ shows the same results as table 2. Appendix B shows the total yearly patent counts issued per country. It shows that for most countries the total patent counts increases a little before the crisis, but that the total patent counts decreases during the crisis. Appendix C also shows this observation, where the total yearly patent counts issued per country are summed in each period.
Table 2: Descriptive statistics
Variables Obs. Minimum Maximum Mean Standard deviation Variance
Country 160 1 20 10.5 5.784386 33.45912
Period 160 2004 2011 2007.5 2.298482 5.283019
Patent 160 50.3635 149826.4 13728.71 32031.29 1.03E+09
VCD 160 1 5937 418.4063 1099.382 1208640
VCE 160 0.33 105992.1 3497.838 11139.26 1.24E+08
R&D exp. 160 295.848 3.82E+07 1925479 6225654 3.88E+13
Start-‐up 160 19 2612220 153739.3 283360.5 8.03E+10
Crisis 160 0 1 0.5 0.5015699 0.2515723
Notes: VCD refers to the total number of completed venture capital deals, VCE refers to the total venture capital expenses, start-‐up refers to the total number of start-‐up firms and R&D exp. refers to the total R&D expenditures.
The mean of the variables, reported in the 5th column of table 2, varies between
0.5 and 1925479, and the variation in the various variables can be seen from the standard deviation, reported in the 6th column, which varies between 0.5015699
variable and the 3rd and 4th columns present the smallest and largest
observations per variable, indicated by the minimum and maximum.
Table 3: Cross-‐correlations
Country Period Patent VCD VCE R&D exp. Start-‐up Crisis
Country 1.0000 Period 0.0000 1.0000 Patent -‐0.5104 *** -‐0.0142 1.0000 VCD -‐0.4791 *** 0.0068 0.8952 *** 1.0000 VCE -‐0.3536 *** -‐0.0196 0.7329 *** 0.7799 *** 1.0000 R&D exp. -‐0.1694 ** 0.0553 0.1720 ** -‐0.0519 -‐0.0537 1.0000 Start-‐up -‐0.3743 *** 0.0947 0.6767 *** 0.7801 *** 0.5316 *** -‐0.1083 1.0000 Crisis 0.0000 0.8729 *** -‐0.0155 -‐0.0030 -‐0.0658 0.0460 0.0803 1.0000
Notes: VCD refers to the total number of completed venture capital deals, VCE refers to the total venture capital expenses, start-‐up refers to the total number of start-‐up firms and R&D exp. refers to the total R&D expenditures. ** and *** indicate significance at 5% and 1% respectively.
Table 3 presents the correlations between the several variables. The smallest correlation is between the variables ‘crisis’ and ‘VCE’, this correlation is negative and is -‐00658. This indicates a strong negative relation. The largest correlation is between the variables ‘VCD’ and ‘patent’, this correlation is positive and is
0.8952. This indicates a strong positive relation. The correlations between ‘patent’ and the two venture capital variables are significant at the 1% level.
4. Empirical results 4.1. Empirical Results
This section presents the main results.
First of all, both the fixed effects regression and the random effects regression are done in order to test the hypothesis of the Hausman test. The Hausman test tests the more efficient model (the random effects model) against the less efficient, but consistent model (the fixed effects model), to ensure that the more efficient model also has consistent results. The null hypothesis is that the coefficients estimated by the efficient random effects estimator are the same as the coefficients estimated by the consistent fixed effects estimator. Table 4
presents the results of the Hausman test. The p-‐value of the test is significant at the 1% level. This means that the null hypothesis is rejected. Therefore, it is appropriate to use the fixed effects model instead of the random effects model.
Table 4: The Hausman test and regressions
Dependent variable: Patent
FE
regression regression RE
(1) t-‐value p-‐value (2) z-‐value p-‐value
VCD 2.760804*** (.9332649) 2.96 0.004 6.131728*** (1.148741) 5.34 0.000 VCE .1006053*** (.0226214) 4.45 0.000 .0902458*** (.0297052) 3.04 0.002 R&D exp. .0002333** (.0001) 2.33 0.021 .0003228** (.0001274) 2.53 0.011 Start-‐up -‐.0044318*** (.0010849) -‐4.08 0.000 -‐.0040565*** (.0014241) -‐2.85 0.004 Crisis (305.3891) -‐775.84** -‐2.54 0.012 -‐8,571,541** (401.0604) -‐2.14 0.033 Constant 12841.72*** (475.4759) 27.01 0.000 11278.16*** (3180.198) 3.55 0.000 R2 0.8131 0.8695 rho .99609603 .97018645 F-‐test 0.0000*** 0.0000*** Hausman test chi2(2) = 61.89 Prob>chi2 0.000***
Notes: Standard errors are in parentheses. ** and *** indicate significance at 5% and 1% respectively. VCD refers to the total number of completed venture capital deals, VCE refers to the total venture capital expenses, start-‐up refers to the total number of start-‐up firms and R&D exp. refers to the total R&D expenditures.
