International Review of Financial Analysis 31 (2014) 13–24

Contents lists available at ScienceDirect

International Review of Financial Analysis

Does cross-border syndication affect venture capital risk and return?☆

Susanne Espenlaub a, Arif Khurshed a, Abdulkadir Mohamed b,⁎a Manchester Business School, University of Manchester, Manchester, UKb School of Management, University of Liverpool, Liverpool, UK

☆ We are grateful to the editor, Brian Lucey and an anothat significantly improved the paper. We also thankMichael Brennan, Murray Dalziel, Olan Henry, Gary CoPawel Bilinski, Brahim Saadouni and Axel Buchner forsuggestions. The authors also wish to thank their colleSchool, Liverpool University and the participants of the Fin(Denver 2011) for their helpful comments on earlier drafts⁎ Corresponding author. Fax: +44 151 794 2000.

E-mail addresses: Susanne.Espenlaub@mbs.ac.uk (S. EArif.Khurshed@mbs.ac.uk (A. Khurshed), Abdulkadir.Moh

1057-5219/$ – see front matter © 2013 Elsevier Inc. All rihttp://dx.doi.org/10.1016/j.irfa.2013.10.003

a b s t r a c t

a r t i c l e i n f o

Article history:Received 19 March 2012Received in revised form 10 October 2013Accepted 12 October 2013Available online 21 October 2013

JEL classification:G24G32

Keywords:Venture capitalCross borderRisk and returnSyndication

Venture capital (VC) cross-border syndication has increased significantly in recent years. This study examines therisk and returns of investments of US–European cross-border syndicates in US portfolio companies. We use alarge sample of investments across four financing stages, and highlight several noteworthy differences betweencross-border syndicates and previous US-only evidence. By comparison, US–European syndicates are larger thanUS-only syndicates, involve younger VCs, and focus more on later financing stages. Controlling for sampleselection bias caused by the endogenous choices of exit route and exit timing, we examine the risk and returnsof investments backed by cross-border syndicates. Consistent with evidence from US-only syndicates, alphaand beta decrease monotonically from the earliest (start-up) stage to the later stages of financing.

© 2013 Elsevier Inc. All rights reserved.

1. Introduction

In recent years, cross-border venture capital (VC) investmentshave increased substantially in terms of the amounts of capital,numbers of deals and range of industries involved (Guler & Guillen,2010). Typically, in an international VC syndicate, local investors teamup with foreign investors in order to invest in a local entrepreneurialcompany (Meuleman & Wright, 2011). Cross-border syndication mayprovide access to more complementary skills and capabilities thandomestic syndicates. VC firms may play an important role in the inter-nationalization of their portfolio companies through their respectivehome-country product and capital markets (Tykvova & Schertler,2011). During the period from 2000 to 2008, more than one third ofportfolio companies globally received financing from foreign VC firms(Tykvova & Schertler, 2011). Previous studies of VC syndication focuson motives for syndication, and document that VC firms syndicate inorder to spread the investment risk, including risks due to the illiquidity

nymous referee for suggestionsRanko Jelic, Norman Strong,

ok, Martin Lozanno, Ning Gao,their valuable comments andagues at Manchester Businessancial Management Associationof this paper.

spenlaub),amed@liv.ac.uk (A. Mohamed).

ghts reserved.

of the VC investment (Lerner, 1994). Amit (2002) concludes thatsyndication can add value to the investment, based on his findingthat the more VC firms participate in the financing, the greater arethe overall benefits to backers and entrepreneurs. Recent studies oncross-border syndicates outside the US investigate why VC firms rely oncross-border syndicates when they internationalize their investments,and emphasize the importance of the legal and institutional frameworks(Lu & Huang, 2010; Tykvova & Schertler, 2011). Wright, Pruthi, andLockett (2005) examineAsianVCmarkets and point out that investmentssyndicated between local and foreign VC firms require further study.

Dimov and Milanov (2009) find that more than 73% of VC invest-ments in the US are syndicated. To date, little is known about theprevalence of cross-border syndicates between US VC firms and foreignVC providers. Also, to our knowledge, there are no prior studies of theperformance of VC investments in US portfolio companies financed bysuch cross-border syndicates. Prior studies of syndication, both domesticand cross-border, examine portfolio companies located outside theUS. Bycontrast, this study analyzes VC investments in US portfolio companiesmade by US VCs syndicating with VCs from outside the US (specificallyEurope).

Measuring the risk and returns of VC investments is challenging.Valuations of VC investments can only be observed when a portfoliocompany goes public or is acquired. Those companies within a VCportfolio that manage to go public or are acquired are likely to be aselect subsample of comparatively good performers. In addition, thetiming of exits, and the resulting return observations, are likely tovary systematically depending on the characteristics of the investmentand the portfolio company, and on the nature and success of the VC

Table 1Previous studies of venture capital risk and returns. The table summarizes the findings of the previous studies on venture capital investments. Only the studies by Cochrane (2005) andKorteweg and Sorensen (2010) evaluate the risk and returns of investment rounds.

Authors Country of venturecapital firm

Sample and period Methodology Returns (Ri) STD (σ) Systematicrisk (β)

Seppä and Laamanen (2001) US VC fund level (1998–1999) Binomial/OLS model NR NR NRChen, Baierl, and Kaplan (2002) US VC fund level (1999) Method of moments

(repeated sale approach)45% all stages 115.6% NR

Ljungqvist and Richardson (2003) US VC fund level (1981–1993) Proportional hazard model 19.81% all stages 22.29% 1.09 all stagesWoodward (2004) US Company data (1990–2003) Method of moments

(building an index)NR 50% 2.0 all stages

Cochrane (2005) US Company data (1987–2000) Selection bias (maximum likelihood) 71% Start-up65% Early stage60% Expansion50% Later stage

96%98%98%99%

1.1 Start up0.9 Early stage0.7 Expansion0.5 Later stage

Hege et al. (2008) US & Europe Company data (1997–2003) OLS 62% (US)106% (EU)

NR NR

Korteweg and Sorensen (2010)ab US Company data (1985–2005) Selection bias 2246% Seed60.10% Early10.9% Late85.41% Mezz

117%134%148%135%

0.74142.7422.6285.888

NR: not reported.a The returns are annualized and based on Table 8.b The standard deviations are annualized and based on Table 5.

14 S. Espenlaub et al. / International Review of Financial Analysis 31 (2014) 13–24

investments. For instance, Gompers and Lerner (2000) highlight thetendency of VCs to delay the liquidations of poor investments andthe consequent realization and reporting of negative returns. Theresulting sample selection biases with regard to observed exit routesand timing is a fundamental problem in evaluating the risk and returnsof VC investments, and our study seeks to address and control for thesebiases.

We study the risk and returns of syndicated financing roundsbetween US and European VC firms in US portfolio companies.1 Wefocus on portfolio companies that received VC financing from two ormore VC firms, where at least one VC firm is European and at leastone is based in the US. We chose Europe because it is the second largestVC industry after the US. US–European cross-border syndicates are aninteresting case whose study minimizes the impact of differences inlegal and institutional settings on cross-border risk and returns. Thisfacilitates the comparison of our results with those of prior studies ofthe risk-return trade-off of VC investments by US-only VCs and VCsyndicates (Cochrane, 2005; Korteweg & Sorensen, 2010).

We document some interesting similarities and differences betweenUS VCs and US–European cross-border syndicates, in terms of the invest-ment behavior and performance. Comparable to studies of US VCs, wefind that US–European cross-border syndicates face a monotonicallydecreasing relationship between both abnormal performance (alpha)and systematic risk (beta), on the one hand, and the stage of VCfinancing,on the other hand. That is, VC investments in earlierfinancing stages havehigher abnormal returns and higher systematic risk than correspondinginvestments in later stages. We document significant evidence of sampleselection bias. Nevertheless, our findings on the risk and returns of VCinvestments are remarkably robust across alternative methods of dealingwith selection bias.

The rest of the paper is organized as follows: Section 2 reviews theliterature on the returns of VC investments; Section 3 describes themotivation, data sources and the methodology employed; Section 4discusses the empirical findings; Section 5 presents the conclusion.

1 There are a handful of studies of VC returns that analyze the returns at the fund levelrather than the returns of each investment round (Chen et al., 2002; Hege et al., 2008;Ljungqvist & Richardson, 2003; Woodward, 2004). Typically, there are more financingrounds than VC funds.

2. Literature review

Sahlman (1990) is a seminal study of the structure of the US VCindustry and the relationship between investors and VC firms.Gompers (1995) investigates the structure of staged VC investments inthe presence of agency and monitoring costs. Hellmann and Puri (2002)investigate the contributions of VC firms to start-up companies locatedin Silicon Valley.

