May 2012 Forthcoming, Journal of Economic Studies

The Internationalization of Venture Capital *

Joshua Aizenman and Jake Kendall

UCSC and the NBER; The Gates Foundation

Abstract

We investigates the factors which affect the market for international venture capital investments, relying on comprehensive deal-level data sources, covering three decades and about 100 countries. We document major shifts in the nature of international flows. A gravity analysis indicates that distance, common language, and colonial ties may have been significant factors in directing these flows. The presence of high-end human capital, a better business environment, military expenditure, and deeper financial markets are important local factors that appear to attract international venture capital. There is some evidence indicating network effects and/or fixed costs of entry may be at work. France, Israel, Canada, India and China were consistent net importers of VC deals, with China emerging as the largest net importer of VC.

Joshua Aizenman Jake Kendall Economics E2, UCSC The Gates Foundation Santa Cruz, CA, 95064 PO Box 23350, Seattle, WA 98102 jaizen@ucsc.edu Jake.Kendall@gatesfoundation.org

Keywords: Venture capital, Gravity equation, path dependence, cross border flows JEL classification: F15, F21

* We would like to thank the following people for their help and insights into the venture capital and private equity world: Brad Maclean of the Asian Venture Capital Journal, Oliver Roupe of the Founder’s Fund, Mawuli Ababibo of the African Venture Capital Association, and various staff members at Thomson VentureXpert. We thank Dennis Novy and Thomas Hellman for useful comments. Any errors are ours.

2

"The Great Divide -- A Narrow Majority of VCs Go Global While the Rest Stay Focused Stateside, the United States is Still the Top Destination of Choice." Global Trends in Venture Capital 2006 Survey, Report Sponsored by Deloitte & Touche USA LLP 1. Introduction

Venture capital (VC) is long-term, hands-on equity investment in privately held, high-growth-potential companies, initiated and managed by professional investors [see Lerner (2001) and Gompers and Lerner (2001) for evaluation and reviews of the recent literature]. While venture capital firms and the companies they own are not significant in terms of employment generation or capital under management, the role of VC in the economy should not be underestimated. Many of the top US technology companies got started with VC funds and some of these same companies even use VC-style investments to enter into foreign markets (see, e.g. Dow Jones (2009)). In addition, VC directed investment seems to be especially potent in the innovation process. This role was confirmed by Kortum and Lerner (2000), who examined the influence of venture capital on patented inventions in the United States across twenty industries over three decades. They found that venture capital investments in an industry are associated with significantly higher patenting rates than regular R&D by almost a 7 to 1 ratio. Countries around the world have taken note, and attempted to replicate the success of the US VC industry on their home soil, both by encouraging inward investment from foreign players and by growing capacity locally.

While the VC sector in the United States has been large and vibrant for many years, the rest of the world saw very little growth in VC activity until the mid 1990s.1 Since that time, the internationalization of the VC sector has been driven both by general factors driving globalization as well as historical facts specific to VC. As Figure 1 shows, a great deal of growth in international activity occurred starting in the late 1990s around the time of the US-centered technology bubble, some of which was driven by an oversupply of funds to US VC firms who may have gone international in the search of deal flow. A lesser known fact, visible in Figure 2 is that while US VC firms went abroad more actively over this period, foreign VC firms also came to the US for investments in US-based companies in greater numbers. While a bubble may have driven the growth VC funding out of and into the US, the destinations of these flows have not been random, and the continuing significance of cross-border flows as well as the continued growth of domestic VC capacity in countries which had almost none previous to the late 1990’s is further evidence that the globalization of the VC industry is here to stay.2

The purpose of this paper is to investigate the increasing internationalization of venture capital investments in recent years and to assess the factors which determine the destination of cross-border VC investment flows. To address these issues, we rely primarily on the Thomson VentureXpert database -- a comprehensive data source for deals done worldwide that includes over 30 years of historical data. It draws its information from quarterly surveys of private equity firms; government filings; public news releases; and Thomson reporters as they interact with

1 The roots of the modern VC industry go back to the aftermath of World War II, including the foundation in 1946 of American Research and Development Corporation. ARDC is credited with the first major venture capital IT success story, when its 1957 seed investment in Digital Equipment Corporation (DEC) delivered annual return of nearly 100% after the company's initial public offering in 1968 [see Wilson (1985)]. 2 Recent survey evidence from Deloitte (2006) indicates that the internationalization of VC and PE activities for firms based in the US and Europe may just be taking off. More than 50% of the firms they survey are planning to expand globally.

3

private equity practitioners.3 Each deal in our sample includes the names of the investors and the locations of their offices, the name of the target company and their location, as well as some details on the stage of the deal, the industry in which the company operates, the total value of the deal (amount of cash invested), and the timing.4 With this information, we can calculate the volume of deals done in a given country in a given year, and the bilateral flows of deals between countries, defining as the source country the home country of the investment firm and the destination country as the home country of the target company. All total, there are just over 100 countries represented in the dataset, both as source and destination countries [see Data Appendix, available at NBER Working Paper No. 14344, and Sections 2 and 3 for further details].

Before continuing, a few definitions are in order. The terms private equity and venture capital are not well defined and have conflicting uses in the financial press. Much of the practitioner community uses the term “private equity” to refer to both venture financing and other private placement deals such as buyouts and acquisitions of private companies. Our definition of venture capital is in line with that used at Thompson VentureXpert, the European Venture Capital Association, and the American National Venture Capital Association, three definitive industry sources. We define venture capital (abbreviated: VC) by the stage of the investment and include seed, startup, early stage, and expansion stage deals.5 These are the preliminary stages a company goes through in its initial journey to becoming a functioning corporation with steady revenues and a well defined product and customer base that has the potential to issue shares through an IPO. The rest of the deals in the VentureXpert dataset are private equity (abbreviated: PE). These include buyout, buy-in, merger, acquisition, late stage, turnaround/rescue, PIPE, and many other types of deals that are not focused on funding startup and growth.6 We do not use these deals in the current analysis.

We start in Section 2 by evaluating the overall trends, inferred by aggregating the deal-level data. The internalization of VC is an ongoing trend but prior to the early 1990s, as a first order approximation, the VC industry was confined to the US. Hence, we focus on the years 1990-2007, and explore the internationalization of both VC through the cycle of the 1990s IT bubble, the subsequent crash, and the more recent period of activity occurring after 2002.

In summary, we find that the late 90’s bubble’s effect on deal volumes was to increase the total volume by roughly 4-5X in our three series: domestic only deals, domestic deals with some foreign VC funding, and foreign deals with some US funding. The percentage of deals in the US which featured any foreign participation increased dramatically from around 10% to over 25% as foreign investors were drawn in at the height of the bubble. However, this ratio was slow to recede to its pre-bubble level and was still over 15% in 2007. The bubble may have also driven US firms to go more abroad more aggressively, especially into Asia and Europe. This trend also has not completely reversed after the bubble (though it’s not clear to what extent this trend is a facet of the ongoing globalization of IT vs. an element of the late 1990’s bubble). The

3 Thompson, and many industry practitioners, refers to venture capital as a form of private equity and the VentureXpert database tracks both traditional VC as well as other forms of private equity. 4 Unfortunately, we do not know the share of the total investment that was contributed by each investor which makes it impossible for us to calculate actual flow volumes. 5 Expansion stage deals generally occur after the venture has achieved some limited success with its product and are often a way to allow founders to achieve some liquidity by selling part of their stakes to new VC partners. 6 We attempted to analyze the PE data in a previous version of the paper but subsequent conversations with heavy users of the data led us to drop this approach as the PE data seems only more recently to have been a priority for VentureXpert.

