Venture capitalists and the dual-track harvest strategy. Maher Kooli*, Thomas Walker, and Aoran Zhang ABSTRACT

Transcription

1 Venture capitalists and the dual-track harvest strategy Maher Kooli*, Thomas Walker, and Aoran Zhang ABSTRACT This paper explores venture capitalists (VCs ) double exit strategy for their portfolio start-ups. Using a sample of VC backed IPOs from 2007 to 2013, we show that VC backed IPOs with more innovative outputs are less prone to employ the double exit strategy. Compared to independent venture capitalists (IVCs), corporate venture capitalists (CVCs) are less likely to push their portfolio companies for a second exit. Specifically, our research reveals that it is CVCs superior capability of nurturing innovation that allows their portfolio companies to stand alone in the public market. Furthermore, we find that VCs double exit strategy is influenced by the market conditions. Those entrepreneurial firms go public during hot public offering (IPO) periods therefore they are anticipated to experience trade sales shortly after the IPO. * Université du Québec à Montréal, Professor of Finance, School of Management and CDPQ Research Chair in Portfolio Management, 315 Rue Sainte-Catherine Est, Montréal, Québec, Canada H2X 3X2, Phone: , ext. 2082, Fax: , Kooli.Maher@uqam.ca Concordia University, Professor of Finance, John Molson School of Business Building, 1450 Rue Guy, Montreal, Quebec, Canada H3G 1M8, Phone: , ext. 2387, Fax: , Thomas.Walker@concordia.ca PhD Candidate in Finance, John Molson School of Business Building, 1450 Rue Guy, Montreal, Quebec, Canada H3G 1M8, Aoran.Zhang@concordia.ca We gratefully acknowledge the financial support of the CDPQ Research Chair in Portfolio Management of the School of Management (UQAM). We are responsible for any remaining errors.

2 1. Introduction Over the recent decades there s been an unparalleled growth in the role of venture capitalists (VCs) in the nurturing of start-ups, allowing them to gain access to the public capital market. A milestone of the entrepreneurial firm, as well as a successful exit strategy for VCs, is the ability of going public via IPO and moving from private to public ownership. Nevertheless, a striking phenomenon for such nascent firms shows that an important number of IPO firms maintain their public ownership for a short time period and leave the stock exchange by means of mergers and acquisitions. For instance, Chemmanur, Signori, and Vismara (2014) report that approximately 15% IPOs in both North America and Europe become acquisition targets within three years of their IPO. De and Jindra (2012) also find that 15.3% of the 1980 to 2006 U.S. IPOs are acquired within three years after going public. After being acquired, the entrepreneur shortly loses the control benefits of a standing-alone public firm. This fact is seemingly a contradiction of the entrepreneur s original intention of going public. Why do certain VC backed IPO firms follow a double exit strategy to get acquired soon after their IPO, while others prefer to remain stand-alone firms? Existing literature mainly sheds light on the issue of the exit mechanisms of VC backed private firms at pre-ipo stage (see. e.g. Bayar and Chemmanur (2011); Poulsen and Stegemoller (2008)) and the valuation premium stemming from VCs sponsorship at the IPO stage (see. e.g. Ivanov and Xie (2010)). Otherwise post-ipo financial performance of the VC backed start-ups is also being examined in current literature (see. e.g. Cao, Jiang, and Ritter (2015); Brav and Gompers (1997)). Conversely, the conundrum of why IPO firms become acquisition targets in a short run, has only been partially unveiled by a very limited number of researchers. (see. e.g. Gill and Walz (2016); Chemmanur, Signori, and Vismara (2014); De and Jindra (2012)). Gill and Walz (2016) use U.S data to find that a VC sponsorship significantly increases the likelihood of an IPO firm becoming

3 an acquisition target, they also confirm that in most cases, VC-backed IPOs can be interpreted as delayed trade sales. However, Chemmanur, Signori, and Vismara (2014) s result is entirely at odds with the one from Gill and Walz (2016). Chemmanur, Signori, and Vismara (2014) examine the exit vehicle of European private firms and confirm that some private firms do follow a double exit strategy. That is, these firms employ IPOs as a primary exit strategy, and shortly after the IPO stage they pursue an acquisition as a second exit. Their empirical results indicate, however, that VC backed private firms are more likely to stand alone within the first five years after the IPO, rather than to implement a double exit strategy. These paradoxical results reveal that other compelling factors, beyond those previously documented in the academic literature, should be considered in the analysis of the double exit strategy for VC firms. The objective of our paper is to formulate and test different hypotheses regarding the double exit question for VCs. These hypotheses are not necessarily mutually exclusive. By measuring the innovation output as the patents an IPO firm possesses, we can say that patent possession represents the uniqueness of a firm and increases its viability against fierce competition on the market. We demonstrate that VC backed IPOs with higher innovation outputs are more capable of standing alone in the stock exchange. On the contrary the VC backed IPOs without patents are more likely to end up with delayed trade sales. In addition, we differentiate a specific subset of VC from the traditional form of VC, which is corporate venture capital (CVC). Unlike the traditional VC, who has a limit partnership and limited lifespan, CVC is a subsidiary of a corporation and has a much longer investment horizon. In addition, CVC s incentive of investing in the portfolio companies is focused mostly on strategic reasoning, instead of pure financial return.