Lerner (1994) studies (US-only) VC syndication in biotechnologyentrepreneurial companies and investigates the rationales for syn-dication.Hefinds that experiencedVCs syndicatewith other experiencedbackers in early-stage investments, while in later-stage investments ex-perienced VCs syndicate with both experienced and inexperienced ones.Lockett and Wright (2001) report that small VC firms have difficultyachieving optimal portfolio diversification, and syndication allows themto achieve a well diversified portfolio. According to Arberk, Filatotchev,and Wright (2006), the fact that VC investments are illiquid and syn-dication is used as a risk-sharing mechanism means that studying VCsyndication is crucial to understanding the activities of VC firms.

The aim of this study is not to contribute to our understanding of themotives for syndication (Bent, Williams, & Gilber, 2004; Brander, Amit,& Antweiler, 2002; Casamatta & Haritchabalet, 2007; Fleming, 2004;Lockett & Wright, 2003). Rather we aim to investigate the risk andreturns of US–European cross-border syndicated investments (takingthe composition of these syndicates as given).

There are a number of studies that examine the risk and returns ofVC investments (see Cochrane, 2005; Korteweg & Sorensen, 2010;Woodward, 2004), but these studies focus on investments by US-onlyVCs and VC syndicates. As we document below, US VC firms havebeen co-investing with European VCs since the late 1980s. Yet, littleis known about the syndicates. To our knowledge, the risk and returns ofEuropean VCs co-investing with US players is as yet unexplored. Table 1summarizes the results of previous studies of VC risk and returns. OnlyCochrane (2005) and Korteweg and Sorensen (2010) investigate therisk and returns of VC investments at the level of the individual portfoliocompany as opposed to the performance of the VC fund overall.

Seppä and Laamanen (2001) investigate the risk and return profilesof investments at the VC fund level (correcting for sample selection biasusing a binomial model). For a sample of 597 investment roundsbetween 1998 and 1999, they find that early-stage investments havehigher returns and implied volatility than other investment rounds.Manigart et al. (2002) examine the determinants of VC returns in five

2 We are grateful to the referee for drawing our attention to the limitations ofcalculating returns based on IPO prices.

3 The weights (1/3, 2/3) are based on the average rate of post-IPO VC exit documented,e.g., by Cumming (2008). For comparison, we report returns calculated (i) on theassumption that the entire VC holding is sold through IPO (at the offer price), and(ii) on the assumption that it is sold immediately at the end of the lock-up period;see Appendix A for details.

4 Share of the old venture investor I ¼ 0:33� 40−15ð Þ40 � 100 ¼ 21%:Weare aware that

some VCs have ‘ratchet agreements’ protecting them from ownership dilution that mightarise due to subsequent investment rounds. However, VentureXpert provides noinformation on the type and extent of these agreements for our sample firms.

15S. Espenlaub et al. / International Review of Financial Analysis 31 (2014) 13–24

markets: the US, the UK, the Netherlands, Belgium and France from1994 to 1997. The authors show that VC firms require higher returnsfor early-stage investments than expansion or later-stage financing. Inaddition to depending on the financing stage, the required returnsdepend on the size of the investment and the country of origin of theportfolio company.

Chen, Baierl, and Kaplan (2002) examine the long-term risk andreturn characteristics of VC investments, and their role in long-termstrategic asset allocation. They use 148 VC funds that had liquidated asof 1999. They report an average arithmetic return of 45%, and a standarddeviation of 115.6%. They conclude that one should only allocate between2% and 9% of a portfolio to VC for theminimum-variance portfolio, due tothe high volatility of VC returns.

Cochrane (2005) examines the mean, standard deviation, alpha(abnormal performance) and beta (systematic risk) of VC investmentsin the US. Based on a sample of 16,613 investment rounds between1987 and 2000, the results show that the expected returns for the firstround are 71%, for the second 65%, for the third round 60% and for thefourth round 51%. The betas corresponding to the stages are 1.1, 0.9,0.7 and 0.5, respectively. Similarly, Korteweg and Sorensen (2010)examine the alpha and beta of VC investments in entrepreneurialcompanies. The authors focus on US VC firms and examine the riskand returns from 1985 to 2005. Correcting for sample selection biasdue to the endogenous timing of VC exits (‘dynamic selection bias’),they find that monthly alphas range from 3.3% to 3.5%, while the betasare more than 1 using the Capital Asset Pricing (CAPM) and three-factor models.

With the exception of Cochrane (2005) and Korteweg and Sorensen(2010), previous studies evaluate the risk and returns of VC investmentsat the fund level (e.g., Kaplan & Schoar, 2005), often using proprietarydata on the portfolios of companies held by specific VC firms or funds(e.g., Kaplan & Stromberg, 2003, 2004). At the fund level, it is difficultto analyze changes over time and variations across industries, whileour study (along with Cochrane, 2005; Korteweg & Sorensen, 2010) isable to examine these aspects using data at the level of the individualdeal (or portfolio company). An added advantage of using deal-leveldata is that there are many more observations of deals than of VC funds.A larger sample improves the reliability of the statistical estimates inour risk-return analysis.

This is the first study to examine the risk and returns of co-investments by US and European VC firms, as most previous studiesfocus on separate samples of either US or European firms. For instance,Hege, Palomino, and Schwienbacher (2008) compare the performanceof VC investments between the US and Europe and report internalrates of return (IRR) of 62% for US VCs and 106% for European VCs.Our analysis sheds light on the differences and similarities betweenUS–European cross-border syndicates and US-only (or European-only)VCs and syndicates in terms of their characteristics and performance.

3. Data and methodology

3.1. Data

From VentureXpert we extract all investments made during 1985–2008 by US–European cross-border syndicates of VC firms, and identifywhether they were exited through an IPO, M&A (including trade sales)or liquidation, or not exited, during 1990–2010. From the database, wecollect information on the value of the backed company after receivingVC financing (post-VC financing value), the country of origin of the VCfirms, the choice of exit route, the number of rounds a portfolio companyreceived, and the number of VCs in the syndicate.

To be included in our sample, a portfolio company should havereceived financing from at least one US and one European VC. Further,we require that data need to be available on (i) the amount investedby the VCs, (ii) the post-VC financing company value, and (iii) thefounding dates of the VCs and the portfolio company. These filter rules

result in a final sample of 26,865 investment rounds made by 2641US–European VC syndicates. Of these investments, 6774 result in IPOexits, 13,907 in M&A exits, 2152 in liquidations and 4032 had notbeen exited by the end of the sample period (i.e. 31 December 2010).Our risk-return calculations are based on the subsample of IPOs andM&As for which realized values are available (not available in the caseof liquidations and non-exited investments).

3.2. Calculation of holding period returns

We compute the holding period return (RT) for each financing roundusing the following equation:

RT ¼ ITIt0

−1 where IT ¼ It0Vt0

� �W

� �MT : ð1Þ

IT is the amount a VC firm receives at the time of exit; It0 is the initialinvestment in a target company; Vt0 is the value of a portfolio companyimmediately after receiving VC funding at t0;W is an adjustment factorcorrecting for changes (dilution) in VC ownership due to changes inthe portfolio company's valuation across successive investment rounds,as discussed below; MT is the market value of the portfolio companyupon exit. The market value (MT) of the portfolio company is the keyfactor in the return calculation, but this information is available onlyfor IPO and M&A exits.

For investments exited through an IPO, we need to take into accountthe fact that VC backers usually do not sell their entire stake at the timeof the IPO (see Cumming, 2008), but instead exit gradually over the firsttwo post-IPO years.2 To account for this, we compute the market valueof the portfolio company not only at the time of the IPO but also in theaftermarket up to two years post-IPO. To calculate the market value atIPO, we multiply the number of shares outstanding after the IPO bythe IPO offer price. To calculate the aftermarket values, we multiplyoutstanding shares by the end-of-month share price for each of the17 months starting from the end of the lock-up agreement (typically6 months after the IPO) and ending roughly 24 months after the IPO.For each of these 17months, we compute VC returns adjusting for thedifferent holding periods. Finally, we compute the VC returns on theinvestment exited (gradually) through an IPO as the weighted averageof one third of the return at the IPO and two thirds of the averagemonthly returns over the 17-month aftermarket period.3 We apply asimilar method to the market benchmark to ensure synchronicity inthe price observations and the matching of holding periods.

Note that this adjustment only applies to IPO exits. For M&A exits, weassume that the VCs fully realize their investments in a single transaction,and we use transaction values to calculate the holding period return.

VentureXpert does not provide information on VC holdings in aportfolio company. We infer VC holdings using an approach similar tothat of Cochrane (2005) and Korteweg and Sorensen (2010). Thisapproach is best explained using an example. Assume that a portfoliocompany receives $5 m from a VC firm and is valued at $15 m post-financing. From these values we can infer that the VC firm owns 33%of the company. If the company were to receive an additional $15 mfrom a different VC firm and was then valued at $40m, the second VCwould own 38%, while the first VC's ownership would be reduced to21%.4

6 Bourguignon, Fournier, and Gurgand (2007) present a survey of the selection biasmodels available in the literature, including Lee (1983) and Dubin and McFadden(1984). Using aMonte Carlo experiment, they conclude that Dahl'smodel performs betterthan both Lee's (1983) and Dubin and McFadden (1984).