Regression 1 in table 4 presents the results of the fixed effects regression and regression 2 in table 4 presents the results of the random effects regression. The remainder of this section will explain the results of the fixed effects regression, because the random effects model is rejected by the Hausman test and the use of the fixed effects regression is advised.
Regression 1 in table 4 presents the results of the fixed effects regression. All the variables in the model are statistical significant, most of them at the 1% level. First of all, the two independent variables are discussed. Both variables
‘venture capital deals’ and ‘venture capital expenses’ are significant at the 1% level, controlling for R&D expenditures and the number of start-‐up firms, and have a positive relation with the dependent variable ‘patent’. They have
considerable explanatory power for the total number of patents. This means that when the number of venture capital deals increases with one deal, the number of patents increases with 2.7608 and this also means that when the total venture capital expenses increases with 1 million, the number of patents increases with 0.1006. According to these estimates, good financing opportunities with regard to venture capital are associated with more innovation. Therefore, when there is more venture capital in circulation, there are more financial capabilities to finance innovative opportunities. So, there could be more patents granted and the total of innovation increases. Venture capital is especially important for start-‐up firms, because these firms perceive difficulties in gaining the necessary financial assets. With the venture capital they have the opportunity to innovate and to develop. So, when they get access to venture capital funding, the number of patents increases.
Second, the first control variable ‘R&D expenditures’ is discussed. This variable is significant at the 5% level and also has a positive effect on the
dependent variable ‘patent’. When the R&D expenditures increase with 1 million in national currency, then the number of patents increases with 0.0002.
Technological opportunities and hence R&D expenditures are often associated with higher innovation. When researchers spend more on R&D expenditures, innovation will increase (Lanjouw and Schankerman, 2004). However, according to these estimates, the effect on patents is relative small.
Third, the other control variable ‘start-‐up firms’ is discussed. This variable is significant at the 1% level and has a negative effect on the dependent variable ‘patent’. When the number of start-‐up firm’s increases with one, the number of patents decreases with -‐0.0044. Although other researchers suggest a possible positive relation with innovation, this result shows a negative effect. An
explanation could be that when the number of start-‐up firms increases, it is harder to be innovative. Therefore there will be fewer patents granted when there are more start-‐up firms in a certain period. The more start-‐up firms enter an area, the less opportunity to develop or innovate.
Finally, the last variable is discussed, the crisis dummy. This variable is significant at the 5% level. The result of the dummy variable indicates a negative effect on patents due to the current financial crisis. This means that in an
economic downturn, such as the financial crisis, there are fewer patents granted and there is less innovation spurred by venture capital.
The remaining results in the table indicate a high R2, a high rho and a low
F-‐test. The regression R2 is the fraction of the sample variance of Yi,t explained by
(or predicted by) the regressors (Stock and Watson, 2012, p. 235). This means that 81.31% of the sample variance of ‘patent’ is explained by the regressors used in this model. The rho in the regression is known as the intraclass correlation. It is the fraction of the variance due to differences across panels (Stock and Watson, 2012, pp. 133-‐134). In this model is 99.61% of the variance due to differences across the several countries. The F-‐test is a test to see whether all the coefficients in the model are different than zero. It tests the joint
hypothesis that all the slope coefficients are zero (Stock and Watson, 2012, pp. 263-‐265). The p-‐value of the F-‐test is 0.0000, which means that the model is significant at the 1% level.
However, some researchers have concerns about possible lags between venture capital funding and patenting (Kortum and Lerner, 2000). The venture capital funds have to be invested before the firms can innovate, thus venture capital has a lagged effect on patents. Hall et al. (1986) suggest that R&D spending and patenting are contemporaneous and that there is a reason why the lags between venture capital funding and patenting should not be long. They state that
companies who obtained venture capital experience pressure to commercialise products quickly. In the following regression the same fixed effects model is used, but now with a lag of 1 year of the ‘venture capital deals’ and ‘venture capital expenses’ variables. The same data is used, but now losing one period because of the lag. The lag is only presented in the data of the two venture capital variables, because the variables ‘R&D expenditures’ and ‘start-‐up firms’ are not involved with a lag.
Table 5: Fixed effects regression with a lag of 1 year
Dependent variable: Patent
1 year lag FE regression (1) t-‐value p-‐value VCD 6.104425*** (1.0465) 5.83 0.000 VCE (0.0249852) 0.11174*** 4.47 0.000 R&D exp. (0.0001147) 0.0000206 0.18 0.857 Start-‐up -‐0.0062317*** (0.0011504) -‐5.42 0.000 Crisis -‐1081.362*** (330.285) -‐3.27 0.001 Constant 17450.55*** (573.3327) 30.44 0.000 R2 0.7849 rho 0.99773788 F-‐test 0.0000***
Notes: Standard errors are in parentheses. *** indicates significance at 1%. VCD refers to the total number of completed venture capital deals, VCE refers to the total venture capital expenses, start-‐up refers to the total number of start-‐up firms and R&D exp. refers to the total R&D expenditures.