4 Hence, once CVC brings their portfolio companies to public, they are not in a rush to cash out their investment, contrary to what an independent VC s would do. Therefore, we propose that CVC backed IPOs are less likely to get acquired immediately after the IPO stage. Moreover, we find that CVC is superior to IVC in terms of maintaining the start-ups public status, because we hypothesize that CVC emphasizes in nurturing innovation. Our empirical results show that CVC funded IPOs with a patent have a better chance to keep their public status. Whereas, when there s no patents involved in their portfolio companies, it appears there is no statistical significant difference in the double exit strategy between the CVC and IVC firms. Furthermore, our analysis reveals that VC s double exit strategy relates to the market condition. The start-ups going public during the hot IPO period are more likely to double exit. This finding indicates that venture capitalists tend to take advantage of high market demand period to maximize their harvest value in the short term, by bringing their portfolio start-ups to public, even though these firms are not necessarily capable of fending themselves on the public market in the long run. Our paper contributes to the corporate finance literature in the following aspects. Primarily, our research closely relates to the literature examining the reason why IPO firms get acquired shortly after going public (See. e. g. Chemmanur, Signori, and Vismara (2014); De and Jindra, 2012 and Jain and Kini, 1999). Moreover, the closest study to our research is the one conducted by Gill and Walz (2016), which examines the impact of VC involvement on the likelihood of IPO firms being taken over. They find that VC backed IPOs are more likely to eventually get acquired due to the corporate governance costs stemming from losing an effective monitor, which happens when the VC s firms cash out from portfolio companies. These costs outweigh the benefits of standing alone on the public market. Our research complements this study by focusing on the new determinants of the double exit strategy and testing new hypotheses.

5 Secondly, our research also adds valuable results to the literature linking innovation, venture capitals and entrepreneurial firms. Cao, Jiang, and Ritter (2015) show that the financial performance of VC backed IPO firms depends on their innovation output. They measure innovation output by each IPO firm s successfully filed patents and find that VC backed start-ups having at least one patent at the time of going public, significantly outperform the VC backed ones without patents. However, their study is stock performance based. To our best knowledge, so far, no study has examined whether VC successfully nurturing innovation relates to the fate of IPOs on the public market. Moreover, our research also relates to the difference between traditional form of venture capitalists and corporate venture capitalists. (See. e.g. Chemmanur, Loutskina, and Tian, 2014; Ivanov and Xie, 2010). We fill an important gap in the literature by showing that CVC sponsorship could affect the survival of their portfolio companies on the stock exchange after IPO, differently from how it would affect independent VC. The fourth strand of literature that our paper is related to is focused on the exit choice for VC firms. Going public is a favorable exit choice for venture capitalists (See e.g. Bayar and Chemmanur (2011); Poulsen and Stegemoller (2008)). Bayar and Chemmanur (2011) point out that as VCs have short term investment horizon and tend to sell out a large portion of their stake at, or soon after, the IPO stage. They are more motivated to cash out immediately through IPOs. These results could also be explained by VC s limited partnership nature. VCs either need to liquidate their investment to fund new project or return the funds back to their limited partners. Additionally, Poulsen and Stegemoller (2008) study a sample of 1,074 IPOs and 735 sellouts from 1995 to 2004 and find that the median market value of IPO firms ($170 million) is significantly higher than sellout firms ($130 million). The result is even stronger for the subsample of VC funded firms.

6 They argue that VC backing up helps to reduce the information asymmetry of their portfolio firms. Our study complements this strand of literature by examining the impact of VC s gradual exit after the IPO stage. VCs have no real incentive to stay long term involved in the exchange with the IPO firm. To the best of our knowledge few studies have examined the possible impact of VC s exit on the IPO firms fate on the exchange. The empirical study that resembles the most to this paper, is the one by Chemmanur, Signori, and Vismara (2014). These authors examine the dynamic exit mechanism of European private firms. Unlike previous studies that consider private firms exit decisions as a one-time choice between IPO and acquisition, their study focuses on IPO firms second time exit. Specifically, some private firms first conduct an IPO and then seek to be acquired by third parties within the three years after the IPO. They also investigate the determinants of private firms double exit decision to find that IPO firms with higher sales growth, higher growth opportunities and lower cases of information asymmetry, draw the attention of bidders. On the other hand, IPO firms with large private benefits of control are less likely to be acquired. However, they find that VC sponsorship decreases the likelihood of European IPO firms being acquired, contradicting the empirical findings of Gill and Walz (2016). By using U.S. data, Gill and Walz (2016) argue that VC-backed IPOs are more likely to become acquisition targets. Hence, whether a VC sponsorship increases or decreases the likelihood of the second-time exit remains an open question and further research should be conducted. Finally, we contribute to the strand of literature related to the link between market condition and start-ups exit mode (See e. g. Brau, Francis, and Kohers (2003); Poulsen and Stegemoller (2008); Ball, Chiu, and Smith (2011); and Chiu and Kini (2014)). These studies solely examine how market condition affects the first-time exit choice between IPO and M&A. Our specific