7 Dahl (2002) recommends the transformation of second- or third-order polynomialapproximations for the correction term. Using a third-order polynomial function doesnot change the results and hence we report the second-order (polynomial function)

16 S. Espenlaub et al. / International Review of Financial Analysis 31 (2014) 13–24

Finally, we annualize the holding period returns using

RAnnual ¼ 1þ RTð Þ1t−1 ð2Þ

where t is the period over which investments are held (in years).5

3.3. Sample selection bias

In the absence of sample selection bias, systematic risk for a givenfinancing stage can be estimated using the following equation:

Rj ¼ Xβ j þ ε j ð3Þ

where Rj is the excess return in stage j, X is themarket excess returns, andβj is the systematic risk in stage j, measuring changes in the investmentreturns in stage j relative to the market excess returns.

However, due to the systematic sample selection criterion that invest-ment returns can only be observed for investments exited through anIPO or M&A, the βj estimate is likely to be biased. Therefore, a sampleselectionmodel is required to provide unbiased estimates. If the objectivewere to control for the binary outcome of inclusion versus exclusion fromthe sample, theHeckman (1979)modelwould be an appropriatemethodby which to control for sample selection bias. However, VC firms do notface the binary choice of an IPO exit or a non-exit. Instead, they chooseto exit (or not) through one of several routes, including IPO, M&A orliquidation. This complicates the estimation of βj, as documented inprevious studies (Cochrane, 2005, Korteweg & Sorensen, 2010).

To address the problem of selectivity in the presence of multipleoutcomes, we use a sample selection model proposed by Dahl (2002).The model extends the classic Heckman (1979) model in two ways:first, it models various choices as opposed to a single choice in theselection equation. Second, the probabilities of these choices aretransformed through a polynomial function.

To overcome the issue of exit-route selectivity, we use the Dahl(2002) model. Consider the following two equations:

R1 ¼ Xβ1 þ μ1 ð4Þ

P�k ¼ Zγk þ ηk k ¼ 1;2;…m ð5Þ

where Eq. (4) is a linear regressionmodel and Eq. (5) is a probability (orchoice) model. R1 is an excess return for stage 1, and the variable Xcontains market excess returns, industry and year dummies. μ1 is thedisturbance term of the outcome equation and satisfies E(μ1|X,Z) = 0and V(μ1|X,Z)=σ2. P*k is the probability of observing no exits throughIPO, M&A and liquidation. k is a categorical variable and describes mchoices (m = 1, IPO, m = 2, M&A, m = 3, liquidation and m = 4, noexit). The variable Z contains control variables for the probabilities ofexits. These variables include the following: age of VCs, amount investedby VCs, VCs syndicated (measured as the number of VCs in thesyndicate), age of portfolio companies, number of rounds a portfoliocompany has received, financing stage, and year and industry dummies.ηk is the residual of the selection model, independently and identicallydistributed (i.i.d.). The stage return (R1) is observed only when the VCinvestors make an investment and then exit through an IPO. Thiscondition is expressed as

P�kN Max

K≠1P�k

� �: ð6Þ

5 Our estimate of the annualized return is based on valuation estimates of the portfoliocompany pre-IPO. These estimates are clearly subject to all the biases and inaccuraciestypically arising in valuations, particularly in the context of VC portfolio companies. Wethank the referee for highlighting this issue and for noting that this is a problem commonto all empirical research into VC.

The conditional probability (P⁎k) is estimated through amultinomiallogit using

P j InvjExitIPO;M&A;Liq0n;no exit

� ¼

exp Zγ j

� X

kexp Zγkð Þ : ð7Þ

Pj is the probability of exit j (or no exit) and is estimated through amultinomial logit. It is transformed using a polynomial approximation:

λ Pjj

� ¼ λ MjP

njj

� ð8Þ

where Mj is a dummy variable that equals one for choice j and zerootherwise, n is the degree of the polynomial function for Pj. λ(Pj) is asample correction term, which depends on the probability Pj and isequivalent to the inverse Mills in the Heckman model. The correctionfunction λ() is assumed to have the same form for all financing stagesthat are exited through the IPO route.6

A consistent estimate of systematic risk for the financing round isestimated using pooled cross-sectional regression and given by

Rj ¼ Xβ j þ λ P j

� þ ζ j j ¼ 1 ð9Þ

where Rj is the excess return for stage j (j=1 start-up stage, j=2 earlystage, j=3 expansion stage, and j=4 the later stage), X contains themarket return, and year and industry dummies. λ(Pj) is the samplecorrection term and includes first- and second-order polynomialapproximations of the probability of an IPO exit.7 Eqs. (7) to (9) areestimated simultaneously. We use clustered standard errors to adjustfor the cross-sectional dependence of VC investments.

In addition to adjusting for selection bias due to the choice of exitroute and the availability of data only for IPO and M&A exits (and notfor other routes), we also need to address the endogenous nature ofthe timing of the VC exit and the resulting idiosyncratic variation inthe holding period across VC investments. In estimating the risk andreturns of VC investments it is clearly essential to control for the factthat VC returns are realized in a discrete and lumpy fashion over widelyvarying holding periods.8 To do so, we adopt an approach similar to thatof Korteweg and Sorensen (2010), who deal with the sample selectionbias due to the endogenous timing of observed returns. While theyaddress this ‘dynamic selection bias’ using Markov Chain Monte Carlo(MCMC) estimation, we estimate an amended Heckman model withthe first stage involving a duration Cox (1972) model.

In Stage I of our model, we estimate, using a Cox (1972) model, thetime from investment to exit by the VC firms. This allows us to modeland control for the endogeneity of the exit-timing decision. In Stage II,we estimate the systematic risk by including, in the form of an inverseMill's ratio, the predicted hazard of exit (i.e., the instantaneous prob-ability of exit at a given point), modeled in Stage I as an additionalexplanatory variable. This adjusts the estimated systematic risk for thedynamic sample selection (due to endogenous timing). A significantcoefficient for the inverse Mill's ratio would indicate the presence ofdynamic selection bias.9

results.8 We are grateful to the referee for drawing our attention to this issue.9 The results of our analysis of the risk and returns of VC investments below need to be

interpreted bearing in mind that they are only generalizable to the extent that ourmethods for correcting the sample selection biases are fully valid. Unfortunately, as theselection biases outlined above are an inherent feature of any VC dataset, the resultinglimitations are unavoidable.

Table 2Definition of the variables.

Variables Definition and unit of measurements Data source

Early stage returns Annualized returns for the early stage rounds. VentureXpertExpansion returns Annualized returns for the expansion stage

roundsVentureXpert

Later stage returns Annualized returns for the later stage rounds. VentureXpertStart-up returns Annualized returns for the start-up rounds. VentureXpertBenchmarkreturns(marketreturns)

Is the annualized returns computed fromMorganStanley World Capital Index (MSCI). S&P 500andMSCI Europe.

Bloomberg

VC-age Age of venture capital firm invested in a portfoliocompany and measured in years as thedifferencebetween founding date and date of investment.

VentureXpert

FFind1–FFind12 Fama and French industry classification VentureXpertRisk free rate Annualized risk free rate DataStreamInvest size($m) Is the amount invested by venture capital firm in

a given round measured in millions of dollarsVentureXpert

Comp age Age of a portfolio company, measured (in years)as the difference between the founding date andthe date of investment.

VentureXpert

Synd (#) Is thenumber of venture capitalfirms syndicatedin financing a portfolio company

VentureXpert

17S. Espenlaub et al. / International Review of Financial Analysis 31 (2014) 13–24

3.4. Determinants of the selection equation

The determinants included in the two steps are motivated as follows:

(a) VC age: The age of the VC company is measured as the differencebetween founding date and investment date. Previous studies(Gompers, 1996; Lee & Wahal, 2004) document that the IPOexit is an important exit method and that VCs improve theirreputations by exiting through this route. Generally, young VCswill rush to exit through an IPO so as to gain a reputation, asreported by Gompers (1996). Hence, the probability of an exitof this type is influenced by the age of the VCs.

(b) Investment size: The size of the VC investment in a portfoliocompany is expected to influence the decision to exit the portfoliocompany. VCs tie up substantial amounts of capital in illiquidinvestments in portfolio companies. The amount of capitalinvested influences the expected value; hence a large invest-ment reflects a VC's confidence in the future success of a portfoliocompany. If VCs have any predictive ability, we expect the prob-ability of a successful exit (through IPO orM&A) to increase withinvestment size.

