- Email: [email protected]
Information production within the venture capital market: Implications for economic growth and development
Accepted Manuscript Title: Information production within the venture capital market: implications for economic growth and development Author: Oghenovo Obrimah PII: DOI: Reference:
S0148-6195(16)30026-1 http://dx.doi.org/doi:10.1016/j.jeconbus.2016.05.003 JEB 5751
To appear in:
Journal of Economics and Business
Received date: Revised date: Accepted date:
25-5-2015 19-2-2016 26-5-2016
Please cite this article as: Oghenovo Obrimah, Information production within the venture capital market: implications for economic growth and development, (2016), http://dx.doi.org/10.1016/j.jeconbus.2016.05.003 This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
ip t
cr
Information production within the venture capital market:
us
implications for economic growth
an
and development Oghenovo Obrimah
d
M
February 19, 2016
te
Abstract
I …nd venture capitalists’(VCs’) information production activities help decrease
Ac ce p
uncertainty about asset valuations within public equity markets, resulting in price convergence within the cross-section of Initial Public O¤erings (IPOs) that are backed by di¤erent classes of VCs and price convergence between VC and non-VC backed IPOs. These …ndings provide evidence that venture capital …nancing can be a component of an e¤ective or e¢ cient price discovery process within public equity markets that yields bene…cial externalities in so far as the pricing of non-VC backed IPOs are concerned. Given a decrease in valuation uncertainty ultimately results in lower costs of capital for issuing …rms, larger stock markets, and improvements in market liquidity or e¢ ciency,
Department of Economics, Banking, and Finance, Babcock University, Ogun State Nigeria. I am grateful to journal editors and two anonymous reviewers for valuable comments. Thanks to participants at the 2013 Accounting and Finance Research Association Conference for valuable comments. Corresponding Author: Dr. Oghenovo Obrimah; Tel: 234 (0903) 920-2317; Primary E-mail: [email protected]; Secondary E-mail: [email protected]; Address: Babcock Business School, Babcock University PMB 21244 Lagos Nigeria.
1 Page 1 of 50
empirical …ndings provide evidence that VCs’information production activities are
induce e¢ cient allocation of resources within stock markets).
cr
JEL Classi…cation: G24
ip t
bene…cial for economic growth and economic development (price discovery processes that
Keywords: Venture Capital, Initial Returns, Innovation, information asymmetry, IPO,
Ac ce p
te
d
M
an
us
Economic Growth & Development
2 Page 2 of 50
I
Introduction
ip t
It is well established that market e¢ ciency - the extent to which prices re‡ect all publicly available information - is an important indicator of stock market development (see for
cr
example, Fama, 1991). In light of the importance of market e¢ ciency for stock market
us
development, economic agents that facilitate price discovery within stock markets contribute directly to stock market development and by extension to economic development. In Leland
an
and Pyle (1977), information production that induces price discovery is an important objective of …nancial intermediation in markets such as venture capital markets. In
M
Campbell and Kracaw (1980), …nancial intermediation is most credible when information generated by …nancial intermediaries (e.g. venture capitalists) is utilized in alternate markets,
d
with the stock market qualifying as an alternate market to the venture capital market.
te
Suppose as in Gompers (1996) we distinguish between ‘incumbent’and ‘new entrant’
Ac ce p
venture capitalists (VCs). If VCs’information production activities are bene…cial for price discovery within stock markets, it is reasonable to expect di¤erences in the pricing of IPOs backed by the two classes of VCs identi…ed in Gompers (1996) attenuate over time, resulting in price convergence. Empirical evidence that VCs’information production activities improve the pricing of IPOs that do not have venture capital backing (non-VC backed IPOs) is evidence, however, that VCs’activities yield bene…cial externalities within stock markets; that is, are directly bene…cial for price discovery, market e¢ ciency, stock market development, and economic development. Studies such as Lee and Wahal (2004), Megginson and Weiss (1991), and Belghitar and Dixon (2012) provide evidence of cross-sectional di¤erences in initial returns to VC and non-VC backed IPOs. Empirical evidence that
3 Page 3 of 50
…nancial market development, which includes stock market development, is an important
ip t
driver of economic growth is provided in studies such as King and Levine (1993), Luintel and Khan (1999), and Rajan and Zingales (1998).
cr
In this study, I examine whether venture capitalists’(VCs’) information production activities a¤ect the time series variability of initial returns to IPOs, resulting in price
us
convergence within public equity markets. In order to demonstrate the robustness of the
an
evidence on convergence of initial returns to IPOs, I show cross-sectional variation in initial returns to IPOs is consistent with the positive relation between initial returns to IPOs and
M
the severity of information asymmetry problems in studies such as Beatty and Ritter (1986), Ibbotson et al. (1994); Ritter (1984), or Rock (1986).
d
With respect to time series convergence, empirical results show, consistent with …ndings
te
in Gompers (1996), that initial returns to IPOs backed by new entrant VCs start o¤ systematically lower than initial returns that accrue to IPOs backed by incumbent VCs.
Ac ce p
Over the course of three periods (years), however, initial returns to IPOs backed by new entrant or incumbent VCs converge to some equilibrium level, such that there do not exist any signi…cant di¤erences in initial returns between the two groups of IPOs. These …ndings provide empirical validation for a three-year reputation development time frame in Nahata (2009), Krishnan and Masulis (2009), and Obrimah (2015). In comparisons of initial returns to VC and non-VC backed IPOs, I …nd di¤erences in initial returns between the two classes of IPOs converge but are not fully attenuated over a period of three venture capital …nancing cycles (9 years), with convergence shown to be induced by the arrival of new and favorable information about VC backed IPOs. Given stock markets cannot be e¢ cient in the absence of price discovery mechanisms or processes, the 4 Page 4 of 50
positive externality from VCs’information production activities - improvements in the
ip t
valuations of non-VC backed IPOs - is evidence that venture capital has a direct impact on market e¢ ciency (stock market development) and by extension, economic development. My
cr
…nding that the slowest convergence rates between initial returns to VC or non-VC backed IPOs relate to the innovation dimension of VCs’portfolio activities provides support for a
us
characterization of innovation as an important or unique facet of venture capital activity in
an
studies such as Aggarwal and Hsu (2013) or Lerner, Sorenson, and Stromberg (2011). Markets within which prices have converged are characterized by less severe valuation
M
uncertainty in relation to markets within which prices have yet to converge. In Huang (2008), uncertainty aversion (aversion to valuation uncertainty) results in slower economic
d
growth, but is mitigated by information production. In Brockman, Liebenberg, and Schutte
te
(2010) there exists a link between swings in information production and the incidence of business cycles. In light of …ndings in Huang (2008) and Brockman et al. (2010), empirical
Ac ce p
results in this study provide evidence that VCs’portfolio activities have bene…cial e¤ects on uncertainty aversion within stock markets. Studies such as Loughran and Ritter (2004), Kirkulak and Davis (2004), Dimovski, Philavanh, and Brooks (2011), Chemmanur and Fulghieri (1994), and Megginson and Weiss (1991) provide mixed evidence on relations between initial returns to IPOs and market reputation. In this study, the formal theory I develop for the assessment of interactions between market reputation and initial returns to IPOs predicts market reputation can be associated with relatively higher initial returns to IPOs whenever the arrival of new innovative projects induces reputable intermediaries to bring their superior information production skills to bear on the pricing of new innovations. Comparisons of time series 5 Page 5 of 50
trends in initial returns to VC or non-VC backed IPOs provide support for the predictions of
ip t
the formal theory. Signalling theories (see for example, Allen and Faulhaber, 1989; Grinblatt and Hwang,
cr
1989; Welch, 1989), information asymmetry theories, and litigation risk theories of IPO
underpricing (Tinic, 1988; Alexander, 1993; Drake and Vetsuypens, 1993; Lowry and Shu,
us
2010) all imply di¤erences in initial returns between high risk and low risk IPOs do not
an
attenuate over time. My …nding that initial returns to IPOs converge fastest among the most innovative ventures, resulting in initial returns that are comparable to those which
M
accrue to the least innovative ventures provides evidence that independent of bene…cial e¤ects of information production, neither of information asymmetry, signalling, nor litigation
d
risk theories are well adapted to explaining time series variation in initial returns to IPOs.
te
My …ndings complement studies which provide evidence of direct, as opposed to indirect bene…ts from venture capital for economic growth (Total Factor Productivity) and
Ac ce p
development (Innovation or Patent Rates). These studies include Kortum and Lerner (2000) or Abdou and Varela (2009) and Chemmanur, Krishnan, and Nandy (2011), which …nd, respectively, that venture capital …nancing contributes to growth rates of innovation (patents) and improvements in Total Factor Productivity (TFP) of private …rms. Consistent with externalities from venture capital, Samila and Sorenson (2011) …nd the arrival of venture capital …nancing within a geographical area is associated with a signi…cant increase in non-venture capital backed entrepreneurial activity. Since the arrival of venture capital …nancing can itself be induced by improvements in arrival rates of good projects within opportunity sets, both venture capital and non-venture capital activity can be induced by the richness of the opportunity set, but with venture capital activity helping to re…ne information 6 Page 6 of 50
sets or prices. Empirical …ndings in this study provide evidence of a mechanism -
ip t
information production that induces re…nements in prices - via which the arrival of venture capital within an opportunity set can induce increases in investment activities not backed by
cr
venture capitalists.
The rest of the paper proceeds as follows. Section II develops the theoretical framework.
us
Section III discusses the matching methodology adopted in the implementation of the study.
an
Section IV describes data, empirical variables, and the structure of empirical tests. Section V reports results from empirical tests. Section VI reports results from robustness tests that
M
establish the directionality of externalities from information production. Section VII
Theoretical Structure
te
II
d
concludes the study.
Ac ce p
Venture capitalists specialize in information production via their investments in young or new ventures with little or no track record, income, or established products. Whenever VCs o¤er portfolio projects for sale in public equity markets for instance, investors gain insights into the sizes and distributions of opportunity sets that are generating new innovative assets within an economy. VCs’ability to grow portfolio projects into …rms that can be exited (sold to other investors) either via IPOs or third party acquisitions enables VCs to raise new capital and continue to produce information in their capacity as fund managers. In what follows, I develop a formal theory that generates testable predictions about time series trends in distribution of information asymmetry that obtain in the presence of heterogeneity or homogeneity in market reputation, with market reputation derived solely
7 Page 7 of 50
from information production. This generic approach to the formulation of the formal theory
ip t
ensures the model is robust to alternative proxies for the severity of information asymmetry between IPO …rms and investors. I leverage on theoretical and empirical evidence that
cr
establishes underpricing as a proxy for the severity of information asymmetry between
investors and IPO issuers to apply the predictions of the formal theory to time series trends
Information Production and Time Series Trends in Distribution of
an
A
us
in initial returns to IPOs.
M
Information Asymmetry
Suppose we assume information asymmetry with mean severity
d
information asymmetry among market participants
at some origin time t, with
and dispersion in
( t ) for …rms classi…ed within the t
the standard deviation of information
te
same general risk class
t
t
asymmetry. Since information production mitigates di¤erences in information or improves
on
Ac ce p
the precision of information, information production has a …rst order e¤ect on
via it’s e¤ect
and may or may not have a second order decreasing e¤ect on the average realization of
information asymmetry, .
In the presence of information production that mitigates information asymmetry within risk class
, we have
t
>
t+1
t+n
!
t
>
t+1
t+n ,
where n is the terminal future
period at which any additional information produced no longer has any signi…cant impact on the dispersion of information asymmetry (the terminal period at which di¤erences in information are no longer signi…cant). If information production merely induces a decrease in the dispersion of information asymmetry for …rms in risk class
,
t
t+1
t+n ,
implying mean severity of information asymmetry is time invariant and independent of
t
.
8 Page 8 of 50
Suppose, however, that
t
= g ( t ), then since the severity of information asymmetry
decreases in response to information production,
>
t+1
t+n
and
t+n
= g(
t+n )
imply
ip t
t
a well behaved g that is activated by information production is an increasing function of . can be divided into two groups
1
&
2
cr
Assume …rms in the same general risk class
based on the market reputation of IPO backers; with mean severity of information 1 t
2; t
&
and dispersion
1 t
2 t.
&
Studies such as Loughran and Ritter (2004),
us
asymmetry
an
Kirkulak and Davis (2004), Fernando, Gatchev, and Spindt (2005), Dimovski, Philavanh, and Brooks (2011), Chemmanur and Fulghieri (1994) or Megginson and Weiss (1991) show,
that either of
1 t
<
2 t
or
1 t
2 t
>
M
conditional on di¤erences in market reputation that are derived from information production, is feasible within the same general risk class or
that occur in response to information
d
the studies examine, however, transitions in
. None of
te
production by reputable intermediaries.
Propositions 1 through 3 generate equilibrium conditions expected to hold within the
Ac ce p
context of externalities from information production by …nancial intermediaries (IPO backers) that are similar or dissimilar in reputation.
Proposition 1 Let
1 t
&
2 t
denote mean realizations of information asymmetry (initial
returns to IPOs) for IPO …rms in groups t
= g ( t ) : Assume …rms in
1
&
2
1
&
2;
respectively, with
1 t
<
2 t
and
are backed by …nancial intermediaries that are similar
in market reputation in so far as the quality of information production is concerned; then equilibrium conditions that govern attenuation of information asymmetry are: 1 t
1 t+n
<0
t+n
future convergence date;
1 t+n t+n
; and
2 t+n
t+n
0<
2 t
2 t+n
where n > 1 is the
is the mean realization of information asymmetry for the
9 Page 9 of 50
universe of IPOs in
&
1
2;
and f is a continuous and invertible function of .
ip t
Proof. See Appendix. Proposition 1 states, in the presence of two classes of IPO …rms backed by …nancial
cr
intermediaries that do not di¤er with respect to market reputation, that equilibrium initial
us
returns to IPOs induced by information production are the outcome of complementarity in
The E¤ect of Di¤erences in Reputation of Financial Intermediaries
Proposition 2 Let
1 t
&
2 t
denote mean realizations of information asymmetry (initial
M
B
an
attenuation of information asymmetry between the two classes of …rms.
returns to IPOs) for IPO …rms in groups
1
&
2;
respectively, with
Assume …nancial intermediaries associated with …rms in
>
2 t
and
t
= g ( t ).
are more reputable than …nancial
d
1
1 t
intermediaries associated with …rms in
in so far as the quality of information production
te
2
Ac ce p
is concerned; then equilibrium conditions that govern convergence of initial returns to IPOs in the well de…ned scenario ( and
2 t
2 t+n
=j (
1 t
1 t
>
1 t+n )
2) t
1 t+n
are:
1; t
<
2 t+n
2; t
2 t
2 t+n
where n > 1 is the future convergence date;
realization of initial returns to the universe of IPOs in but non-invertible functions of (
<
1 t
1 ) t+n
or (
1 t
1
1 t+n );
&
2;
= j( t+n
1 t
1 ); t+n
is the mean
and j or j are increasing
respectively.
Proof. See Appendix.
Proposition 2 states, in the presence of …nancial intermediaries in reputable than …nancial intermediaries in
2,
1
that are more
that equilibrium initial returns to IPOs
decrease with information production. Convergence, however, is a function of changes in initial returns that accrue to …rms in
1,
as opposed to a function of the level of initial
10 Page 10 of 50
1 t
investors requires
>
2 t
1.
This is the case because the imposition of rationality on 1 t+n
implies
more reputable than those in
2,
2 . t+n
Given …nancial intermediaries in
any improvements in the risk pro…les of …rms in
2 t+n
= j(
1 t
1 ). t+n
In equilibrium, the pricing of …rms in
improvements in risk pro…les of …rms in
1
2
2
remain
are expected to induce changes in the
an
2
induce
can be altered by
even if the risk pro…les of …rms in
unaltered. Since changes in the pricing of …rms in
1
resulting in
us
2 t
2;
cr
demands on the risk pro…les, or equivalently, the pricing of …rms in
are
1
ip t
returns that accrue to …rms in
risk pro…les of these …rms, ultimately both the pricing and the risk of the …rms in
2
are
M
altered. In equilibrium, however, risk may lead pricing or pricing may induce changes in risk pro…les. 1 t
&
2 t
denote mean realizations of information asymmetry (initial
d
Proposition 3 Let
te
returns to IPOs) for IPO …rms in groups
1
&
2;
respectively, with
Ac ce p
Assume …nancial intermediaries associated with …rms in intermediaries associated with …rms in
2
1
1 t
<
2 t
and
t
= g ( t ).
are more reputable than …nancial
in so far as the quality of information production
is concerned; then equilibrium conditions that govern convergence of initial returns to IPOs are indeterminate. We can say, however, that 1 t+1
2 t+1
>0&
1 t+1
2 t+1
1 t
2 t
<0&
1 t
2 t
< 0 followed by
0 is evidence of a strategic information production
response.
Proof. See Appendix Proposition 3 predicts mixed empirical evidence on relations between initial returns to IPOs and market reputation can be explained by a strategic information production response from reputable intermediaries in contexts within which they currently back relatively low risk 11 Page 11 of 50
IPOs. Speci…cally, if reputable …nancing intermediaries currently associated with relatively
ip t
low risk projects bring their superior information production skills to bear on the production of information that relates to new innovative projects, information production can result in 1. t
cr
III
>
Matching Methodology
us
1 t+1
an
In line with discussions in Section II, comparisons of the evolution of information asymmetry that are credible must relate to …rms in the same risk class
, with the risk class su¢ ciently
M
rich to support dispersion in severity of information asymmetry problems. In this study, initial returns to IPOs are context-speci…c measures of the severity of information asymmetry
d
problems. In order to facilitate the interpretation of empirical results within the context of
te
equilibrium conditions established in Propositions 1 through 3, and as is evident in studies
Ac ce p
such as Gompers (1996) or Purnanandam and Swaminathan (2004), the matching of sample IPOs along speci…c risk dimensions is of critical importance. I implement two di¤erent sets of matching algorithms using sample IPO …rms. The …rst set of matching algorithms utilize IPO …rms brought to market by VCs that enter the venture capital market during adjacent years to generate two groups of …rms backed by …nancial intermediaries (VCs) that likely do not di¤er in so far as market reputation derived from information production is concerned. For ease of exposition, I refer to VCs that enter at time t as incumbent VCs, and VCs that enter at time t + 1 as new entrant VCs. Empirical results derived from these matched pairs of IPO …rms map to any of Propositions 1 through 3. Empirical …ndings in Megginson and Weiss (1991) indicate VCs contribute reputation
12 Page 12 of 50
during IPOs not attributable to IPO underwriters. The second set of matching algorithms
ip t
utilize VC and non-VC backed IPOs to generate two groups of …rms backed by …nancial intermediaries shown to di¤er either in market reputation (Megginson and Weiss, 1991) or
cr
the risk of IPO …rms (Lee and Wahal, 2004). Empirical results derived from these matched pairs of IPO …rms map to either of Propositions 2 or 3. Analyzed in juxtaposition with
us
Propositions 1 through 3, comparisons of empirical …ndings derived from the two sets of
A
an
matching algorithms serve as checks on the robustness of study conclusions.
Convergence and the number of matching dimensions
M
If IPO …rms are matched along k > m dimensions, the probability of convergence in initial
d
returns to IPOs increases as the number of dimensions increases from m to k. This increase
te
in probability of convergence with the number of matching variables implies studies of convergence yield the most credible inferences if IPO …rms are matched using a parsimonious
Ac ce p
set of matching variables. Since the probability of observing convergence increases with the number of matching variables, the most credible evidence in favor of convergence is derived from the most parsimonious number of matching dimensions.
B
Matching algorithm for VC backed IPOs
Venture Capitalists manage portfolios of projects and produce information within the context of portfolio management. In light of this, I focus on matches of IPO …rms that re‡ect VCs’ portfolio activities. This matching algorithm ensures empirical tests capture VCs’ information as derived from actual asset allocations as opposed to some arti…cial imposition of SIC code preferences. For robustness, I utilize several di¤erent measures of portfolio
13 Page 13 of 50
activities, which simultaneously double as measures of portfolio risk in the implementation of
ip t
the matching algorithm. These measures are: investment stage specialization Her…ndahls, industry specialization Her…ndahls, average syndicate size, and portfolio innovation focus. I
cr
describe the construction of each of these measures of portfolio risk in Section IV. Given matched pairs of VCs are adjacent to each other in so far as entry year into the venture
us
capital market is concerned and seek to exit portfolio projects via IPOs at the earliest time
an
possible (Gompers, 1996), in order to capture reputation e¤ects implicit in Propositions 1 through 3, I do not match new entrant and incumbent VCs’portfolios by IPO year.