4

number of VC partners per deal and average deal size grow over the course of the bubble with both sets of series staying noticeably higher after the bubble crash.

In addition to the changes in investment patterns over the course of the bubble, crash, and recovery, we also look at systematic differences between domestic and international deals. Domestic-only deals appear to be smaller on average, with foreign funded US deals and US funded foreign deals successively larger (a pattern that holds up before, during, and after the bubble). US VC firms going abroad tend to have more investor partners (over 4 on average) compared to 2-3 for US based deals with or without foreign participation (we don’t know if the foreign funded deals done in the US feature more investor partners than deals done in the foreign market without international participation, due to very limited data on these deals). Finally, foreign funded deals in the US are the most likely to have been in the high-tech sector followed closely by US domestic only deals. Since both these ratios are over 80% for most of the sample, they are not much increased by the bubble; however, in foreign deals with some US funding, which started with less that 60% of deal volume occurring in high-tech deals caught up rapidly, increasing to near 80% where it remained after the crash. This final trend may be partly accounted for by the apparent increase in US participation in Europe and Asia after the crisis (see Figure 6), which experienced strong bio-tech and IT based growth respectively.

We continue in Section 3 with an econometric assessment of the factors explaining the number of international venture capital deals between countries. We apply various versions of the gravity model to the patterns of flows. The empirical gravity structure has been widely applied in the international trade literature [see the review by Anderson and Van Wincoop (2004)] and to model international financial flows of FDI [Aizenman and Marion (2004)], portfolio equity [Portes and Rey (2005)], and sovereign lending [Rose and Spiegel (2004)].7

We find in the full sample that distance is negatively related to deal volume and to the number of venture capital deals, indicating the location bias in these types of investment activity. This result is not surprising in light of the intensive monitoring and advisory roles that these types of investments require. Conversations with practitioners indicate that even purely US-based VC’s will often choose only a few cities in which to operate, sourcing several deals in each city so as to economize on travel to that area. Language connections between countries and colonial relationships are also significant predictors of venture capital deal counts, possibly being a proxy for the extent of social and business networking between the source and destination countries as well as for the similarity of institutions, levels of trust, and communication costs. We find evidence of persistence in venture capital flows even when controlling for source and destination country effects indicating the possibility of network spillover effects, herding, and/or “beachhead” effects related to fixed costs of commencing activity in a given country. We confirm that venture capital tends to flow to countries with more high-end human capital, with a better institutional environment, and deeper financial markets (especially stock markets, which facilitate exit from venture investments via IPO). We find that these same factors are also associated with increased flows from the country. There is also evidence defense expenditure and venture capital activity may be complementary.8

In regressions with the US as the sole source or destination of the deal, the patterns are largely similar, with some interesting differences. When the US is source, distance is actually positively and significantly associated with deal flow, and distance is insignificant in the

7 While empirically useful, the theoretical foundations for applying the gravity to FDI and other flows of capital remains wanting and debatable [see Blonigen (2005) for a review of related issues]. 8 Plenty of anecdotal evidence indicates that the military played a central role in the development of the US and Israeli VC sectors [see Wilson (1983)].

5

regression with US as destination. Largely this reflects the US’s unique geographic position, distant from most of the VC deal centers in Europe and Asia. Dummy variables for the source country having been a colonizer, and any shared language also have opposite the expected sign in these two regressions. The US may also be special in regard to these two variables. Only the Philippines has the US as a colonizer in our sample; and few of the developed countries where VC deals are done have English as an official language despite its wide spread use among business professionals.

Section 4 concludes with a summary of the results and the implications for future work. 2. Trends and Features of the Data

The patterns of the international movement of venture capital funds are interesting in their own right and have not gotten much attention in the academic literature. Here we attempt to document the aggregate patterns as they appear in our data. We focus on three types of deals: US based deals with international VC funding, foreign based deals with US VC funding, and US based deals funded purely from domestic based VC funds, mainly to provide basis for comparison. Except for Figure 6 where we attempt to estimate actual cross-border flows, we generally calculate total volumes of funds invested in deals including funds from both foreign and domestic investors.9 Since most deals include both domestic and foreign investors, these volumes are not necessarily good indicators of the level of cross-border investment flows into these deals but the variation in these flows is likely highly correlated with cross-border flows. 2.1 Data Set and Data Limitations The data set we use is derived solely from the Thomson VentureXpert database which has some important limitations that constrain our analysis. First, we limit all regressions to years greater than 1995, and most graphical analysis to 1992 and later in a few cases. While VentureXpert contains deals going back to the 1980’s, our conversations with VentureXpert staff indicate it was only in the mid 1990’s that a concerted effort was made to make the database a comprehensive source for international deals. Second, while Thomson did make significant investments in collecting international data in the 1990’s, we received some anecdotal reports that until more recently the dataset was still not as comprehensive for deals done outside the US without the participation of US VCs as it is for deals done with participation by US firms. The cited reason for this was that, despite investments in collecting non-US data, VentureXpert remained more closely connected (and more familiar) to US firms who report their deals more regularly. In light of this, we restrict most of our analysis of cross border deal flows to those which have one leg in the US, either flows of US VC participation in non-US-based deals, or flows of international VC participation in US-based deals. We also report some regressions using bilateral relationships between all source and destination countries, but warn the reader that the data behind them may be biased, mainly due to lower reporting of deals by non-US VC firms. We also restrict the sample to the set of countries which had more than 10 deals as destination over the full time period (1992-2007) to focus on countries with at least a minimum of VC activity. The set of countries this restriction implies is listed in the Appendix, available at NBER Working Paper No. 14344.

9 We take this approach partly do to the fact that we do not have information on the contribution of each investor to the deal; we only have data on the number and location of investors, and the total deal size.

6

2.2 Trends: Bubble, Crash, and Post-Bubble Figure 1 charts the total volume of US-based VC deals in the VentureXpert dataset over

time as well as the volume of those deals which feature international participation and the amount of US. One can see clearly the boom in the volume of venture capital money associated with the late 1990’s internet bubble which then crashes precipitously in 2001-2002. In Figure 2, one can see the rise in the proportion of deals in the US which are funded with some cross-border participation. This ratio goes from less than 15% in the early 1990’s to over 25% in 2001 and 2002, falling back to 15% only in 2007. This two figures document the more than 4x rise in the volume of US based VC deals, the almost identical rise in the volume of deals funded, at least in part, by foreigners, and in the volume of deals done by US VC firms overseas. The bubble was not just in US based deals but also drove US based firms to develop a higher volume of deals overseas and increased the percentage of US based deals with foreign participation, a trend which persisted long after the bubble burst. An increase in cross-border participation may have been a long term consequence of firms making investments in entering international markets at the height of the boom. 2.3 Differences in the Different Types of International and Domestic Deals