7 contribution is the hypothesis that the market timing also has an impact on entrepreneurial firms double exit strategy. 2. Related literature and hypotheses development In this section, we briefly review the literature and establish the framework for testing our hypotheses Innovation hypothesis VCs play an important role in nurturing innovation. Kortum and Lerner (2000) consider 20 industries from 1965 to 1992 and examine the influence of VC on patented inventions. They find that VC activity has a strong impact on innovation. For instance, they find that during the decade from 1983 to 1992, although venture capitals account for less than 3% of corporate R&D on average, they contribute to about 8% of innovation output in the U.S. Bayar and Chemmanur (2011) develop a theoretical model to predict the post-ipo fate of the entrepreneurial firms. In their model, they show that a firm s quality depends on whether it has a viable business model to compete against other firms on the stock exchange. In equilibrium, their model suggests that high quality firms with a viable business mode are more likely to stand alone. Whereas low quality firms tend to employ a double exit strategy to get acquired soon after the IPO, especially when they observe competition has established a toehold on the market. Following these results, we argue that the successful research and development (R&D) expenses yield the patents an IPO firm owns, and these can represent better the uniqueness of a nascent firm. Thus, IPO firms with patents should be more viable against the market competition and should be capable of fending themselves against competition, on the exchange market in a long term basis. In the same vein, Cao, Jiang, and Ritter (2015) examine whether patents can predict long-term performance of 2,254 VC backed IPO firms in the U.S. They find that VC backed IPO firms with at least one

8 successfully granted patent before and at the IPO year, significantly outperform VC sponsored IPOs without any patents. They further show that the long-term performance of VC backed IPOs without patents is similar to that of a non-vc backed IPOs. They conclude that patent, as an innovation proxy, has a significant predictive power of IPOs long-term performance. They note, however, that investors underestimate the economic importance of patents, particularly, for those patents that are hard to value. Chemmanur, Signori, and Vismara (2014) find that IPO firms are attractive acquisition target if it is easier for the acquirer to value the firm. They find that an IPO firm would draw the attention from potential bidders if the firm was brought public by a more reputable underwriter, if it had a greater analyst coverage, as well as a higher extent of IPO oversubscription, and if it showed greater liquidity. That is, potential acquirers are prone to buy firms with lower extent of information asymmetry. In this context, we argue that patents increase the IPO firms information asymmetry and potential acquirers may not want to buy such firms given the fact that there is a greater difficulty in valuing them. Hence, our first hypothesis is: Hypothesis 1: VC-backed IPO firms with patents are more likely to stand alone on the stock exchange, while those without patents are more likely to be acquired shortly after the IPO Longer investment horizon hypothesis Corporate venture capital (CVC) has its unique features comparing to independent venture capital (IVC). First, CVC is structured as a subsidiary of a corporation, whereas IVC employs a limited partnership which has a limited investment lifespan. Within a limited time period, IVC has to liquidate its investments and return funds to its partners. By contrast, CVC has a longer investment horizon due to its hybrid mode. Second, the investment goal of IVC is to seek pure financial returns, and the fund manager s compensation is tied with IVC s financial performance. CVC are, however, looking for more strategic achievements rather than pure pecuniary gains. Chemmanur, Loutskina,

9 and Tian (2014) investigate U.S. IPO firms sponsored by IVC and CVC respectively. They find that CVC is superior to IVC in terms of nurturing innovation in their portfolio companies, but they are less profitable and riskier in the short term. They conclude that CVC s greater accomplishment in nurturing innovation stems from improved technological fit between CVC parent and start-ups. Masulis and Nahata (2009) also show that CVC backed start-ups have a better technological fit with CVC parents. They point out that this advantage arises from one of CVC s investment motivations, which is to help its parent, form a strategic alliance with the CVC backed ventures. These ventures could augment the competitive advantage of the CVC s parent by bringing new ideas and technologies. Masulis and Nahata (2009) further argue that such a strategic alliance is a double-edged sword. The corporate parents tend to prioritize their own profitability at the expense of their CVC backed ventures. Hence, the alliance between CVC-funded start-ups and CVC s parent could impede the attractive acquisition bids made by the competitors of CVC s corporate parent, because CVC aim to protect the competitive advantage of its corporate parent. In addition, the CVC toeholds could decrease the value of an otherwise attractive target to potential acquirers because of the strategic alliance between start-ups and corporate parents. Therefore, a CVC backed IPO might be somewhat forced to stand alone, even if it is not a value maximizing decision. Furthermore, because of CVC s almost unconstrained investment horizon and its abundant funds supply, as well as the fact that investors tend to invest in more innovative projects which have a long-term horizon. These innovative projects may not be able to generate immediate positive cash flows but have great upside potential of yielding innovation outputs which can significantly benefit start-ups viability in the market (Chemmanur, Loutskina, and Tian (2014)). Hence, the myopic acquirers might not be able to detect the potential value of these CVC backed start-ups and are less