M
Kaplan and Schoar (2005) and Schweinbacher (2008) …nd …nancing cycles occur every three years within the venture capital market. Nahata (2009), Krishnan and Masulis (2009),
d
and Obrimah (2015) utilize three-year cycles for measurement of VCs’market reputation.
te
Combined, these studies imply three-year reputation cycles within the venture capital market and indicate, independent of proxies for reputation, that activities of venture capitalists that
Ac ce p
do not enter the venture capital market during the same reputation cycle likely are not directly comparable. Empirical …ndings in Lerner (1994) that VCs with similar experience tend to syndicate together is consistent with the existence of reputation cycles within venture capital markets. In light of evidence of reputation cycles, I utilize a convergence time frame of three years in the implementation of the algorithm for matching new entrant and incumbent venture capitalists. Starting with VCs that enter the venture capital market at time t = 0, the implementation of the matching algorithm requires the matching of VCs that enter the market during time periods, t = 0 & t = 1, t = 1 & t = 2, and t = 2 & t = 3, resulting in three sets of matches that require four cohort VC years. Formally, let yj
2,
yj
1
& yj denote three consecutive years or periods bounded by four 14 Page 14 of 50
points in time (four cohort VC years), tj
3 , t j 2 , tj 1
& tj ; fj
3,
fj
2,
fj
1
and fj denote, 3,
nj
2,
nj
1
ip t
respectively, funds that commenced business during the four cohort years; nj
and nj denote, respectively, the number of funds that commenced business during the four
matches obtained for each pi by fj
3j
fj
2,
fj
2j
fj
1;
and fj
1j
cr
cohort years; while pi denote a measure of portfolio activity or risk. Denote adjacent
fj . Since new entrant funds
us
are matched to incumbent funds, the number of matched pairs are determined by nj
an
and nj , respectively. This implies the number of matched pairs in cohort fj unique and di¤erent from the number of matched pairs in cohort fj 1
6= nj . This matching strategy and the fact that nj
matches fj
3j
fj
2,
fj
2j
fj
1;
and fj
1j
1
is
fj whenever
d
the inclusion of fj
2
in fj
3j
fj
2, 2
while matches in
and fj
2j
fj
1
te
1,
1
1;
fj do not consist of exactly the same funds or
number of funds. Since matches in period one are determined by fj period two are determined by fj
fj
nj
6= nj for all j ensures adjacent
M
nj
1j
2j
2;
decreases the probability of convergence in initial returns to IPOs across periods one and
Ac ce p
two. Similar arguments apply to the probability of convergence across periods two and three and demonstrate the matching algorithm is not biased towards producing evidence of convergence in initial returns to VC backed IPOs.
C
The matching of non-VC and VC backed IPOs
While the literature on the pricing of IPOs that have or do not have venture capital backing yields mixed results in so far as the directionality of relations between initial returns to VC and non-VC backed IPOs are concerned, all studies conclude VC backing has certi…cation value during IPOs. These studies include Barry, Muscarella, Peavy, and Vetsuypens (1990), Megginson and Weiss (1991), and Lee and Wahal (2004). In light of the fact that non-VC 15 Page 15 of 50
backed IPOs are not necessarily brought to market within the context of portfolio
ip t
management, I match non-VC backed IPOs to matched pairs of VC backed IPOs derived from the matching algorithm in the preceding sub-section using 1-digit industry SIC codes
cr
and the year in which an IPO occurred. The restriction of industry classi…cations to the 1-digit SIC code is necessary because the VC backed IPOs are not as numerous as the
us
non-VC backed IPOs. In the absence of any direct competition e¤ects, and given substantial
an
yearly variability in initial returns to IPOs, the introduction of IPO year as a matching variable is of paramount importance. While the utilization of only two variables - the
M
industry classi…cation and issue year - as matching variables results in a rather parsimonious matching algorithm, as discussed in Section III.A, empirical evidence of intertemporal price
Data, Variables, and Empirical Structure
Ac ce p
IV
te
d
convergence becomes less robust with the number of matching variables.
Empirical evidence that levels of initial returns change signi…cantly over time in studies, such as Ritter (1984), Lowry and Schwert (2002), or Pastor and Veronesi (2005) imply empirical results for one three-year cycle cannot be overlaid on results for some other three-year cycle to arrive at average time series e¤ects. That is, results obtained from matches of VCs that enter the venture capital market between 1981 through 1984 cannot be combined with results obtained from matches based on data from 1985 through 1988 because the levels of initial returns to IPOs during these periods can be very di¤erent. In order to generate time series inferences, I limit the matching methodology in the preceding sub-section to funds that commenced business within four consecutive years only, commencing with the …rst feasible
16 Page 16 of 50
year in the sample; that is, 1981. The limitation to four consecutive years results in sample
ip t
VCs commencing business in 1981 (the base year), 1982, 1983, or 1984. Given data from about 1980 is regarded as the earliest year with robust coverage of the venture capital market
cr
in VentureXpert (Gompers and Lerner, 2004), the empirical structure utilizes the longest
A
us
series of IPO activity available in the study.
Constraints on sample of non-VC backed IPOs
an
If VCs’information production induce externalities for non-VC backed IPOs, this inference cannot be restricted to VCs that commenced business between 1981 and 1984. In light of
M
restrictions discussed in the preceding sub-section and in order to provide evidence from a
d
di¤erent set of funds, I compare initial returns that accrue to IPOs backed by VCs which
te
commenced business in 1986 or 1987 with initial returns that accrue to comparable IPOs that
B
Ac ce p
do not have venture capital backing.
Measures of VCs’Portfolio Activities
I construct four measures of VCs’portfolio activities (risk) that are motivated within the venture capital literature. These portfolio risk measures are: investment stage specialization Her…ndahls, industry specialization Her…ndahls, average syndicate size, and portfolio innovation focus.
B.1
Portfolio Specialization Her…ndahls
Fulgheiri and Sevilir (2009), Gompers, Kovner, and Lerner (2009), and Knill (2009) all link VCs’investment performance with portfolio specialization, with specialization typically
17 Page 17 of 50
measured as industry or investment stage specialization. I construct investment stage
ip t
specialization Her…ndahls (Her…ndahl-Hirschman) as the sum of squares of the percentage of total investments in (i) start-up, (ii) early stage, and (iii) expansion stage ventures, with
cr
classi…cations as indicated by the VentureXpert Database of SDC Platinum.
I construct industry specialization Her…ndahls as
us
Industry Specialization Her…ndahls
the sum of squares of the percentage of total investments in the following three broad
an
industries: (1) High Technology; (2) Medical/Health Related; and (3) Non-high Technology. The term “high technology” refers to industries that are not technologically intensive within
M
the context of web based or internet related businesses. All classi…cations are as indicated by
Syndicate Size
te
B.2
d
VentureXpert.
Ac ce p
Hochberg et al. (2007) …nd VCs’exit performance is an increasing function of the size of their investment networks. Cassamatta and Haritchabalet (2007) develop a theoretical model within which syndicate size is a proxy for the size of VCs’investment networks. I construct syndicate size as the average number of VC …rms that participate in each …rst round …nancing deal that ultimately results in an IPO.
B.3
Portfolio Innovation Focus
Measures of portfolio specialization or syndication activity are not peculiar to the venture capital market in so far as attempts at the benchmarking of portfolio risk are concerned. These measures of portfolio risk also are utilized in studies of banking activity (see for
18 Page 18 of 50
example, Berger and Hannan, 1989; Esty and Megginson, 2003). Obrimah (2015) develops a
ip t
measure of portfolio risk that is speci…c to the venture capital market and shown to be a better predictor of market reputation than portfolio specialization or syndication activity
cr
variables. This measure of portfolio activity captures the expectation or …ndings in studies such as Kortum and Lerner (2000) that VCs specialize in the …nancing of innovation. This
us
variable is termed “portfolio innovation focus”. For a discussion of theoretical rationales for
The construction of innovation focus
an
this measure of the risk of VCs’portfolio activities, see Obrimah (2015).
As is the case in Obrimah (2015), I normalize
M
VCs’capital disbursements to projects by adjusting capital disbursements for changes in in‡ation using GDP de‡ators from the World Development Indicators database that is
d
managed by the World Bank. I divide normalized capital disbursements for the entire
te
sample into percentiles and de…ne median innovation ventures as projects whose capital
Ac ce p
disbursements lie between the 25th and 75th percentile of capital disbursements; low innovation ventures as projects whose capital disbursements lie above the 75th percentile of capital disbursements; and high innovation ventures as projects whose capital disbursements lie below the 25th percentile of capital disbursements. I utilize the proportion of total capital disbursements that accrue to low, median, or high innovative ventures as measures of the extent to which these VCs’portfolio activities are focused on innovation. This construction of portfolio innovation focus is consistent with the prediction in Madan and Yen (2004) that the dollar amount invested in an asset is positively related to expected returns, negatively related to the variance of expected returns, and positively related to the skewness of expected returns, providing evidence of an investment decision rule or choice function that maps dollar
19 Page 19 of 50
amounts invested in portfolio projects to the moments of portfolio expected returns.
Data
ip t
C
cr
The venture capital data consist of funds that commenced operations between 1980 and 2005, but which are not a¢ liated with other …nancial or non-…nancial institutions.1 This
us
restriction ensures VCs within the sample are producing information for investors, as opposed to producing information for their parent or a¢ liated …rms. I refer to these VCs as
an
“independent VCs”. In order to focus on the quality of information (deal screening information) as opposed to the quantity of information produced (project management or
M
monitoring information), I construct measures of VCs’portfolio activities using capital
d
disbursed during …rst …nancing rounds only.
te
I restrict the venture capital data to include: transactions where the identities of the funds providing …nancing are disclosed along with the amounts invested; venture capital
Ac ce p
funds with fund size greater than US$5 million. The venture capital investments data (including initial returns) are obtained from VentureXpert, while data on initial returns to IPOs for non-VC backed IPOs are from the equity issues database of SDC Platinum. I report summary statistics for the funds included in my data sample in Table I, which show measures of VCs’portfolio activities (industry specialization, investment stage specialization, and portfolio innovation focus) are signi…cantly correlated at the 1% con…dence level. The positive correlation between the proportion of all projects that are high innovation ventures and either of industry or investment stage specialization Her…ndahls is evidence that 1
Gompers and Lerner (2004) provide rationales for the exclusion of pre-1980 data from empirical studies of venture capital.
20 Page 20 of 50
innovation focus is a measure of portfolio risk within the venture capital market.
Empirical Results: Information Production and
us
Convergence of Initial Returns to IPOs
cr
V
ip t
Insert Table I About Here
an
In this section, I report empirical results that examine the e¤ects of VCs’information production activities on the time series evolution of initial returns to IPOs over time.
M
Empirical results in this section are organized as follows. In Table II, I compare initial returns that accrue to IPOs backed by new entrant or incumbent VCs using my entire sample
d
of venture capital funds; that is, funds that commenced business between 1980 and 2005.
te
These empirical results demonstrate that cross-sectional variations in initial returns to VC backed IPOs are explained by the information asymmetry rationale for IPO underpricing
Ac ce p
proposed in Rock (1986).
In Table III, I compare three-year trends in initial returns that accrue to IPOs backed by new entrant or incumbent VCs. Since two di¤erent groups of VCs either are identical or dissimilar in so far as market reputation is concerned, these empirical tests generate results that can be interpreted within the contexts of Propositions 1, 2, or 3. In Table IV, I utilize a matched sample of VC and non-VC backed IPOs to test for price convergence between VC and non-VC backed IPOs. In light of empirical evidence of di¤erences in the market reputation of VC and non-VC backers of IPOs in Megginson and Weiss (1991) and di¤erences in levels of initial returns to IPOs in Lee and Wahal (2004), empirical tests in Table IV generate results that can be interpreted within the context of 21 Page 21 of 50
Propositions 2 or 3 only. In Tables V and VI (Section VI), I provide empirical evidence that
ip t
price convergence observed in Table IV can be attributed to the arrival of new favorable information about VC backed IPOs.
The Cross-section of VC-backed IPOs
cr
A
portfolio activity at a time. Let
n
and
e
us
The matching algorithm matches new and incumbent VCs relative to only one measure of denote, respectively, average initial returns to
(Ho) in Tables II and III is: Ho :
n
e
an
IPOs backed by new entrant or incumbent VCs, then the null hypothesis for each time period = 0. Let
n
and
e
denote, respectively, the
M
distributions of initial returns to IPOs backed by new entrant or incumbent VCs, then the 'e = 0; that is,
d
test for equality of distributions in the same tables examines whether 'n
te
examines whether there exist signi…cant di¤erences in dispersion of initial returns to IPOs between new and incumbent VCs’portfolios. These tests for equality of distributions are
Ac ce p
implemented using the Wilcoxon signed rank test. In light of the validation of industry or investment stage specialization and syndication activity as measures of the risk of VCs’portfolios in prior studies, empirical tests reported in Table II focus on the validation of portfolio innovation focus as a measure of the severity of information asymmetry problems in so far as VCs’portfolios are concerned. Empirical results in Table II show initial returns to IPOs range between 28:65% for portfolios characterized by low innovation focus to 37:49% for portfolios characterized by high innovation focus. Given the data in this study includes IPOs that occurred between 2001 and 2005, these levels of initial returns to IPOs compare favorably with average levels of 26:82% in Lee and Wahal (2004) for IPOs that occurred between 1980 and 2000. 22 Page 22 of 50
A comparison of initial returns that accrue to new entrant funds shows initial returns
ip t
increase with innovation focus from 29:14% to 32:25%, and then to 36:72 percent. Among incumbent funds, initial returns increase from 28:15% to 35:31%, and then to 38:26% as
cr
innovation focus increases from low to high innovation focus. Since portfolio risk increases from low to high innovation focus, these trends in initial returns to IPOs provide evidence
us
that portfolio innovation focus is a measure of the severity of information asymmetry
an
problems between VC backed issuers of IPOs and investors.
Empirical results generated by the Wilcoxon signed rank tests show the distributions of
M
initial returns that accrue to IPOs backed by new entrant or incumbent VCs di¤er signi…cantly. This …nding indicates new entrant and incumbent VCs’portfolio companies
d
have unique innovation characteristics that are re‡ected in initial returns during IPOs.
te
Combined, empirical …ndings in Tables I and II provide evidence that portfolio innovation focus is a good measure of portfolio risk within the cross-section of the venture capital
Ac ce p
market and a good measure of the severity of information asymmetry between VC backed issuers and investors.
Insert Table Il About Here
B
Time trends in initial returns to IPOs
I examine trends in initial returns that accrue to funds which commenced business between 1981 and 1984. As discussed in Section III, the choice of a three-year convergence time frame is predicated on the utilization of three-year time frames for measuring VCs’ reputation in Nahata (2009), Krishnan and Masulis (2011), and Obrimah (2015). Given the venture capital industry in the USA was still quite young in the early 1980s, and the 23 Page 23 of 50
expectation that the market power or in‡uence of VCs grows over time (see for example,
ip t
Nahata (2009)), the choice of the “…rst three years” in my sample for the examination of time trends in initial returns provides the lowest probability of observing price convergence in
cr
initial returns to IPOs.
In Table III, I report three-year trends in initial returns to IPOs. Since average initial
us
returns to IPOs associated with funds that commenced business between 1981 and 1984
an
cannot be constrained to be equal to average initial returns over the entire sample period of 1980 through 2005, the levels of initial returns in Table III are not comparable to those in
M
Table II. In discussing the results in Table III then, only time trends in initial returns are important.
d
The results in Panels A, B, and C of Table III are for time trends in initial returns to
te
IPOs for portfolios matched relative to portfolio specialization or syndicate size. Regardless of the speci…c portfolio activity, the results show initial returns that accrue to IPOs backed
Ac ce p
by new or incumbent VCs converge over time. The fastest convergence rate is relative to industry specialization, while the slowest convergence rate is relative to syndicate size. Viewed within the context of the formal theory in Section II, empirical results in Panels A through C are consistent with equilibrium conditions established in Proposition 1. This is evident in trends within Panel A for instance, which show that while average initial returns remain relatively stable, with realizations between 18:42% (Period 3) and 19:73% (Period 2), initial returns for new entrant VCs decrease from a high of 24:09% to 18:66%, and initial returns for incumbent VCs increase from a low of 13:01% to 18:17% between periods one and three. Empirical results in Panel A show equilibrium conditions, 1 t
1 t+n
<0
t+n
1 t+n
; and
2 t+n
t+n
0<
2 t
2 t+n
derived in
24 Page 24 of 50
Proposition 1, which result in the relations, (13:01 18:66
0 < (24:09
18:66+18:17 2
18:17 ; and
18:66) are satis…ed. These …ndings demonstrate, in so
ip t
18:66+18:17 2
18:17) < 0
far as information production activities are concerned, that new entrant and incumbent VCs
cr
are similar in market reputation.
Consistent with the prediction in Proposition 1 that di¤erences in dispersion of initial
us
returns converge along with mean realizations of initial returns, Wilcoxon signed rank tests
an
show di¤erences in distributions fully converge alongside di¤erences in mean realizations of initial returns to IPOs by the end of Period Three. Given convergence of distributions is a
M
necessary condition within the context of Proposition 1, the absence of full convergence in distributions by the second year and the presence of full convergence in distributions by the
te
within venture capital markets.
d
third year provides empirical validation for a three-year reputation development time frame
Ac ce p
Insert Table IlI About Here
In Panels D, E, and F of Table III, I report time trends in initial returns relative to the three di¤erent groupings of portfolio innovation focus. The results show initial returns that accrue to IPOs backed by new entrant or incumbent VCs converge over time. In Panels E and F, convergence results in di¤erences in initial returns that are economically and statistically insigni…cant within one period. These di¤erences decrease from economically signi…cant levels of between 9:66% and 10:47% in Panels E and F (Period One) to become economically and statistically insigni…cant by Period Three. In Panel D, while di¤erences in initial returns remain signi…cant during Period Three, they attenuate signi…cantly from an absolute value of 5:46% during the second period to an absolute value of 2:02% during the 25 Page 25 of 50
third period.
ip t
With respect to the distribution of initial returns, Wilcoxon signed rank tests in Panels D, E, and F show that while the distributions of initial returns that accrue to new entrant or
cr
incumbent VCs start o¤ statistically di¤erent, these di¤erences are totally attenuated by
Period 3. This …nding indicates the unique components of VCs’innovations get priced over
us
time, such that new entrant and incumbent VCs’innovations do not appear any di¤erent
an
from each other within three periods. The pricing of the unique characteristics of VCs’ innovations within three time periods ensures VCs have to continue to innovate in order to
M
earn innovation rents from their portfolio companies.
As is the case in Panels A through C, empirical results in Panels D through F are
d
consistent with equilibrium conditions established in Proposition 1. While the distribution of
te
di¤erences in initial returns reveals some divergence between Periods 1 and 2 in Panels A through C, however, the distribution of di¤erences in initial returns converges between
Ac ce p
Periods 1 and 2, and fully converges by Period 3 in Panels D through F. These …ndings indicate information produced by VCs relates more to the innovation dimension of portfolio …rms, as opposed to the diversi…cation bene…ts of portfolio …rms, with diversi…cation re‡ected in investment stage or industry specialization and syndication activity. Consistent with these inferences, a comparison of empirical results in Panels A through F shows initial returns to IPOs converge the fastest in empirical tests within which portfolio activity is de…ned relative to median or high innovation focus. In light of the demonstrated importance of diversi…cation for viability of commercial banking (see for example, Diamond (1984)), empirical results provide evidence that relative to diversi…cation strategies, innovation focus is a more important driver of portfolio asset allocations or portfolio management within 26 Page 26 of 50
venture capital markets.
The universe of IPOs
ip t
C
cr
Empirical tests in this sub-section examine three-year time trends in initial returns to IPOs over three fund-raising or investment cycles for an aggregate time period of 9 years. Studies
us
that …nd VC backed IPOs have lower initial returns in relation to non-VC backed IPOs include Megginson and Weiss (1991) and Belghitar and Dixon (2012). Studies which provide
an
evidence that VC backed IPOs have higher initial returns in relation to non-VC backed IPOs include Lee and Wahal (2004). Studies that provide mixed evidence on relations between
M
underwriter reputation and initial returns to IPOs include Loughran and Ritter (2004),
d
Kirkulak and Davis (2004), Dimovski, Philavanh, and Brooks (2011), and Chemmanur and
te
Fulghieri (1994). In light of these …ndings and theoretical predictions in Proposition 3, ‡ips in relations between initial returns to VC or non-VC backed IPOs in Table IV are consistent
Ac ce p
with outcomes of information production.
Panel A of Table IV reports empirical results from the matching of non-VC backed IPOs to VC backed IPOs within the same industry, regardless of whether the VC backed IPOs are associated with new entrant or incumbent VC funds. Empirical results show initial returns to VC backed IPOs start o¤ much lower than initial returns to non-VC backed IPOs, with the starting di¤erential amounting to about 97% of initial returns to non-VC backed IPOs during the …rst period (1987 through 1989). During the second period (1990 through 1992), the di¤erence in initial returns to IPOs between the two groups attenuates to 32 percent. By the third period under consideration (1993 through 1995), the di¤erence amounts to only 7.40%. Viewed within the context of Proposition 3, empirical results in Panel A provide 27 Page 27 of 50
evidence of a strategic information production response on the part of VCs between the …rst
ip t
and second periods, and continuation of information production between the second and third periods. Speci…cally, while transitions between Period 1 and Period 2 satisfy the
1 t+1
2 t+1
1 t
2 t
>0&
<0&
1 t
1 t+1
2 t+1
2 t
< 0 followed by
cr
condition, “
0 ” in Proposition 3, and while transitions between
or
between Period 2 and Period 3 are not
an
asymmetry problems, transitions in either of
us
Period 2 and Period 3 reveal evidence of information production that mitigates information
well behaved. Consistent with continuation of information production between Period and
during Period 2 to 8.28 during Period 3.