(c) Syndication allowsVCs todiversify their investment risk. Syndicatesize typically increases as the portfolio company develops and itsneed for VC funds increases. Giot and Schwienbacher (2007)argue that larger syndicates facilitate successful trade-sale andIPO exits by providing valuable contacts, connections and qualitycertification. Hence, we expect VC syndicate size to increasethe probability of exit through M&A or IPO. At the same time,larger syndicate size may also increase the probability ofliquidation. A larger syndicate has greater bargaining powervis-à-vis the entrepreneur(s) and can exert more pressureon the entrepreneur(s) to liquidate an unsuccessful venture(e.g., Sharifzadeh & Walz, 2012).

(d) Company age: The age of the backed company is measured inyears from founding date to investment date. VCs invest inyoung companies, and their investment decision might beinfluenced by the characteristics of the backed company. Forinstance, a young company with high growth opportunitiesmight be more attractive to VC firms than a mature companywith low growth opportunities.

(e) Market excess return (Rm−Rf) is the main variable of interest,in that its coefficient is interpreted as a measure of systematicrisk (beta). Like all investors, VC firms are likely to benchmarkthe returns of their investments against market excess returns.According to the CAPM, VC backers will require a return that isin direct relation to the beta of the investment and to the marketreturn.We use the US risk free rate for US VC firms, while for theEuropean VC firms we use the average risk free rate of Europeancountries in the sample.

We use industry and year dummies as additional control variables.Table 2 shows the definitions of the independent variables.

4. Descriptive statistics, results and analysis

4.1. Descriptive statistics

Table 3 shows the descriptive statistics of the VC and portfoliocompany characteristics by exit route. Panel A showsVC-related variables,while Panel B shows portfolio company characteristics.

For IPO and M&A exits, we find that the mean age of the VC firmsin US–European cross-border syndicates ranges from 21 to 23 years,with the median ranging from 17 to 20 years. For comparison, Giotand Schwienbacher (2007) report for US-only syndicates averageages of 28.8 and 27.6 for IPO and M&A exits, respectively. This mayin part be explained by the greater focus of our cross-border syn-dicates on expansion-stage investments, which tend to be backed

by less experienced VCs (Lerner, 1994). We find that approximately45 to 46% of the investments by our cross-border syndicates aremade at the expansion stage, while start-ups attract the smallestproportion of investments (around 11 to 12%); see also Table 4. Bycontrast, previous studies on US VCs show a much smaller proportionof expansion-stage investments and a higher proportion of early-stageinvestments. For instance, Cochrane (2005) reports that early-stageinvestments account for 46% of his sample, compared to 14.7% forexpansion rounds. Thus, while US VC's investments tend to cluster inearly-stage financing, our findings highlight that US–European cross-border syndicates focus on mature investments, where the risk tendsto be lower than in the earlier stages (as we will show below).

We find that the size of US–European cross-border syndicates (at 10members for IPOs and 8 for M&As) is around double that reported byGiot and Schwienbacher (2007) for US-only VC syndicates (5 for IPOsand 4 forM&As). The larger syndicate sizemay be related to the greaterfocus of US–European cross-border syndicates on expansion-stageinvestments and the younger average age of the syndicate members(Lerner, 1994).

Interestingly, we find that syndicate size is larger for investmentsthat are liquidated (12 members) than for investments exited throughother routes. This may be due to the relative inexperience (low averageage) of the VCs and VC syndicates involved in investments that go on tobe liquidated. Another possible explanation may be that, in particularlylarge syndicates, the costs of syndication (e.g. due to free-riding amongVC members) offset the benefits (in terms of risk sharing, etc.). Free-riding by self-interested syndicate members may cause each individualVC to spend too little time and effort on screening and monitoring theportfolio companies, and this may make liquidation more likely. For adiscussion of the costs and benefits of syndication, see, e.g., Sharifzadehand Walz (2012).

At over 50%, the proportion of investments exited through M&As ishigher than that for all other routes. This is followed by IPOs witharound 25 to 30% of the investments. Similarly, previous studies (Giot& Schwienbacher, 2007) document that the M&A route is the mostcommon exit method for VCs, followed by IPOs.

Comparing IPOs andM&As, we find that portfolio companies that areexited through the IPO route tend to receive relatively larger investmentsthan those exited via M&As (the average investment sizes are $5.90 m

Table 3Descriptive statistics by exit route. The table shows the descriptive statistics for the sample (1990–2010) based on the exit method. Panel A shows VC firms' characteristics while nel B shows the portfolio company statistics. VC-age is the age ofventure capitalfirm investing in a portfolio company and ismeasured as the difference between founding date anddate of investment (in years). Invest size is the amount invested b venture capital firm in a given round ($m). Synd (#) is the numberof venture capital firms syndicated in financing a portfolio company. Comp age is the age of a portfolio company and is measured as the difference between founding date and da of investment (in years). Start-up, Early, Expansion and Later aredummy variables taking a value of 1 for each stage and zero otherwise. The industries are dummies based on the Fama and French 12 industry classifications.

IPO exits M&A exits Un-exited Liquidation exit

Mean Median STD Max Min Mean Median STD Max Min Mean Median STD Ma Min Mean Median STD Max Min

Panel A: Venture capital firms characteristicsVC-age 20.670 16.953 16.944 189.069 4.000 22.759 19.912 16.528 192.751 3.000 19.367 17.000 15.147 148 00 0.000 16.162 15.000 14.774 135.000 0.000Invest size ($m) 5.909 2.532 16.853 620.000 0.300 4.767 2.877 6.706 334.995 0.100 2.959 2.110 2.920 30 00 0.300 2.770 2.000 2.719 30.792 1.000Synd (#) 10 9 5 35 2 8 7 5 31 2 12 11 6 28 2 12 11 6 33 2

Panel B: Portfolio company characteristicsComp age 8.392 6.789 8.293 24.340 1.000 9.008 7.690 6.901 20.049 1.000 10.118 9.000 7.419 19 00 2.000 5.720 5.000 3.559 21.000 1.000Start-ups 0.125 0 0.331 1 0 0.114 0.000 0.317 1.000 0.000 0.057 0.000 0.231 1 00 0.000 0.079 0.000 0.271 1.000 0.000Early 0.203 0 0.402 1 0 0.254 0.000 0.435 1.000 0.000 0.163 0.000 0.369 1 00 0.000 0.148 0.000 0.355 1.000 0.000Expansion 0.454 0 0.498 1 0 0.448 0.000 0.497 1.000 0.000 0.506 1.000 0.500 1 00 0.000 0.513 1.000 0.500 1.000 0.000Later 0.218 0 0.413 1 0 0.184 0.000 0.388 1.000 0.000 0.275 0.000 0.446 1 00 0.000 0.260 0.000 0.439 1.000 0.000Consumer non-durable 0.010 0 0.099 1 0 0.012 0.000 0.111 1.000 0.000 0.006 0.000 0.074 1 00 0.000 0.001 0.000 0.038 1.000 0.000Consumer durable 0.006 0 0.078 1 0 0.006 0.000 0.080 1.000 0.000 0.001 0.000 0.022 1 00 0.000 0.011 0.000 0.103 1.000 0.000Manufacturing 0.026 0 0.159 1 0 0.018 0.000 0.135 1.000 0.000 0.022 0.000 0.146 1 00 0.000 0.017 0.000 0.129 1.000 0.000Energy, oil & gas 0.003 0 0.056 1 0 0.002 0.000 0.042 1.000 0.000 0.000 0.000 0.016 1 00 0.000 0.000 0.000 0.000 0.000 0.000Chemicals and allied products 0.008 0 0.090 1 0 0.003 0.000 0.054 1.000 0.000 0.001 0.000 0.036 1 00 0.000 0.003 0.000 0.053 1.000 0.000Business equipment 0.463 0 0.499 1 0 0.655 1.000 0.475 1.000 0.000 0.490 0.000 0.500 1 00 0.000 0.558 1.000 0.497 1.000 0.000Telecom 0.041 0 0.199 1 0 0.035 0.000 0.184 1.000 0.000 0.043 0.000 0.202 1 00 0.000 0.086 0.000 0.280 1.000 0.000Utilities 0.002 0 0.043 1 0 0.002 0.000 0.045 1.000 0.000 0.005 0.000 0.071 1 00 0.000 0.000 0.000 0.000 0.000 0.000Wholesale & retail 0.047 0 0.212 1 0 0.035 0.000 0.185 1.000 0.000 0.040 0.000 0.196 1 00 0.000 0.113 0.000 0.317 1.000 0.000Healthcare 0.251 0 0.433 1 0 0.128 0.000 0.334 1.000 0.000 0.263 0.000 0.440 1 00 0.000 0.100 0.000 0.300 1.000 0.000Money finance 0.020 0 0.139 1 0 0.019 0.000 0.137 1.000 0.000 0.010 0.000 0.098 1 00 0.000 0.018 0.000 0.132 1.000 0.000Others 0.123 0 0.328 1 0 0.084 0.000 0.277 1.000 0.000 0.120 0.000 0.325 1 00 0.000 0.094 0.000 0.292 1.000 0.000No of obs 6774 13,907 4032 2152

18S.Espenlaub

etal./InternationalReviewofFinancialA

nalysis31

(2014)13

–24

Payte

x

.0

.5

.0

.0

.0

.0

.0

.0

.0

.0

.0

.0

.0

.0

.0

.0

.0

.0

.0

10 While we do not report a return averaged across all exit routes, such an estimatewould have to take into account the number of and returns on liquidations.