10 likely to make acquisition bids on such firms in the short run. Our second hypothesis will be the following: Hypothesis 2: CVC backed IPO firms are more likely to stand alone on the stock exchange, while IVC backed IPO firms are more likely to be acquired shortly after the IPO. Hypothesis 3: CVC backed IPO firms with patent are more likely to maintain the public status Harvesting s effect hypothesis Brau, Sutton, and Hatch (2010) examine 679 takeovers from 1995 to 2004 and find that public firms that employ a double exit strategy earn significant higher harvest value than private firms that employ a single-track sell out. Specifically, public dual-track sell-outs receive 18% to 21% higher premiums than single-track sell-outs. They also find that VC backed firms are more likely to conduct dual-track sell outs. Pan and Zhou (2015) also investigate U.S. post-ipo acquisitions from 1980 to 2007 and find that IPO firms are more likely to become acquisition targets than seasoned firms. In addition, IPO firms that employ a double exit strategy receive significant higher takeover premium than seasoned targets. Chemmanur, Signori, and Vismara (2014) discuss the motivation of the double exit strategy in Europe. They point out that most IPO firms that opt for a dual-track sell-out originally expected to remain on the stock exchange as stand-alone firms. However, they later choose to be acquired given the attractive value of the acquisition bid they get offered. Based on the above literature, we believe that the harvest value plays a crucial role in the double exit strategy. It is possible that for some IPO firms, the gross proceeds received from going public are not as high as expected. Hence, venture capitalists are not satisfied with the harvest value from the IPO stage. Under this scenario, venture capitalists would seek attractive acquisition offers for their portfolio companies to further maximize the harvest value. If there is an acquisition offer

11 whose valuation is higher than the benefits of standing alone for such IPO firms, venture capitalists would push those IPO firms to double exit by accepting the acquisition bid. Thus, we propose the following hypothesis: Hypothesis 4: There is a negative relation between IPO gross proceeds and the likelihood of employing a double exit strategy Market timing hypothesis Brau, Francis, and Kohers (2003) s empirical work reveals that the exit mode of private firms is affected by market conditions. They analyze the exit choice of 9,500 U.S. private firms and find that the exit vehicle of private firms is bounded up with the macroeconomic market condition. Specifically, when the economy follows a hot period and the T-bill rate is high, IPOs are more likely to occur, while during cold periods, sellouts are more likely to occur. Poulsen and Stegemoller (2008) confirm these results and find that exit mechanisms of private firms are affected by the aggregate capital demand. They point out that these results are strongest within VC backed firms, given their expertise in maximizing their harvest value through the optimal exit route. Ball, Chiu, and Smith (2011) investigate whether VCs can time the market by analyzing the exit mechanisms of VC backed start-ups. Using a sample of 8,163 VC backed firms over three decades, they find that the exit strategy of venture capitalists is in line with the market timing hypothesis. Specifically, their empirical results indicate that market run-up before the exit is positively related to the choice of going public. In addition, their results suggest that the preference of VC s exit choice depends on the capital demand, measured by the one-year Treasury bill rate. The IPO waves occur when the capital demand is high, which means that there are plenty of investment opportunities on the market, and VCs prefer to take advantage of the hot market to exit through an IPO. On the other side, when the capital demand is high, firms on the market have rich menus of

12 their own growth opportunities and are less prone to acquire others. Hence, during hot IPO period, potential acquirers are scarce and acquisition will not be a favorable exit choice for venture capitalists. Chiu and Kini (2014) also confirm the relation between market conditions and private firms exits. They use the aggregate equity mutual fund flows as a proxy of the rational component of the aggregate demand for equity. When the equity mutual fund flows are higher they find that more firms conduct IPO and fewer firms withdraw their IPOs. In line with the aforementioned literature, we argue that VC backed firms do not always prefer to conduct dual-track sell outs. By timing the market, VC s choice for their exit route is based on maximizing their harvest value. During hot IPO period, venture capitalists are able to raise more capital through conducting IPO. Meanwhile, potential acquirers are hard to find when the market is hot (Ball, Chiu and Smith, 2011). Therefore, VC can make the start-ups that are not necessarily able to fend themselves on the public market, public, to be able to earn a higher capital during the hot market period. Then, VC can seek for potential bidders for such firms for a second-time exit. On the other hand, if VC choose to bring their portfolio firms to public during cold IPO period, it means that VC believes such firms are the real gold and are able to stand alone as a public company making IPO the best exit choice for them. Thus, we propose the following hypothesis: Hypothesis 5: VC backed IPO firms are more (less) likely to employ a double exit strategy if they go public during the hot (cold) period. 3. Data and Sample: To establish our research sample, we have collected data across various database. 3.1 IPO and double exit data We identify the VC-sponsored IPO firms by using Thomson-Reuters Securities Data Company (SDC) new issues database. We exclude IPOs with offer price less than $5 or missing, unit

13 offerings, spinoffs, closed end funds, American depositary receipts, and real estate investment trusts. To identify the double exit, we check IPO firms public status up to the third year after issuing year. We consider a VC-sponsored IPO firm employs the double exit strategy if this firm is acquired within the three years after the IPO year. We use the SDC Mergers & Acquisitions database as well as CRSP delisting codes to identify the acquisition targets. We exclude acquisitions with deal value smaller than $1 million or acquisitions of less than 50% of target equity. VC-backed IPOs who are delisted by other means than M&A are not considered in our sample because we specifically focus on VC s double harvesting strategy. Finally, there are 1,318 VC-sponsored IPOs in our sample from 1980 to Because we want to trace whether an IPO firm gets acquired within the three years after going public, our sample periods ends in 2013 given the data availability. 3.2 Patent data We collect the patent data from the latest version of the National Bureau of Economic Research (NBER) patent database created by Hall, Jaffe, and Trajtenberg (2001). The NBER patent database provides detailed information on more than three million patents that were granted by the United States Patent and Trademark. We match the NBER patent database to VC-backed companies using GVKEY and CUSIP identifiers (Cao, Jiang, and Ritter (2015)). We then check the grant date of each IPO s filed patents. We define the variable of Patent by counting the granted number of patents before IPO year. For robustness check, we also examine the quality of patents, reflected by the number of citations each granted patent receives 1. While patents protection starts from the application dates and innovation begins once they appear (Hall, Jaffe, and Trajtenberg, 2001), it takes on average two years for a patent to be granted by 1 We count the number of citations by the same method as counting the number of patent.