M
Period 3, however, the di¤erence in distributions attenuates by about 20:01% from 10.36
d
Empirical results in Panels B through G are for IPOs matched by industry 1-digit SIC
te
codes and year of IPO, but with the resulting matches reported using segmentations that relate to the di¤erent measures of portfolio risk; that is using the industry specialization
Ac ce p
(Panel B), investment stage specialization (Panel C), syndicate size (Panel D), low innovation (Panel E), median innovation (Panel F), and high innovation (Panel G) characteristics of the comparable venture capital portfolios. Given the segmentation is only possible with the venture capital backed IPOs, average initial returns to non-VC backed IPOs remain the same in each period, regardless of the measure of portfolio risk. The results in Panels B through G provide evidence of convergence in initial returns to IPOs in relation to all of the di¤erent facets of portfolio activities. Viewed within the context of conditions in Propositions 3, convergence in initial returns to IPOs evidences a strategic response between periods one and two and an information production response between periods two and three only in relation to median or high innovation focus and 28 Page 28 of 50
syndication activity. This …nding implies trends in Panel A for the entire sample of VC
ip t
backed IPOs largely re‡ect the innovation dimensions of VCs’portfolios. These …ndings are consistent with empirical results in Table III, which show information production induces the
cr
fastest convergence speeds with respect to the median or high innovation focus dimensions of portfolio activities. Combined, empirical …ndings in Tables III and IV provide evidence that
us
portfolio innovation focus is an important dimension for understanding VCs’portfolio
management of portfolio risk-return trade-o¤s.
an
activities, and implications of these activities for portfolio risk, portfolio strategies, and
The source of convergence in initial returns to VC and
d
VI
M
Insert Table IV About Here
te
non-VC backed IPOs
Ac ce p
Combined, empirical results in Tables III and IV provide evidence consistent with price convergence induced by venture capitalists’information production activities. If the observed convergence in initial returns to IPOs are induced by VCs’information production activities, alternative measures of IPO risk that are not simultaneously measures of the severity of information asymmetry between issuers and investors will indicate VC backed IPOs are less risky than their non-VC backed counterparts; that is, will indicate the steady increase in initial returns to VC backed IPOs in Table IV is directly attributable to VCs’ information production activities. In order to compare the relative riskiness of VC and non-VC backed IPOs, I utilize a regression framework to examine the relation between underwriting spreads, gross spreads, 29 Page 29 of 50
the proportional change between the mid…le price and the o¤er price, post-IPO price-to-book
ip t
values of sample IPOs, and di¤erences in initial returns between VC and non-VC backed IPOs. Underwriting or gross spreads have been shown to increase with IPO issuer risk in
cr
studies such as Altinkilic and Hansen (2000), Chemmanur and Fulghieri (1994), Chen and
Ritter (2000), and Eckbo, Masulis, and Norli (2007). The proportional change between the
us
mid…le price and o¤er price has been utilized as a measure of IPO issuer risk in Benveniste
an
and Spindt (1989) and Loughran and Ritter (2002). The conventional assumption in securities valuation that in the presence of …rms that are otherwise similar, …rms with lower
M
price-to-book ratios are better buys than those with higher price-to-book ratios implies post-IPO price-to-book ratios are good indicators of risk within stock markets, with risk an
d
increasing function of price-to-book ratios.
te
The regression models are speci…ed as:
0
+
0 initialrtnsprdi
+ "0i
(1a)
grossi =
1
+
1 initialrtnsprdi
+ "1i
(1b)
midtof f eri =
2
+
2 initialrtnsprdi
+ "2i
(1c)
p=bi =
3
+
3 initialrtnsprdi
+ "3i
(1d)
Ac ce p
writeri =
where writer are underwriting spreads; gross are gross spreads; midto¤ er are proportional changes between the mid…le and o¤er price; and p/b are price-to-book ratios for the venture capital backed IPOs.2 Let vc and nvc denote VC and non-VC backed IPOs, 2
Consistent with stylized evidence, gross spreads lie between 7.00% and 8.00% and exhibit signi…cant clustering evident in standard deviations that range between 1.56% and 1.80%. The proportional change between the mid-…le and o¤er price lies between -2.80% and 2.80%, and price-to-book ratios lie between 3.00 and 5.00.
30 Page 30 of 50
respectively, while initialrtn denotes initial returns to IPOs, then the variable initialrtnsprd
ip t
(initial return spread ) is de…ned as the di¤erence between initial returns to VC backed and
initialrtnnvc :
us
initialrtnsprd = initialrtnvc
cr
non-VC backed IPOs. That is,
The speci…cations of equations (1a) through (1d) re‡ect predictions or …ndings in Yeoman
an
(2001) and Tinic (1988) that underwriting spreads and initial returns to IPOs are jointly determined (during the book building process) prior to their realizations on the o¤er date of
M
an IPO. Given the jointness of determination, it is immaterial whether initial return spread is stipulated as a dependent or independent variable within a univariate regression framework.
d
Using the matched sample of VC and non-VC backed IPOs from Table IV, the results in
te
Panels A and B of Table V show - wherever signi…cant - that an increase in initial return
Ac ce p
spread is associated with a decrease in underwriting spreads, gross spreads, and post-IPO price-to-book ratios associated with VC backed IPOs. These negative relations demonstrate increases in initial returns to VC backed IPOs in Table IV, which are accompanied by shrinkage in initial returns spreads, are induced by increases in the risk of VC backed IPOs. This increase in the risk of VC backed IPOs is consistent with a strategic response from VCs in so far as the application of their information production skills are concerned. The book building theory of Benveniste and Spindt (1989) and the con‡ict of interest theory of Loughran and Ritter (2002) both predict the IPO o¤er price is not fully adjusted from the midpoint of the …le price range whenever underwriters receive favorable information These distributional characteristics are available from the author upon request.
31 Page 31 of 50
(see also, Welch and Ritter (2002)) regarding an IPO.
Consistent with the arrival of new
ip t
favorable information in respect of VC backed IPOs, the proportional change between the IPO mid…le and o¤er prices decreases with initial return spread. In so far as venture capital
cr
backed IPOs are concerned then, increases in initial returns to VC backed IPOs (in relation to initial returns that accrue to non-VC backed IPOs) are consistent with the arrival of new
us
favorable information about VC backed IPOs. Since the arrival of new favorable information
an
implies a decrease in severity of information asymmetry, empirical results in Table V provide evidence that convergence in initial returns in Table IV is predicated on VCs’information
M
production activities.
d
Insert Table V About Here
te
In Gompers (1996), IPOs backed by new entrant VCs attract higher initial returns to IPOs in relation to IPOs backed by incumbent VCs, with this phenomenon induced by
Ac ce p
“grandstanding” on the part of new entrant VCs. Empirical results that obtain during Period I of Table III are consistent with the grandstanding hypothesis in that initial returns that accrue to IPOs backed by new entrant VCs are larger than those that accrue to IPOs backed by incumbent VCs. While this …nding provides support for the cross-sectional predictions of the grandstanding hypothesis, increases in initial returns to IPOs backed by reputable VCs between Periods 1 and 2 in Table III and increases in initial returns to the universe of VC backed IPOs in Table IV simultaneously demonstrate the inapplicability of the grandstanding hypothesis to explanations of time series variability in underpricing of VC backed IPOs.
32 Page 32 of 50
VII
Conclusions
ip t
I …nd venture capitalists’(VCs’) information production activities induce time series convergence in initial returns that accrue to IPOs backed by new entrant or incumbent VCs
cr
within a three-year span. These …ndings provide empirical validation for the utilization of a
us
three-year time frame for assessing market reputation within the venture capital market in Nahata (2009), Krishnan and Masulis (2009), and Obrimah (2015). In empirical tests that
an
compare time series trends in initial returns to IPOs, I …nd VCs’information production activities induce convergence in initial returns that accrue to VC and non-VC backed IPOs,
M
with convergence directly attributable to the arrival of new and favorable information about VC backed IPOs. These …ndings provide empirical evidence that VCs’information
d
production activities induce improvements in the pricing of non-VC backed IPOs, indicating
te
VCs’information production activities induce price discovery within stock markets. Given
Ac ce p
stock prices are expected to be higher in markets characterized by price discovery processes, empirical …ndings show venture capital can have a direct impact on economic growth via it’s e¤ect on the sizes of stock markets. In light of the fact that market e¢ ciency requires price discovery processes and induces market development (more e¢ cient allocation of resources), empirical …ndings demonstrate VCs’information production activities are directly bene…cial for economic development.
33 Page 33 of 50
Appendix
1 t
2, t
=
1
&
2;
equivalence of reputation of …nancial intermediaries does not imply
and as such does not violate the assumption,
1 t
2. t
<
In the presence of
cr
…rms in
ip t
Proof of Proposition 1. Conditional on within-class di¤erences in initial risk pro…les of
1 t+n
that 1 t+n
2 t+n
2 t+n
<
<
1 t
1 t
2 t 2 t
a¤ects the pricing of …rms in
1
. Suppose
t
, implying (since
1 t
<
1
M
1 t+n
2 t+n
<
and
2 t+n
<
2. t
We have
2 t
(2)
d
t+n
1 t
>
is the mean realization of severity of information asymmetry for the universe
and
3 2.
Equation (2) implies,
Ac ce p
of …rms in
t+n
1 t+n
2) t
te
where
<
and vice-versa, such
= g ( t ) : It must be the case that
then that: 1 t
2,
an
produced in relation to …rms in
us
equivalence of reputation of …nancial intermediaries, it must be the case that information
1 t
1 t+n
t+n
2 t+n
< 0 0<
t+n 2 t
1 t+n
(3)
2 t+n
(4)
QED.
Proof of Proposition 2. Suppose …nancial intermediaries associated with …rms in more reputable than …nancial intermediaries associated with …rms in
2
1
are
in so far as the
3 In the presence of similarities in reputation, 1t < 2t can be interpreted as a strategic outcome. Suppose otherwise, then holding reputation …xed and assuming investor rationality, a strategic response results in the …nancial intermediary in 1 going after riskier …rms and the …nancial intermediary in 2 going after less risky 2 2 …rms. This results in 1t < 1t+n t+n t+n < t ; that is, exactly the same condition induced by information production.
34 Page 34 of 50
quality of information production is concerned. Then it must be the case that information
implies
1 t+n
=g
1 t+n
2 t+n
and
a¤ects the pricing of …rms in
1
1 t+n
=h
non-invertible function of . Since changes in = (
2 ) t+n
1 t+n
= v(h(g
1
but not vice-versa. This
); with h an increasing but
also can e¤ect
2,
cr
2 t+n
1 t+n
= h(g
2,
ip t
produced in relation to …rms in
)), with v a decreasing, but non-invertible function of . Given 2 t+n
1 t+n ,
1 t+n
implying the condition
2 t+n
need not 1 t
<
2 t
can be
an
necessarily converge towards
us
v is a decreasing function, while both h and g are increasing functions,
violated within a well behaved equilibrium.
M
In light of evidence that higher reputation …nancial intermediaries can at di¤erent points in time be associated with …rms characterized by either of lower or higher severity of
d
information asymmetry (see for example, Loughran and Ritter, 2004; Kirkulak and Davis,
necessarily imply
1 t
<
2. t
te
2004; and Dimovski et al., 2011), more reputable …nancial intermediaries in Suppose
1 t
1 t+n
<
2 t+n
Ac ce p
information production induces <
2 t.
However, since
1 t+n
<
1 t
>
1 t.
and
2. t t
1
does not
Then in the well behaved scenario, 1 t+n
= g ( t ) implies
2 t+n
<
2 t
imply
1 t
>
<
t+n
1 t
!
and
2 t
1 t
2 t+n
<
>
t+n
> 2 t
2 t
and
!
equilibrium conditions in equations (3) and (4) are violated and the only inviolate conditions are:
i t+n
<
i t
8i
(5)
i t+n
<
i t
8i
(6)
2 t
2 t+n
= j(
2 t
2 t+n
= j (
1 t 1 t
1 t+n ) 1 t+n )
(7) (8)
35 Page 35 of 50
where j and j are increasing but non-invertible functions of ( 1
1 ) t+n
or (
1 t
1 t+n ).
are more reputable than …nancial intermediaries in
implies only …nancial intermediaries in
1
can credibly switch strategies. Since 1 t+n
>
2; t
2. t
…nancial intermediary can result in either of
Strategic positioning by the more reputable 1 t+n
1 t
1 t+n
an
<
us
QED. 1 t
2, t
>
that is, exactly the same equilibrium
conditions in equations (5) through (8).
Proof of Proposition 3. Suppose
2 t
>
cr
1 t
however, a switch in strategy leads to
1 t
2, 1 t
Since
ip t
…nancial intermediaries in
1 t
<
(information production) or
>
1 t
M
(strategic application of superior information production capabilities to a riskier subset of …rms within the same general risk class), the number of equilibrium conditions that can be 1 t
<
2 t
1 t
are potentially in…nite. For concreteness,
<
2 t
d
generated within the context of
te
indicates investors are willing to tolerate risk more severe than risk currently borne by 1 t+n
reputable …nancial intermediaries. This implies
>
1 t
is feasible; that is, reputable
Ac ce p
…nancial intermediaries can take on higher risk projects to bring their reputation or superior information production skills to bear on the pricing of projects with information asymmetry 2. t
levels of
Since this action can induce less reputable …nancial intermediaries to switch
strategies in order to remain competitive; that is, induce a switch to projects with risk rating 1, t
1 t
<
we can observe a switch in ordering of initial returns to IPOs; that is,
1 t+n
2 t+n
2 t
<
(convergence) is just as feasible as
1 t+n
>
2 t
>
1 t
2 t+n
(divergence). In equilibrium then, we are unable to predict relations between or the location of equilibrium risk parameter, 1 t
2 t
<0&
1 t
2 t
t+n . 1 t+1
< 0 followed by
1 t+n
and
2 t+n
We can say, however, that 2 t+1
>0&
1 t+1
2 t+1
> 0 is
evidence of a strategic information production response. QED. 36 Page 36 of 50
References
ip t
Abdou, K., and O. Varela, 2009. Is there a puzzle in the failure of venture capital backed portfolio companies? Applied Financial Economics 19, 1439-1452.
cr
Aggarwal, V.A., and D.H. Hsu, 2013. Entrepreneurial exits and innovation, Management
us
Science 60, 867-887.
an
Alexander, J., 1993. The lawsuit avoidance theory of why Initial Public O¤erings are underpriced. UCLA Law Review 41, 17-73.
M
Allen, F., and G. Faulhaber, 1989. Signaling by underpricing in the IPO market, Journal of Financial Economics 23, 303-323.
te
d
Altinkilic, O., & Hansen, R.S., 2000. Are there economies of scales in underwriting fees? Evidence of raising external …nancing costs, Review of Financial Studies 13: 191–218.
Ac ce p
Barry, Christopher B., Muscarella, Chris J., Peavy, John W., and Michael R. Vetsuypens, 1990. The role of venture capital in the creation of public companies: evidence from the going-public process, Journal of Financial Economics 27, 447-71. Beatty, R., and J. Ritter, 1986. Investment banking, reputation, and the underpricing of initial public o¤erings, Journal of Financial Economics 15, 213-232. Belghitar, Y., and R. Dixon, 2012. Do venture capitalists reduce underpricing and underperformance of IPOs? Applied Financial Economics 22, 33-44. Benveniste, L.M., and P.A. Spindt, 1989. How investment bankers determine the o¤er price and allocation of new issues, Journal of Financial Economics 24, 343-362. 37 Page 37 of 50
Berger, A.N., and T.H. Hannan, 1989. The price-concentration relationship in Banking, The
ip t
Review of Economics and Statistics 71, 291-299. Brockman, P., Liebenberg, I., and M. Schutte, 2010. Comovement, information production,
cr
and the business cycle, Journal of Financial Economics 97, 107-129.
us
Campbell, T.S., and W.A. Kracaw, 1980. Information Production, Market Signalling, and the
an
Theory of Financial Intermediation, Journal of Finance 35, 863-882.
Casamatta, C., and C. Haritchabalet, 2007. Experience, screening, and syndication in venture
M
capital investments, Journal of Financial Intermediation 16, 368-398. Chemmanur, T.J., and Fulghieri, P., 1994. Investment Bank Reputation, Information
d
Production, and Financial Intermediation. Journal of Finance. 49: 57-79.
te
Chemmanur, T.J., Krishnan, K., and D.K. Nandy, 2011. How does venture capital …nancing
Ac ce p
improve e¢ ciency in private …rms? A look beneath the surface, Review of Financial Studies 24, 4037-4090.
Chen, H.C., and J.R. Ritter, 2000. The seven percent solution, Journal of Finance 55, 1105-1131.
Diamond, D., 1984. Financial Intermediation and Delegated Monitoring. Review of Economic Studies 51, 393-414. Dimovski, W., Philavanh, S. and R. Brooks, 2011. Underwriter reputation and underpricing: evidence from the Australian IPO market, Review of Quantitative Finance and Accounting 37, 409-426.
38 Page 38 of 50
Drake, P., Vetsuypens, M., 1993. IPO underpricing and insurance against legal liability
ip t
Financial Management 22, 64-73.. Eckbo, B.E., Masulis R.W., & Norli, O. 2007. Security o¤erings. In: E. Eckbo, Editor,
cr
Handbook of Corporate Finance: Empirical Corporate Finance, Elsevier, North-Holland,
us
Amsterdam.
Esty, B.C., and W.L. Megginson, 2003. Creditor rights, enforcement, and debt ownership
an
structure: Evidence from Global Syndicated Loan Market, Journal of Financial and
M
Quantitative Analysis 38, 37-59.
Fama, E.F., 1991. E¢ cient Capital Markets: II, Journal of Finance 46, 1575-1617.
d
Fernando C.S., Gatchev, V.A. and P. Spindt, 2005. Wanna dance? How …rms and
te
underwriters choose each other. Journal of Finance 60, 2437-2469.
Ac ce p
Fulghieri, P., and M. Sevilir, 2009. Size and focus of a venture capitalists’portfolio. Review of Financial Studies. 22, 4643-80.
Gompers, P., 1996. Grandstanding in the Venture Capital Industry. Journal of Financial Economics 42, 133-156.
Gompers, P., Kovner, and J. Lerner, 2009. Specialization and Success: Evidence from Venture Capital, Journal of Economics and Management Strategy 18, 817-844. Gompers, P., and J. Lerner, 2004. The venture capital cycle, 2nd edition, Cambridge: MIT Press.
39 Page 39 of 50
Grinblatt, M., and C. Hwang, 1989. Signaling and the pricing of new issues, Journal of
ip t
Finance 44, 393-420. Hochberg, Y., Ljungqvist, A., and Y. Lu, 2007. Whom you know matters: Venture capital
cr
networks and investment performance. Journal of Finance. 62, 251-301.
us
Huang, R.R., 2008. Tolerance for uncertainty and the growth of informationally opaque
an
industries, Journal of Development Economics 87, 333-353.
Ibbotson, R., Sindelar, J., Ritter, J., 1994. The market’s problems with the pricing of initial
M
public o¤erings. Journal of Applied Corporate Finance 7, 66–74.
King, Robert G., and Ross Levine, 1993. Finance and growth, Schumpeter might be right,
d
Quarterly Journal of Economics 108, 717-737.
te
Kirkulak, B., and C. Davis, 2004. Underwriter reputation and underpricing: Evidence from
Ac ce p
the Japanese IPO market, Paci…c-Basin Finance Journal 13, 451-470. Knill, A., 2009. Should venture capitalists put all their eggs in one basket? Diversi…cation versus pure play strategies in venture capital. Financial Management. 38, 441-486. Kortum, S. and J. Lerner, 2000. Assessing the contribution of venture capital to innovation, RAND Journal of Economics vol. 31(4), 674–92. Krishnan, C.N.V., and R.W. Masulis, 2011. Venture Capital Reputation: A Survey in “The Oxford Handbook of Venture Capital ”, Dougas J. Cummings, ed., Oxford University Press. Lee, P M., and S. Wahal, 2004. Grandstanding, certi…cation and the underpricing of venture capital backed IPOs, J ournal of Financial Economics 73, 375-407. 40 Page 40 of 50
Leland, H.E., and D.H. Pyle, 1977. Informational Asymmetries, Financial Structure, and
ip t
Financial Intermediation, Journal of Finance 32, 371-387. Lerner, J., 1994. The Syndication of Venture Capital Investments, Financial Management 23,
cr
16-27.
an
case of innovation. Journal of Finance 66(2): 445-477.
us
Lerner, J., M. Sorensen, P. Stromberg. 2011. Private equity and long-run investment: The
Loughran, T., and J.R. Ritter, 2002. Why don’t issuers get upset about leaving money on the
M
table in IPOs? Review of Financial Studies 15, 413-443.
Loughran, T., and J. Ritter, 2004. Why has IPO underpricing changed over time? Financial
d
Management 33, 5-37.
te
Lowry, M., and W.G. Schwert, 2002. IPO Market Cycles: Bubbles or Sequential Learning?,
Ac ce p
Journal of Finance 52, 1171-1200.
Lowry, M., and S. Shu, 2010. Litigation risk and IPO underpricing, Journal of Financial Economics 65, 309-335.
Luintel, K.B., and M. Khan, 1999. A quantitative reassessment of the …nance-growth nexus: evidence from a multivariate VAR, Journal of Development Economics 60, 381-405. Madan, D., and J. J. Yen, 2004. Asset allocation with multivariate non-Gaussian returns, Chapter 23 in Handbooks in operations research and management science: Financial Engineering 15, Eds. John Birge, and Vadim Linetsky, North-Holland.
41 Page 41 of 50
Megginson, William L., and Kathleen A. Weiss, 1991. Venture capitalist certi…cation in
ip t
initial public o¤erings, Journal of Finance 46, 879-903. Myers, S. C., 1977. “Determinants of Corporate Borrowing,” Journal of Financial Economics
cr
5, 147-175.
us
Nahata, R., 2009. Venture capital reputation and investment performance. Journal of
an
Financial Economics 90. 127-51.
Obrimah, O.A., 2015, How important is innovationn for venture capitalists’(VCs’) market
M
reputation, Quarterly Review of Economics and Finance, forthcoming, 10.1016/j.qref.2015.11.007.
d
Pastor, L., and P. Veronesi, 2005. Rational IPO waves, Journal of Finance 60, 1713-1757.
te
Purnanandam, A. K., and B. Swaminathan, 2004. “Are IPOs Really Underpriced?” Review
Ac ce p
of Financial Studies, 17, 811-848.