Table 4Descriptive statistics byfinancing round. The table shows descriptive statistics for the sample (1990–2010) based onfinancing stage. VC-age is the age of venture capitalfirm investing in aportfolio company and ismeasured as the difference between founding date and date of investment (in years). Invest size is the amount invested by venture capital firm in a given round($m). Synd (#) is the number of venture capital firms syndicated in financing a portfolio company. Comp age is the age of a portfolio company and ismeasured as the difference betweenfounding date and date of investment (in years). IPO is the number of investments exited through IPO route. MA is the number of investments exited through mergers and acquisitions.Liqu'd is the number of investments exited through liquidation. Private is the number of investments not exited by the end of the sample period. We use Fama and French 12 industryclassifications.

Variable Start-up stage Early stage Expansion stage Later stage

Mean Median STD Mean Median STD Mean Median STD Mean Median STD

VC-age (years) 21.323 18.000 15.848 22.755 20.000 17.292 20.998 18.000 16.553 20.128 17.000 15.403Invest size ($m) 9.913 2.083 29.406 14.796 3.396 45.651 23.040 5.000 47.869 32.540 6.749 59.392Synd (#) 8 7 5 8 6 5 9 8 5 11 10 6Comp age (years) 7.562 6.488 7.061 7.545 6.499 5.742 9.091 7.636 7.554 9.859 8.742 7.739IPO 0.303 0 0.460 0.236 0 0.425 0.250 0 0.433 0.261 0 0.439MA 0.556 1 0.497 0.598 1 0.490 0.499 0 0.500 0.448 0 0.497Liqu'd 0.060 0 0.238 0.054 0 0.226 0.088 0 0.284 0.098 0 0.297No exit 0.081 0 0.272 0.111 0 0.314 0.163 0 0.369 0.194 0 0.395Consumer non-durable 0.012 0 0.107 0.008 0 0.089 0.011 0 0.102 0.009 0 0.097Consumer durable 0.003 0 0.056 0.006 0 0.079 0.007 0 0.085 0.003 0 0.058Manufacturing 0.022 0 0.147 0.019 0 0.138 0.022 0 0.145 0.020 0 0.139Energy, oil & gas 0.001 0 0.038 0.001 0 0.032 0.003 0 0.052 0.001 0 0.023Chemicals and allied products 0.004 0 0.065 0.004 0 0.060 0.006 0 0.075 0.001 0 0.026Business equipment 0.524 1 0.500 0.594 1 0.491 0.571 1 0.495 0.585 1 0.493Telecom 0.033 0 0.180 0.040 0 0.196 0.044 0 0.204 0.044 0 0.204Utilities 0.002 0 0.046 0.001 0 0.032 0.002 0 0.046 0.004 0 0.063Wholesale & retail 0.045 0 0.208 0.032 0 0.176 0.050 0 0.218 0.048 0 0.213Healthcare 0.233 0 0.423 0.178 0 0.383 0.163 0 0.370 0.178 0 0.383Money finance 0.017 0 0.129 0.019 0 0.136 0.018 0 0.134 0.016 0 0.126Others 0.103 0 0.303 0.097 0 0.296 0.104 0 0.306 0.091 0 0.288No of obs 2843 5897 12,500 5625

19S. Espenlaub et al. / International Review of Financial Analysis 31 (2014) 13–24

and $4.76 m, respectively). By contrast, companies exited through IPOstend to be younger at the time of investment than those exited throughM&As and those not exited. Companies that are liquidated are theyoungest (at around 5–6 years on average). We find little variation interms of the investment stages between IPO and M&A exits.

As expected, VC investments are clustered in the business equipmentand healthcare industries, regardless of the exit route. This is consistentwith previous studies (Cochrane, 2005; Giot & Schwienbacher, 2007).

There are differences in the VC characteristics between investmentsthat end up being liquidated and others. The results show that liquidatedinvestments were backed by young VCs, and generally the portfoliocompanies too were younger on average than those portfolio com-panies in the sample that were yet to be exited. Nearly half of theliquidated investments are from the expansion-stage round. Giot andSchwienbacher (2007) report similar shares of liquidations betweenthe early and expansion stages. Interestingly, the syndicate size istypically larger for liquidated investments than for portfolios exited viaan IPO or M&A. The association between syndicate size and liquidationsuggests that VC syndication is not necessarily good news for a portfoliocompany, possibly due to free-riding by VC firms resulting in too littleVC screening and monitoring and a greater chance of liquidation.

Table 4 shows the characteristics of the VC firms and portfoliocompanies by stage of financing. VC experience (as measured by theaverage age of the syndicate members) is quite similar across all thedifferent financing stages, ranging from 21–22 years in the start-upand early stages to 20–21 years in the expansion and later stages. Thisis an interesting observation, as it has previously been found that start-up and early-stage rounds are typically financed by experienced VCs,while expansion or later stages are funded by less experienced ones.For instance, Lerner (1994) finds that experienced US VCs syndicateonly with other experienced VCs in the early-stage round. By contrast,we find that, among our US–European cross-border syndicates, theaverage age of the syndicatemembers varies little across all the financingstages.

Consistent with Giot and Schwienbacher (2007), we find that thesize of investment is smaller in the start-up and early stages than in

the expansion and later stages. Again, as in Giot and Schwienbacher(2007), who report that syndicates are smaller in the start-up andearly stages than in the expansion and later stages, we find that syndicatesize rises to 11 in later stages compared to 8 in the earlier stages. Onaverage, portfolio companies that have received financing at the start-up or early stage are younger than those that have received expansionand later-stage financing.

Cochrane (2005) reports that around 9% of investments withfunding from US VCs in the early and expansion stages are liquidated.We find a similar proportion of liquidations among investments backedby US–European syndicates in the expansion and later stages. However,surprisingly, we find amuch lower liquidation rate of only 5 to 6% in thestart-up and early stages.

In conclusion, our preliminary analysis identifies a number ofinteresting similarities and differences between US VCs and US–European cross-border syndicates.

4.2. Results and analysis

Table 5 provides information on the average annualized returns forVC investments (without correcting for any sample selection biases)exited through IPOs andM&As. The IPO figures in Panel A are calculatedso as to account for gradual post-IPO exit by the VC (as outlined in Dataand methodology Section 3.2 above) by basing returns on a weightedaverage of the IPO offer price and aftermarket prices. The returns onthe stock indices shown in Table 5 are calculated based on staggeredholding periods to match the calculation of the returns on investmentsexited (gradually) through and after IPOs. Panel B shows the returns forinvestments exited through the M&A route. While Table 5 reports onlythe returns on investments exited through IPOs or M&As, it is reasonableto assume that the corresponding returns on the 2152 investments in oursample that were exited by liquidation are−100% in each case.10

Table 5IPO/MAannualized returns. The table showsmean,median, standarddeviation,maximumandminimumvalues of venture capital returns exited through IPO andM&A exits. Returnsare annualized and computed over the period 1990 through 2010 using a weightedaverage of IPO and aftermarket prices. The rows entitled ‘Start-up stage’, ‘Early stage’,‘Expansion stage’ and ‘Later stage’ show returns for the respective financing stages.‘MSCIWorld’ is the annualized MSCI world index returns. ‘MSCI Europe’ is the annualizedMSCI Europe index returns. ‘S&P 500’ is the annualized S&P 500 composite returns. ‘Riskfree’ is the annualized risk-free rate for US and Europe, respectively.

Mean Median STD Max Min

Panel A: Annualized returns for IPO exits (corrected for gradual VC exit post-IPO)Start-up stage 2.532 0.402 7.554 31.148 −0.959Early stage 1.632 0.301 6.253 27.597 −0.967Expansion stage 1.228 0.146 4.533 24.947 −0.977Later stage 0.987 0.114 3.531 21.871 −0.927MSCI world 0.116 0.104 0.091 0.311 −0.119MSCI Europe 0.121 0.114 0.131 0.398 −0.162S&P 500 0.141 0.122 0.110 0.370 −0.098FTSE all shares 0.105 0.091 0.108 0.322 −0.134Risk free rate (US) 0.045 0.049 0.017 0.059 0.000Risk free rate (Europe) 0.061 0.055 0.019 0.147 0.033No of obs 6774

Panel B: Annualized returns for M&A exitsStart-up stage 1.315 0.282 3.241 19.287 −0.921Early stage 1.091 0.228 3.016 18.166 −0.912Expansion stage 0.669 0.052 2.629 16.194 −0.879Later stage 0.563 0.008 2.428 15.188 −0.795MSCI world 0.037 0.045 0.109 0.236 −0.223MSCI Europe 0.058 0.071 0.135 0.336 −0.235S&P 500 0.045 0.042 0.117 0.293 −0.211FTSE all shares 0.041 0.047 0.108 0.263 −0.223Risk free rate (US) 0.030 0.033 0.019 0.080 0.000Risk free rate (Europe) 0.053 0.055 0.011 0.146 0.033No of obs 13,907

20 S. Espenlaub et al. / International Review of Financial Analysis 31 (2014) 13–24

Comparing Panels A and B, we find that IPO returns are higher thanM&A returns in each of the four financing stages. This is consistent withprior studies that document that IPOs are the preferred and mostprofitable form of exit (e.g., Black & Gilson, 1998).