14 USPTO. To avoid the look-ahead bias due to patent application-grant lag, following Hirshleifer, Hsu, and Li (2013), we hence date all patents by their grant date instead of application date. To measure the number of patent, we calculate the natural logarithm of number of patent for each IPO firm 23. Similarly, we calculated the natural logarithm of number of citation for each start-up. Given the fact that the latest version of NBER patent database has the data available until 2006, our subsample examining the effect of innovation on double exit strategy is from 1980 to Our subsample consists of 1,100 VC-backed IPOs. 3.3 Corporate venture capital data We define corporate venture capital (CVC) by checking the SDC VentureXpert database. VentureXpert database provides the information whether an IPO receives funding from CVC. We verify whether a CVC is indeed an investment arm of a corporation by using a variety of source (Bloomberg, Google, etc.). We find some VC cannot be considered CVC because they are actually financial institutions, partnerships, or have unknown corporate parents. Finally, 354 IPOs in our sample receive funding from CVC. We use different measures to capture the CVC involvement. First, we create a CVC dummy (CVCD) which takes the value of one if an IPO receives funding from at least one CVC before IPO. Second, we find many firms are funded by more than one CVC. We then calculate the share percentage of CVC in a VC syndicate (CVCS). In our Poisson regression, we also count the number of CVC in an investing VC syndicate (CVCN). 3.4 IPO proceeds data 2 Griliches, Hall, and Pakes (1987) indicate that the distribution of the value of patents is extremely skewed because the patents are clusted in a small number of firms. 3 When taking the natural logarithm, we add one to actual values to avoid losing observations with zero patent and/or citation (Chemmanur, Loutskina and Tian, 2014)

15 The IPO gross proceeds data is obtained from SDC new issues database, which is the dollar amount of capital raised from conducting IPO. We calculate the natural logarithm of the dollar amount. 3.5 Market condition data Following Gill and Walz (2016), we create two measures for market condition. The first one is IPO number, which measures the aggregate number of IPOs in the issuing year. We obtain the number of IPOs in each year from Dr. Jay Ritter s website. The second measure is Market return, which is total annual returns of the market index composite, including NYSE, AMEX, and NASDAQ of the issuing year. Market return information is collected from CRSP. 3.6 Control variables Following the IPO and venture capital literature, we include the following control variables in our baseline sample: the percentage of annual gross proceeds raised by the lead VC (VC reputation), Carter and Manaster s underwriter rank (Underwriter rank), profitability (ROA), firm size (Size), leverage (LEV), free cash flow (FCFF), capital expenditure (CAPEX), growth opportunities (Tobin s q), sales increase (Sales growth), trading volume (Volume), firm age (Age), stock performance (Stock return), Herfindahl index (HHI), dummy variable indicating IPO bubble period (Bubble period), and dummy variable indicating IT firms (IT dummy). Detailed information of our control variables is explained in Table 1. Please insert Table 1 about here 4. Methodology In this section, we introduce the applied models to study VC backed IPO s double exit strategy. 4.1 Innovation and double exits To examine how innovation can influence the likelihood of VC-sponsored IPO s second time exit, we carry out the following logit regression analysis:

16 1/0, (1) where is a binary variable that equals 1 if a VC-backed IPO firm becomes acquisition target within the three years after issuing year, and 0 otherwise; innovation is measure by the variable of Patent (see Table 1 for detailed definition); is a vector of control variables (VC reputation, Underwriter rank, ROA, Size, LEV, FCFF, CAPEX, Tobin s q, Sales growth, Volume, Age, Stock return, HHI, Bubble period, IT dummy); are industry fixed effects 4 ; and are year fixed effects. If innovation enhances a VC-funded IPO s survival on the public market, we expect the coefficients on Patent to be negative. Put it differently, we expect innovation diminishes the likelihood of a second time exit for VC-funded IPO. For robustness check, we replace Patent with Citation to investigate whether the quality of innovation relates to VCsponsored IPO s double exit strategy. Because of the availability of patent data explained in Section 3, we study Hypothesis 1 by using our subsample from 1980 to Corporate venture capital and double exits To investigate our investment horizon hypothesis, we implement the following logit model by adding a key variable measuring CVC involvement: 1/0, (2) CVC involvement is captured by two alternative measures, CVC dummy (CVCD) and CVC share (CVCS) (see Table 1 for detailed definition); is a vector of control variables (VC reputation, Underwriter rank, ROA, Size, LEV, FCFF, CAPEX, Tobin s q, Sales growth, Volume, Age, Stock return, HHI, Bubble period, IT dummy); are industry fixed effects; and are year fixed 4 Industry dummy variables are constructed based on the 49 Fama-French industry classification.