Figure 3 plots average deal sizes (in $millions) for deals done by US VC firms abroad (possibly with foreign VC participation), deals done by foreign VC firms in the US (possibly with US VC participation), and deals done in the US without foreign participation for comparison. We can see that deal sizes did increase during the bubble, then decrease again afterwards in all three series. Additionally, there seems to be a hierarchy where domestic only deals are the smallest, deals done by US firms abroad are the next biggest, and US based deals with foreign participation are the largest. This ordering is more or less stable for the years before, during, and after the bubble. In Figure 4 we can see that US based deals with some foreign VC participation also feature the most investors per deal (around 4 on average, though this measure rises to 5 around the peak of the bubble). The domestic only deals and the deals done by US VC firms in foreign countries tend to feature similar numbers of investors, starting around 2 in the early 1990s and moving up around 3 at the peak of the bubble and staying there.10

In Figure 5, we plot the percent deals which are in companies in high-tech sectors (including IT or biotech) vs. all other lower-tech sectors. The percent of US deals where foreigners participate is very high, nearly 90% or above in every year, and is always higher than the percent of high-tech deals in domestic only deals, which are just over 80% in most years (the percentages tend to rise slightly during the bubble peak and fall back again after). For deals done by US based VC firms going abroad, the percent of high-tech deals is at first much lower than the other two series, though the gap closes appreciably in 1998-2000 and does not open back up again after the crash in 2001. It was around this time that the IT sectors in many countries outside the US began to develop which may account for some of the persistence of the increase in high-tech deals.

A survey by Deloitte (2006) indicates that cross-border VC firms often partner with local firms to obtain inside information and greater savvy in navigating local conditions, while often providing needed liquidity to fund larger deals. Wright, Pruthi, and Lockett (2005) find that international investors often take consistently different positions and act on different investment criteria than local VC firms. Figures 3-5 confirm the idea that there are observable differences 10 The number of investors participating in a given deal is asked of any firm who reports data on the deal, thus even if, e.g., only 1 of 3 deal participants reports data on the deal to venture expert, the number of participants we count would be 3 rather than just 1 as would be the case if we counted deal entries only.

7

between foreign deals and their domestic counterparts including greater deal size for deals with foreign participation, more investors (at least for deals done in the US with foreign participation), and a greater percentage of low-tech deals for deals done outside the US by US firms. 2.4 Some Regional Patterns

Historically, VC was largely a US grown phenomena, which helped the takeoff of the US IT sector. The globalization of IT activities induced the US venture capital to mature, and to start exporting their unique skills as VC managers, often following IT growth elsewhere while continuing to dominate the VC of the US domestic market. This is reflected in the fast diffusion of VC from the US to other destinations. Figure 6 summarizes the flow of VC funds from the US to selected areas over time, comparing them with the volume of funds from domestic investors in these areas.11 We take the growth of the volume of funds from the domestic data as a rough indication of growth in the size and capacity of the local VC sector, though the VentureXpert data likely misses a large fraction of the true total domestic volume in many countries. Areas depicted are Europe (ex. UK), Asia (inc. Japan), Latin America, Central and Eastern Europe, India, and China. A diffusion process may entail learning -- in countries where the human capital and institutional and financial infrastructure is mature, the diffusion process leads to the solidifying of a VC takeoff, with growing involvement of domestic VC. This seems to be the case in Europe, where the domestic VC took off around the time of entry of US VC and remained high after US flows to Europe subsided. In countries with less mature infrastructure, the diffusion may entail the dominance of the US during the initial VC diffusion process (as apparently has been the case so far in China and, to a lesser extent, India). To the extent that local firms will learn and grow their capacity through international partnerships, the local market may develop as a result. Increases in the volume of purely domestic funded investments in India and China in the past few years indicate the slow emergence of a viable domestic VC base. In some cases, if the learning associated with the diffusion provides disappointing results, after a period of establishing a “beachhead” we find negative diffusion, as apparently has been the case in Latin America. The bottom two panels of Figure 6 summarize the patterns of VC in the US, tracing the volume of deals in the US funded exclusively by US domestic investors (solid line, bottom left panel) as well as estimates of the flow of VC funds from the Rest of the World (ROW) to the US market and from the US to the ROW, (bottom right panel). It confirms that most of the VC in the US is self financed, and that US exports of VC took off in the mid 1990s.12

In summary, the late 90’s bubble’s effect on deal volumes was to increase the total volume by roughly 4-5X in all three series (domestic only deals, domestic deals with some foreign VC funding, and foreign deals with some US funding). The percentage of deals in the US which featured any foreign participation increased dramatically from around 10% to over 25% as foreign investors were drawn in at the height of the bubble. However, this ratio was slow to recede to its pre-bubble level and was still over 15% in 2007. The bubble may have also driven US firms to go more abroad more aggressively, especially into Asia and Europe. This trend also had not completely reversed after the bubble. The number of partners and average deal size grow

11 We have data on total deal size and the location of each VC firm participating in the deal; we do not have information on the actual amount contributed by each VC firm. Thus the dollar volumes of these regional flows are rough estimates calculated by assuming each VC participant contributes an equal share. 12 The series depicted in the bottom right panel of Figure 6 differ from their counterparts in Figure 1 in that they are estimates of the total cross-border flow of investment funds rather than the volume of deals which featured cross-border participation

8

over the course of the bubble with both sets of series staying noticeably higher after the bubble crash.

Domestic-only deals appear to be smaller on average, with foreign funded US deals and US funded foreign deals successively larger (a pattern that holds up before, during, and after the bubble). US VC firms going abroad tend to have more investor partners (over 4 on average) compared to 2-3 for US based deals with or without foreign participation (we don’t know if the foreign funded deals done in the US feature more investor partners than deals done in the foreign market without international participation, due to very limited data on these deals). Finally, foreign funded deals in the US are the most likely to have been in the high-tech sector followed closely by US domestic only deals. Since these series are mostly over 80% they are not much impacted by the bubble; however, the foreign deals with some US funding, which started with less that 60% of deal volume occurring in high-tech deals caught up rapidly, increasing to near 80% where it remained after the crash. This final trend may be partly accounted for by the apparent increase in US participation in Europe and Asia after the crisis (see Figure 6), which experienced strong bio-tech and IT based growth respectively.

Figure 7 provides a measure of the “venture capital surplus” during 5-year windows, defined as the number of times a domestic VC firm participated in a deal outside the country, minus the number times a foreign firm participated in a deal within the country. It reveals the emerging global dominance of the US as net exporter of VC deals through the 1990s. The US was a net exporter of 14 deals during 1992-1997. This number is practically zero given that the gross cross-border deal flow into and out of the US was over 20,000 each way. The US surplus grew to just over 3500 during the IT boom years (1998-2002), leveling to about 3000 during 2003-7. While in the early 1990s there were four other countries with significant net export position (UK, Singapore, Hong Kong and Luxemburg), the US has now became the dominant VC net exporter, dwarfing all the other countries. The UK switched from being a sizable net exporter of VC deals (more than 250 during 1992-7) to the second largest net importer during 2003-2007 (more than 350 deals). France, Israel, Canada, India and China were consistent net importers of VC deals, with China emerging as the largest net importer of VC deals (approaching 800 during 2003-2007).