Rajan, Raghuram G., and Luigi Zingales, 1998. Financial Dependence and Growth, American Economic Review 88, 559-586. Ritter, J.R., 1984. The ‘hot issue market’of 1980, Journal of Business 57, 215-239. Rock, K., 1986. “Why New Issues are Underpriced,” Journal of Financial Economics, 15, 187-212. Samila, S., and O. Sorenson, 2011. Venture capital, entrepreneurship, and economic growth, Review of Economics and Statistics 93, 338-349.
42 Page 42 of 50
Schweinbacher, A., 2008. Innovation and venture capital exits, The Economic Journal 118,
ip t
1888-1916. Tinic, S., 1988. Anatomy Of initial public o¤erings of common stock, Journal of Finance 43,
cr
789-822.
us
Welch, I., 1989. Seasoned o¤erings, imitation costs and the underpricing of initial public
an
o¤erings, Journal of Finance 44, 421-449.
Yeoman, J.C., 2001. The optimal spread and o¤ering price for underwritten securities,
Ac ce p
te
d
M
Journal of Financial Economics 62, 169-198.
43 Page 43 of 50
44
Page 44 of 50
Table I: Summary statistics for VCs’portfolio activities
d
te
Ac ce p
us
an
M
cr
ip t
This table reports summary statistics for measures of portfolio activity utilized in the matching of new entrant and incumbent venture capital funds. Year is the year a fund commenced business; # of funds is the number of funds that commenced business in a particular year; industry and investment stage specialization are constructed using an Her…ndahl-Hirschman methodology; that is, as the sum of squares of an industry or investment stage’s share in aggregate capital disbursements to each fund’s portfolio companies. The industry and investment stage classi…cations are from VentureXpert and are (1) Non-hitech; (2) medical/health; and (3) hitech for industry classi…cations and (a) start-up; (b) early stage and (c) expansion stage for investment stage classi…cations. Syndicate Size is constructed as the average size of …nancing syndicates that involve a VC i. Each VC’s portfolio ventures are classi…ed as (i) low, (ii) median, or (iii) high innovation focus ventures using proportions of capital disbursements within the following percentiles: low innovation ventures lie within the top 25% of disbursements; median innovation ventures lie between the 25th and 75th percentiles of disbursements, while high innovation ventures lie in the bottom 25% of capital disbursements. The venture capital data are from VentureXpert and are for funds that commenced business between 1980 and 2005. Industry Investment Stage Innovation Focus Year # of funds Specializaton Her…ndahls Specialization Her…ndahls Low Median High Syndicate Size 1980 13 0.67 0.48 0.14 0.75 0.11 4.25 1981 29 0.70 0.35 0.00 0.69 0.31 4.67 1982 42 0.57 0.41 0.00 0.94 0.06 4.50 1983 53 0.54 0.47 0.24 0.65 0.11 3.27 1984 63 0.56 0.49 0.29 0.64 0.13 1.73 1985 47 0.73 0.63 0.45 0.49 0.06 4.30 1986 41 0.50 0.50 0.00 0.74 0.26 3.10 1987 66 0.91 0.66 0.43 0.54 0.03 2.50 1988 45 0.58 0.72 0.00 0.48 0.52 2.13 1989 53 1.00 0.45 0.00 0.91 0.09 2.73 1990 29 0.59 0.46 0.00 0.91 0.09 2.22 1991 22 0.59 0.39 0.00 0.77 0.23 1.50 1992 42 0.93 0.35 0.00 0.89 0.11 2.20 1993 50 1.00 1.00 0.00 1.00 0.00 2.00 1994 79 0.99 0.36 0.00 0.91 0.09 1.93 1995 87 0.51 0.77 0.00 1.00 0.00 3.00 1996 96 1.00 0.45 0.95 0.04 0.01 1.25 1997 148 1.00 0.43 0.76 0.24 0.00 2.00 1998 174 1.00 1.00 0.00 0.76 0.23 4.00 1999 241 1.00 0.69 0.68 0.32 0.00 3.60 2000 298 0.51 0.42 0.81 0.18 0.00 3.00 2001 thro 2005 957 0.73 0.60 0.84 0.15 0.00 2.71 Global Mean 0.71 0.54 0.27 0.59 0.14 2.64 c c c c Correlations with High Innovation 0.2486 0.1488 -0.8700 -0.4669 1.0000 0.0077 c indicates signi…cance at the 1% con…dence level
45
Page 45 of 50
Table II: Cross-sectional di¤erences in initial returns to IPOs
e
n
e
'e n
e
'n
'e
n
e
n
e
High Innovation 38.26c -1.535b (0.03) 9439
ip t
9439
36.72c
cr
-2.153b (0.03)
us
an
n
Median Innovation 35.31c -3.062c (0.00) 5085
e
M
Low Innovation Initial Returns 29.14c 28.15c 0.983 2.118b 32.25c (0.08) (0.05) # of observations 10575 10575 5085 b & c denote signi…cance at the 5% and 1% con…dence levels, respectively.
n
'n
d
te
Ac ce p
'e -2.006b (0.05)
'n
This table reports empirical results derived from comparisons of initial returns that accrue to IPOs backed by venture capital funds that commenced business between 1980 and 2005. The cross-sectional comparisons are implemented using matched samples of VC funds that commenced business during immediately adjacent years, with measures of the funds’portfolio activities (portfolio risk) as matching dimensions. The matching algorithm is implemented as follows. Let yj 2 , yj 1 & yj denote three consecutive years or periods bounded by four points in time, tj 3 , tj 2 , tj 1 & tj ; fj 3 , fj 2 , fj 1 and fj denote, respectively, funds that commenced business during the four cohort years; nj 3 , nj 2 , nj 1 & nj denote, respectively, the number of venture capital funds that commenced business during the four cohort years; while pi denote a measure of portfolio activity or risk. Denote adjacent matches generated during the three cohort years by fj 1 jf j , and fj jf j+1 .Since new entrant funds are matched to incumbent funds, the number of matches in the two cohorts are determined by nj 1 and nj , respectively. This implies the number of matches in each cohort is unique whenever nj 1 6= nj . Combined, the matching algorithm and the fact that nj 1 6= nj for all j ensure adjacent matches fj 1 jf j , and fj jf j+1 do not consist of the same funds from year cohort yj . The matching algorithm is implemented relative to one measure of portfolio risk at a time. The measures of portfolio risk in this table are: (A) Low Innovation; (B) MedianInnovation; and (C) High Innovation focus. I describe the construction of each of these measures of portfolio risk in Table I. The venture capital data (including initial returns) are from VentureXpert, while other data on initial returns to IPOs are from the equity issues database of SDC Platinum. Di¤erences in means ( n e ) are two-sided tests of signi…cant di¤erences between new entrant and established VCs. ('n 'e ) are estimates of the extent to which the distribution of initial returns to IPOs for the two matched samples are identical, and are implemented using the Wilcoxon signed rank test. p-values for statistical signi…cance are in parentheses below reported parameters. New Older New Older New Older Fundsy Fundsy 1 Fundsy Fundsy 1 Fundsy Fundsy 1
46
Page 46 of 50
d
te
e
n
e
'n
'e n
M n
e
Period 2 12.79c 9.548c (0.00) 207
Panel B: Di¤erence in Means: 3-year time trends for Investment Stage Specialization Period 1 Period 2 c c c b c Initial Returns 25.69 14.01 11.68 2.163 19.34 20.88c -1.537 (0.00) (0.03) (0.45) # of observations 271 271 628 628
Panel C: Di¤erence in Means: 3-year time trends for Syndicate Size Period 1 Initial Returns 24.54c 9.30c 15.24c 2.184b 22.33c (0.00) (0.03) # of observations 78 78 207 b & c denote signi…cance at the 5% and 1% con…dence levels, respectively.
'e
3.556b (0.00)
0.194 (0.85)
n
e
Period 3 15.16c 4.930c (0.00) 414
1197
1197
414
e
Period 3 0.816 (0.51) 18.31c
19.13c
20.09c
n
Period 3 18.17c 0.489 (0.60) 1906
ip t
1906
18.66c
cr
3.395c (0.00)
'n
us
an
e
Panel A: Di¤erence in Means: 3-year time trends for Industry Specialization Period 1 Period 2 Initial Returns 24.09c 13.01c 11.08c 3.148c 20.53c 18.92c 1.610 (0.00) (0.00) (0.30) # of observations 442 442 987 987
n
'e
1.846 (0.07)
0.920 (0.36)
0.770 (0.44)
'n
This table compares three-year time trends in initial returns to IPOs that accrue to IPOs backed by venture capital (VC) funds that commenced business between 1981 and 1984. Comparisons of initial returns to IPOs are implemented using matched pairs of venture capital funds that commenced business during immediately adjacent years. The matching variables are six di¤erent measures of the risk of VCs’fund portfolios, with the matching algorithm implemented relative to one measure of portfolio risk at a time. I describe the matching algorithm that pairs funds that enter the market at time t + 1 to funds that entered the market at time t in Table II. The measures of portfolio risk are: (1) industry specialization; (2) investment stage specialization; (3) syndicate size; and three measures of innovation focus for each fund de…ned as (4) low innovation; (5) median innovation; and (6) high innovation. I describe the construction of each measure of portfolio activity in Table I. Period 1 results are for year cohort f81 jf 82 ; Period 2 results augment f81 jf82 with f82 jf83 ; while Period 3 results augment the Period 2 sample with f83 jf84 . The venture capital data (including initial returns) are from VentureXpert, while other data on initial returns are from the equity issues database in SDC Platinum. Di¤erence in means ( n e ) are two-sided tests of signi…cant di¤erences between new entrant and established VCs. ('n 'e ) are estimates of the extent to which the distribution of initial returns to IPOs for the two matched samples are identical, and are implemented using the Wilcoxon signed rank test. p-values for statistical signi…cance are reported in parentheses. New Older New Older New Older Fundsy Fundsy 1 Fundsy Fundsy 1 Fundsy Fundsy 1
Ac ce p
Table III: VCs’portfolio activities and attenuation of di¤erences in initial returns to IPOs: three-year time trends for VC backed IPOs only
47
Page 47 of 50
e
1
n
e
'n
Ac ce p
Older Fundsy
'e n
te
1180
1 n
e
Period 2 22.97c -5.456c (0.01) 1180
e
Older Fundsy
'e
1 n
e
Period 3 16.78c -0.616 (0.96) 826
Period 3 17.55c 1.085 (0.48) 652
Period 3 19.61c -2.020c (0.00) 338
e
Older Fundsy
ip t
826
16.72c
652
18.63c
338
17.59c
n
New Fundsy
cr
1.788 (0.07)
1.343 (0.18)
-0.804 (0.42)
'n
us
Period 2 16.92c 1.846 (0.34) 373
an
M
Panel F: Di¤erence in Means: 3-year time trends for High Innovation Focus Period 1 Initial Returns 21.87c 11.40c 10.47c 2.878c 18.77c (0.08) (0.00) # of observations 139 139 373 c denotes signi…cance at the 1% con…dence level.
d
Panel E: Di¤erence in Means: 3-year time trends for Median Innovation Focus Period 1 Period 2 Initial Returns 21.59c 11.92c 9.664c 2.542c 19.32c 19.58c -0.263 (0.01) (0.01) (0.92) # of observations 142 142 338 338
# of observations
Panel D: Di¤erence in Means: 3-year time trends for Low Innovation Focus Period 1 Initial Returns Not enough data 17.51c
n
New Fundsy
New Fundsy
'e
-0.649 (0.52)
1.155 (0.25)
-0.603 (0.55)
'n
Table III (contd): VCs’portfolio activities and attenuation of di¤erences in initial returns to IPOs: three-year time trends for VC backed IPOs only
48
Page 48 of 50
d
te
cj
v c)
c
dif f %
'v
'c v
'v
Period 2: 90-92 8.301c 32%c (0.00) 816
Panel C: Di¤erence in Means: 3-period time trends for Investment Stage Specialization Period 1: 87-89 Period 2: 90-92 Initial Returns -1.957c 54.41c 104%c -4.386c 10.40c 8.301c 20% (0.01) (0.00) (0.14) # of observations 41 41 136 136 b & c denote signi…cance at the 10%, 5%, and 1% con…dence levels, respectively.
3.180c (0.00)
1.158 (0.25)
10.36c (0.00)
v
847
13.92c
847
34.15c
c
dif f %
Period 3: 93-95 21.62c 7.4%c (0.01) 5082
Period 3: 93-95 21.62c 36%c (0.01) 847
Period 3: 93-95 21.62c 37%c (0.00) 847
ip t
5082
20.03c
cr
'c
us
dif f %
an c
Panel B: Di¤erence in Means: 3-period time trends for Industry Specialization Period 1: 87-89 Period 2: 90-92 c c b c c Initial Returns 4.90 54.41 100% -3.130 6.685 8.301c 19% (0.03) (0.00) (0.25) # of observations 41 41 136 136
Panel A: Di¤erence in Means: 3-period time trends for Entire Sample Period 1: 87-89 Initial Returns 1.439c 54.41c 97%b -9.440c 12.19c (0.00) (0.00) # of observations 246 246 816
v
M
-3.173c (0.00)
10.41c (0.00)
8.278c (0.00)
'v
'c
) are two-sided tests of signi…cant di¤erences between venture and non-
venture backed IPOs. ('v 'c ) are estimates of the extent to which the distribution of initial returns to IPOs for the venture and non-venture backed samples are identical, and are implemented using the Wilcoxon signed rank test. p-values for statistical signi…cance are reported in parentheses. VC Non-VC VC Non-VC VC Non-VC Backedy Backedy Backedy Backedy Backedy Backedy
v proportional di¤erences for ease of exposition and computed as dif f % = ( max(
j
Using three sequential periods consisting of three years each, this table compares time trends in initial returns that accrue to venture or non-venture backed IPOs. The non-venture backed IPOs are matched to IPOs backed by venture capital funds that commenced business in 1986 or 1987 using industry one-digit SIC codes and the year of the IPO. The choice of VCs that commenced business in 1986 or 1987 is predicated on the fact that company speci…c initial returns for non-VC backed IPOs are available in the IPOs database of SDC Platinum only from 1986. The choice of three sequential time periods consisting of three years each results in sample IPOs occurring between 1987 and 1995. Prior to the matching of venture backed and non-venture backed IPOs by industry and IPO year, sample funds’portfolios are matched using six di¤erent measures of VCs’portfolio activities (portfolio risk), which are: (1) industry specialization; (2) investment stage specialization; (3) syndicate size; (4) low innovation; (5) median innovation; and (6) high innovation. I discuss each measure of portfolio risk in Table I and describe the matching algorithm in Table II. The matching algorithm is implemented relative to one matching variable at a time. Period 1 results are for matched IPOs that occurred between 1987 and 1989; Period 2 results are for matched IPOs that occurred between 1990 and 1992; & Period 3 results are for matched IPOs that occurred between 1993 and 1995. The venture capital data (including initial returns) are from VentureXpert - an SDC Platinumbdatabase, while other data on initial returns to IPOs are from the equity issues database in SDC Platinum. Di¤erences in means - expressed as
Ac ce p
Table IV: VCs’portfolio activities and attenuation of di¤erences in initial returns to IPOs: VC versus non-VC backed IPOs
49
Page 49 of 50
dif f %
'v
'c v
dif f %
Period 2: 90-92 8.301c 27%b (0.03) 136
c
Non-VC Backedy
'c
Panel G: Di¤erence in Means: 3-period time trends for High Innovation Focus Period 1: 87-89 Period 2: 90-92 Initial Returns -0.079 54.41c 100%b -4.166c 15.98c 8.301c 48%c (0.02) (0.00) (0.00) # of observations 41 41 136 136 b & c denote signi…cance at the 5%, and 1% con…dence levels, respectively.
6.111c (0.00)
847
22.80c
dif f %
Period 6: 93-95 21.62c 5.2%c (0.01) 847
Period 6: 93-95 21.62c 12% (0.06) 847
Period 3: 93-95 21.62c 40%c (0.00) 847
Period 3: 93-95 21.62c 3.0% (0.77) 847
c
Non-VC Backedy
ip t
847
18.97c
847
12.92c
847
22.06c
v
VC Backedy
cr
7.001c (0.00)
3.352b (0.02)
(0.00)
3.665c
'v
us
an
M
Panel F: Di¤erence in Means: 3-period time trends for Median Innovation Focus Period 1: 87-89 Period 2: 90-92 Initial Returns 7.299c 54.41c 87%b -2.728c 18.00c 8.301c 54%c (0.04) (0.01) (0.00) # of observations 41 41 136 136
d
Panel E: Di¤erence in Means: 3-period time trends for Low Innovation Focus Period 1: 87-89 Period 2: 90-92 Initial Returns -1.454c 54.41c 103%c -4.335c 10.61c 8.301c 22% (0.01) (0.00) (0.10) # of observations 41 41 136 136
te
Panel D: Di¤erence in Means: 3-period time trends for Syndicate Size Period 1: 87-89 Initial Returns -.0.079 54.41c 100%b -4.166c 11.33c (0.02) (0.00) # of observations 41 41 136
c
v
Ac ce p
Non-VC Backedy
VC Backedy
VC Backedy
Table IV (cont’d): VCs’portfolio activities and attenuation of di¤erences in initial returns to IPOs: VC versus non-VC backed IPOs
'c
6.597b (0.00)
2.998c (0.00)
-4.820c (0.00)
4.446c (0.00)
'v
50
Page 50 of 50
Table V: IPO issuer risk and di¤erences in initial returns between VC and non-VC backed IPOs
R-squared 0.1649 Model p-value 0.0288 BIC 81.26 # of observations 29 b & c denote signi…cance at
0.0042 0.6965 -4.839
0.0151 0.4886 169.2
0.0022 (0.49) 3.5936c (0.00)
0.0007 0.7770 117.3
-0.0006 (0.78) 1.4168c (0.00)
0.0190 0.1111 469.7
-0.0108 (0.11) 7.0710c (0.00)
136
0.0530 0.0070 -53.29
-0.0027c (0.01) -0.0350c (0.00)
99
0.0000 0.9873 391.3
0.0002 (0.99) 2.3388c (0.00)
0.0429 0.0398 386.9 risk
us
0.0169 0.0413 0.0944 0.0591 0.4642 0.0260 0.0003 0.0044 169.1 112.32 458.9 -54.17 Innovation Focus dimension of VCs’portfolio
an
M
40 39 34 120 135 the 5%, and 1% con…dence levels, respectively.
0.0001 0.9611 225.1
R-squared 0.1590 0.0004 0.0044 Model p-value 0.0322 0.9018 0.6898 BIC 81.47 225.0 -4.846 Panel B: Matched pairs obtained relative to High Initial Return Spread -0.0034b -0.0003 9.50e-05 (0.03) (0.96) (0.70) c c Constant 1.9606 8.8369 0.0377 (0.00) (0.00) (0.31)
d
te
Ac ce p
cr
713
0.0008 0.4609 1090
847
0.0308 0.0000 2932
-0.0060c (0.00) 7.4997c (0.00)
ip t -0.0004 (0.46) 1.6297c (0.00)
0.0000 0.9938 1090
0.0190 0.0001 2942
846
0.0491 0.0000 -252.6
-0.0012c (0.00) 0.0353c (0.00)
0.0432 0.000 -247.45
566
0.0018 0.3193 4931
-0.0181 (0.32) 3.2449c (0.00)
0.0018 0.3097 4931
This table reports structural relations derived within univariate regression frameworks for relations between measures of IPO risk, and di¤erences in initial returns between matched samples of venture backed and non-venture backed IPOs. Let rvc and rnvc denote, respectively, initial returns to venture backed and non-venture backed IPOs. Then the initial return spread between comparable IPOs is constructed as: (rvc rnvc ). The measures of IPO risk are: underwriting spreads ( writer ); gross spreads (gross), comprising of underwriting spreads and book management spreads; the proportional change between the IPO mid…le price and the o¤er price (midto¤ er ); and the price-to-book value (P=B ) achieved subsequent to the o¤er. The non-venture backed IPOs are matched to IPOs backed by venture capital funds that commenced business in 1986 or 1987 using industry one-digit SIC codes and the year of the IPO. Matched pairs are obtained relative to the (1) industry specialization; and (2) high innovation focus dimensions of VCs’portfolio risk. These measures of portfolio activity are described in Section IV of the study and Table I. Tables II, III, and IV discuss matching algorithms that generate matched pairs of VC and non-VC backed IPOs. The choice of three sequential time periods consisting of three years each results in sample IPOs occurring between 1987 and 1995. Period 1 results are for matched IPOs that occurred between 1987 and 1989; Period 2 results are for matched IPOs that occurred between 1990 and 1992; and Period 3 results are for matched IPOs that occurred between 1993 and 1995. The venture capital data (including initial returns) are from VentureXpert - an SDC Platinum database, while other data on initial returns to IPOs are from the equity issues database of SDC Platinum. I report p-values for statistical signi…cance in parentheses. writer gross midto¤er P/B writer gross midto¤er P/B writer gross midto¤er P/B Panel A: Matched pairs obtained relative to Industry Specialization dimension of VCs’portfolio risk Period 1: 87-89 Period 2: 90-92 Period 3: 93-95 Initial Return Spread -0.0034b -0.0007 9.88e-05 -0.0024 -0.0057b -0.0254c -0.0030c -0.0233b 3.96e-06 -0.0041c -0.0010c -0.0164 (0.03) (0.90) (0.69) (0.46) (0.03) (0.00) (0.00) (0.04) (0.99) (0.00) (0.00) (0.31) c c c c c c c c c c Constant 1.9810 8.8228 0.0374 3.5904 1.4074 6.9456 -0.0602 2.3096 1.6287 7.5382 0.0440 3.3885c (0.00) (0.00) (0.31) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00)
S0148-6195(16)30026-1 http://dx.doi.org/doi:10.1016/j.jeconbus.2016.05.003 JEB 5751
To appear in:
Journal of Economics and Business
Received date: Revised date: Accepted date:
25-5-2015 19-2-2016 26-5-2016
Please cite this article as: Oghenovo Obrimah, Information production within the venture capital market: implications for economic growth and development, (2016), http://dx.doi.org/10.1016/j.jeconbus.2016.05.003 This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
ip t
cr
Information production within the venture capital market:
us
implications for economic growth
an
and development Oghenovo Obrimah
d
M
February 19, 2016
te
Abstract
I …nd venture capitalists’(VCs’) information production activities help decrease
Ac ce p
uncertainty about asset valuations within public equity markets, resulting in price convergence within the cross-section of Initial Public O¤erings (IPOs) that are backed by di¤erent classes of VCs and price convergence between VC and non-VC backed IPOs. These …ndings provide evidence that venture capital …nancing can be a component of an e¤ective or e¢ cient price discovery process within public equity markets that yields bene…cial externalities in so far as the pricing of non-VC backed IPOs are concerned. Given a decrease in valuation uncertainty ultimately results in lower costs of capital for issuing …rms, larger stock markets, and improvements in market liquidity or e¢ ciency,
Department of Economics, Banking, and Finance, Babcock University, Ogun State Nigeria. I am grateful to journal editors and two anonymous reviewers for valuable comments. Thanks to participants at the 2013 Accounting and Finance Research Association Conference for valuable comments. Corresponding Author: Dr. Oghenovo Obrimah; Tel: 234 (0903) 920-2317; Primary E-mail: [email protected]; Secondary E-mail: [email protected]; Address: Babcock Business School, Babcock University PMB 21244 Lagos Nigeria.