For both IPO and M&A exits, the table shows that expansion andlater-stage returns are on average lower than start-up and early-stagereturns, based on both means and medians returns.11 On invest-ments exited through an IPO, VC firms earn median returns of40.2% on investments in the start-up stage, 30.1% in the early stage,14.6% in the expansion stage, and 11.4% in the later stages. Thecross-sectional standard deviations (which can be interpreted astotal risk) corresponding to these stage are 755%, 625%, 453% and353%, respectively.

The corresponding median returns (standard deviations) for theM&A exits are 28.2 (324)% in the start-up stage, 22.8 (301)% for theearly stage, 5.2 (262)% for the expansion stage, and 0.8 (242)% for thelater stages.

For both IPO and M&A exits, total risk decreases significantly fromthe start-up to the later stages of financing. This is to be expected asportfolio companies mature from earlier to later stages (see Table 4),reducing the level of risk and thus the expected return.

As noted above, the returns shown in Table 5 are not corrected forsample selection bias. Table 6 examines the impact of controlling forexit-route selection bias on returns. By year of (IPO) exit, it shows theVC returns on investments exited through an IPO with and withoutcorrection for sample selection bias. The returns are corrected using

11 The following discussion focuses onmedian returns. A large difference betweenmeanandmedian values suggests that returns are skewed anddrivenby a few investmentswithvery high returns. Skewness is to be expected for VC investments, where a few dealstypically provide very high returns to compensate for large numbers of unsuccessful andliquidated investments. We thank the anonymous referee for highlighting this issue.

the Dahl model (outlined in Section 3.3).12 Following Cochrane (2005),we deal with the issue of return skewness, noted above, by calculatinglogarithmic returns.

It is evident from the table that in many exit years the uncorrectedreturns are higher than the corrected returns. These differences areoften statistically significant at 5% or higher. VC's returns are positivein the vast majority of exit years and financing rounds. However, thereturns are particularly high in the bubble year of 1999. After thecredit crisis in 2007–8, returns for the expansion and later stagesfell dramatically by around 80% from their average levels in 2006.While returns did not decline as much for those investments made inthe earlier stages, exit activity declined. In fact, no IPO exits occurredfor start-up investments during 2008–9.

As Table 6 shows, adjusting for selection bias has a major impact onreported returns. From this we conclude that it is important to controlfor selectionbiaswhen estimating the risk-return trade-off.13 As outlinedin Section 3.3,we control for exit-route selection bias usingDahl's (2002)two-stage model. This involves a multinomial logit model in the firststage, predicting the chosen exit route, and in the second stage, a linearregression of excess return on the VC investment on the market riskpremium (proxied by the excess return on the S&P 500 index). Wecalculate both VC and benchmark returns taking into account the factthat VCs do not sell all of their holdings at the time of the IPO(i.e. correcting returns for gradual VC exit). Table 7 shows theestimated parameters of thismodel. As expected,we find clear evidenceof selection bias, as shownby the statistical significance of the estimatedselection-bias parameters, λ(P1) and λ(P2).

We find that returns on investments in all stages are positivelyrelated to the market benchmark (i.e. they have positive betas that arestatistically significant at 5% or above). As is to be expected, start-upand early-stage investments have higher systematic risk (higher betas)than expansion and later-stage investments. The start-up and early-stage betas of 1.316 and1.122, respectively,mean that returns on invest-ments made in these stages are expected to change by 1.31 and 1.12%,respectively, for every one-percent change in the market benchmark(S&P 500). The betas for expansion and later-stage investments are1.079 and 0.797, respectively.

For comparison, Cochrane (2005) reports betas of 1.1 for start-upinvestments, 0.9 for early-stage, 0.7 for expansion and 0.5 for later-stage investments. Both Cochrane's and our results demonstrate thatVC investors are exposed to higher systematic risk in earlier stages offinancing, and that systematic risk (beta) declines continually, thelater the stage of investment.

This may be due to the difficulty of selling a portfolio company in itsearly phase of development. Previous studies have suggested that thehigher returns in earlier stages are required to attract and compensatethemore experienced VCfirms that aremore likely than less experiencedVCs to invest in these riskier stages. In linewith Cochrane (2005), wefindhigher risk and returns in earlier stages, but we find little variation in thelevel of experience of US–European cross-border syndicates acrossfinancing stages (see Table 4).

Abnormal performance, asmeasured by the coefficient of the constantin our second-stage regression, is shown in Panel B of Table 7. Thismeasure is the so-called alpha. We estimate alpha as 41.6%, 35.2%,28.1%, and 23.5% for the start-up, early, expansion and later stages,respectively. Similar to beta, we find alpha to decline monotonically, thelater the financing stage. This is comparable to Cochrane, who reportsalphas of 71% for start-up, 65% for early, 60% for expansion, and 50% forlater-stage investments.

12 The Dahl method (similar to the method used by Cochrane) gives us estimates of themean of the (logarithmic) return. For comparison,we focus similarly on themeans (ratherthan the medians) of the uncorrected returns.13 In interpreting the figures in Table 6, it is important to bear inmind that theymay stillprovide a biased estimate of the typical return on a VC investment. The returns reportedhere are only representative to the extent that our correction for sample selection bias isvalid.

Table 6Venture capital returns corrected for (exit-route) sample-selection bias. The table shows the descriptive statistics of annualized log returns for the Start-up stage, Early, Expansion and Later stage round based on the subsample of investments exitedthrough an IPO. The statistics are tabulated by the year of IPO.We report both corrected and uncorrected returns for sample selection bias. The corrected returns for sample selection bias are estimated using Dahl (2002) semi-parametricmethod. T-test shows if the difference in means between corrected and un-corrected returns is statistically significant.

Start-up returns Early stage returns Expansion stage returns Later stage returns

Un-corrected Corrected Un-corrected Corrected Un-corrected Corrected Un-corrected Corrected

IPO-year Mean Std Mean STD T-test Mean Std Mean STD T-test Mean Std Mean STD T-test Mean Std Mean STD T-test

1990 −0.834 1.081 −0.250 0.893 −3.623 0.204 0.575 0.281 0.620 −0.789 −0.295 1.005 0.261 0.564 −4.334 0.169 0.235 0.253 0.383 −1.6561991 0.046 0.098 0.452 0.899 −4.988 0.570 0.750 0.582 0.725 −0.100 0.023 0.090 0.298 0.200 −11.611 0.457 0.375 0.326 0.261 2.5221992 −0.344 0.910 −0.404 0.682 0.457 0.273 0.593 0.415 0.391 −1.767 −0.010 0.080 0.029 0.059 −3.352 0.328 0.417 0.378 0.307 −0.8561993 0.083 0.998 0.453 0.748 −2.601 0.326 0.443 0.366 0.482 −0.539 0.037 0.779 0.157 0.051 −1.771 0.387 0.524 0.278 0.259 1.7051994 −0.175 1.088 −0.483 0.705 2.101 0.274 0.581 0.326 0.718 −0.492 0.158 0.225 0.130 0.217 0.769 0.480 0.457 0.349 0.304 2.1071995 0.496 0.921 0.436 0.692 0.455 0.482 0.623 0.480 0.643 0.021 0.420 0.795 0.518 0.750 −0.783 0.434 0.330 0.574 0.228 −3.0741996 0.293 0.952 0.577 0.881 −1.898 0.625 0.737 0.640 0.134 −0.218 0.213 0.839 0.521 0.538 −2.734 0.402 0.727 0.484 0.401 −0.8921997 0.111 0.905 0.645 0.678 −4.131 0.547 0.720 0.623 0.438 −0.798 0.420 0.576 0.707 0.636 −2.905 0.320 0.582 0.382 0.268 −0.8971998 0.709 0.796 0.788 0.979 −0.544 0.895 0.962 0.836 0.665 0.443 0.602 0.784 0.749 0.679 −1.236 – – – – –