As a crude simplification, VC was a US grown phenomena, which helped the takeoff of the US IT sector. The globalization of IT activities induced the US venture capital to mature, and to start exporting their unique skills as VC managers, often following IT growth elsewhere while continuing to dominate the VC of the US domestic market. This is reflected in the fast diffusion of VC from the US to other destinations. Figure 8 summarizes the flow of VC funds from the US to selected areas over time, comparing them with the volume of funds from domestic investors in these areas. We take the growth of the volume of funds from the domestic base as a rough indication of growth in the size and capacity of the local VC sector. Areas depicted are Europe (ex. UK), Asia (inc. Japan), Latin America, Central and Eastern Europe, India, and China. A diffusion process may entail learning -- in countries where the human capital and institutional and financial infrastructure is mature, the diffusion process leads to the solidifying of a VC takeoff, with growing involvement of domestic VC. This seems to be the case in Europe, where the domestic VC took off around the time of entry of US VC and remained high after US flows to Europe subsided. In countries with less mature infrastructure, the diffusion may entail the dominance of the US during the initial VC diffusion process (as apparently has been the case so far in China and, to a lesser extent, India). To the extent that local firms will learn and grow their capacity through international partnerships, the local market may develop as a result. Increases in the volume of purely domestic funded investments in India

9

and China in the past few years indicate the slow emergence of a viable domestic VC base. In some cases, if the learning associated with the diffusion provides disappointing results, after a period of establishing a “beachhead” we find negative diffusion, as apparently has been the case in Latin America. The bottom two panels of Figure 8 summarize the patterns of VC in the US, tracing the volume of deals in the US funded exclusively by US domestic investors (solid line, bottom left panel) as well as the flow of VC funds from the Rest of the World (ROW) to the US market and from the US to the ROW, (bottom right panel). It confirms that most of the VC in the US is self financed, and that US exports of VC took off in the mid 1990s. These dollar volume numbers confirm the trend seen in cross-border US deal count from Figure 7.

3. Regression Analysis

In the previous section we have documented the increase in cross border venture capital flows and volumes, and the unique nature of cross-border deals (i.e. larger and involving more investors). The globalization of venture capital is clearly well underway and has potential to be an important source of funds for funding innovation and financial reengineering around the world. In this section of the paper we seek to measure what fundamental factors are associated with VC outflows, inflows, and better connections between countries (i.e. gravity variables).

3.1 Regression Specifications

We employ a basic gravity model structure to measure the correlates of cross border flows of venture capital funds. As with any standard gravity analysis, we include measures of the relationship between countries such as distance, whether or not they share a common language, and colonial legacy. In some specifications we also include variables which measure source and destination country characteristics; the choice of which was guided by intuition and a reading of the other literature, but considerably constrained by data availability.

We use negative binomial regression techniques to test two sets of gravity specifications.13 Both specifications feature counts of the deal flow between countries, and include the log of source and destination country GDP, as well as the log of the distance between the countries, as is standard in gravity models.

The first set of specifications features both source and destination country dummies and tests the role of variables that measure the relationship between countries such as distance, common language, and a legacy of colonial dependence between the two. This gravity model is captured by the following equation:

tjitjijitjitji Xy ,,,,,,, ')distanceln( (1)

Where tjiy ,, is the log of deal volume or, in the Poisson regressions, the count of the

number of deals from source country j to destination country i in year t. The vector X, are measures of the relation between the source and destination countries (such as the presence of a common language among their respective populations). Within this specification some regressions include variables which measure path dependence such as the lagged average of bilateral deal flows.

13 The discreet count nature of the data indicates a negative binomial or Poisson specification. The dispersion parameters for the negative binomial distribution are reported for each regression (labeled “alpha”) along with t-statistic. Since the data exhibits greater dispersion than would a standard Poisson distribution, we choose the negative binomial as a better model.

10

The second set of specifications drops the destination and/or source country dummies and includes country characteristics such as measures of the strength of institutions, financial development, technical infrastructure, market size and sophistication, and the availability of high-end human capital. This specification is designed to shed light on the question of what factors attract international venture capital investments and what factors are associated with source of VC funding.

tjitjijitjtji Xy ,,,,,,, ')distanceln( (2)

Where now, i has been dropped and tjiX ,, has been augmented by variables which measure

characteristics of the economy of i. Within these two gravity model specifications, we also restrict the sample in various ways to see how the parameters change and what factors gain or lose statistical significance.

Our main dependent variable is the count of venture capital deals involving participation of any firm from source country j in a deal done in source country i over the course of year t. Since deals may have backers from multiple countries a single deal may generate more than one count. For example, consider a deal done in Belgium with participation from two US based VC firms, one German based firm, and two Belgian firms. This deal would add one to the total from Germany to Belgium for that year and one to the total from the US to Belgium (since there is only one deal). In Table 3 we use the number of firms from country j participating in deals done in country i as the RHS variable (in which case, the previous example would count a one from Germany to Belgium and a two from the US to Belgium since two US firms participated.)

3.2 Data Set and Data Limitations

In the following section we explore our regression results. The data set we use is derived solely from the Thomson VentureXpert database which has some important limitations that constrain our analysis. For reasons outlined in Section 2 above, we rest First, we limit all regressions to years greater than 1995. While VentureXpert contains deals going back to the 1980’s, our conversations with VentureXpert staff indicate it was only in the mid 1990’s that a concerted effort was made to make the database a comprehensive source for international deals. Second, while Thomson did make significant investments in collecting international data in the 1990’s, we received some anecdotal reports that until more recently the dataset was still not as comprehensive for deals done outside the US without the participation of US VCs as it is for deals done with participation by US firms. The reason for this was hypothesized to be that, despite investments in collecting non-US data, VentureXpert remained more closely connected (and more familiar) to US firms who report their deals more regularly. In light of this, we restrict most of our analysis of cross border deal flows to those which have one leg in the US, either flows of US VC participation in non-US-based deals, or flows of international VC participation in US-based deals. We also report some regressions using all bilateral relationships, but warn the reader that the data behind them may be biased due to lower reporting of deals by non-US VC firms. We also restrict the sample to the set of countries which had more than 10 deals as destination over the full time period (1992-2007) to focus on countries with at least a minimum of VC activity. The set of countries this restriction implies is listed in the Appendix, available at NBER Working Paper No. 14344.

11

3.3 Results of Regression Analysis

In this section we explore our regression results from Tables 1-4.

Basic Gravity Specifications with Both Source and Destination Country Fixed Effects Table 1, shows the results of the gravity specification with both source and destination

country dummies and year dummies. There are three regressions reported, one limited to international deals with the US as the source of VC funds (labeled US Source), one limited to deals done in the US with international funding (labeled US Destination), and one with all bilateral relationships between all countries in the sample (labeled All Pairs). These samples - which share the common feature that at least one party to the deal is based in the US - are used in the subsequent regressions and tables and were chosen because of our belief that VentureXpert has best coverage of US based VC firms.

Conversations with practitioners indicate that venture capitalists factor in travel costs when making decisions about where to invest, especially in light of the intensive monitoring and advisory roles that these types of investments require. Often, VC’s will choose a few cities in which to operate, sourcing several deals in each city so as to economize on travel to that area. It is also likely that distance proxies for several other factors including limited local information, dissimilarity of institutions, and lack of social/business networks in distant countries [e.g. Sorenson and Stuart (2001) also document a very strong location bias in domestic US firms due to inter-firm networking effects]. These considerations indicate distance should (in general) be negatively related to deal flow, which is also what one would expect from other gravity literature [e.g. Portes and Rey (2005)].