These results appear quite robust. Both the two venture capital variables have a positive relation and the same significance level as the regression without the time lag. Although the higher impact of the crisis, the coefficients of the ‘venture capital deals’ and ‘venture capital expenses’ variables are also higher. A reason might be that the time lag is taken into account in this regression. Only the control variable ‘R&D expenditures’ is not statistically significant in this regression.
Overall, the several regressions give the same results, namely a positive and highly significant result for the two venture capital variables and a negative impact of the current financial crisis.
The last part of this section discusses some implications of the findings. The results suggest that there is a positive relation between venture capital funding and innovation. The higher coefficient on ‘venture capital expenses’ compared to ‘R&D expenditures’ means that money invested in venture capital is more potent
in stimulating innovation than money spend on R&D. The model controls for R&D expenditures and the number of start-‐up firms, and the crisis dummy indicates that the current financial crisis has a negative impact on the number of patents. This means that, although there is a positive relation between venture capital and innovation, the impact of the crisis ensures that this positive relation is smaller during the crisis than this particular relation before the crisis.
4.2 Robustness check
This section presents some robustness checks and additional results.
There are two different robustness checks implemented with respect to the crisis. First of all, the data is divided into two different time periods. The first period is 2004-‐2007, representing the period before the financial crisis and the second period is 2008-‐2011, representing the period during the financial crisis. The following table presents the results of the first robustness check.
Table 6: Fixed effects regression with to different time periods
Dependent variable: Patent
FE regression FE regression
2004-‐2007 t-‐value p-‐value 2008-‐2011 t-‐value p-‐value
VCD 5.650075** (2.225657) 2.54 0.014 3.360216*** (.9626131) 3.49 0.001 VCE (0.0480309) 0.0276541 0.58 0.567 (0.0537506) 0.0161227 0.30 0.765 R&D exp. 0.0007305** (0.0002919) 2.50 0.015 (0.0001191) 0.000156 1.31 0.195 Start-‐up (0.0043561) 0.0084477* 1.94 0.058 (0.0009099) -‐0.0008349 -‐0.92 0.363 Constant 9670.667*** (1056.499) 9.15 0.000 9912.848*** (549.7961) 18.03 0.000 R2 0.8073 0.8704 rho 0.99678012 0.99779161 F-‐test 0.0000*** 0.0000***
Notes: Standard errors are in parentheses. *, ** and *** indicate significance at 10%, 5% and 1% respectively. VCD refers to the total number of completed venture capital deals, VCE refers to the total venture capital expenses, start-‐up refers to the total number of start-‐up firms and R&D exp. refers to the total R&D expenditures.
In table 6 are the results of the fixed effects regressions of the two different time periods. Derived from the coefficients, all the variables have a smaller impact during the crisis than before the crisis. However, the ‘venture capital deals’
variable has both before and during the crisis a significant impact on
innovations, while the ‘venture capital expenses’ variable has in neither period a significant impact. The insignificance of the ‘venture capital expenses’ variable could be caused by the small sample size, because the period 2004-‐2011 is divided into two different time periods, each representing only 4 years.
The second robustness check takes, besides the crisis, the concerns about possible lags between venture capital funding and patenting into account. In the following regression the same two time periods are used, but now with a lag of 1 year of the ‘venture capital deals’ and ‘venture capital expenses’ variables.
Table 7: Fixed effects regression with two different time periods and a lag of 1 year
Dependent variable: Patent
1 year lag 1 year lag
FE regression FE regression
2004-‐2007 t-‐value p-‐value 2008-‐2011 t-‐value p-‐value
VCD 16.59311*** (3.109579) 5.34 0.000 8.006799*** (1.890951) 4.23 0.000 VCE (.0796435) .217704*** 2.73 0.006 .1289256*** (.0166306) 7.75 0.000 R&D exp. (.0004533) .0003589 0.79 0.429 (.0001248) -‐.0001154 -‐0.92 0.359 Start-‐up (.006926) .0042715 0.62 0.537 -‐.0050706*** (.0016645) -‐3.05 0.004 Constant (4262.121) 7208.914* 1.69 0.091 17189.41*** (1176.18) 14.61 0.000 R2 0.8212 0.8414 rho 0.9860556 0.99926453 F-‐test 0.0000*** 0.0000***
Notes: Standard errors are in parentheses. * and *** indicate significance at 10% and 1%
respectively. VCD refers to the total number of completed venture capital deals, VCE refers to the total venture capital expenses, start-‐up refers to the total number of start-‐up firms and R&D exp. refers to the total R&D expenditures.
Table 7 presents the results of the fixed effects regression with two different time periods and a lag of 1 year. These results also show for all the variables lower coefficients during the crisis than before the crisis and both venture capital variables are significant at the 1% level. This means that the impact of venture capital on innovations is lower during the crisis than before the crisis.