Upload
tyrell-shiner
View
221
Download
2
Tags:
Embed Size (px)
Citation preview
Whom You Know Matters:Venture Capital Networks and
Investment Performance YAEL HOCHBERG
NORTHWESTERN UNIVERSITY
ALEXANDER LJUNGQVISTNEW YORK UNIVERSITY
YANG LUNEW YORK UNIVERSITY
2
MOTIVATION
Networks feature prominently in the venture capital industry
VCs tend to syndicate investments, rather than investing alone (Lerner (1994))
VCs draw on their networks of service providers to help companies succeed (Gorman and Sahlman (1989), Sahlman (1990))
Capital comes from small set of investors with whom VCs have long-standing relationships (Lerner and Schoar (2005))
Performance consequences of this organizational choice remain unknown
Some VCs should have better networks and relationships
Implies differences in clout, opportunity sets, information access
Structure of syndication networks, motivations for use have been looked at, but not performance implications
Do these differences help explain the cross-section of VC investment performance?
3
FOCUS ON SYNDICATION
Syndication relationships are a natural starting point
Good reasons to believe they are vital to VC performance
1. Ability to source high-quality deal flow
Invite others to co-invest in expectation of future reciprocity (Lerner (1994))
Better investment decisions through pooling of correlated signals (Sah and Stiglitz (1986))
Diffuse information across sector boundaries and widen spatial radius of exchange (Stuart and Sorensen (2001))
2. Ability to nurture investments
Facilitate sharing of information, contacts and resources (Bygrave (1988))
Improve chances of securing follow-on funding, widen capital pool
Indirectly gain access to other VCs’ relationships with service providers
4
THE PUNCHLINE
YES – NETWORKS DO MATTER
Funds raised by better-networked VCs have better performance
Portfolio companies of better-networked VCs are more likely to survive
To exit
To future funding rounds
Effects flow through both deal flow access and value-added
5Figure 1. Network of biotech VC firms, 1990-1994
6
MEASURING HOW ‘NETWORKED’ A VC IS
Borrow from mathematical discipline of graph theory
Tools for describing networks at a macro level
Tools for measuring relative importance, or ‘centrality’, of each VC in the network
Access to and control over resources or information are particularly well-suited to measurement by these concepts (Knoke and Burt (1983))
Used before in economics literature: Robinson and Stuart (2004), Stuart, Hoang and Hybels (1999)
Network is represented by a square “adjacency matrix”
Cells represent ties between the VCs
Undirected: ties matter, but not who originated them
Directed: distinguish between originator (lead VC in syndicate) and receiver of ties (non-lead syndicate member)
7
NETWORK ANALYSIS METHODOLOGY
Networks are not static
New entry of VCs, changes in relationships, exit of VCs
Relationships get stale
Construct adjacency matrices over trailing five-year windows
Network measures, lead VC designations change over time
All measures ‘normalized’ (based on network size)
Five measures of centrality:Degree: no. of relationships proxy for access to information, deal flow, expertise, contacts, and pools of capital
Indegree: no. of syndicate invitations access to resources and investment opportunity set
Outdegree: no. of syndicate investment in future reciprocity
Eigenvector: recursive degree access to the best-connected VCs
Betweenness: economic broker
8
MEASURING PERFORMANCE
Performance at the fund level
Ideally, would like to use returns, but data not available
Measure indirectly: Exit rates
Relate to IRRs provided in FOIA requests
Performance at the portfolio company level
Again, data availability prevents us from computing returns
Survival from round to round
Achieving exit (IPO or sale)
Time to exit
9
SAMPLE AND DATA
Thomson Venture Economics
1980-1999 vintage year funds
Venture investments only, by U.S. based VCs
47,705 investment rounds in 16,315 portfolio companies made by 3,469 VC funds managed by 1,974 VC firms
Distinguish between funds, firms, and companies
Most funds organized as ten-year limited partnerships
First three to four years spent selecting investments
Middle years spent nurturing and making follow-on investments
Exit occurs in second half of fund life: IPO, M&A
Funds raised in sequence
10
MODELLING PERFORMANCE (1)
Fund performance = f (fund characteristics, competition for deal flow, investment opportunities, parent experience, network centrality)
Fund characteristics (benchmark model)
Committed capital (fund size)
Fund sequence number
Vintage year
Industry specialization
Stage focus (seed/early stage, later stage)
Competition for deal flow
“Money chasing deals” (Gompers and Lerner (2000)), proxied using aggregate VC fund inflows
Investment opportunities
Investment opportunities proxied using industry average B/M or P/E ratio
Kaplan and Schoar (2004)
11
MODELLING PERFORMANCE (2)
Fund performance = f (fund characteristics, competition for deal flow, investment opportunities, parent experience, network centrality)
Parent experience
Persistence of returns (Kaplan and Schoar (2004)) importance of experience
Length of investment history since inception