In Table 1 we can see that deal flow is positively and significantly related to distance when the US is the source country (regression (1)). While contrary to the expectation that distance should deter VC flows in general, it is not as surprising when considering that many US deals going abroad are oriented toward Europe or Asia rather than nearer to home in Canada, Mexico, or Latin America. Distance is insignificant in the regression (2) with the US as the destination country and thus does not seem to significantly deter VC funding coming to the US, indicating that the US attracts deals from all over the world, regardless of location. This is in contrast with a strongly negative and significant relationship (what would be expected from other gravity literature) when all country pairs are included in the regression (3).

The indicator variables for a past colonial relationships and common languages also have the expected positive sign when all country pairs are included, but are again opposite expectations (negative) when the US is source or destination. Its not clear what would drive such a result, except that with the US as one end of all bilateral pairs, there may could be a variety of historical and economic factors that are correlated with these variables and which overwhelm the effects. For instance, many developing countries adopt English as a first or second official language and the result maybe biased by this fact. Table 2 presents regressions which control for more facets of the source and destination, such as financial development and infrastructure, and the parameters on the gravity variables change in these regressions.

Gravity Specifications with Source and Destination Country Characteristics Table 2 presents results from the second type of specification which drops the source

and/or destination country dummies and includes variables which measure specific country traits. In regression (3), with all source and destination countries, the original gravity variables have fairly similar coefficients than they do in the Table 1 regressions, which have country level fixed effects rather than country traits. In regressions (1) and (2) with US as source or destination

12

country, the new parameters are very different from those in Table 1. The change in the parameter values may be in part due to the new specifications but also due the fact that the inclusion of the source/destination country characteristics necessitates dropping observations with missing values in these new variables, and thus changes the sample of observations. In these regressions with US as source (1) or destination (2), common language now has the expected positive and significant sign, and distance is negative and significant in the US-as-source regression (1). Colonial ties are still negatively and significantly related to deal flow in regression (1) but the parameter value is much smaller.

Turning to the variables representing country traits, we can see broadly that the parameters on source country traits are fairly similar between the regression (2), with US as destination, and (3) with all bilateral pairs of sources and destinations, where as the pattern of significant coefficients are relatively different between regression (1) with US as source, and regression (3). Much of the world’s cross-border VC deal flow comes from a few countries developed countries and the deals going to the US are sourced from the same ones as the deals going to other countries. Haemmig (2003) makes the case that the reasons for engaging in cross border investing by firms from outside the US VC market are to achieve scale and seek deals outside of limited local markets, whereas the impetus of many US based firms was to reach new consumer markets with growth potential and find new technology hotspots.

Regressions (2) and (3) both depict VC exporters as having strong institutions (in terms of investment freedoms and freedom from corruption). This result is in concordance with research that finds institutions a critical determinant of the ability of VCs to operate [see, e.g. Jeng and Wells (2000), Gompers and Lerner (2001), and Guler and Guillen (2005)]. VC flows also seem to originate in countries with well developed stock markets, widespread computer deployment (per 1000 people), and significant military expenditure to GDP.

The pattern of significant coefficients is relatively different between regression (1) with US as source, and regression (3). In regression (1), measures of institutions are insignificant, and stock market size relative to GDP is significant at 10%, measures of technology deployment (which likely correlate with infrastructure development) are significant at 5% and the amount of computers and communication services in exports strongly significant at 1%. VC firms seem to be targeting countries which some level of technological infrastructure development, and which export computers and tech-related services, without much regard to the local institutions. Other authors [see e.g. Black and Gilson (1998)] have confirmed the value of stock markets to VC investors as a main exit mechanism. The measure of market capitalization to GDP of the destination country (where one might assume a VC backed firm would list for an IPO exit) is not significant in either regression (1) or (3), but is highly significant as a source country characteristic in regressions (2) and (3). It appears international investor firms develop in markets with IPO potential but do not necessarily seek deals in countries with developed stock markets. Conversations with practitioners indicate that VC firms are fairly adaptable, and can develop new exit strategies (such as selling to local private equity investors) or taking firms public on regional exchanges outside the country where the deal originates.

The literature on financial development has compared stock market based financial systems to bank based financial systems and found them to be quite different [see Levine and Zervos (1998)]. In our regressions, the size of the domestic banking sector in the source or destination country is not significantly related to VC deal flows.

In regression (3) investment freedom is negatively and significantly associated with being a VC destination, but freedom from corruption is positively and significantly associated. VC firms may see a sweet spot where local restrictions on investment activity have hampered the

13

development of local funding sources but where corruption is not such a problem that a start-up cannot succeed.

High-end human capital in the form of university graduates (who might become engineers, or entrepreneurs) are positively and significantly associated with being a destination for deals in both regressions (1) and (3).

The results regarding the importance of destination country human capital, technological orientation of services, and the importance of technological infrastructure, are in line with other research which finds that the entrepreneurial environment is a critical factor for the development of the VC industry [see, e.g. Romaine and Van Pottelsberghe (2004) and Da Rin, Nicodano, and Sembenelli (2005)].

Military expenditure to GDP is strongly associated with being a VC destination, as it was with being a VC source country. In countries such as Taiwan, Israel, and the US, military spending on R&D is often directed to VC backed startups which may generate a demand for international VC funds as well as foster a local pool of VC firms which eventually go abroad.

Table 3 presents results from a similar set of specifications where the LHS variable is now the number of unique firms which have done deals between country i and j in year t. Conversations with practitioners indicated that [justify why to do this specification…]. The results in Table 3 are nearly identical to Table 2, as the number of deals done and the number of firms transacting between countries are very highly correlated.

Gravity Specifications with Time Lags and Aggregate Flows A study by Guler and Guillen (2005) finds that VC firms tend to replicate the foreign

entry patterns of their syndicate partners. Research by Sorensen and Stuart (2005) confirms the importance of networking in the geographic location of VCs. Table 4 shows modified gravity specifications with source and destination country dummies (as in Table 1) which contain moving averages of the deal counts between the two countries over the previous two years (t-1 and t-2). To asses whether the capital from one source country is complement or substitute to VC capital from other sources, or whether activity in one destination is complement or substitute to activity by the same source country in other destinations, we also include the contemporaneous sum of all deals from the source country to all other destination countries and to the destination country from all other sources. Critically, we include both source and destination country GDP (in logs) and source/destination fixed effects to control for the size and growth rates of the two economies in generating aggregate flows and lagged flows.

These regressions reveal some interesting patterns. First, there does seem to be persistence in venture capital and private equity. In regression (1) – with US as source – and in regression (3) – with all bilateral pair relationships – lagged values of the LHS variable between source and destination are significant and positive, indicating higher flows in year t-1 and t-2 predict higher flows in year t. We believe this is evidence for a path-dependency in VC flows. One candidate mechanism for this type of effect are the fixed costs to establishing a “beach-head” in a foreign country such as travel, learning the local market, establishing relationships with local partners, hiring staff familiar with the foreign market, and possibly setting up satellite offices. A related possibility is that the diffusion of information amongst source country firms engenders a “follow the leader” type of dynamic. In regression (2), where the US is destination, the lag variable is not significant.