17 effects. If CVC indeed are more patient investors and have different investment strategies than IVC, we expect the coefficients on CVC to be negative 5. We next study our third hypothesis pertaining to the effect of CVC nurturing innovation on the likelihood of IPO firms second time exit. We include an interaction term of CVC and innovation (CVCD Patent) or (CVCD Citation) in the logit regression. 1/0, (3) If CVC s advantage in terms of augmenting the survival ability of their portfolio firms on the public market is that CVC do a better job in successfully nurturing innovation output, we expect the coefficient of CVCD Patent to be negative. 4.3 Harvesting effect and double exits With regard to our harvesting effect hypothesis, we carry out the following logit model: 1/0,,,, (4) Harvesting value is measured with IPO proceeds, which is the natural logarithm value of the dollar amount of the gross proceeds raised through conducting IPO. is a vector of control variables (VC reputation, Underwriter rank, ROA, Size, LEV, FCFF, CAPEX, Tobin s q, Sales growth, Volume, Age, Stock return, HHI, Bubble period, IT dummy); are industry fixed effects; and are year fixed effects. If our Hypothesis 4 holds, venture capitalists who are not able to maximize their wealth through the exit route of IPO are more likely to seek for a second chance of harvesting. We expect the coefficient of to be negative. 4.4 Market timing and double exits 5 We also include Patent in the logit model for all the following hypotheses. Due to the availability of patent data, the regression by including Patent is based on our subsample.

18 To test our market timing hypothesis, we further include the two proxies for market condition in the logit model: 1/0, (5) IPO number measures the number of firms going public in each issuing year and Market return measures the total annual returns of the market index composite, including NYSE, AMEX, and NASDAQ of the issuing year. is a vector of control variables (VC reputation, Underwriter rank, ROA, Size, LEV, FCFF, CAPEX, Tobin s q, Sales growth, Volume, Age, Stock return, HHI, Bubble period, IT dummy); are industry fixed effects; and are year fixed effects.in line with our Hypothesis 4, if the start-ups going public during high market demand period are more likely to become acquisition target, we expect and to be positive. 4.5 Poisson regression To check the robustness our innovation hypothesis and investment horizon analysis, we carry out Poisson regression analysis given the fact the innovation outputs and corporate venture capitalists involves count variables. For the robustness check of the effect of CVC on their portfolio companies double exit strategy, we check the number of CVC in each VC-backed IPO s VC investing syndicate. Our dependent variable in the Poisson model is CVCN, which is the count of CVC in the VC syndicate. Our model reads as follows:, (6) Our interest in the Poisson model shown as Equation (6) is the coefficient of Double exit ( ). If more CVC involvement attenuates the likelihood of start-ups second time exit, we would observe the coefficient of to be negative. is a vector of control variables (Patent, IPO proceeds, IPO number, Market return, VC reputation, Underwriter rank, ROA, Size, LEV, FCFF, CAPEX,

19 Tobin s q, Sales growth, Volume, Age, Stock return, HHI, Bubble period, IT dummy); are industry fixed effects; and are year fixed effects. Similarly, for the robustness check for our innovation hypothesis, we propose a similar Poisson framework. The dependent variable is the count of either Patent or Citation. See Table 1 for detailed definition for Patent and Citation. The model is shown as follow: /, (7) Again, if innovation outputs would enhance entrepreneurial firms capability of fending themselves on the public market. The coefficient of Double exit,, is expected to be negative. Control variables, as well as industry and year fixed effected remain the same as Equation (6). 4.6 Survival analysis In this section, we employ Cox hazard model to carry out survival analysis for VC-backed IPOs. To examine whether innovation, CVC sponsorship, harvest value and market condition relate to the likelihood of the hazard event of a second time exit. We propose the following model: h(t)=h0e x'β (8) the Cox hazard model depicts the hazard function h(t) as a vector of explanatory variables x, h0 represents the baseline hazard rate. In line with our study, the hazard rate represent the instantaneous probability of a VC-backed IPO becoming acquisition target within the three years after listing. We are interested in the hazard rate (exp(x'β)) for the variable of Patent (Citation), CVCD, IPO proceeds, IPO number and Market return, as well as the interaction term between Patent (Citation) and CVCD. Based on our hypotheses, we expect that innovation, CVC involvement, harvest value would decrease the hazard rate of double exit. On the other hand, going public during hot market period would increase the hazard rate. 5. Empirical Results