1 Page 1 of 50
empirical …ndings provide evidence that VCs’information production activities are
induce e¢ cient allocation of resources within stock markets).
cr
JEL Classi…cation: G24
ip t
bene…cial for economic growth and economic development (price discovery processes that
Keywords: Venture Capital, Initial Returns, Innovation, information asymmetry, IPO,
Ac ce p
te
d
M
an
us
Economic Growth & Development
2 Page 2 of 50
I
Introduction
ip t
It is well established that market e¢ ciency - the extent to which prices re‡ect all publicly available information - is an important indicator of stock market development (see for
cr
example, Fama, 1991). In light of the importance of market e¢ ciency for stock market
us
development, economic agents that facilitate price discovery within stock markets contribute directly to stock market development and by extension to economic development. In Leland
an
and Pyle (1977), information production that induces price discovery is an important objective of …nancial intermediation in markets such as venture capital markets. In
M
Campbell and Kracaw (1980), …nancial intermediation is most credible when information generated by …nancial intermediaries (e.g. venture capitalists) is utilized in alternate markets,
d
with the stock market qualifying as an alternate market to the venture capital market.
te
Suppose as in Gompers (1996) we distinguish between ‘incumbent’and ‘new entrant’
Ac ce p
venture capitalists (VCs). If VCs’information production activities are bene…cial for price discovery within stock markets, it is reasonable to expect di¤erences in the pricing of IPOs backed by the two classes of VCs identi…ed in Gompers (1996) attenuate over time, resulting in price convergence. Empirical evidence that VCs’information production activities improve the pricing of IPOs that do not have venture capital backing (non-VC backed IPOs) is evidence, however, that VCs’activities yield bene…cial externalities within stock markets; that is, are directly bene…cial for price discovery, market e¢ ciency, stock market development, and economic development. Studies such as Lee and Wahal (2004), Megginson and Weiss (1991), and Belghitar and Dixon (2012) provide evidence of cross-sectional di¤erences in initial returns to VC and non-VC backed IPOs. Empirical evidence that
3 Page 3 of 50
…nancial market development, which includes stock market development, is an important
ip t
driver of economic growth is provided in studies such as King and Levine (1993), Luintel and Khan (1999), and Rajan and Zingales (1998).
cr
In this study, I examine whether venture capitalists’(VCs’) information production activities a¤ect the time series variability of initial returns to IPOs, resulting in price
us
convergence within public equity markets. In order to demonstrate the robustness of the
an
evidence on convergence of initial returns to IPOs, I show cross-sectional variation in initial returns to IPOs is consistent with the positive relation between initial returns to IPOs and
M
the severity of information asymmetry problems in studies such as Beatty and Ritter (1986), Ibbotson et al. (1994); Ritter (1984), or Rock (1986).
d
With respect to time series convergence, empirical results show, consistent with …ndings
te
in Gompers (1996), that initial returns to IPOs backed by new entrant VCs start o¤ systematically lower than initial returns that accrue to IPOs backed by incumbent VCs.
Ac ce p
Over the course of three periods (years), however, initial returns to IPOs backed by new entrant or incumbent VCs converge to some equilibrium level, such that there do not exist any signi…cant di¤erences in initial returns between the two groups of IPOs. These …ndings provide empirical validation for a three-year reputation development time frame in Nahata (2009), Krishnan and Masulis (2009), and Obrimah (2015). In comparisons of initial returns to VC and non-VC backed IPOs, I …nd di¤erences in initial returns between the two classes of IPOs converge but are not fully attenuated over a period of three venture capital …nancing cycles (9 years), with convergence shown to be induced by the arrival of new and favorable information about VC backed IPOs. Given stock markets cannot be e¢ cient in the absence of price discovery mechanisms or processes, the 4 Page 4 of 50
positive externality from VCs’information production activities - improvements in the
ip t
valuations of non-VC backed IPOs - is evidence that venture capital has a direct impact on market e¢ ciency (stock market development) and by extension, economic development. My
cr
…nding that the slowest convergence rates between initial returns to VC or non-VC backed IPOs relate to the innovation dimension of VCs’portfolio activities provides support for a
us
characterization of innovation as an important or unique facet of venture capital activity in
an
studies such as Aggarwal and Hsu (2013) or Lerner, Sorenson, and Stromberg (2011). Markets within which prices have converged are characterized by less severe valuation
M
uncertainty in relation to markets within which prices have yet to converge. In Huang (2008), uncertainty aversion (aversion to valuation uncertainty) results in slower economic
d
growth, but is mitigated by information production. In Brockman, Liebenberg, and Schutte
te
(2010) there exists a link between swings in information production and the incidence of business cycles. In light of …ndings in Huang (2008) and Brockman et al. (2010), empirical
Ac ce p
results in this study provide evidence that VCs’portfolio activities have bene…cial e¤ects on uncertainty aversion within stock markets. Studies such as Loughran and Ritter (2004), Kirkulak and Davis (2004), Dimovski, Philavanh, and Brooks (2011), Chemmanur and Fulghieri (1994), and Megginson and Weiss (1991) provide mixed evidence on relations between initial returns to IPOs and market reputation. In this study, the formal theory I develop for the assessment of interactions between market reputation and initial returns to IPOs predicts market reputation can be associated with relatively higher initial returns to IPOs whenever the arrival of new innovative projects induces reputable intermediaries to bring their superior information production skills to bear on the pricing of new innovations. Comparisons of time series 5 Page 5 of 50
trends in initial returns to VC or non-VC backed IPOs provide support for the predictions of
ip t
the formal theory. Signalling theories (see for example, Allen and Faulhaber, 1989; Grinblatt and Hwang,
cr
1989; Welch, 1989), information asymmetry theories, and litigation risk theories of IPO
underpricing (Tinic, 1988; Alexander, 1993; Drake and Vetsuypens, 1993; Lowry and Shu,
us
2010) all imply di¤erences in initial returns between high risk and low risk IPOs do not
an
attenuate over time. My …nding that initial returns to IPOs converge fastest among the most innovative ventures, resulting in initial returns that are comparable to those which
M
accrue to the least innovative ventures provides evidence that independent of bene…cial e¤ects of information production, neither of information asymmetry, signalling, nor litigation
d
risk theories are well adapted to explaining time series variation in initial returns to IPOs.
te
My …ndings complement studies which provide evidence of direct, as opposed to indirect bene…ts from venture capital for economic growth (Total Factor Productivity) and
Ac ce p
development (Innovation or Patent Rates). These studies include Kortum and Lerner (2000) or Abdou and Varela (2009) and Chemmanur, Krishnan, and Nandy (2011), which …nd, respectively, that venture capital …nancing contributes to growth rates of innovation (patents) and improvements in Total Factor Productivity (TFP) of private …rms. Consistent with externalities from venture capital, Samila and Sorenson (2011) …nd the arrival of venture capital …nancing within a geographical area is associated with a signi…cant increase in non-venture capital backed entrepreneurial activity. Since the arrival of venture capital …nancing can itself be induced by improvements in arrival rates of good projects within opportunity sets, both venture capital and non-venture capital activity can be induced by the richness of the opportunity set, but with venture capital activity helping to re…ne information 6 Page 6 of 50
sets or prices. Empirical …ndings in this study provide evidence of a mechanism -
ip t
information production that induces re…nements in prices - via which the arrival of venture capital within an opportunity set can induce increases in investment activities not backed by
cr
venture capitalists.
The rest of the paper proceeds as follows. Section II develops the theoretical framework.
us
Section III discusses the matching methodology adopted in the implementation of the study.
an
Section IV describes data, empirical variables, and the structure of empirical tests. Section V reports results from empirical tests. Section VI reports results from robustness tests that
M
establish the directionality of externalities from information production. Section VII
Theoretical Structure
te
II
d
concludes the study.
Ac ce p
Venture capitalists specialize in information production via their investments in young or new ventures with little or no track record, income, or established products. Whenever VCs o¤er portfolio projects for sale in public equity markets for instance, investors gain insights into the sizes and distributions of opportunity sets that are generating new innovative assets within an economy. VCs’ability to grow portfolio projects into …rms that can be exited (sold to other investors) either via IPOs or third party acquisitions enables VCs to raise new capital and continue to produce information in their capacity as fund managers. In what follows, I develop a formal theory that generates testable predictions about time series trends in distribution of information asymmetry that obtain in the presence of heterogeneity or homogeneity in market reputation, with market reputation derived solely
7 Page 7 of 50
from information production. This generic approach to the formulation of the formal theory
ip t
ensures the model is robust to alternative proxies for the severity of information asymmetry between IPO …rms and investors. I leverage on theoretical and empirical evidence that
cr
establishes underpricing as a proxy for the severity of information asymmetry between
investors and IPO issuers to apply the predictions of the formal theory to time series trends
Information Production and Time Series Trends in Distribution of
an
A
us
in initial returns to IPOs.
M
Information Asymmetry
Suppose we assume information asymmetry with mean severity
d
information asymmetry among market participants
at some origin time t, with
and dispersion in
( t ) for …rms classi…ed within the t
the standard deviation of information
te
same general risk class
t
t
asymmetry. Since information production mitigates di¤erences in information or improves
on
Ac ce p
the precision of information, information production has a …rst order e¤ect on
via it’s e¤ect
and may or may not have a second order decreasing e¤ect on the average realization of
information asymmetry, .
In the presence of information production that mitigates information asymmetry within risk class
, we have
t
>
t+1
t+n
!
t
>
t+1
t+n ,
where n is the terminal future
period at which any additional information produced no longer has any signi…cant impact on the dispersion of information asymmetry (the terminal period at which di¤erences in information are no longer signi…cant). If information production merely induces a decrease in the dispersion of information asymmetry for …rms in risk class
,
t
t+1
t+n ,
implying mean severity of information asymmetry is time invariant and independent of
t
.
8 Page 8 of 50
Suppose, however, that
t
= g ( t ), then since the severity of information asymmetry
decreases in response to information production,
>
t+1
t+n
and
t+n
= g(
t+n )
imply
ip t
t
a well behaved g that is activated by information production is an increasing function of . can be divided into two groups
1
&
2
cr
Assume …rms in the same general risk class
based on the market reputation of IPO backers; with mean severity of information 1 t
2; t
&
and dispersion
1 t
2 t.
&
Studies such as Loughran and Ritter (2004),
us
asymmetry
an
Kirkulak and Davis (2004), Fernando, Gatchev, and Spindt (2005), Dimovski, Philavanh, and Brooks (2011), Chemmanur and Fulghieri (1994) or Megginson and Weiss (1991) show,
that either of
1 t
<
2 t
or
1 t
2 t
>
M
conditional on di¤erences in market reputation that are derived from information production, is feasible within the same general risk class or
that occur in response to information
d
the studies examine, however, transitions in
. None of
te
production by reputable intermediaries.
Propositions 1 through 3 generate equilibrium conditions expected to hold within the
Ac ce p
context of externalities from information production by …nancial intermediaries (IPO backers) that are similar or dissimilar in reputation.
Proposition 1 Let
1 t
&
2 t
denote mean realizations of information asymmetry (initial
returns to IPOs) for IPO …rms in groups t
= g ( t ) : Assume …rms in
1
&
2
1
&
2;
respectively, with
1 t
<
2 t
and
are backed by …nancial intermediaries that are similar
in market reputation in so far as the quality of information production is concerned; then equilibrium conditions that govern attenuation of information asymmetry are: 1 t
1 t+n
<0
t+n
future convergence date;
1 t+n t+n
; and
2 t+n
t+n
0<
2 t
2 t+n
where n > 1 is the
is the mean realization of information asymmetry for the
9 Page 9 of 50
universe of IPOs in
&
1
2;
and f is a continuous and invertible function of .
ip t
Proof. See Appendix. Proposition 1 states, in the presence of two classes of IPO …rms backed by …nancial
cr
intermediaries that do not di¤er with respect to market reputation, that equilibrium initial
us
returns to IPOs induced by information production are the outcome of complementarity in
The E¤ect of Di¤erences in Reputation of Financial Intermediaries
Proposition 2 Let
1 t
&
2 t
denote mean realizations of information asymmetry (initial
M
B
an
attenuation of information asymmetry between the two classes of …rms.
returns to IPOs) for IPO …rms in groups
1
&
2;
respectively, with
Assume …nancial intermediaries associated with …rms in
>
2 t
and
t
= g ( t ).
are more reputable than …nancial
d
1
1 t
intermediaries associated with …rms in
in so far as the quality of information production
te
2
Ac ce p
is concerned; then equilibrium conditions that govern convergence of initial returns to IPOs in the well de…ned scenario ( and
2 t
2 t+n
=j (
1 t
1 t
>
1 t+n )
2) t
1 t+n
are:
1; t
<
2 t+n
2; t
2 t
2 t+n
where n > 1 is the future convergence date;
realization of initial returns to the universe of IPOs in but non-invertible functions of (
<
1 t
1 ) t+n
or (
1 t
1
1 t+n );
&
2;
= j( t+n
1 t
1 ); t+n
is the mean
and j or j are increasing
respectively.
Proof. See Appendix.
Proposition 2 states, in the presence of …nancial intermediaries in reputable than …nancial intermediaries in
2,
1
that are more
that equilibrium initial returns to IPOs
decrease with information production. Convergence, however, is a function of changes in initial returns that accrue to …rms in
1,
as opposed to a function of the level of initial
10 Page 10 of 50
1 t
investors requires
>
2 t
1.
This is the case because the imposition of rationality on 1 t+n
implies
more reputable than those in
2,
2 . t+n
Given …nancial intermediaries in
any improvements in the risk pro…les of …rms in
2 t+n
= j(
1 t
1 ). t+n
In equilibrium, the pricing of …rms in
improvements in risk pro…les of …rms in
1
2
2
remain
are expected to induce changes in the
an
2
induce
can be altered by
even if the risk pro…les of …rms in
unaltered. Since changes in the pricing of …rms in
1
resulting in
us
2 t
2;
cr
demands on the risk pro…les, or equivalently, the pricing of …rms in
are
1
ip t
returns that accrue to …rms in
risk pro…les of these …rms, ultimately both the pricing and the risk of the …rms in
2
are
M
altered. In equilibrium, however, risk may lead pricing or pricing may induce changes in risk pro…les. 1 t
&
2 t
denote mean realizations of information asymmetry (initial
d
Proposition 3 Let
te
returns to IPOs) for IPO …rms in groups
1
&
2;
respectively, with
Ac ce p
Assume …nancial intermediaries associated with …rms in intermediaries associated with …rms in
2
1
1 t
<
2 t
and
t
= g ( t ).
are more reputable than …nancial
in so far as the quality of information production
is concerned; then equilibrium conditions that govern convergence of initial returns to IPOs are indeterminate. We can say, however, that 1 t+1
2 t+1
>0&
1 t+1
2 t+1
1 t
2 t
<0&
1 t
2 t
< 0 followed by
0 is evidence of a strategic information production
response.
Proof. See Appendix Proposition 3 predicts mixed empirical evidence on relations between initial returns to IPOs and market reputation can be explained by a strategic information production response from reputable intermediaries in contexts within which they currently back relatively low risk 11 Page 11 of 50
IPOs. Speci…cally, if reputable …nancing intermediaries currently associated with relatively
ip t
low risk projects bring their superior information production skills to bear on the production of information that relates to new innovative projects, information production can result in 1. t
cr
III
>
Matching Methodology
us
1 t+1
an
In line with discussions in Section II, comparisons of the evolution of information asymmetry that are credible must relate to …rms in the same risk class
, with the risk class su¢ ciently
M
rich to support dispersion in severity of information asymmetry problems. In this study, initial returns to IPOs are context-speci…c measures of the severity of information asymmetry
d
problems. In order to facilitate the interpretation of empirical results within the context of
te
equilibrium conditions established in Propositions 1 through 3, and as is evident in studies
Ac ce p
such as Gompers (1996) or Purnanandam and Swaminathan (2004), the matching of sample IPOs along speci…c risk dimensions is of critical importance. I implement two di¤erent sets of matching algorithms using sample IPO …rms. The …rst set of matching algorithms utilize IPO …rms brought to market by VCs that enter the venture capital market during adjacent years to generate two groups of …rms backed by …nancial intermediaries (VCs) that likely do not di¤er in so far as market reputation derived from information production is concerned. For ease of exposition, I refer to VCs that enter at time t as incumbent VCs, and VCs that enter at time t + 1 as new entrant VCs. Empirical results derived from these matched pairs of IPO …rms map to any of Propositions 1 through 3. Empirical …ndings in Megginson and Weiss (1991) indicate VCs contribute reputation
12 Page 12 of 50
during IPOs not attributable to IPO underwriters. The second set of matching algorithms
ip t
utilize VC and non-VC backed IPOs to generate two groups of …rms backed by …nancial intermediaries shown to di¤er either in market reputation (Megginson and Weiss, 1991) or
cr
the risk of IPO …rms (Lee and Wahal, 2004). Empirical results derived from these matched pairs of IPO …rms map to either of Propositions 2 or 3. Analyzed in juxtaposition with
us
Propositions 1 through 3, comparisons of empirical …ndings derived from the two sets of
A
an
matching algorithms serve as checks on the robustness of study conclusions.
Convergence and the number of matching dimensions
M
If IPO …rms are matched along k > m dimensions, the probability of convergence in initial
d
returns to IPOs increases as the number of dimensions increases from m to k. This increase
te
in probability of convergence with the number of matching variables implies studies of convergence yield the most credible inferences if IPO …rms are matched using a parsimonious
Ac ce p
set of matching variables. Since the probability of observing convergence increases with the number of matching variables, the most credible evidence in favor of convergence is derived from the most parsimonious number of matching dimensions.
B
Matching algorithm for VC backed IPOs
Venture Capitalists manage portfolios of projects and produce information within the context of portfolio management. In light of this, I focus on matches of IPO …rms that re‡ect VCs’ portfolio activities. This matching algorithm ensures empirical tests capture VCs’ information as derived from actual asset allocations as opposed to some arti…cial imposition of SIC code preferences. For robustness, I utilize several di¤erent measures of portfolio
13 Page 13 of 50
activities, which simultaneously double as measures of portfolio risk in the implementation of
ip t
the matching algorithm. These measures are: investment stage specialization Her…ndahls, industry specialization Her…ndahls, average syndicate size, and portfolio innovation focus. I
cr
describe the construction of each of these measures of portfolio risk in Section IV. Given matched pairs of VCs are adjacent to each other in so far as entry year into the venture
us
capital market is concerned and seek to exit portfolio projects via IPOs at the earliest time
an
possible (Gompers, 1996), in order to capture reputation e¤ects implicit in Propositions 1 through 3, I do not match new entrant and incumbent VCs’portfolios by IPO year.