1999 0.895 1.031 1.010 1.076 −0.671 1.175 0.859 1.113 0.774 0.466 1.084 1.161 1.001 1.278 0.418 0.692 0.722 0.705 0.823 −0.1062000 0.299 0.458 0.649 0.836 −3.319 0.242 0.864 0.716 0.637 −3.868 0.258 0.531 0.169 0.201 1.487 0.220 0.379 0.125 0.157 2.1602001 0.137 0.342 0.840 0.700 −8.269 0.573 0.854 0.770 0.689 −1.568 0.004 0.009 0.037 0.091 −3.946 0.292 0.611 0.376 0.317 −1.1132002 −0.323 0.731 −0.355 0.548 0.308 0.213 0.120 0.451 0.767 −3.280 −0.128 0.258 −0.119 0.172 −0.253 0.023 0.061 0.047 0.040 −2.8232003 0.337 0.870 0.674 0.653 −2.705 0.451 0.391 0.426 0.230 0.482 0.263 0.305 0.438 0.240 −3.941 0.215 0.302 0.301 0.196 −2.1002004 0.218 0.685 0.519 0.403 −3.399 0.503 0.369 0.527 0.305 −0.444 0.302 0.523 0.458 0.430 −2.013 0.264 0.245 0.402 0.245 −3.4362005 0.186 0.782 0.453 0.586 −2.391 0.463 0.498 0.491 0.401 −0.371 0.100 0.173 0.288 0.393 −4.051 0.044 0.052 0.179 0.126 −9.2382006 0.218 0.670 0.405 0.502 −1.958 0.370 0.374 0.396 0.282 −0.488 0.263 0.522 0.422 0.290 −2.400 0.209 0.209 0.423 0.299 −5.1512007 0.003 0.094 0.436 0.234 −16.154 0.107 0.089 0.046 0.039 5.827 0.256 0.495 0.496 0.399 −3.286 0.261 0.346 0.424 0.105 −4.4482008 – – – – – 0.053 0.373 0.228 0.304 −3.156 0.038 0.168 0.126 0.106 −3.952 0.051 0.038 0.045 0.073 0.6602009 – – – – – 0.057 0.166 0.239 0.133 −7.465 0.043 0.375 0.047 1.915 −0.023 0.062 0.075 0.056 0.060 0.5832010 0.388 0.351 0.449 0.753 −0.677 0.068 0.211 0.233 0.101 −6.486 0.066 0.051 0.090 0.077 −2.283 0.033 0.046 0.062 0.095 −2.516No of obs 858 1389 3068 1459

21S.Espenlaub

etal./InternationalReviewofFinancialA

nalysis31

(2014)13

–24

Table 7Risk-return estimation allowing for gradual post-IPO disinvestment by VCs. The estimation of Stage 1 in Panel A uses the full sample, while Stage 2 in Panel B uses only investments exitedthrough IPOs. Returns in Panel B are based on a weighted average of the IPO offer price and post-IPO market prices, as described in Section 3.2. The coefficients are estimated using Dahl's(2002) semi-parametricmethod.Weuse clustered standarderrors as opposed to robust standarderrors allowing for cross sectional dependence of the returns.Market is the excessmarketreturns using MSCI Europe and S&P 500. The correction terms in the Dahl (2002) model are transformed using first and second order polynomial approximations. *** Significant at 1%, **significant at 5% and * significant at 10%.

Panel A: Stage I selection equation

Variables IPO exits M&A exits Liquidation exits

Coeff Odds Coeff Odds Coeff Odds

VC-age 0.032*** 1.033 0.040*** 1.041 −0.007*** 0.993Invest size 0.014*** 1.014 −0.015*** 0.985 −0.004 0.996Synd 0.058*** 1.060 0.020*** 1.020 0.023*** 1.023Comp-age 0.008** 1.008 −0.024*** 0.976 −0.256*** 0.774Start-up −0.672*** 0.511 −1.144*** 0.319 0.266** 1.305Early −0.609*** 0.544 −0.441*** 0.643 −0.388*** 0.678Expansion −0.220*** 0.803 −0.171*** 0.843 0.105** 1.111Cons 3.866*** 4.035*** 1.041 1.859*** 6.417Industry & year Yes Yes YesPseudo R2 0.317No of obs 26,865

Panel B: Stage II outcome equation

Variables Start-up Early stage Expansion stage Later stage

Coeff P-values Coeff P-values Coeff P-values Coeff P-values

Constant 0.416*** 0.000 0.352*** 0.000 0.281** 0.028 0.235*** 0.004Market 1.316*** 0.000 1.122*** 0.000 1.079*** 0.000 0.797** 0.033λ(P1) −0.688* 0.089 −0.461** 0.430 −0.421* 0.096 −0.368* 0.086λ(P2) 0.326 0.257 0.257* 0.085 0.068 0.897 0.189 0.272Industry Yes Yes Yes YesYear Yes Yes Yes YesR2 0.161 0.221 0.181 0.137No of obs 858 1389 3068 1459

22 S. Espenlaub et al. / International Review of Financial Analysis 31 (2014) 13–24

To address the dynamic selection bias that is due to the endogenoustiming of VC exits, we use a Heckman-type two-stage model (outlinedin Section 3.3), where the selection equation is estimated using a Cox(1972) duration model. Table 8 shows the results.

Table 8Cox estimation model. As in Table 7, the estimation of Stage 1 in Panel A uses the full sample, wbased on a weighted average of the IPO offer price and post-IPO market prices, as described inhazard). We use clustered standard errors as opposed to robust standard errors allowing for cEurope and S&P 500. *** Significant at 1%, ** significant at 5% and * significant at 10%.

Panel A: Stage I selection equation

Variable Coeff

VC-age 0.008*Invest-size 0.005*Synd 0.005*Comp-age 0.086*Start-up −0.35Early −0.66Expansion −0.37Industry & year YesPseudo R2 0.387No of obs 26,865

Panel B: Stage II outcome equation

Variables Start-up Early stage

Coeff P-values Coeff P-valu

Constant 0.457*** 0.000 0.418*** 0.000Market 1.403*** 0.000 1.2143*** 0.000Inv. mills 0.031 0.563 −0.297*** 0.000Industry Yes YesYear Yes YesR2 0.163 0.224No of obs 858 1389

We find clear evidence of sample selection bias, as evidenced by thestatistical significance of the coefficient of the selectivity parameter (theinverseMill's ratio), in all stages except the start-up stage. Taking accountof the bias, we find results on the risk and returns of VC investments that

hile Stage 2 in Panel B uses only investments exited through IPOs; returns in Panel B areSection 3.2. The coefficients are estimated using Heckman with Cox model (proportionalross sectional dependence of the returns. Market is the excess market returns using MSCI

Hazard ratio

** 1.008** 1.005** 1.005** 1.0907*** 0.7002*** 0.5167** 0.686

Expansion stage Later stage

es Coeff P-values Coeff P-values

0.273** 0.028 0.212*** 0.0041.102*** 0.000 0.868** 0.033−0.419*** 0.000 0.235*** 0.000Yes YesYes Yes0.189 0.1413068 1459

Mean Median STD Max Min

Panel A: Annualized returns for 100% IPO exitsStart-up stage 2.056 0.544 9.139 51.267 −0.901Early stage 3.058 0.596 11.156 57.267 −0.910Expansion stage 4.594 0.367 16.017 61.267 −0.951Later stage 6.338 0.491 18.772 77.267 −0.931MSCI world 0.101 0.093 0.088 0.296 −0.118MSCI Europe 0.101 0.099 0.120 0.385 −0.157S&P 500 0.122 0.116 0.101 0.360 −0.093FTSE all shares 0.081 0.090 0.092 0.305 −0.122Risk free rate (US) 0.043 0.048 0.016 0.079 0.000Risk free rate (Europe) 0.061 0.055 0.019 0.147 0.033No of obs 6774

Panel B: Annualized returns for post-IPO exits at the end of the IPO lock-up periodStart-up stage 2.367 0.376 7.063 29.123 −0.897Early stage 1.611 0.297 6.173 27.244 −0.955Expansion stage 1.192 0.142 4.402 24.224 −0.949Later stage 0.965 0.111 3.453 21.390 −0.907MSCI world 0.115 0.103 0.090 0.307 −0.118MSCI Europe 0.119 0.112 0.129 0.391 −0.159S&P 500 0.140 0.121 0.109 0.367 −0.097FTSE all shares 0.104 0.090 0.107 0.319 −0.133Risk free rate (US) 0.045 0.049 0.017 0.058 0.000Risk free rate (Europe) 0.060 0.054 0.019 0.144 0.032No of obs 6774

23S. Espenlaub et al. / International Review of Financial Analysis 31 (2014) 13–24

are remarkably robust. The results for alpha (abnormal performance) andbeta (systematic risk) adjusted for dynamic selection bias (in Table 8) areremarkably similar to those adjusted for exit-route selection bias(in Table 7). While the start-up and early stages have higher alphas inthe dynamic selection model than in the exit-route selectionmodel (46 and 42% compared to 42 and 35%, respectively), thereverse holds for the expansion and later stages (27 and 21%compared to 28 and 24%, respectively). Nevertheless, both modelsindicate clearly that abnormal performance is lower, the later thestage of financing.