20 In this section, we report the results of our empirical analysis, starting from univariate analysis and following by multivariate regression analysis. 5.1 Descriptive statistics Table 2 describes the annual number of VC-backed IPOs that remain standing-alone on the stock exchange within the three years after going public, as well as that becoming acquisition targets within the same period after IPO stage across over sample period. In addition, we report the annual percentage of the VC-sponsored IPOs carrying out the double exit strategy. As we can see, in total there are 295 VC-funded IPOs experiencing a second time exit within the three years after listing, taking up to about 20% of all VC-backed IPOs. Moreover, we can observe that the number of double exit firms varies across over sample. In general, Table 2 shows that double exit strategy becomes more common from the last decade of the 20 th century, with the number of double-exit being the highest during the IPO bubble period (1999 to 2000). Table 3 depicts the annual dollar amount of the capital raised by VC-backed start-ups from IPO exit route over our sample period, along with the annual dollar transaction value of a second time exit through M&A transaction. In a nut shell, Table 3 exhibit a similar trend as Table 2. IPO activities are the most active during the IPO bubble period, whereas the IPOs going public during such period are more likely to experience a second exit shortly after listing, regardless of measuring by percentage or dollar amount. This is seemingly to be the first evidence that marking timing is bound up with VC-backed IPO s double exit strategy. Table 4 illustrate the industry classification of VC-backed IPOs in our sample. Industry classification is based on the 49 Fama-French industry classification. We can see the start-ups in our sample are distributed across 35 industries. One interesting fact is that both standing-alone and

21 double exit IPOs are clustered in the IT industry. For this reason, we create an indicator variable of IT dummy for our following multivariate regression analysis. Please insert Table 2, 3, and 4 about here Table 5 report the summary statistics for all the explanatory variables of standing-alone and double exit VC-backed IPOs (see Table 1 for detailed variable definitions). Our sample consists of 1,023 standing-alone VC-sponsored IPOs and 295 double exit ones. Because the patent data is only available up to 2006, there are 851 standing-alone and 249 double exit VC-funded IPOs in our subsample for the study of innovation. For standing-alone firms, we can clearly see that they have more innovation outputs, regardless of which proxy is used (i.e. Patent or Citation). The mean of Patent and Citation is 1.12 and 1.23 for standing-alone firms, which is 86.67% and 46.43% higher than that for double-exit ones. The differences in Patent and Citation between standing-alone VC-backed IPOs and double exit ones are statistically significant at the 1% level, displaying univariate support for our innovation hypothesis (Hypothesis 1), i.e. that VC-sponsored IPOs who have successfully nurtured innovation are less likely to implement the double exit strategy. With regard to our longer investment horizon hypothesis (Hypothesis 2), our univariate results confirm our conjecture that CVC-funded IPOs have a less chance to seek for a second time exit, compared to IVC-backed ones. We find statistically significance in all the measure of CVC involvement (CVCD, CVCN, and CVCS). For instance, for our main CVC dummy variable, it shows that 28% of the standing-alone IPOs have received funding from corporate ventures, whereas only 22% of the double exit IPOs have CVC sponsorship. For our market timing hypothesis, we find that our first measure statistically support our story: IPO firms going public in the issuing year with higher IPO number are more likely to

22 carry out the double exit strategy. The mean value of IPO number for double exit firms is 11.41% higher and the univariate difference between the two groups is statistically significant at 1% level. However, we find no univariate evidence that the issuing year s market return is higher for double exit firms. Also, we did not find univariate evidence that VC-funded IPOs whose gross proceeds are lower are more likely to seek for a second time harvest. We will explore these factors further in our multivariate context. For the control variables, we find that double exit VC-backed IPOs are on average backed by less reputable venture capitalists (VC reputation), have less growth opportunities (CAPEX and Tobin s q), more free cash flow (FCFF), and worse stock performance (Stock return). Also, VC-funded IPOs listed during bubble period (Bubble period) as well as belong to the IT industry (IT dummy) are more likely implement the double exit strategy. The correlation matrices in Table 6 show that the pairwise correlations potential multicollinearity is not a major concern, we also calculate the Variance Inflation Factors (VIF) in our multivariate regressions. In line with our bivariate correlation analysis, multicollinearity does not appear to pose any problems in a multivariate setting 6. Please insert Table 5 and 6 about here 5.2 The effect of innovation on VC-backed IPO s double exit strategy To investigate the link between innovation and the probability of carrying out the double exit strategy, we show the results of logit regression in Table 7. Our baseline results in column 1 indicate that innovation outputs decreases the likelihood of double exit. The coefficient of Patent is and statistically significant at the 1% level. The related marginal effect reveals that on average, the predicted probability of becoming an acquisition target of a VC-backed IPO decreases 6 In all our regression analysis, VIF is less than the critical value of 10 (see Kutner et al., 2005).

23 by 4.94% with a one-unit increase in Patent. Similarly, column 2 reveals that the quality of innovation diminishes the probability of the second time exit. The coefficient of Citation is and statistically significant at the 1% level. The marginal effect shows that on average, a one-unit increase in Citation decreases the likelihood of a second time harvest for VC-funded IPOs by 3.18%. In sum, these results provide support for Hypothesis 1, namely that VC-backed start-ups are not in favor of the double exit strategy. In addition, the results are both statistically and economically significant. In line with Bayar and Chemmanur (2011) s theoretical framework, innovation outputs significantly increases IPO firms capability to fend themselves against the fierce competition on the public market. Under this scenario, both the entrepreneur and VC agree to maintain the public status. Please insert Table 7 about here 5.3 CVC involvement and the double exit strategy Table 8 depicts the empirical results of the effect of CVC sponsorship on their portfolio companies probability of experiencing a second time harvest. Column (1) shows the results by measuring CVC involvement by CVC dummy (CVCD). The coefficient of CVCD is and statistically significant at 1% level. The related marginal effect indicates that the probability of carrying out the double strategy decreases from 24.32% by 40.09% (or 6.96% percentage points). This result is robust when measuring CVC sponsorship by CVC s share in a VC syndicate (CVCS). The coefficient of CVCS is and statistically significant at 5% level, illustrating that a one-unit increase in the CVC s share in a VC syndicate mitigates the likelihood of the double exit strategy by 24.83%. In addition, column (3) and (4) provide the results including innovation outputs by using our subsample and our results remain unchanged 7. The above finding provide strong 7 For robustness check, we also include the measure of innovation by using Citation. The results remain unchanged.