In regression (4), where the US is the source, the number of deals done by US firms in a given destination country i in a given year is negatively correlated with the total number of international deals done by US firms in other countries, and negatively but insignificantly related

14

to the total number of deals bound for that country from non-US sources. Both of these variables are negative (and this time both significant) in regression (5), where the US is the destination country. Neither variable is significant in regression (6) where all bilateral relationships are included in the regression. The evidence appears to indicate that, if anything, deals from a given source country j to destination country i are substitutes to deals from another source countries and to increased investing by country j in other destination countries.

4. Conclusion

A crude simplification of the history of venture capital is that, in its origin, it was largely a US grown phenomena. It helped in facilitating the remarkable takeoff of the IT revolution, and has had the US as it's center. At this point, the venture capital industry has internationalized driven by a variety of forces.

Countries which are more often VC deal destinations do not necessarily feature better institutional environments but have more developed stock markets which may facilitate IPO exits and have more widespread computer and cell phone deployment which may indicate better infrastructure where high-tech firms can flourish.

We find that the countries which have more often been the sources of international deals have better institutional environments, more developed financial markets, high computer penetration, and more university graduates than countries which have lagged in generating deals. These features likely create a conducive environment for financial companies to form and raise capital thus facilitating the supply side of this market.

We also find a positive relationship between military expenditures and VC deals, which is robust even when the US is the destination, implying that US military expenditures are not driving this result. This last result is consistent with the conjectures that public sector involvement stimulating the demand for IT services, and policies aimed at the formation of proper human capital may have a positive role in driving the development of the VC and IT sectors. Studying and testing these conjectures is left for future research.

15

References

Aizenman J. and N. P. Marion (2004), “The merits of horizontal versus vertical FDI in the presence of uncertainty,” Journal of International Economics, 62, pp 125-148.

Anderson E. J. and E Van Wincoop (2004), “Trade Costs,” Journal of Economic Literature, 42, 3, pp. 691-751.

Blonigen, B. A. (2005), “A Review of the Empirical Literature on FDI Determinants,” NBER Working paper # 11299.

Black B. S. and R. J. Gillson (1998), "Venture Capital and the Structure of Capital Markets: Banks versus Stock Markets," Journal of Financial Economics, Vol. 47, pp. 243-277.

Da Rin, N., G. Nicodano and A. Sembenelli (2005), “Public Policy and the Creation of Active Venture Capital Markets,” European Central Bank Working Paper Series, No. 430.

Deloitte (2006), “Global Trends in Venture Capital, 2006 Survey” sponsored by Deloitte & Touche LLP.

Dow Jones (2009), “Cisco Uses Venture Capital To Push Into South Korea” 09/03/2009 Dow Jones VentureWire.

Gompers P. and J. Lerner (2001), “The Venture Capital Revolution,” The Journal of Economic Perspectives, 15, 2, pp. 145-168.

Guler I. and M. F. Guillen (2005), “Knowledge, Institutions, and Foreign Entry: The Internationalization of US Venture Capital Firms”, working paper draft.

Haemmig M. (2003), The Globalization of Venture Capital –Haupt Verlag, Berne, 2003. Jeng L. A. and P. C. Wells (2000), “The Determinants of Venture Capital Funding: Evidence

Across Countries” working paper. Kortum S. and J. Lerner (2000), “Assessing the Contribution of Venture Capital to Innovation”

The RAND Journal of Economics, 31, 4, pp. 674-692. Lerner, J., (2001), Venture Capital, Technological Innovation, and Growth (Boston: Harvard

Business School). Levine, R. and Zervos, S. (1998), “Stock Markets, Banks, and Economic Growth”, American

Economic Review, Vol. 88, No. 3. Melitz M. J. (2003), “The Impact of Trade on Intra-Industry Reallocations and Aggregate

Industry Productivity,” Econometrica, 71, 6, pp. 1695-1725. Portes R. and H. Rey (2005), “The Determinants of Cross-Border Equity Flows,” Journal of

International Economics, 65, pp 269-296. Price Waterhouse Coopers (2006), “Global Private Equity Report”, report by Price Waterhouse

Coopers, LLC. Romain A. and B. Van Pottelsberghe (2004), “The Determinants of Venture Capital: A Panel

Analysis in 16 OECD Countries”, working paper draft. Rose A. K. and M. M. Spiegel (2004) “A Gravity Model of Sovereign Lending: Trade, Default,

and Credit,” IMF Staff Papers, 51, pp. 50-63. Silva J. M. C. S. and S. Tenreyro (2006), “The Log of Gravity,” The Review of Economics and

Statistics, 641-658. Sorenson O. and T. Stuart (2001), “Syndication Networks and the Spatial Distribution of Venture

Capital Investments,” American Journal of Sociology, Volume 106 Number 6. Wilson, R. (1983), The New Ventures – Inside the High Stakes World of Venture Capital,

Addison Wesley Publishing Company. Wright, P. W. Pruthi and A. Lockett (2005), “Do Foreign and Domestic Venture Capital Firms

Differ in Their Monitoring of Investees?” Asia Pacific Journal of Management, 20, pp. 175-204.

16

Tables and Figures

Table 1: Negative binomial regressions. LHS is yearly count of VC deals where funding came from source country j to destination country i. Country dummies for source and destination countries and year dummies not reported. Robust standard errors, significance levels are * <10%, ** < 5%, *** <1%.

LHS = Count VC Deals US Source US

Destination All Bilateral

Pairs

(1) (2) (3) Ln(GDP) destination 1.74*** 1.54*** (0.3) (0.2) Ln(GDP) source 0.64** 0.89*** (0.3) (0.2) ln(distance) 1.69** -0.33 -1.11*** (0.8) (0.4) (0.05) Colonial ties -3.59*** -12.1 0.33*** (1.4) (8.0) (0.1) Common language -57.2*** -12.5* 0.68*** (9.7) (6.5) (0.07) Observations 555 733 26302 Source country FE N Y Y Destination country FE Y N Y alpha -1.16*** -2.94*** -0.046 (0.1) (0.3) (0.08)

Notes: Robust standard errors in parentheses: *** p<0.01, ** p<0.05, * p<0.1; year fixed effects included in all regressions. Alpha is negative binomial distribution over-dispersion parameter (alpha = 0 is Poisson distribution). No constant included in regressions.

17

Table 2: Negative binomial regressions. LHS is yearly count of VC deals where funding came from source country j to destination country i. Dummies for years included but not

reported. Robust standard errors, significance levels are * <10%, ** < 5%, *** <1%.