M
Kaplan and Schoar (2005) and Schweinbacher (2008) …nd …nancing cycles occur every three years within the venture capital market. Nahata (2009), Krishnan and Masulis (2009),
d
and Obrimah (2015) utilize three-year cycles for measurement of VCs’market reputation.
te
Combined, these studies imply three-year reputation cycles within the venture capital market and indicate, independent of proxies for reputation, that activities of venture capitalists that
Ac ce p
do not enter the venture capital market during the same reputation cycle likely are not directly comparable. Empirical …ndings in Lerner (1994) that VCs with similar experience tend to syndicate together is consistent with the existence of reputation cycles within venture capital markets. In light of evidence of reputation cycles, I utilize a convergence time frame of three years in the implementation of the algorithm for matching new entrant and incumbent venture capitalists. Starting with VCs that enter the venture capital market at time t = 0, the implementation of the matching algorithm requires the matching of VCs that enter the market during time periods, t = 0 & t = 1, t = 1 & t = 2, and t = 2 & t = 3, resulting in three sets of matches that require four cohort VC years. Formally, let yj
2,
yj
1
& yj denote three consecutive years or periods bounded by four 14 Page 14 of 50
points in time (four cohort VC years), tj
3 , t j 2 , tj 1
& tj ; fj
3,
fj
2,
fj
1
and fj denote, 3,
nj
2,
nj
1
ip t
respectively, funds that commenced business during the four cohort years; nj
and nj denote, respectively, the number of funds that commenced business during the four
matches obtained for each pi by fj
3j
fj
2,
fj
2j
fj
1;
and fj
1j
cr
cohort years; while pi denote a measure of portfolio activity or risk. Denote adjacent
fj . Since new entrant funds
us
are matched to incumbent funds, the number of matched pairs are determined by nj
an
and nj , respectively. This implies the number of matched pairs in cohort fj unique and di¤erent from the number of matched pairs in cohort fj 1
6= nj . This matching strategy and the fact that nj
matches fj
3j
fj
2,
fj
2j
fj
1;
and fj
1j
1
is
fj whenever
d
the inclusion of fj
2
in fj
3j
fj
2, 2
while matches in
and fj
2j
fj
1
te
1,
1
1;
fj do not consist of exactly the same funds or
number of funds. Since matches in period one are determined by fj period two are determined by fj
fj
nj
6= nj for all j ensures adjacent
M
nj
1j
2j
2;
decreases the probability of convergence in initial returns to IPOs across periods one and
Ac ce p
two. Similar arguments apply to the probability of convergence across periods two and three and demonstrate the matching algorithm is not biased towards producing evidence of convergence in initial returns to VC backed IPOs.
C
The matching of non-VC and VC backed IPOs
While the literature on the pricing of IPOs that have or do not have venture capital backing yields mixed results in so far as the directionality of relations between initial returns to VC and non-VC backed IPOs are concerned, all studies conclude VC backing has certi…cation value during IPOs. These studies include Barry, Muscarella, Peavy, and Vetsuypens (1990), Megginson and Weiss (1991), and Lee and Wahal (2004). In light of the fact that non-VC 15 Page 15 of 50
backed IPOs are not necessarily brought to market within the context of portfolio
ip t
management, I match non-VC backed IPOs to matched pairs of VC backed IPOs derived from the matching algorithm in the preceding sub-section using 1-digit industry SIC codes
cr
and the year in which an IPO occurred. The restriction of industry classi…cations to the 1-digit SIC code is necessary because the VC backed IPOs are not as numerous as the
us
non-VC backed IPOs. In the absence of any direct competition e¤ects, and given substantial
an
yearly variability in initial returns to IPOs, the introduction of IPO year as a matching variable is of paramount importance. While the utilization of only two variables - the
M
industry classi…cation and issue year - as matching variables results in a rather parsimonious matching algorithm, as discussed in Section III.A, empirical evidence of intertemporal price
Data, Variables, and Empirical Structure
Ac ce p
IV
te
d
convergence becomes less robust with the number of matching variables.
Empirical evidence that levels of initial returns change signi…cantly over time in studies, such as Ritter (1984), Lowry and Schwert (2002), or Pastor and Veronesi (2005) imply empirical results for one three-year cycle cannot be overlaid on results for some other three-year cycle to arrive at average time series e¤ects. That is, results obtained from matches of VCs that enter the venture capital market between 1981 through 1984 cannot be combined with results obtained from matches based on data from 1985 through 1988 because the levels of initial returns to IPOs during these periods can be very di¤erent. In order to generate time series inferences, I limit the matching methodology in the preceding sub-section to funds that commenced business within four consecutive years only, commencing with the …rst feasible
16 Page 16 of 50
year in the sample; that is, 1981. The limitation to four consecutive years results in sample
ip t
VCs commencing business in 1981 (the base year), 1982, 1983, or 1984. Given data from about 1980 is regarded as the earliest year with robust coverage of the venture capital market
cr
in VentureXpert (Gompers and Lerner, 2004), the empirical structure utilizes the longest
A
us
series of IPO activity available in the study.
Constraints on sample of non-VC backed IPOs
an
If VCs’information production induce externalities for non-VC backed IPOs, this inference cannot be restricted to VCs that commenced business between 1981 and 1984. In light of
M
restrictions discussed in the preceding sub-section and in order to provide evidence from a
d
di¤erent set of funds, I compare initial returns that accrue to IPOs backed by VCs which
te
commenced business in 1986 or 1987 with initial returns that accrue to comparable IPOs that
B
Ac ce p
do not have venture capital backing.
Measures of VCs’Portfolio Activities
I construct four measures of VCs’portfolio activities (risk) that are motivated within the venture capital literature. These portfolio risk measures are: investment stage specialization Her…ndahls, industry specialization Her…ndahls, average syndicate size, and portfolio innovation focus.
B.1
Portfolio Specialization Her…ndahls
Fulgheiri and Sevilir (2009), Gompers, Kovner, and Lerner (2009), and Knill (2009) all link VCs’investment performance with portfolio specialization, with specialization typically
17 Page 17 of 50
measured as industry or investment stage specialization. I construct investment stage
ip t
specialization Her…ndahls (Her…ndahl-Hirschman) as the sum of squares of the percentage of total investments in (i) start-up, (ii) early stage, and (iii) expansion stage ventures, with
cr
classi…cations as indicated by the VentureXpert Database of SDC Platinum.
I construct industry specialization Her…ndahls as
us
Industry Specialization Her…ndahls
the sum of squares of the percentage of total investments in the following three broad
an
industries: (1) High Technology; (2) Medical/Health Related; and (3) Non-high Technology. The term “high technology” refers to industries that are not technologically intensive within
M
the context of web based or internet related businesses. All classi…cations are as indicated by
Syndicate Size
te
B.2
d
VentureXpert.
Ac ce p
Hochberg et al. (2007) …nd VCs’exit performance is an increasing function of the size of their investment networks. Cassamatta and Haritchabalet (2007) develop a theoretical model within which syndicate size is a proxy for the size of VCs’investment networks. I construct syndicate size as the average number of VC …rms that participate in each …rst round …nancing deal that ultimately results in an IPO.
B.3
Portfolio Innovation Focus
Measures of portfolio specialization or syndication activity are not peculiar to the venture capital market in so far as attempts at the benchmarking of portfolio risk are concerned. These measures of portfolio risk also are utilized in studies of banking activity (see for
18 Page 18 of 50
example, Berger and Hannan, 1989; Esty and Megginson, 2003). Obrimah (2015) develops a
ip t
measure of portfolio risk that is speci…c to the venture capital market and shown to be a better predictor of market reputation than portfolio specialization or syndication activity
cr
variables. This measure of portfolio activity captures the expectation or …ndings in studies such as Kortum and Lerner (2000) that VCs specialize in the …nancing of innovation. This
us
variable is termed “portfolio innovation focus”. For a discussion of theoretical rationales for
The construction of innovation focus
an
this measure of the risk of VCs’portfolio activities, see Obrimah (2015).
As is the case in Obrimah (2015), I normalize
M
VCs’capital disbursements to projects by adjusting capital disbursements for changes in in‡ation using GDP de‡ators from the World Development Indicators database that is
d
managed by the World Bank. I divide normalized capital disbursements for the entire
te
sample into percentiles and de…ne median innovation ventures as projects whose capital
Ac ce p
disbursements lie between the 25th and 75th percentile of capital disbursements; low innovation ventures as projects whose capital disbursements lie above the 75th percentile of capital disbursements; and high innovation ventures as projects whose capital disbursements lie below the 25th percentile of capital disbursements. I utilize the proportion of total capital disbursements that accrue to low, median, or high innovative ventures as measures of the extent to which these VCs’portfolio activities are focused on innovation. This construction of portfolio innovation focus is consistent with the prediction in Madan and Yen (2004) that the dollar amount invested in an asset is positively related to expected returns, negatively related to the variance of expected returns, and positively related to the skewness of expected returns, providing evidence of an investment decision rule or choice function that maps dollar
19 Page 19 of 50
amounts invested in portfolio projects to the moments of portfolio expected returns.
Data
ip t
C
cr
The venture capital data consist of funds that commenced operations between 1980 and 2005, but which are not a¢ liated with other …nancial or non-…nancial institutions.1 This
us
restriction ensures VCs within the sample are producing information for investors, as opposed to producing information for their parent or a¢ liated …rms. I refer to these VCs as
an
“independent VCs”. In order to focus on the quality of information (deal screening information) as opposed to the quantity of information produced (project management or
M
monitoring information), I construct measures of VCs’portfolio activities using capital
d
disbursed during …rst …nancing rounds only.
te
I restrict the venture capital data to include: transactions where the identities of the funds providing …nancing are disclosed along with the amounts invested; venture capital
Ac ce p
funds with fund size greater than US$5 million. The venture capital investments data (including initial returns) are obtained from VentureXpert, while data on initial returns to IPOs for non-VC backed IPOs are from the equity issues database of SDC Platinum. I report summary statistics for the funds included in my data sample in Table I, which show measures of VCs’portfolio activities (industry specialization, investment stage specialization, and portfolio innovation focus) are signi…cantly correlated at the 1% con…dence level. The positive correlation between the proportion of all projects that are high innovation ventures and either of industry or investment stage specialization Her…ndahls is evidence that 1
Gompers and Lerner (2004) provide rationales for the exclusion of pre-1980 data from empirical studies of venture capital.
20 Page 20 of 50
innovation focus is a measure of portfolio risk within the venture capital market.
Empirical Results: Information Production and
us
Convergence of Initial Returns to IPOs
cr
V
ip t
Insert Table I About Here
an
In this section, I report empirical results that examine the e¤ects of VCs’information production activities on the time series evolution of initial returns to IPOs over time.
M
Empirical results in this section are organized as follows. In Table II, I compare initial returns that accrue to IPOs backed by new entrant or incumbent VCs using my entire sample
d
of venture capital funds; that is, funds that commenced business between 1980 and 2005.
te
These empirical results demonstrate that cross-sectional variations in initial returns to VC backed IPOs are explained by the information asymmetry rationale for IPO underpricing
Ac ce p
proposed in Rock (1986).
In Table III, I compare three-year trends in initial returns that accrue to IPOs backed by new entrant or incumbent VCs. Since two di¤erent groups of VCs either are identical or dissimilar in so far as market reputation is concerned, these empirical tests generate results that can be interpreted within the contexts of Propositions 1, 2, or 3. In Table IV, I utilize a matched sample of VC and non-VC backed IPOs to test for price convergence between VC and non-VC backed IPOs. In light of empirical evidence of di¤erences in the market reputation of VC and non-VC backers of IPOs in Megginson and Weiss (1991) and di¤erences in levels of initial returns to IPOs in Lee and Wahal (2004), empirical tests in Table IV generate results that can be interpreted within the context of 21 Page 21 of 50
Propositions 2 or 3 only. In Tables V and VI (Section VI), I provide empirical evidence that
ip t
price convergence observed in Table IV can be attributed to the arrival of new favorable information about VC backed IPOs.
The Cross-section of VC-backed IPOs
cr
A
portfolio activity at a time. Let
n
and
e
us
The matching algorithm matches new and incumbent VCs relative to only one measure of denote, respectively, average initial returns to
(Ho) in Tables II and III is: Ho :
n
e
an
IPOs backed by new entrant or incumbent VCs, then the null hypothesis for each time period = 0. Let
n
and
e
denote, respectively, the
M
distributions of initial returns to IPOs backed by new entrant or incumbent VCs, then the 'e = 0; that is,
d
test for equality of distributions in the same tables examines whether 'n
te
examines whether there exist signi…cant di¤erences in dispersion of initial returns to IPOs between new and incumbent VCs’portfolios. These tests for equality of distributions are
Ac ce p
implemented using the Wilcoxon signed rank test. In light of the validation of industry or investment stage specialization and syndication activity as measures of the risk of VCs’portfolios in prior studies, empirical tests reported in Table II focus on the validation of portfolio innovation focus as a measure of the severity of information asymmetry problems in so far as VCs’portfolios are concerned. Empirical results in Table II show initial returns to IPOs range between 28:65% for portfolios characterized by low innovation focus to 37:49% for portfolios characterized by high innovation focus. Given the data in this study includes IPOs that occurred between 2001 and 2005, these levels of initial returns to IPOs compare favorably with average levels of 26:82% in Lee and Wahal (2004) for IPOs that occurred between 1980 and 2000. 22 Page 22 of 50
A comparison of initial returns that accrue to new entrant funds shows initial returns
ip t
increase with innovation focus from 29:14% to 32:25%, and then to 36:72 percent. Among incumbent funds, initial returns increase from 28:15% to 35:31%, and then to 38:26% as
cr
innovation focus increases from low to high innovation focus. Since portfolio risk increases from low to high innovation focus, these trends in initial returns to IPOs provide evidence
us
that portfolio innovation focus is a measure of the severity of information asymmetry
an
problems between VC backed issuers of IPOs and investors.
Empirical results generated by the Wilcoxon signed rank tests show the distributions of
M
initial returns that accrue to IPOs backed by new entrant or incumbent VCs di¤er signi…cantly. This …nding indicates new entrant and incumbent VCs’portfolio companies
d
have unique innovation characteristics that are re‡ected in initial returns during IPOs.
te
Combined, empirical …ndings in Tables I and II provide evidence that portfolio innovation focus is a good measure of portfolio risk within the cross-section of the venture capital
Ac ce p
market and a good measure of the severity of information asymmetry between VC backed issuers and investors.
Insert Table Il About Here
B
Time trends in initial returns to IPOs
I examine trends in initial returns that accrue to funds which commenced business between 1981 and 1984. As discussed in Section III, the choice of a three-year convergence time frame is predicated on the utilization of three-year time frames for measuring VCs’ reputation in Nahata (2009), Krishnan and Masulis (2011), and Obrimah (2015). Given the venture capital industry in the USA was still quite young in the early 1980s, and the 23 Page 23 of 50
expectation that the market power or in‡uence of VCs grows over time (see for example,
ip t
Nahata (2009)), the choice of the “…rst three years” in my sample for the examination of time trends in initial returns provides the lowest probability of observing price convergence in
cr
initial returns to IPOs.
In Table III, I report three-year trends in initial returns to IPOs. Since average initial
us
returns to IPOs associated with funds that commenced business between 1981 and 1984
an
cannot be constrained to be equal to average initial returns over the entire sample period of 1980 through 2005, the levels of initial returns in Table III are not comparable to those in
M
Table II. In discussing the results in Table III then, only time trends in initial returns are important.
d
The results in Panels A, B, and C of Table III are for time trends in initial returns to
te
IPOs for portfolios matched relative to portfolio specialization or syndicate size. Regardless of the speci…c portfolio activity, the results show initial returns that accrue to IPOs backed
Ac ce p
by new or incumbent VCs converge over time. The fastest convergence rate is relative to industry specialization, while the slowest convergence rate is relative to syndicate size. Viewed within the context of the formal theory in Section II, empirical results in Panels A through C are consistent with equilibrium conditions established in Proposition 1. This is evident in trends within Panel A for instance, which show that while average initial returns remain relatively stable, with realizations between 18:42% (Period 3) and 19:73% (Period 2), initial returns for new entrant VCs decrease from a high of 24:09% to 18:66%, and initial returns for incumbent VCs increase from a low of 13:01% to 18:17% between periods one and three. Empirical results in Panel A show equilibrium conditions, 1 t
1 t+n
<0
t+n
1 t+n
; and
2 t+n
t+n
0<
2 t
2 t+n
derived in
24 Page 24 of 50
Proposition 1, which result in the relations, (13:01 18:66
0 < (24:09
18:66+18:17 2
18:17 ; and
18:66) are satis…ed. These …ndings demonstrate, in so
ip t
18:66+18:17 2
18:17) < 0
far as information production activities are concerned, that new entrant and incumbent VCs
cr
are similar in market reputation.
Consistent with the prediction in Proposition 1 that di¤erences in dispersion of initial
us
returns converge along with mean realizations of initial returns, Wilcoxon signed rank tests
an
show di¤erences in distributions fully converge alongside di¤erences in mean realizations of initial returns to IPOs by the end of Period Three. Given convergence of distributions is a
M
necessary condition within the context of Proposition 1, the absence of full convergence in distributions by the second year and the presence of full convergence in distributions by the
te
within venture capital markets.
d
third year provides empirical validation for a three-year reputation development time frame
Ac ce p
Insert Table IlI About Here
In Panels D, E, and F of Table III, I report time trends in initial returns relative to the three di¤erent groupings of portfolio innovation focus. The results show initial returns that accrue to IPOs backed by new entrant or incumbent VCs converge over time. In Panels E and F, convergence results in di¤erences in initial returns that are economically and statistically insigni…cant within one period. These di¤erences decrease from economically signi…cant levels of between 9:66% and 10:47% in Panels E and F (Period One) to become economically and statistically insigni…cant by Period Three. In Panel D, while di¤erences in initial returns remain signi…cant during Period Three, they attenuate signi…cantly from an absolute value of 5:46% during the second period to an absolute value of 2:02% during the 25 Page 25 of 50
third period.
ip t
With respect to the distribution of initial returns, Wilcoxon signed rank tests in Panels D, E, and F show that while the distributions of initial returns that accrue to new entrant or
cr
incumbent VCs start o¤ statistically di¤erent, these di¤erences are totally attenuated by
Period 3. This …nding indicates the unique components of VCs’innovations get priced over
us
time, such that new entrant and incumbent VCs’innovations do not appear any di¤erent
an
from each other within three periods. The pricing of the unique characteristics of VCs’ innovations within three time periods ensures VCs have to continue to innovate in order to
M
earn innovation rents from their portfolio companies.
As is the case in Panels A through C, empirical results in Panels D through F are
d
consistent with equilibrium conditions established in Proposition 1. While the distribution of
te
di¤erences in initial returns reveals some divergence between Periods 1 and 2 in Panels A through C, however, the distribution of di¤erences in initial returns converges between
Ac ce p
Periods 1 and 2, and fully converges by Period 3 in Panels D through F. These …ndings indicate information produced by VCs relates more to the innovation dimension of portfolio …rms, as opposed to the diversi…cation bene…ts of portfolio …rms, with diversi…cation re‡ected in investment stage or industry specialization and syndication activity. Consistent with these inferences, a comparison of empirical results in Panels A through F shows initial returns to IPOs converge the fastest in empirical tests within which portfolio activity is de…ned relative to median or high innovation focus. In light of the demonstrated importance of diversi…cation for viability of commercial banking (see for example, Diamond (1984)), empirical results provide evidence that relative to diversi…cation strategies, innovation focus is a more important driver of portfolio asset allocations or portfolio management within 26 Page 26 of 50
venture capital markets.
The universe of IPOs
ip t
C
cr
Empirical tests in this sub-section examine three-year time trends in initial returns to IPOs over three fund-raising or investment cycles for an aggregate time period of 9 years. Studies
us
that …nd VC backed IPOs have lower initial returns in relation to non-VC backed IPOs include Megginson and Weiss (1991) and Belghitar and Dixon (2012). Studies which provide
an
evidence that VC backed IPOs have higher initial returns in relation to non-VC backed IPOs include Lee and Wahal (2004). Studies that provide mixed evidence on relations between
M
underwriter reputation and initial returns to IPOs include Loughran and Ritter (2004),
d
Kirkulak and Davis (2004), Dimovski, Philavanh, and Brooks (2011), and Chemmanur and
te
Fulghieri (1994). In light of these …ndings and theoretical predictions in Proposition 3, ‡ips in relations between initial returns to VC or non-VC backed IPOs in Table IV are consistent
Ac ce p
with outcomes of information production.
Panel A of Table IV reports empirical results from the matching of non-VC backed IPOs to VC backed IPOs within the same industry, regardless of whether the VC backed IPOs are associated with new entrant or incumbent VC funds. Empirical results show initial returns to VC backed IPOs start o¤ much lower than initial returns to non-VC backed IPOs, with the starting di¤erential amounting to about 97% of initial returns to non-VC backed IPOs during the …rst period (1987 through 1989). During the second period (1990 through 1992), the di¤erence in initial returns to IPOs between the two groups attenuates to 32 percent. By the third period under consideration (1993 through 1995), the di¤erence amounts to only 7.40%. Viewed within the context of Proposition 3, empirical results in Panel A provide 27 Page 27 of 50
evidence of a strategic information production response on the part of VCs between the …rst
ip t
and second periods, and continuation of information production between the second and third periods. Speci…cally, while transitions between Period 1 and Period 2 satisfy the
1 t+1
2 t+1
1 t
2 t
>0&
<0&
1 t
1 t+1
2 t+1
2 t
< 0 followed by
cr
condition, “
0 ” in Proposition 3, and while transitions between
or
between Period 2 and Period 3 are not
an
asymmetry problems, transitions in either of
us
Period 2 and Period 3 reveal evidence of information production that mitigates information
well behaved. Consistent with continuation of information production between Period and
during Period 2 to 8.28 during Period 3.