In terms of the systematic risk, measured by beta, we similarlyfind that both selection models show beta to drop continually fromthe start-up stage to the later stage. The estimated betas are between3 and 10% higher in the dynamic selection model than in the exit-route (Dahl) selection model. The beta for the start-up stage is1.316 in the exit-route model compared to 1.403 in the dynamicselection model. The later-stage beta is around 40% lower than thestart-up beta, at 0.797 in the exit route model and 0.868 in thedynamic selection model.

5. Conclusion

This study examines exit choices and the risk and returns of invest-ments backed by US–European cross-border syndicates. We control forgradual VC exit, post-IPO, and two types of sample selection bias: onedue to the endogenous choice of exit route, and another due to theendogenous timing of exits.

Overall, our study highlights some interesting similarities anddifferences between US VCs and US–European cross-border syn-dicates, in terms of investment behavior and performance. We findthat, compared to US VCs, US–European syndicates on averagecomprise younger VCs investing relatively more in later stages offinancing. Also, the size of the US–European cross-border syndicatesis double than that of US VC syndicates, exhibiting a higher tendencyto spread investment risk. The level of experience of the US–European syndicates tends to vary little across financing stages, incontrast to US-only syndicates where earlier stages attract moreexperienced VCs.

For our sample of US–European cross-border syndicates, we finda monotonically decreasing relationship between both abnormal per-formance (alpha) and systematic risk (beta), on the one hand, and thestage of VC financing, on the other hand. These findings are broadly inlinewith those of Cochrane (2005) for US VCs.We document significantevidence of sample selection bias. Nevertheless, our findings on therisk and returns of VC investments are robust to controlling for twoalternative sources of sample selection bias. Our results are broadlycomparable to those reported by Cochrane (2005) for investmentsby US VCs.

Appendix A

This table complements the results presented in Table 5. It showsmean, median, standard deviation, maximum and minimum values ofventure capital returns exited through IPO exits. As in Table 5, returnsare annualized and computed over the period 1990 through 2010.Unlike in Table 5, Panel A of this table reports returns calculated onthe assumption that the 100% of the VC holding is sold through theIPO (at the offer price); while Panel B reports the correspondingreturns on the assumption that the VC holding is sold immediatelyat the end of the IPO lock-up period. The rows entitled ‘Start-upstage’, ‘Early stage’, ‘Expansion stage’ and ‘Later stage’ show returnsfor the respective financing stages. ‘MSCI World’ is the annualizedMSCI world index returns. ‘MSCI Europe’ is the annualized MSCIEurope index returns. ‘S&P 500’ is the annualized S&P 500composite returns. ‘Risk free’ is the annualized risk-free rate forUS and Europe, respectively.

References

Amit, B. (2002). Syndication among venture capitalists: When and with whom. Researchin Finance, 19, 285–321.

Arberk, M., Filatotchev, I., & Wright, M. (2006). Venture capitalists, syndication andgovernance in initial public offerings. Small Business Economics, 26, 337–350.

Bent, N., Williams, K., & Gilber, E. (2004). The syndication of private equity investments inSouth Africa. Journal of Business Managements, 35, 39–48.

Black, S., & Gilson, J. (1998). VC and the structure of capital markets: Banks versus stockmarkets. Journal of Financial Economics, 47, 243–277.

Bourguignon, F., Fournier, M., & Gurgand, M. (2007). Selection bias corrections based onthe multinomial logit model: Monte Carlo comparisons. Journal of Economic Surveys,21, 175–205.

Brander, J., Amit, R., & Antweiler,W. (2002). Venture capital syndication: Improved ventureselection vs. the value added hypothesis. Journal of Economics & Management Strategy,11, 423–452.

Casamatta, C., & Haritchabalet, C. (2007). Experience, screening and syndication inventure capital investments. Journal of Financial Intermediation, 16(3), 368–398.

Chen, P., Baierl, T., & Kaplan, P. (2002). Venture capital and its role in strategicasset allocation. Journal of Portfolio Management, 28, 83–90.

Cochrane, J. (2005). The risk and return of venture capital. Journal of Financial Economics,75, 3–52.

Cox, D. (1972). Regression models and life-tables. Journal of Statistics Society, 34, 187–220.Cumming, D. (2008). Contracts and exits in venture capital finance. Review of Financial

Studies, 21, 1947–1982.Dahl, G. (2002). Mobility and the returns to education: Testing Roy model with multiple

markets. Econometrica, 70, 2364–2420.Dimov, D., & Milanov, H. (2009). The interplay of need and opportunity in venture capital

investment syndication. Journal of Business Venturing, 25, 331–348.Dubin, J., & McFadden, D. (1984). An econometric analysis of residential electric appliance

holdings and consumption. Econometrica, 52, 345–362.Fleming, G. (2004). Venture capital returns in Australia. Venture Capital, 6, 23–45.Giot, P., & Schwienbacher, A. (2007). IPOs, trade sales and liquidation: Modelling venture

capital exits using exits using survival analysis. Journal of Banking& Finance, 31, 697–702.Gompers, P. (1995). Optimal investment, monitoring and the staging of venture capital.

Journal of Finance, 50, 1461–1489.Gompers, P. (1996). Grandstanding in the venture capital industry. Journal of Financial

Economics, 42, 133–156.Gompers, P., & Lerner, J. (2000). Money chasing deals? The impact of fund inflows on

private equity valuations. Journal of Financial Economics, 55, 281–325.Guler, I., & Guillen, F. (2010). Institutions and the internationalization of US venture

capital firms. Journal of International Business Studies, 41, 185–205.Heckman, J. (1979). Sample selection bias as a specification error. Econometrica, 47, 153–161.Hege, U., Palomino, F., & Schwienbacher, A. (2008). Venture capital performance: The

disparity between Europe and the United States. Working paper.Hellmann, T., & Puri, M. (2002). Venture capital and the professionalization of start-up

firms: Empirical evidence. Journal of Finance, 57, 169–197.Kaplan, S., & Schoar, A. (2005). Private equity performance: Returns, persistent, and

capital flows. Journal of Finance, 60, 1791–1823.Kaplan, S., & Stromberg, P. (2003). Financial contracting theory meets the real world: An

empirical analysis of venture capital contracts. Review of Economic Studies, 70, 281–316.Kaplan, S., & Stromberg, P. (2004). Characteristics, contracts and actions: Evidence from

venture capitalists analyses. Journal of Finance, 59, 2173–2206.

24 S. Espenlaub et al. / International Review of Financial Analysis 31 (2014) 13–24

Korteweg, A., & Sorensen, M. (2010). Risk and return characteristics of venturecapital backed entrepreneurial companies. Review of Financial Studies, 10,3738–3772.

Lee, L. (1983). Generalized econometric models with selectivity. Econometrica, 5,507–512.

Lee, M., & Wahal, S. (2004). Grandstanding, certification and the underpricing of venturecapital backed IPOs. Journal of Financial Economics, 73, 375–407.

Lerner, J. (1994). The syndication of venture capital investments. Financial Management, 23,17–22.

Ljungqvist, A., & Richardson, M. (2003). The cash flow, return and risk characteristics ofprivate equity. Stern University working paper 9454.

Lockett, A., & Wright, M. (2001). The syndication of venture capital investments. TheInternational Journal of Management Science, 29, 375–390.

Lockett, A., & Wright, M. (2003). The structure and management of alliances:Syndication in the venture capital industry. Journal of Management Studies, 40,2073–2102.

Lu, Q., & Huang, P. (2010). The impact of liability of foreignness on internationalventure capital firms in Singapore. Asia Pacific Journal of Management, 27,81–91.

Manigart, S., Waele, K., Wright, M., Robbie, K., Desbrieres, P., Sapienz, H., et al. (2002).Determinants of required return in venture capital investments: A five-countrystudy. Journal of Business Venturing, 17, 291–312.

Meuleman,M., &Wright,M. (2011). Cross-border private equity syndication: Institutionalcontext and learning. Journal of Business Venturing, 26, 35–48.

Sahlman, W. (1990). The structure and governance of venture capital organization.Journal of Financial Economic, 27, 473–524.

Seppä, T., & Laamanen, T. (2001). Valuation of venture capital investments: Empiricalevidence. R&D Management, 31, 215–231.

Sharifzadeh, A., &Walz, U. (2012). Syndication of venture capital investments, governanceand contract design. In D. Cumming (Ed.), The Oxford handbook of venture capital.Oxford University Press.

Tykvova, T., & Schertler, A. (2011). Venture capital and internationalization. InternationalBusiness Review, 20, 423–439.

Woodward, S. (2004). Measuring risk and performance of private equity. Sand Hill workingpaper.

Wright, M., Pruthi, S., & Lockett, A. (2005). International venture capital research: Fromcross country comparison to crossing countries. International Journal of ManagementReview, 7, 135–166.