24 evidence for Hypothesis 2 that CVC are more patient investors and have different investing motives from IVC by exhibiting longer investment horizon. Therefore, IPO firms with CVC involvement are less likely to be delisted by M&A transactions. Please insert Table 8 about here 5.4 The effect of CVC nurturing innovation and the double exit strategy Table 9 shows the empirical results for testing our Hypothesis 3. Our main interest is the interaction term between CVC involvement and innovation outputs (CVCD Patent) or (CVCD Citation). The coefficient of CVCD Patent is negative and statistically significant at 1% level. We interpret it as the supporting evidence for our Hypothesis 3 that CVC is advantageous over IVC on the ground that CVC do a greater job in terms of nurturing innovation for their portfolio companies. Therefore, CVC-backed entrepreneurial firms with patent are more able to stand alone on the public market and do not need to seek for a second time exit. Similarly, we do find statistically significant evidence that CVC-backed IPOs with better quality of the innovation outputs are more likely to stand alone. To sum up, we find robust evidence to support our Hypothesis 3 that CVCbacked start-ups with innovation outputs are less likely to employ the double exit strategy. Please insert Table 9 about here 5.5 Harvesting effect and the double exit strategy The relationship between harvesting value of VC-backed IPOs double exit are shown in Table 10. Nonetheless, regardless of using our entire sample or our subsample. The coefficient of IPO proceeds, which measures the harvesting effect, is not statistically significant at conventional levels. Hence, we cannot find empirical support of our Hypothesis 4. Please insert Table 10 about here 5.6 Market timing and the double exit strategy

25 The empirical evidence of the impact of market condition on venture capitalists and entrepreneurs double exit strategy is displayed in Table 11. The coefficient of IPO number in both column 1 and column 2 (our entire sample and subsample of including innovation) is positive and statistically significant at 1% level. For instance, the related marginal effect of IPO number in column 1 (our entire sample) reveals that a one-unit increase in the number of IPO in each issuing year is associated with a 0.03% increase in a second time exit within three years after listing for VCbacked IPOs. This finding is in line with our Hypothesis 4: during hot IPO period, both venture capitalists and entrepreneurs want to time the market to achieve higher harvesting value, so they both agree to bring the start-ups to public instead of trade-sales. However, such firms are not necessarily able to fend themselves as public firms in them lone term. Thus, market-timing IPOs are more likely to seek for another exit in the short run. Please insert Table 11 about here 5.7 Control variables In various multivariate regressions, we find some statistically significant evidence for some control variables. Firstly, we find that more profitable firms (ROA) are more likely to double exit. This finding is consistent with that of De and Jindra (2012) that good financial performance IPOs are more likely to draw acquirers attention. In addition, we find that large firms (Size) as well as firms with more growth options (Tobin s q) are less likely to get acquired shortly after IPO. Morck et al. (1990) show that firms with a rich menu of investment opportunities have higher demand for funding from public markets and such firms are prone to keep their public status. Moreover, firms with better stock performance (Stock return) are less likely to experience a second time exit. Probably for such firms shareholders are not happy for their stock performance, it is better to seek for a second time exit to increase shareholders wealth. Furthermore, we find that firms in the IT

26 industry (IT dummy) and get listed during the IPO bubble period (Bubble period) are more likely to employ the double exit strategy. We argue that this fact somehow supports our market timing hypothesis. As we know the majority of IPOs during the bubble period are IT firms, those firms intend to make use of the hot market period to maximize their wealth. Therefore, they rush into the public market without considering how long they are able to stand alone on the public market. 6. Robustness checks 6.1 Poisson analysis For our first robustness check, we investigate whether the existence of more corporate venture capitalists in a VC investing syndicate results in a higher chance for their portfolio companies to stand alone as a public firm. Similarly, we examine whether more patents (citation) counts decreases the likelihood of VC-sponsored IPO s double exit. We employ the Poisson model to study the above questions. Table 12 and 13 provide the empirical results for the Poisson regression analysis. In Table 12, we can clearly see that the coefficient of Double exit in all three specifications (the entire sample and subsample by including innovation information) is negative and statistically significant at 5% level. This result further support our Hypothesis 2 that CVC involvement augments portfolio companies survival chance on the public market. Interestingly, the Poisson regression analysis also unveils that more corporate venture capitalists in a VC syndicate results in higher innovation outputs, regardless of patents themselves or citations of granted patents. This result is echoing the find of (Chemmanur, Loutskina, and Tian, 2014) that CVC do a better job in successfully nurturing innovation. Table 13 describes the result of Poisson regression for the link between innovation and the likelihood of double exit. We can see that regardless of measuring innovation by Patent or Citation, firms with more patent (citation) count