LHS = Count VC Deals US Source US

Destination All ij Pairs

(1) (2) (3) Ln(GDP) destination 0.69*** 0.42*** (0.04) (0.06) Ln(GDP) source 0.89*** 0.73*** (0.09) (0.07) ln(distance) -0.30*** -0.12 -0.96*** (0.07) (0.1) (0.05) Colonial ties -1.11*** -0.039 0.45*** (0.4) (0.3) (0.2) Common language 0.47*** 1.31*** 1.35*** (0.09) (0.2) (0.2)

Destination country variables

Investment Freedom Index -0.0011 -0.0093** (0.003) (0.004) Freedom from Corruption Index 0.0034 0.025*** (0.004) (0.005) Domestic credit/GDP -0.00098 -0.00081 (0.001) (0.002) Market cap. (% of GDP) 0.0018* 0.00035 (0.0009) (0.0008) Cellphone subscribers/cap. 0.0073** -0.0046 (0.003) (0.003) Computers (per 100 people) 0.011** -0.0079* (0.004) (0.004) Computer & comm. Exports 0.019*** 0.0017 (0.003) (0.005) Military expenditure/GDP 0.19*** 0.26*** (0.03) (0.03) University Students/pop 0.000071** 0.00010** (0.00004) (0.00005)

(continued on next page)

18

(Table 2, continued from previous page)

Source country variables

Investment Freedom Index 0.015*** 0.015*** (0.005) (0.004) Freedom from Corruption Index 0.014*** 0.017*** (0.005) (0.005) Domestic credit/GDP -0.00026 0.000045 (0.001) (0.001) Market cap. (% of GDP) 0.0038*** 0.0054*** (0.001) (0.0007) Cellphone subscribers/cap. 0.0052 -0.0052 (0.005) (0.003) Computers (per 100 people) 0.032*** 0.014*** (0.006) (0.004) Computer & comm. Exports 0.011*** -0.0011 (0.004) (0.003) Military expenditure/GDP 0.14*** 0.089** (0.05) (0.04) University Students/pop 0.000087** 0.0000043 (0.00004) (0.00004) Constant -16.0*** -25.9*** -30.6*** (1.4) (2.2) (3.7) Observations 417 495 13823 alpha -0.87*** -0.18 0.58** (0.1) (0.2) (0.2)

Notes: Robust standard errors in parentheses: *** p<0.01, ** p<0.05, * p<0.1; year fixed effects included in all regressions. Alpha is negative binomial distribution overdispersion parameter (alpha = 0 is Poisson distribution).

19

Table 3: Negative binomial regressions. LHS is yearly count of active VC firms from source country j funding a deal(s) in destination country i. Dummies for years included but not

reported. Robust standard errors, significance levels are * <10%, ** < 5%, *** <1%.

LHS = Count, Active VC Firms

US Source US

Destination All ij Pairs

(1) (2) (3) Ln(GDP) destination 0.72*** 0.42*** (0.05) (0.06) Ln(GDP) source 0.91*** 0.73*** (0.09) (0.07) ln(distance) -0.39*** -0.099 -0.98*** (0.07) (0.1) (0.05) Colonial ties -1.25*** 0.032 0.36** (0.4) (0.3) (0.2) Common language 0.46*** 1.48*** 1.45*** (0.1) (0.2) (0.2)

Destination country variables

Investment Freedom Index 0.0035 -0.0090** (0.003) (0.004) Freedom from Corruption Index 0.0024 0.024*** (0.004) (0.005) Domestic credit/GDP -0.00056 -0.0011 (0.001) (0.002) Market cap. (% of GDP) 0.0024** 0.00041 (0.001) (0.0008) Cellphone subscribers/cap. 0.0059* -0.0039 (0.004) (0.003) Computers (per 100 people) 0.011** -0.0085* (0.004) (0.004) Computer & comm. Exports 0.017*** 0.0018 (0.003) (0.005) Military expenditure/GDP 0.21*** 0.26*** (0.03) (0.03) University Students/pop 0.000096** 0.00011** (0.00004) (0.00005)

(continued on next page)

20

(Table 3, continued from previous page)

Source country variables

Investment Freedom Index 0.013** 0.015*** (0.005) (0.004) Freedom from Corruption Index 0.013** 0.015*** (0.005) (0.005) Domestic credit/GDP -0.00036 -0.000052 (0.002) (0.001) Market cap. (% of GDP) 0.0032** 0.0055*** (0.001) (0.0007) Cellphone subscribers/cap. 0.0081 -0.0038 (0.005) (0.003) Computers (per 100 people) 0.033*** 0.015*** (0.007) (0.004) Computer & comm. Exports 0.011** -0.00050 (0.004) (0.003) Military expenditure/GDP 0.15*** 0.093** (0.05) (0.04) University Students/pop 0.000081* 0.0000063 (0.00004) (0.00004) Constant -15.7*** -26.4*** -30.5*** (1.5) (2.3) (3.7) Observations 417 495 13823 alpha -0.63*** -0.030 0.72*** (0.10) (0.2) (0.2)

Notes: Robust standard errors in parentheses: *** p<0.01, ** p<0.05, * p<0.1; year fixed effects included in all regressions. Alpha is negative binomial distribution overdispersion parameter (alpha = 0 is Poisson distribution).

21

Table 4: Negative binomial regressions. LHS is yearly count of VC deals where funding came from source country j to destination country i. Regressions (1)-(3) feature lags of some deal flows; all lags are 2 year moving average lags (t-1 & t-2). Country dummies for source and destination countries and year dummies not reported. Robust standard errors, significance levels are * <10%, ** < 5%, *** <1%.

LHS = Count VC Deals US Source US

Destination All Bilateral

Pairs US Source

US Destination

All Bilateral Pairs

(1) (2) (3) (4) (5) (6) Ln(GDP) destination 1.67*** 1.49*** 1.34*** 1.52*** (0.3) (0.2) (0.2) (0.2) Ln(GDP) source 0.67** 0.81*** 0.39 0.93*** (0.3) (0.2) (0.3) (0.2) ln(distance) 1.82** -0.33 -1.08*** 1.94** -0.28 -1.11*** (0.9) (0.4) (0.04) (0.8) (0.4) (0.05) Colonial ties -3.43** -17.7** 0.32** -2.59** 1.23 0.33*** (1.4) (7.9) (0.1) (1.2) (0.9) (0.1) Common language -56.8*** -12.7* 0.65*** 2.26 -0.34 0.68*** (10) (7.6) (0.07) (1.7) (0.9) (0.07) MA Lag of VC deals, j->i 0.0020*** -0.0011 0.0049*** (0.0007) (0.0008) (0.001) Σ VC deals j->all (not i) -0.013*** -0.0040*** -0.000075 (0.002) (0.001) (0.0001) Σ VC deals all (not j) ->i -0.0019 -0.0079*** 0.000065 (0.003) (0.002) (0.0002) Observations 555 733 26302 555 733 26302 Source country FE N Y Y N Y Y Destination country FE Y N Y Y N Y alpha -1.18*** -2.91*** -0.058 -1.67*** -3.54*** -0.045 (0.1) (0.3) (0.08) (0.2) (0.6) (0.08)

Notes: Robust standard errors in parentheses: *** p<0.01, ** p<0.05, * p<0.1; year fixed effects included in all regressions. Alpha is negative binomial distribution over-dispersion parameter (alpha = 0 is Poisson distribution). No constant is included in the regressions.

22

Figure 1: Total volume of venture capital deals into and out of US from VetureXpert database. “Foreign w/ Some US Funding” is the dollar value of deals located outside the US that featured any cross-border participation by US-based VC firms. “US w/ Some Foreign Funding” is the dollar value of deals located inside the US that featured any cross-border

participation by foreign-based VC firms.

$0 Mn

$5,000 Mn

$10,000 Mn

$15,000 Mn

$20,000 Mn

$25,000 Mn

1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007

$0 Bn

$10 Bn

$20 Bn

$30 Bn

$40 Bn

$50 Bn

$60 Bn

$70 Bn

$80 Bn

$90 Bn

$100 Bn

Foreign w/ US Funding (left) US w/ Foreign Funding (left) All US Based Deals (right)

23

Figure 2: Total volume of domestic venture capital deals in US from VetureXpert database (thick dashed line) and percent of those which featured any cross-border participation by

foreign-based VC firms (thin line).