M
Period 3, however, the di¤erence in distributions attenuates by about 20:01% from 10.36
d
Empirical results in Panels B through G are for IPOs matched by industry 1-digit SIC
te
codes and year of IPO, but with the resulting matches reported using segmentations that relate to the di¤erent measures of portfolio risk; that is using the industry specialization
Ac ce p
(Panel B), investment stage specialization (Panel C), syndicate size (Panel D), low innovation (Panel E), median innovation (Panel F), and high innovation (Panel G) characteristics of the comparable venture capital portfolios. Given the segmentation is only possible with the venture capital backed IPOs, average initial returns to non-VC backed IPOs remain the same in each period, regardless of the measure of portfolio risk. The results in Panels B through G provide evidence of convergence in initial returns to IPOs in relation to all of the di¤erent facets of portfolio activities. Viewed within the context of conditions in Propositions 3, convergence in initial returns to IPOs evidences a strategic response between periods one and two and an information production response between periods two and three only in relation to median or high innovation focus and 28 Page 28 of 50
syndication activity. This …nding implies trends in Panel A for the entire sample of VC
ip t
backed IPOs largely re‡ect the innovation dimensions of VCs’portfolios. These …ndings are consistent with empirical results in Table III, which show information production induces the
cr
fastest convergence speeds with respect to the median or high innovation focus dimensions of portfolio activities. Combined, empirical …ndings in Tables III and IV provide evidence that
us
portfolio innovation focus is an important dimension for understanding VCs’portfolio
management of portfolio risk-return trade-o¤s.
an
activities, and implications of these activities for portfolio risk, portfolio strategies, and
The source of convergence in initial returns to VC and
d
VI
M
Insert Table IV About Here
te
non-VC backed IPOs
Ac ce p
Combined, empirical results in Tables III and IV provide evidence consistent with price convergence induced by venture capitalists’information production activities. If the observed convergence in initial returns to IPOs are induced by VCs’information production activities, alternative measures of IPO risk that are not simultaneously measures of the severity of information asymmetry between issuers and investors will indicate VC backed IPOs are less risky than their non-VC backed counterparts; that is, will indicate the steady increase in initial returns to VC backed IPOs in Table IV is directly attributable to VCs’ information production activities. In order to compare the relative riskiness of VC and non-VC backed IPOs, I utilize a regression framework to examine the relation between underwriting spreads, gross spreads, 29 Page 29 of 50
the proportional change between the mid…le price and the o¤er price, post-IPO price-to-book
ip t
values of sample IPOs, and di¤erences in initial returns between VC and non-VC backed IPOs. Underwriting or gross spreads have been shown to increase with IPO issuer risk in
cr
studies such as Altinkilic and Hansen (2000), Chemmanur and Fulghieri (1994), Chen and
Ritter (2000), and Eckbo, Masulis, and Norli (2007). The proportional change between the
us
mid…le price and o¤er price has been utilized as a measure of IPO issuer risk in Benveniste
an
and Spindt (1989) and Loughran and Ritter (2002). The conventional assumption in securities valuation that in the presence of …rms that are otherwise similar, …rms with lower
M
price-to-book ratios are better buys than those with higher price-to-book ratios implies post-IPO price-to-book ratios are good indicators of risk within stock markets, with risk an
d
increasing function of price-to-book ratios.
te
The regression models are speci…ed as:
0
+
0 initialrtnsprdi
+ "0i
(1a)
grossi =
1
+
1 initialrtnsprdi
+ "1i
(1b)
midtof f eri =
2
+
2 initialrtnsprdi
+ "2i
(1c)
p=bi =
3
+
3 initialrtnsprdi
+ "3i
(1d)
Ac ce p
writeri =
where writer are underwriting spreads; gross are gross spreads; midto¤ er are proportional changes between the mid…le and o¤er price; and p/b are price-to-book ratios for the venture capital backed IPOs.2 Let vc and nvc denote VC and non-VC backed IPOs, 2
Consistent with stylized evidence, gross spreads lie between 7.00% and 8.00% and exhibit signi…cant clustering evident in standard deviations that range between 1.56% and 1.80%. The proportional change between the mid-…le and o¤er price lies between -2.80% and 2.80%, and price-to-book ratios lie between 3.00 and 5.00.
30 Page 30 of 50
respectively, while initialrtn denotes initial returns to IPOs, then the variable initialrtnsprd
ip t
(initial return spread ) is de…ned as the di¤erence between initial returns to VC backed and
initialrtnnvc :
us
initialrtnsprd = initialrtnvc
cr
non-VC backed IPOs. That is,
The speci…cations of equations (1a) through (1d) re‡ect predictions or …ndings in Yeoman
an
(2001) and Tinic (1988) that underwriting spreads and initial returns to IPOs are jointly determined (during the book building process) prior to their realizations on the o¤er date of
M
an IPO. Given the jointness of determination, it is immaterial whether initial return spread is stipulated as a dependent or independent variable within a univariate regression framework.
d
Using the matched sample of VC and non-VC backed IPOs from Table IV, the results in
te
Panels A and B of Table V show - wherever signi…cant - that an increase in initial return
Ac ce p
spread is associated with a decrease in underwriting spreads, gross spreads, and post-IPO price-to-book ratios associated with VC backed IPOs. These negative relations demonstrate increases in initial returns to VC backed IPOs in Table IV, which are accompanied by shrinkage in initial returns spreads, are induced by increases in the risk of VC backed IPOs. This increase in the risk of VC backed IPOs is consistent with a strategic response from VCs in so far as the application of their information production skills are concerned. The book building theory of Benveniste and Spindt (1989) and the con‡ict of interest theory of Loughran and Ritter (2002) both predict the IPO o¤er price is not fully adjusted from the midpoint of the …le price range whenever underwriters receive favorable information These distributional characteristics are available from the author upon request.
31 Page 31 of 50
(see also, Welch and Ritter (2002)) regarding an IPO.
Consistent with the arrival of new
ip t
favorable information in respect of VC backed IPOs, the proportional change between the IPO mid…le and o¤er prices decreases with initial return spread. In so far as venture capital
cr
backed IPOs are concerned then, increases in initial returns to VC backed IPOs (in relation to initial returns that accrue to non-VC backed IPOs) are consistent with the arrival of new
us
favorable information about VC backed IPOs. Since the arrival of new favorable information
an
implies a decrease in severity of information asymmetry, empirical results in Table V provide evidence that convergence in initial returns in Table IV is predicated on VCs’information
M
production activities.
d
Insert Table V About Here
te
In Gompers (1996), IPOs backed by new entrant VCs attract higher initial returns to IPOs in relation to IPOs backed by incumbent VCs, with this phenomenon induced by
Ac ce p
“grandstanding” on the part of new entrant VCs. Empirical results that obtain during Period I of Table III are consistent with the grandstanding hypothesis in that initial returns that accrue to IPOs backed by new entrant VCs are larger than those that accrue to IPOs backed by incumbent VCs. While this …nding provides support for the cross-sectional predictions of the grandstanding hypothesis, increases in initial returns to IPOs backed by reputable VCs between Periods 1 and 2 in Table III and increases in initial returns to the universe of VC backed IPOs in Table IV simultaneously demonstrate the inapplicability of the grandstanding hypothesis to explanations of time series variability in underpricing of VC backed IPOs.
32 Page 32 of 50
VII
Conclusions
ip t
I …nd venture capitalists’(VCs’) information production activities induce time series convergence in initial returns that accrue to IPOs backed by new entrant or incumbent VCs
cr
within a three-year span. These …ndings provide empirical validation for the utilization of a
us
three-year time frame for assessing market reputation within the venture capital market in Nahata (2009), Krishnan and Masulis (2009), and Obrimah (2015). In empirical tests that
an
compare time series trends in initial returns to IPOs, I …nd VCs’information production activities induce convergence in initial returns that accrue to VC and non-VC backed IPOs,
M
with convergence directly attributable to the arrival of new and favorable information about VC backed IPOs. These …ndings provide empirical evidence that VCs’information
d
production activities induce improvements in the pricing of non-VC backed IPOs, indicating
te
VCs’information production activities induce price discovery within stock markets. Given
Ac ce p
stock prices are expected to be higher in markets characterized by price discovery processes, empirical …ndings show venture capital can have a direct impact on economic growth via it’s e¤ect on the sizes of stock markets. In light of the fact that market e¢ ciency requires price discovery processes and induces market development (more e¢ cient allocation of resources), empirical …ndings demonstrate VCs’information production activities are directly bene…cial for economic development.
33 Page 33 of 50
Appendix
1 t
2, t
=
1
&
2;
equivalence of reputation of …nancial intermediaries does not imply
and as such does not violate the assumption,
1 t
2. t
<
In the presence of
cr
…rms in
ip t
Proof of Proposition 1. Conditional on within-class di¤erences in initial risk pro…les of
1 t+n
that 1 t+n
2 t+n
2 t+n
<
<
1 t
1 t
2 t 2 t
a¤ects the pricing of …rms in
1
. Suppose
t
, implying (since
1 t
<
1
M
1 t+n
2 t+n
<
and
2 t+n
<
2. t
We have
2 t
(2)
d
t+n
1 t
>
is the mean realization of severity of information asymmetry for the universe
and
3 2.
Equation (2) implies,
Ac ce p
of …rms in
t+n
1 t+n
2) t
te
where
<
and vice-versa, such
= g ( t ) : It must be the case that
then that: 1 t
2,
an
produced in relation to …rms in
us
equivalence of reputation of …nancial intermediaries, it must be the case that information
1 t
1 t+n
t+n
2 t+n
< 0 0<
t+n 2 t
1 t+n
(3)
2 t+n
(4)
QED.
Proof of Proposition 2. Suppose …nancial intermediaries associated with …rms in more reputable than …nancial intermediaries associated with …rms in
2
1
are
in so far as the
3 In the presence of similarities in reputation, 1t < 2t can be interpreted as a strategic outcome. Suppose otherwise, then holding reputation …xed and assuming investor rationality, a strategic response results in the …nancial intermediary in 1 going after riskier …rms and the …nancial intermediary in 2 going after less risky 2 2 …rms. This results in 1t < 1t+n t+n t+n < t ; that is, exactly the same condition induced by information production.
34 Page 34 of 50
quality of information production is concerned. Then it must be the case that information
implies
1 t+n
=g
1 t+n
2 t+n
and
a¤ects the pricing of …rms in
1
1 t+n
=h
non-invertible function of . Since changes in = (
2 ) t+n
1 t+n
= v(h(g
1
but not vice-versa. This
); with h an increasing but
also can e¤ect
2,
cr
2 t+n
1 t+n
= h(g
2,
ip t
produced in relation to …rms in
)), with v a decreasing, but non-invertible function of . Given 2 t+n
1 t+n ,
1 t+n
implying the condition
2 t+n
need not 1 t
<
2 t
can be
an
necessarily converge towards
us
v is a decreasing function, while both h and g are increasing functions,
violated within a well behaved equilibrium.
M
In light of evidence that higher reputation …nancial intermediaries can at di¤erent points in time be associated with …rms characterized by either of lower or higher severity of
d
information asymmetry (see for example, Loughran and Ritter, 2004; Kirkulak and Davis,
necessarily imply
1 t
<
2. t
te
2004; and Dimovski et al., 2011), more reputable …nancial intermediaries in Suppose
1 t
1 t+n
<
2 t+n
Ac ce p
information production induces <
2 t.
However, since
1 t+n
<
1 t
>
1 t.
and
2. t t
1
does not
Then in the well behaved scenario, 1 t+n
= g ( t ) implies
2 t+n
<
2 t
imply
1 t
>
<
t+n
1 t
!
and
2 t
1 t
2 t+n
<
>
t+n
> 2 t
2 t
and
!
equilibrium conditions in equations (3) and (4) are violated and the only inviolate conditions are:
i t+n
<
i t
8i
(5)
i t+n
<
i t
8i
(6)
2 t
2 t+n
= j(
2 t
2 t+n
= j (
1 t 1 t
1 t+n ) 1 t+n )
(7) (8)
35 Page 35 of 50
where j and j are increasing but non-invertible functions of ( 1
1 ) t+n
or (
1 t
1 t+n ).
are more reputable than …nancial intermediaries in
implies only …nancial intermediaries in
1
can credibly switch strategies. Since 1 t+n
>
2; t
2. t
…nancial intermediary can result in either of
Strategic positioning by the more reputable 1 t+n
1 t
1 t+n
an
<
us
QED. 1 t
2, t
>
that is, exactly the same equilibrium
conditions in equations (5) through (8).
Proof of Proposition 3. Suppose
2 t
>
cr
1 t
however, a switch in strategy leads to
1 t
2, 1 t
Since
ip t
…nancial intermediaries in
1 t
<
(information production) or
>
1 t
M
(strategic application of superior information production capabilities to a riskier subset of …rms within the same general risk class), the number of equilibrium conditions that can be 1 t
<
2 t
1 t
are potentially in…nite. For concreteness,
<
2 t
d
generated within the context of
te
indicates investors are willing to tolerate risk more severe than risk currently borne by 1 t+n
reputable …nancial intermediaries. This implies
>
1 t
is feasible; that is, reputable
Ac ce p
…nancial intermediaries can take on higher risk projects to bring their reputation or superior information production skills to bear on the pricing of projects with information asymmetry 2. t
levels of
Since this action can induce less reputable …nancial intermediaries to switch
strategies in order to remain competitive; that is, induce a switch to projects with risk rating 1, t
1 t
<
we can observe a switch in ordering of initial returns to IPOs; that is,
1 t+n
2 t+n
2 t
<
(convergence) is just as feasible as
1 t+n
>
2 t
>
1 t
2 t+n
(divergence). In equilibrium then, we are unable to predict relations between or the location of equilibrium risk parameter, 1 t
2 t
<0&
1 t
2 t
t+n . 1 t+1
< 0 followed by
1 t+n
and
2 t+n
We can say, however, that 2 t+1
>0&
1 t+1
2 t+1
> 0 is
evidence of a strategic information production response. QED. 36 Page 36 of 50
References
ip t
Abdou, K., and O. Varela, 2009. Is there a puzzle in the failure of venture capital backed portfolio companies? Applied Financial Economics 19, 1439-1452.
cr
Aggarwal, V.A., and D.H. Hsu, 2013. Entrepreneurial exits and innovation, Management
us
Science 60, 867-887.
an
Alexander, J., 1993. The lawsuit avoidance theory of why Initial Public O¤erings are underpriced. UCLA Law Review 41, 17-73.
M
Allen, F., and G. Faulhaber, 1989. Signaling by underpricing in the IPO market, Journal of Financial Economics 23, 303-323.
te
d
Altinkilic, O., & Hansen, R.S., 2000. Are there economies of scales in underwriting fees? Evidence of raising external …nancing costs, Review of Financial Studies 13: 191–218.
Ac ce p
Barry, Christopher B., Muscarella, Chris J., Peavy, John W., and Michael R. Vetsuypens, 1990. The role of venture capital in the creation of public companies: evidence from the going-public process, Journal of Financial Economics 27, 447-71. Beatty, R., and J. Ritter, 1986. Investment banking, reputation, and the underpricing of initial public o¤erings, Journal of Financial Economics 15, 213-232. Belghitar, Y., and R. Dixon, 2012. Do venture capitalists reduce underpricing and underperformance of IPOs? Applied Financial Economics 22, 33-44. Benveniste, L.M., and P.A. Spindt, 1989. How investment bankers determine the o¤er price and allocation of new issues, Journal of Financial Economics 24, 343-362. 37 Page 37 of 50
Berger, A.N., and T.H. Hannan, 1989. The price-concentration relationship in Banking, The
ip t
Review of Economics and Statistics 71, 291-299. Brockman, P., Liebenberg, I., and M. Schutte, 2010. Comovement, information production,
cr
and the business cycle, Journal of Financial Economics 97, 107-129.
us
Campbell, T.S., and W.A. Kracaw, 1980. Information Production, Market Signalling, and the
an
Theory of Financial Intermediation, Journal of Finance 35, 863-882.
Casamatta, C., and C. Haritchabalet, 2007. Experience, screening, and syndication in venture
M
capital investments, Journal of Financial Intermediation 16, 368-398. Chemmanur, T.J., and Fulghieri, P., 1994. Investment Bank Reputation, Information
d
Production, and Financial Intermediation. Journal of Finance. 49: 57-79.
te
Chemmanur, T.J., Krishnan, K., and D.K. Nandy, 2011. How does venture capital …nancing
Ac ce p
improve e¢ ciency in private …rms? A look beneath the surface, Review of Financial Studies 24, 4037-4090.
Chen, H.C., and J.R. Ritter, 2000. The seven percent solution, Journal of Finance 55, 1105-1131.
Diamond, D., 1984. Financial Intermediation and Delegated Monitoring. Review of Economic Studies 51, 393-414. Dimovski, W., Philavanh, S. and R. Brooks, 2011. Underwriter reputation and underpricing: evidence from the Australian IPO market, Review of Quantitative Finance and Accounting 37, 409-426.
38 Page 38 of 50
Drake, P., Vetsuypens, M., 1993. IPO underpricing and insurance against legal liability
ip t
Financial Management 22, 64-73.. Eckbo, B.E., Masulis R.W., & Norli, O. 2007. Security o¤erings. In: E. Eckbo, Editor,
cr
Handbook of Corporate Finance: Empirical Corporate Finance, Elsevier, North-Holland,
us
Amsterdam.
Esty, B.C., and W.L. Megginson, 2003. Creditor rights, enforcement, and debt ownership
an
structure: Evidence from Global Syndicated Loan Market, Journal of Financial and
M
Quantitative Analysis 38, 37-59.
Fama, E.F., 1991. E¢ cient Capital Markets: II, Journal of Finance 46, 1575-1617.
d
Fernando C.S., Gatchev, V.A. and P. Spindt, 2005. Wanna dance? How …rms and
te
underwriters choose each other. Journal of Finance 60, 2437-2469.
Ac ce p
Fulghieri, P., and M. Sevilir, 2009. Size and focus of a venture capitalists’portfolio. Review of Financial Studies. 22, 4643-80.
Gompers, P., 1996. Grandstanding in the Venture Capital Industry. Journal of Financial Economics 42, 133-156.
Gompers, P., Kovner, and J. Lerner, 2009. Specialization and Success: Evidence from Venture Capital, Journal of Economics and Management Strategy 18, 817-844. Gompers, P., and J. Lerner, 2004. The venture capital cycle, 2nd edition, Cambridge: MIT Press.
39 Page 39 of 50
Grinblatt, M., and C. Hwang, 1989. Signaling and the pricing of new issues, Journal of
ip t
Finance 44, 393-420. Hochberg, Y., Ljungqvist, A., and Y. Lu, 2007. Whom you know matters: Venture capital
cr
networks and investment performance. Journal of Finance. 62, 251-301.
us
Huang, R.R., 2008. Tolerance for uncertainty and the growth of informationally opaque
an
industries, Journal of Development Economics 87, 333-353.
Ibbotson, R., Sindelar, J., Ritter, J., 1994. The market’s problems with the pricing of initial
M
public o¤erings. Journal of Applied Corporate Finance 7, 66–74.
King, Robert G., and Ross Levine, 1993. Finance and growth, Schumpeter might be right,
d
Quarterly Journal of Economics 108, 717-737.
te
Kirkulak, B., and C. Davis, 2004. Underwriter reputation and underpricing: Evidence from
Ac ce p
the Japanese IPO market, Paci…c-Basin Finance Journal 13, 451-470. Knill, A., 2009. Should venture capitalists put all their eggs in one basket? Diversi…cation versus pure play strategies in venture capital. Financial Management. 38, 441-486. Kortum, S. and J. Lerner, 2000. Assessing the contribution of venture capital to innovation, RAND Journal of Economics vol. 31(4), 674–92. Krishnan, C.N.V., and R.W. Masulis, 2011. Venture Capital Reputation: A Survey in “The Oxford Handbook of Venture Capital ”, Dougas J. Cummings, ed., Oxford University Press. Lee, P M., and S. Wahal, 2004. Grandstanding, certi…cation and the underpricing of venture capital backed IPOs, J ournal of Financial Economics 73, 375-407. 40 Page 40 of 50
Leland, H.E., and D.H. Pyle, 1977. Informational Asymmetries, Financial Structure, and
ip t
Financial Intermediation, Journal of Finance 32, 371-387. Lerner, J., 1994. The Syndication of Venture Capital Investments, Financial Management 23,
cr
16-27.
an
case of innovation. Journal of Finance 66(2): 445-477.
us
Lerner, J., M. Sorensen, P. Stromberg. 2011. Private equity and long-run investment: The
Loughran, T., and J.R. Ritter, 2002. Why don’t issuers get upset about leaving money on the
M
table in IPOs? Review of Financial Studies 15, 413-443.
Loughran, T., and J. Ritter, 2004. Why has IPO underpricing changed over time? Financial
d
Management 33, 5-37.
te
Lowry, M., and W.G. Schwert, 2002. IPO Market Cycles: Bubbles or Sequential Learning?,
Ac ce p
Journal of Finance 52, 1171-1200.
Lowry, M., and S. Shu, 2010. Litigation risk and IPO underpricing, Journal of Financial Economics 65, 309-335.
Luintel, K.B., and M. Khan, 1999. A quantitative reassessment of the …nance-growth nexus: evidence from a multivariate VAR, Journal of Development Economics 60, 381-405. Madan, D., and J. J. Yen, 2004. Asset allocation with multivariate non-Gaussian returns, Chapter 23 in Handbooks in operations research and management science: Financial Engineering 15, Eds. John Birge, and Vadim Linetsky, North-Holland.
41 Page 41 of 50
Megginson, William L., and Kathleen A. Weiss, 1991. Venture capitalist certi…cation in
ip t
initial public o¤erings, Journal of Finance 46, 879-903. Myers, S. C., 1977. “Determinants of Corporate Borrowing,” Journal of Financial Economics
cr
5, 147-175.
us
Nahata, R., 2009. Venture capital reputation and investment performance. Journal of
an
Financial Economics 90. 127-51.
Obrimah, O.A., 2015, How important is innovationn for venture capitalists’(VCs’) market
M
reputation, Quarterly Review of Economics and Finance, forthcoming, 10.1016/j.qref.2015.11.007.
d
Pastor, L., and P. Veronesi, 2005. Rational IPO waves, Journal of Finance 60, 1713-1757.
te
Purnanandam, A. K., and B. Swaminathan, 2004. “Are IPOs Really Underpriced?” Review
Ac ce p
of Financial Studies, 17, 811-848.