$0 Bn

$20 Bn

$40 Bn

$60 Bn

$80 Bn

$100 Bn

1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007

0%

5%

10%

15%

20%

25%

30%

All US Based Deals (left)

US w/ For. Funding (left)% All US Deals w/ For. Funding (right)

24

Figure 3: Average sizes of deals done in the US with no foreign funding (dashed line), deals with a foreign target company and at least one US based venture investor (thin solid), deals with a US based target company and at least one foreign based venture investor (thick solid).

$0 Mn

$5 Mn

$10 Mn

$15 Mn

$20 Mn

$25 Mn

1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007

US w only US Funding Foreign with US Funding US with Foreign Funding

25

Figure 4: Average number of investors per deal in deals done in the US with no foreign funding (dashed), deals with a foreign target company and at least one US based venture investor (thin solid), deals with a US based target company and at least one foreign based venture investor (thick solid).

1.00

2.00

3.00

4.00

5.00

6.00

1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007

US w only US Funding Foreign with US Funding US with Foreign Funding

26

Figure 5: Percent deals in biotechnology and IT (vs. non-high tech deals) of deals done in the US with no foreign funding (dashed), deals with a foreign target company and at least one US based venture investor (thin solid), deals with a US based target company and at least one foreign based venture investor (thick solid).

40%

60%

80%

100%

1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007

US w only US Funding Foreign with US Funding US with Foreign Funding

27

Figure 6: Graphs of the flow of VC funds from the US to selected areas over time as the dashed line, compared with the volume of funds from local, domestic investors as the solid line (data on local funds in VentureXpert data base are not likely comprehensive but are shown here to illustrate trends). Areas are Europe (ex. UK), Asia (inc. Japan, India, and China), Latin America, Central and Eastern Europe, India, and China.

Europe (ex. UK)

0

0.5

1

1.5

2

2.5

3

3.5

4

4.5

5

1990 1992 1994 1996 1998 2000 2002 2004 2006

$US

Bn

VC ($Bn) Euro->Euro VC ($Bn), US->Euro

Asia

0

0.5

1

1.5

2

2.5

3

1990 1992 1994 1996 1998 2000 2002 2004 2006

$US

Bn

VC ($Bn) Asia->Asia VC ($Bn), US->Asia

Latin America

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007

$U

S B

n

VC ($Bn) LatAm->LatAm VC ($Bn), US->LatAm

Central and Eastern Europe

0

0.05

0.1

0.15

0.2

0.25

0.3

1993 1995 1997 1999 2001 2003 2005 2007

$US

Bn

VC ($Bn) CEE->CEE VC ($Bn), US->CEE

India

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007

$U

S B

n

VC ($Bn) IND->IND VC ($Bn), US->IND

China

0

0.2

0.4

0.6

0.8

1

1.2

1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007

$U

S B

n

VC ($Bn) CHN->CHN VC ($Bn), US->CHN

USA (domestic )

0

10

20

30

40

50

60

70

80

90

1990 1992 1994 1996 1998 2000 2002 2004 2006

$US

Bn

VC ($Bn) USA->USA

USA (Inflows and Outflows)

0

2

4

6

8

10

12

14

16

1990 1992 1994 1996 1998 2000 2002 2004 2006

$US

Bn

VC ($Bn), ROW->USA VC ($Bn), USA->ROW

28

Figure 7: Venture capital deal “surplus”. Bar height indicates the number of times the country's local VC firms participated in a deal outside the country during the 5-year window, minus the number of times foreign firms participated in a local deal. The sample of countries was restricted to only those countries which were observed in all three time periods so the sums will not necessarily equal zero. The difference of the sum from zero is less than 10% of total in all three cases. The USA is essentially zero in the first period given that the gross number of cross-border deals is >40,000.

1992-1997

-100

-50

0

50

100

150

200

250

300

GBR

SGP

HKG

LUX

USA

DEU

BEL

CHE

KOR

FIN

GRC

NOR

RUS

TW

CMR

CZE

DNK

HUN

LVA

PRT

NLD

ZAF

ITA

SW

AUT

POL

IRL

ESP

JPN

NZL

PHL

THA

CAN

AUS

MYS

BRA

IND

ISR

FRA

CHN

Num

ber of Dea

ls

1998-2002

-1500

-1000

-500

0

500

1000

1500

2000

2500

3000

3500

4000

USA

TW

N

JPN

HKG

CHE

SGP

NLD

GRC

BEL

PHL

LVA

CMR

RUS

THA

POL

ZAF

PRT

CZE

GBR

LUX

MYS

NZL

HUN

AUT

DNK

NOR

DEU

FIN

AUS

ITA

ISR

BRA

ESP

IND

SW

E

IRL

CHN

KOR

CAN

FRA

Num

ber of Dea

ls

2003-2007

-1000

-500

0

500

1000

1500

2000

2500

3000

3500

USA

JPN

HKG

SGP

TW

N

CHE

NOR

LUX

NZL

GRC

PHL

MYS

CZE

PRT

CMR

AUS

NLD

LVA

ZAF

POL

BRA

THA

HUN

RUS

ITA

BEL

AUT

KOR

ISR

ESP

DEU

DNK

FIN IRL

IND

FRA

CAN

GBR

SW

E

CHN

Num

ber of Dea

ls

29

Figure 8: Graphs of the flow of VC funds from the US to selected areas over time as the dashed line, compared with the volume of funds from local, domestic investors as the solid line. Areas are Europe (ex. UK), Asia (inc. Japan, India, and China), Latin America, Central and Eastern Europe, India, and China.

Europe (ex. UK)

0

0.5

1

1.5

2

2.5

3

3.5

4

4.5

5

1990 1992 1994 1996 1998 2000 2002 2004 2006

$U

S B

n

VC ($Bn) Euro->Euro VC ($Bn), US->Euro

Asia

0

0.5

1

1.5

2

2.5

3

1990 1992 1994 1996 1998 2000 2002 2004 2006

$U

S B

n

VC ($Bn) Asia->Asia VC ($Bn), US->Asia

Latin America

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007

$US

Bn

VC ($Bn) LatAm->LatAm VC ($Bn), US->LatAm

Central and Eastern Europe

0

0.05

0.1

0.15

0.2

0.25

0.3

1993 1995 1997 1999 2001 2003 2005 2007

$US

Bn

VC ($Bn) CEE->CEE VC ($Bn), US->CEE

India

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007

$U

S B

n

VC ($Bn) IND->IND VC ($Bn), US->IND

China

0

0.2

0.4

0.6

0.8

1

1.2

1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007

$U

S B

n

VC ($Bn) CHN->CHN VC ($Bn), US->CHN

USA (domestic )

0

10

20

30

40

50

60

70

80

90

1990 1992 1994 1996 1998 2000 2002 2004 2006

$US

Bn

VC ($Bn) USA->USA

USA (Inflows and Outflows)

0

2

4

6

8

10

12

14

16

1990 1992 1994 1996 1998 2000 2002 2004 2006

$US

Bn

VC ($Bn), ROW->USA VC ($Bn), USA->ROW