Rajan, Raghuram G., and Luigi Zingales, 1998. Financial Dependence and Growth, American Economic Review 88, 559-586. Ritter, J.R., 1984. The ‘hot issue market’of 1980, Journal of Business 57, 215-239. Rock, K., 1986. “Why New Issues are Underpriced,” Journal of Financial Economics, 15, 187-212. Samila, S., and O. Sorenson, 2011. Venture capital, entrepreneurship, and economic growth, Review of Economics and Statistics 93, 338-349.
42 Page 42 of 50
Schweinbacher, A., 2008. Innovation and venture capital exits, The Economic Journal 118,
ip t
1888-1916. Tinic, S., 1988. Anatomy Of initial public o¤erings of common stock, Journal of Finance 43,
cr
789-822.
us
Welch, I., 1989. Seasoned o¤erings, imitation costs and the underpricing of initial public
an
o¤erings, Journal of Finance 44, 421-449.
Yeoman, J.C., 2001. The optimal spread and o¤ering price for underwritten securities,
Ac ce p
te
d
M
Journal of Financial Economics 62, 169-198.
43 Page 43 of 50
44
Page 44 of 50
Table I: Summary statistics for VCs’portfolio activities
d
te
Ac ce p
us
an
M
cr
ip t
This table reports summary statistics for measures of portfolio activity utilized in the matching of new entrant and incumbent venture capital funds. Year is the year a fund commenced business; # of funds is the number of funds that commenced business in a particular year; industry and investment stage specialization are constructed using an Her…ndahl-Hirschman methodology; that is, as the sum of squares of an industry or investment stage’s share in aggregate capital disbursements to each fund’s portfolio companies. The industry and investment stage classi…cations are from VentureXpert and are (1) Non-hitech; (2) medical/health; and (3) hitech for industry classi…cations and (a) start-up; (b) early stage and (c) expansion stage for investment stage classi…cations. Syndicate Size is constructed as the average size of …nancing syndicates that involve a VC i. Each VC’s portfolio ventures are classi…ed as (i) low, (ii) median, or (iii) high innovation focus ventures using proportions of capital disbursements within the following percentiles: low innovation ventures lie within the top 25% of disbursements; median innovation ventures lie between the 25th and 75th percentiles of disbursements, while high innovation ventures lie in the bottom 25% of capital disbursements. The venture capital data are from VentureXpert and are for funds that commenced business between 1980 and 2005. Industry Investment Stage Innovation Focus Year # of funds Specializaton Her…ndahls Specialization Her…ndahls Low Median High Syndicate Size 1980 13 0.67 0.48 0.14 0.75 0.11 4.25 1981 29 0.70 0.35 0.00 0.69 0.31 4.67 1982 42 0.57 0.41 0.00 0.94 0.06 4.50 1983 53 0.54 0.47 0.24 0.65 0.11 3.27 1984 63 0.56 0.49 0.29 0.64 0.13 1.73 1985 47 0.73 0.63 0.45 0.49 0.06 4.30 1986 41 0.50 0.50 0.00 0.74 0.26 3.10 1987 66 0.91 0.66 0.43 0.54 0.03 2.50 1988 45 0.58 0.72 0.00 0.48 0.52 2.13 1989 53 1.00 0.45 0.00 0.91 0.09 2.73 1990 29 0.59 0.46 0.00 0.91 0.09 2.22 1991 22 0.59 0.39 0.00 0.77 0.23 1.50 1992 42 0.93 0.35 0.00 0.89 0.11 2.20 1993 50 1.00 1.00 0.00 1.00 0.00 2.00 1994 79 0.99 0.36 0.00 0.91 0.09 1.93 1995 87 0.51 0.77 0.00 1.00 0.00 3.00 1996 96 1.00 0.45 0.95 0.04 0.01 1.25 1997 148 1.00 0.43 0.76 0.24 0.00 2.00 1998 174 1.00 1.00 0.00 0.76 0.23 4.00 1999 241 1.00 0.69 0.68 0.32 0.00 3.60 2000 298 0.51 0.42 0.81 0.18 0.00 3.00 2001 thro 2005 957 0.73 0.60 0.84 0.15 0.00 2.71 Global Mean 0.71 0.54 0.27 0.59 0.14 2.64 c c c c Correlations with High Innovation 0.2486 0.1488 -0.8700 -0.4669 1.0000 0.0077 c indicates signi…cance at the 1% con…dence level
45
Page 45 of 50
Table II: Cross-sectional di¤erences in initial returns to IPOs
e
n
e
'e n
e
'n
'e
n
e
n
e
High Innovation 38.26c -1.535b (0.03) 9439
ip t
9439
36.72c
cr
-2.153b (0.03)
us
an
n
Median Innovation 35.31c -3.062c (0.00) 5085
e
M
Low Innovation Initial Returns 29.14c 28.15c 0.983 2.118b 32.25c (0.08) (0.05) # of observations 10575 10575 5085 b & c denote signi…cance at the 5% and 1% con…dence levels, respectively.
n
'n
d
te
Ac ce p
'e -2.006b (0.05)
'n
This table reports empirical results derived from comparisons of initial returns that accrue to IPOs backed by venture capital funds that commenced business between 1980 and 2005. The cross-sectional comparisons are implemented using matched samples of VC funds that commenced business during immediately adjacent years, with measures of the funds’portfolio activities (portfolio risk) as matching dimensions. The matching algorithm is implemented as follows. Let yj 2 , yj 1 & yj denote three consecutive years or periods bounded by four points in time, tj 3 , tj 2 , tj 1 & tj ; fj 3 , fj 2 , fj 1 and fj denote, respectively, funds that commenced business during the four cohort years; nj 3 , nj 2 , nj 1 & nj denote, respectively, the number of venture capital funds that commenced business during the four cohort years; while pi denote a measure of portfolio activity or risk. Denote adjacent matches generated during the three cohort years by fj 1 jf j , and fj jf j+1 .Since new entrant funds are matched to incumbent funds, the number of matches in the two cohorts are determined by nj 1 and nj , respectively. This implies the number of matches in each cohort is unique whenever nj 1 6= nj . Combined, the matching algorithm and the fact that nj 1 6= nj for all j ensure adjacent matches fj 1 jf j , and fj jf j+1 do not consist of the same funds from year cohort yj . The matching algorithm is implemented relative to one measure of portfolio risk at a time. The measures of portfolio risk in this table are: (A) Low Innovation; (B) MedianInnovation; and (C) High Innovation focus. I describe the construction of each of these measures of portfolio risk in Table I. The venture capital data (including initial returns) are from VentureXpert, while other data on initial returns to IPOs are from the equity issues database of SDC Platinum. Di¤erences in means ( n e ) are two-sided tests of signi…cant di¤erences between new entrant and established VCs. ('n 'e ) are estimates of the extent to which the distribution of initial returns to IPOs for the two matched samples are identical, and are implemented using the Wilcoxon signed rank test. p-values for statistical signi…cance are in parentheses below reported parameters. New Older New Older New Older Fundsy Fundsy 1 Fundsy Fundsy 1 Fundsy Fundsy 1
46
Page 46 of 50
d
te
e
n
e
'n
'e n
M n
e
Period 2 12.79c 9.548c (0.00) 207
Panel B: Di¤erence in Means: 3-year time trends for Investment Stage Specialization Period 1 Period 2 c c c b c Initial Returns 25.69 14.01 11.68 2.163 19.34 20.88c -1.537 (0.00) (0.03) (0.45) # of observations 271 271 628 628
Panel C: Di¤erence in Means: 3-year time trends for Syndicate Size Period 1 Initial Returns 24.54c 9.30c 15.24c 2.184b 22.33c (0.00) (0.03) # of observations 78 78 207 b & c denote signi…cance at the 5% and 1% con…dence levels, respectively.
'e
3.556b (0.00)
0.194 (0.85)
n
e
Period 3 15.16c 4.930c (0.00) 414
1197
1197
414
e
Period 3 0.816 (0.51) 18.31c
19.13c
20.09c
n
Period 3 18.17c 0.489 (0.60) 1906
ip t
1906
18.66c
cr
3.395c (0.00)
'n
us
an
e
Panel A: Di¤erence in Means: 3-year time trends for Industry Specialization Period 1 Period 2 Initial Returns 24.09c 13.01c 11.08c 3.148c 20.53c 18.92c 1.610 (0.00) (0.00) (0.30) # of observations 442 442 987 987
n
'e
1.846 (0.07)
0.920 (0.36)
0.770 (0.44)
'n
This table compares three-year time trends in initial returns to IPOs that accrue to IPOs backed by venture capital (VC) funds that commenced business between 1981 and 1984. Comparisons of initial returns to IPOs are implemented using matched pairs of venture capital funds that commenced business during immediately adjacent years. The matching variables are six di¤erent measures of the risk of VCs’fund portfolios, with the matching algorithm implemented relative to one measure of portfolio risk at a time. I describe the matching algorithm that pairs funds that enter the market at time t + 1 to funds that entered the market at time t in Table II. The measures of portfolio risk are: (1) industry specialization; (2) investment stage specialization; (3) syndicate size; and three measures of innovation focus for each fund de…ned as (4) low innovation; (5) median innovation; and (6) high innovation. I describe the construction of each measure of portfolio activity in Table I. Period 1 results are for year cohort f81 jf 82 ; Period 2 results augment f81 jf82 with f82 jf83 ; while Period 3 results augment the Period 2 sample with f83 jf84 . The venture capital data (including initial returns) are from VentureXpert, while other data on initial returns are from the equity issues database in SDC Platinum. Di¤erence in means ( n e ) are two-sided tests of signi…cant di¤erences between new entrant and established VCs. ('n 'e ) are estimates of the extent to which the distribution of initial returns to IPOs for the two matched samples are identical, and are implemented using the Wilcoxon signed rank test. p-values for statistical signi…cance are reported in parentheses. New Older New Older New Older Fundsy Fundsy 1 Fundsy Fundsy 1 Fundsy Fundsy 1
Ac ce p
Table III: VCs’portfolio activities and attenuation of di¤erences in initial returns to IPOs: three-year time trends for VC backed IPOs only
47
Page 47 of 50
e
1
n
e
'n
Ac ce p
Older Fundsy
'e n
te
1180
1 n
e
Period 2 22.97c -5.456c (0.01) 1180
e
Older Fundsy
'e
1 n
e
Period 3 16.78c -0.616 (0.96) 826
Period 3 17.55c 1.085 (0.48) 652
Period 3 19.61c -2.020c (0.00) 338
e
Older Fundsy
ip t
826
16.72c
652
18.63c
338
17.59c
n
New Fundsy
cr
1.788 (0.07)
1.343 (0.18)
-0.804 (0.42)
'n
us
Period 2 16.92c 1.846 (0.34) 373
an
M
Panel F: Di¤erence in Means: 3-year time trends for High Innovation Focus Period 1 Initial Returns 21.87c 11.40c 10.47c 2.878c 18.77c (0.08) (0.00) # of observations 139 139 373 c denotes signi…cance at the 1% con…dence level.
d
Panel E: Di¤erence in Means: 3-year time trends for Median Innovation Focus Period 1 Period 2 Initial Returns 21.59c 11.92c 9.664c 2.542c 19.32c 19.58c -0.263 (0.01) (0.01) (0.92) # of observations 142 142 338 338
# of observations
Panel D: Di¤erence in Means: 3-year time trends for Low Innovation Focus Period 1 Initial Returns Not enough data 17.51c
n
New Fundsy
New Fundsy
'e
-0.649 (0.52)
1.155 (0.25)
-0.603 (0.55)
'n
Table III (contd): VCs’portfolio activities and attenuation of di¤erences in initial returns to IPOs: three-year time trends for VC backed IPOs only
48
Page 48 of 50
d
te
cj
v c)
c
dif f %
'v
'c v
'v
Period 2: 90-92 8.301c 32%c (0.00) 816
Panel C: Di¤erence in Means: 3-period time trends for Investment Stage Specialization Period 1: 87-89 Period 2: 90-92 Initial Returns -1.957c 54.41c 104%c -4.386c 10.40c 8.301c 20% (0.01) (0.00) (0.14) # of observations 41 41 136 136 b & c denote signi…cance at the 10%, 5%, and 1% con…dence levels, respectively.
3.180c (0.00)
1.158 (0.25)
10.36c (0.00)
v
847
13.92c
847
34.15c
c
dif f %
Period 3: 93-95 21.62c 7.4%c (0.01) 5082
Period 3: 93-95 21.62c 36%c (0.01) 847
Period 3: 93-95 21.62c 37%c (0.00) 847
ip t
5082
20.03c
cr
'c
us
dif f %
an c
Panel B: Di¤erence in Means: 3-period time trends for Industry Specialization Period 1: 87-89 Period 2: 90-92 c c b c c Initial Returns 4.90 54.41 100% -3.130 6.685 8.301c 19% (0.03) (0.00) (0.25) # of observations 41 41 136 136
Panel A: Di¤erence in Means: 3-period time trends for Entire Sample Period 1: 87-89 Initial Returns 1.439c 54.41c 97%b -9.440c 12.19c (0.00) (0.00) # of observations 246 246 816
v
M
-3.173c (0.00)
10.41c (0.00)
8.278c (0.00)
'v
'c
) are two-sided tests of signi…cant di¤erences between venture and non-
venture backed IPOs. ('v 'c ) are estimates of the extent to which the distribution of initial returns to IPOs for the venture and non-venture backed samples are identical, and are implemented using the Wilcoxon signed rank test. p-values for statistical signi…cance are reported in parentheses. VC Non-VC VC Non-VC VC Non-VC Backedy Backedy Backedy Backedy Backedy Backedy
v proportional di¤erences for ease of exposition and computed as dif f % = ( max(
j
Using three sequential periods consisting of three years each, this table compares time trends in initial returns that accrue to venture or non-venture backed IPOs. The non-venture backed IPOs are matched to IPOs backed by venture capital funds that commenced business in 1986 or 1987 using industry one-digit SIC codes and the year of the IPO. The choice of VCs that commenced business in 1986 or 1987 is predicated on the fact that company speci…c initial returns for non-VC backed IPOs are available in the IPOs database of SDC Platinum only from 1986. The choice of three sequential time periods consisting of three years each results in sample IPOs occurring between 1987 and 1995. Prior to the matching of venture backed and non-venture backed IPOs by industry and IPO year, sample funds’portfolios are matched using six di¤erent measures of VCs’portfolio activities (portfolio risk), which are: (1) industry specialization; (2) investment stage specialization; (3) syndicate size; (4) low innovation; (5) median innovation; and (6) high innovation. I discuss each measure of portfolio risk in Table I and describe the matching algorithm in Table II. The matching algorithm is implemented relative to one matching variable at a time. Period 1 results are for matched IPOs that occurred between 1987 and 1989; Period 2 results are for matched IPOs that occurred between 1990 and 1992; & Period 3 results are for matched IPOs that occurred between 1993 and 1995. The venture capital data (including initial returns) are from VentureXpert - an SDC Platinumbdatabase, while other data on initial returns to IPOs are from the equity issues database in SDC Platinum. Di¤erences in means - expressed as
Ac ce p
Table IV: VCs’portfolio activities and attenuation of di¤erences in initial returns to IPOs: VC versus non-VC backed IPOs
49
Page 49 of 50
dif f %
'v
'c v
dif f %
Period 2: 90-92 8.301c 27%b (0.03) 136
c
Non-VC Backedy
'c
Panel G: Di¤erence in Means: 3-period time trends for High Innovation Focus Period 1: 87-89 Period 2: 90-92 Initial Returns -0.079 54.41c 100%b -4.166c 15.98c 8.301c 48%c (0.02) (0.00) (0.00) # of observations 41 41 136 136 b & c denote signi…cance at the 5%, and 1% con…dence levels, respectively.
6.111c (0.00)
847
22.80c
dif f %
Period 6: 93-95 21.62c 5.2%c (0.01) 847
Period 6: 93-95 21.62c 12% (0.06) 847
Period 3: 93-95 21.62c 40%c (0.00) 847
Period 3: 93-95 21.62c 3.0% (0.77) 847
c
Non-VC Backedy
ip t
847
18.97c
847
12.92c
847
22.06c
v
VC Backedy
cr
7.001c (0.00)
3.352b (0.02)
(0.00)
3.665c
'v
us
an
M
Panel F: Di¤erence in Means: 3-period time trends for Median Innovation Focus Period 1: 87-89 Period 2: 90-92 Initial Returns 7.299c 54.41c 87%b -2.728c 18.00c 8.301c 54%c (0.04) (0.01) (0.00) # of observations 41 41 136 136
d
Panel E: Di¤erence in Means: 3-period time trends for Low Innovation Focus Period 1: 87-89 Period 2: 90-92 Initial Returns -1.454c 54.41c 103%c -4.335c 10.61c 8.301c 22% (0.01) (0.00) (0.10) # of observations 41 41 136 136
te
Panel D: Di¤erence in Means: 3-period time trends for Syndicate Size Period 1: 87-89 Initial Returns -.0.079 54.41c 100%b -4.166c 11.33c (0.02) (0.00) # of observations 41 41 136
c
v
Ac ce p
Non-VC Backedy
VC Backedy
VC Backedy
Table IV (cont’d): VCs’portfolio activities and attenuation of di¤erences in initial returns to IPOs: VC versus non-VC backed IPOs
'c
6.597b (0.00)
2.998c (0.00)
-4.820c (0.00)
4.446c (0.00)
'v
50
Page 50 of 50
Table V: IPO issuer risk and di¤erences in initial returns between VC and non-VC backed IPOs
R-squared 0.1649 Model p-value 0.0288 BIC 81.26 # of observations 29 b & c denote signi…cance at
0.0042 0.6965 -4.839
0.0151 0.4886 169.2
0.0022 (0.49) 3.5936c (0.00)
0.0007 0.7770 117.3
-0.0006 (0.78) 1.4168c (0.00)
0.0190 0.1111 469.7
-0.0108 (0.11) 7.0710c (0.00)
136
0.0530 0.0070 -53.29
-0.0027c (0.01) -0.0350c (0.00)
99
0.0000 0.9873 391.3
0.0002 (0.99) 2.3388c (0.00)
0.0429 0.0398 386.9 risk
us
0.0169 0.0413 0.0944 0.0591 0.4642 0.0260 0.0003 0.0044 169.1 112.32 458.9 -54.17 Innovation Focus dimension of VCs’portfolio
an
M
40 39 34 120 135 the 5%, and 1% con…dence levels, respectively.
0.0001 0.9611 225.1
R-squared 0.1590 0.0004 0.0044 Model p-value 0.0322 0.9018 0.6898 BIC 81.47 225.0 -4.846 Panel B: Matched pairs obtained relative to High Initial Return Spread -0.0034b -0.0003 9.50e-05 (0.03) (0.96) (0.70) c c Constant 1.9606 8.8369 0.0377 (0.00) (0.00) (0.31)
d
te
Ac ce p
cr
713
0.0008 0.4609 1090
847
0.0308 0.0000 2932
-0.0060c (0.00) 7.4997c (0.00)
ip t -0.0004 (0.46) 1.6297c (0.00)
0.0000 0.9938 1090
0.0190 0.0001 2942
846
0.0491 0.0000 -252.6
-0.0012c (0.00) 0.0353c (0.00)
0.0432 0.000 -247.45
566
0.0018 0.3193 4931
-0.0181 (0.32) 3.2449c (0.00)
0.0018 0.3097 4931
This table reports structural relations derived within univariate regression frameworks for relations between measures of IPO risk, and di¤erences in initial returns between matched samples of venture backed and non-venture backed IPOs. Let rvc and rnvc denote, respectively, initial returns to venture backed and non-venture backed IPOs. Then the initial return spread between comparable IPOs is constructed as: (rvc rnvc ). The measures of IPO risk are: underwriting spreads ( writer ); gross spreads (gross), comprising of underwriting spreads and book management spreads; the proportional change between the IPO mid…le price and the o¤er price (midto¤ er ); and the price-to-book value (P=B ) achieved subsequent to the o¤er. The non-venture backed IPOs are matched to IPOs backed by venture capital funds that commenced business in 1986 or 1987 using industry one-digit SIC codes and the year of the IPO. Matched pairs are obtained relative to the (1) industry specialization; and (2) high innovation focus dimensions of VCs’portfolio risk. These measures of portfolio activity are described in Section IV of the study and Table I. Tables II, III, and IV discuss matching algorithms that generate matched pairs of VC and non-VC backed IPOs. The choice of three sequential time periods consisting of three years each results in sample IPOs occurring between 1987 and 1995. Period 1 results are for matched IPOs that occurred between 1987 and 1989; Period 2 results are for matched IPOs that occurred between 1990 and 1992; and Period 3 results are for matched IPOs that occurred between 1993 and 1995. The venture capital data (including initial returns) are from VentureXpert - an SDC Platinum database, while other data on initial returns to IPOs are from the equity issues database of SDC Platinum. I report p-values for statistical signi…cance in parentheses. writer gross midto¤er P/B writer gross midto¤er P/B writer gross midto¤er P/B Panel A: Matched pairs obtained relative to Industry Specialization dimension of VCs’portfolio risk Period 1: 87-89 Period 2: 90-92 Period 3: 93-95 Initial Return Spread -0.0034b -0.0007 9.88e-05 -0.0024 -0.0057b -0.0254c -0.0030c -0.0233b 3.96e-06 -0.0041c -0.0010c -0.0164 (0.03) (0.90) (0.69) (0.46) (0.03) (0.00) (0.00) (0.04) (0.99) (0.00) (0.00) (0.31) c c c c c c c c c c Constant 1.9810 8.8228 0.0374 3.5904 1.4074 6.9456 -0.0602 2.3096 1.6287 7.5382 0.0440 3.3885c (0.00) (0.00) (0.31) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00)