Number of completed rounds since inception
Total $$ invested since inception
Number of portfolio companies since inception
Network centrality
degree, outdegree, indegree
eigenvector
betweenness
13
FUND-LEVEL RESULTS (1)
Benchmark determinants of fund performance
Replicate Kaplan and Schoar’s fund performance model
Positive, concave relationship between size and performance
First time funds have worse performance
“Money chasing deals” has expected negative effect
Better investment opportunities has expected positive effect
More experienced VC parent firms enjoy better performance
Controlling for these effects, network measures are positively and significantly related to fund performance
15
FUND-LEVEL RESULTS (2)
Performance persistence
There is considerable performance persistence in exit rates as well as IRRs
Maybe better-networked VCs are simply the ones with better past performance
Re-estimate with additional control for the exit rate of most recent past fund
Three of the five network measures continue to be positively and significantly related to fund performance; similar economic significance
Reverse causality
Could argue that superior performance enables VCs to improve their network positions, rather than vice versa
Timeline should mitigate concerns of reverse causality
‘Network centrality’ measured prior to fund vintage
Results are robust when controlling for past performance
Find no evidence of this when we model evolution of network
17
FUND-LEVEL RESULTS (3)
Exit rates and internal rates of return
Sample of fund IRRs recently disclosed by limited partners (LPs) under FOIA
Available for 188 of the 3,469 funds in our sample
Exit rates are a useful but noisy proxy (correlation = 0.42)
Re-estimate models using sub-sample for which we have IRRs
indegree and eigenvector remain significant; very large economic effects
Regress IRRs on exit rates
Estimated relation is nearly one-to-one (point estimate = 1.046)
If we assume relation remains one-to-one in overall sample, implies we can translate economic effect on exit rates into IRR gains on same basis
2 pct point increase in exit rate roughly equivalent to 2 pct point increase in IRR (from mean of 15%)
18Figure 3.
19
COMPANY-LEVEL RESULTS
Round-by-round survival models
Network measures significantly and positively related to company survival
Experience measures lose significance
Pooled panel survival models
Network measures significantly and positively related to company survival
Experience measures have negative effect
Time-to-exit models
Controlling for state of exit markets, network measures significantly and negatively related to time-to-exit
23
ROBUSTNESS
Exit rates, survival probabilities may only reflect a better-networked VC’s ability to “push out” even poor quality portfolio companies
Look at M&A and IPOs separately
Look at financials of companies at time of IPO (positive net earnings)
Look at delisting probability post-IPO
Results don’t support this alternative hypothesis
Similar results for M&A rates alone
Portfolio companies of well-networked VCs more likely to be in the black at IPO
Portfolio companies of well-networked VCs less likely to delist post-IPO
Syndication vs. Networking
Robust to controlling for whether deal is syndicated
Result remains in the sub-sample of non-syndicated deals
24
LOCATION/INDUSTRY SPECIFIC NETWORKS
So far, network measures assumed each VC in U.S. potentially syndicates with every other U.S. VC
If VCs geographically concentrated, or industry focused, we may underestimate a VC’s network centrality
e.g., biotech VC may be central in network of biotech VCs, but lack connections to non-biotech VCs
e.g., Silicon Valley VC may be well connected in CA but not in network that includes East Coast VCs
Repeat the analysis for
Six industry-specific networks
California VCs
Same positive and significant effect; larger economic magnitude
25
HOW DOES NETWORKING EFFECT PERFORMANCE?
Deal flow is important, but networking also positively affects ability to provide value-added:
1. Proxy and control for deal flow access
Classify firms as above or below median indegree, interact with other networking measures
For eigenvector, degree - effects are stronger when indegree is lower: Networking boosts performance precisely when the VC does not have good access to deals
2. Networking with “value-added” (corporate) VCs
Construct separate measures of centrality based on networking with CVCs
Reduce effect of deal flow access: 2nd round deals, lead managed by new VCs, with no CVCs involved
Companies financed by new VCs that are well-networked to CVCs are more likely to survive to next round
26
EVOLUTION OF NETWORK POSITIONS
If being networked has such high pay-off, how do you become networked?
Emerging track record more desirable syndication partner in future
For a rookie VC, a track record consists of exits and arm’s-length follow-on rounds
Network centralityi,t = f(exitsi,t-1, follow-on roundsi,t-1, experiencei,t-1, IPO underpricingi,t-1, log # new fundst, centralityi,t-1)
Results
Controlling for persistence and unobserved VC-specific heterogeneity, VC firms improve their network position, …
…the more experience they become
… the more arm’s-length follow-on rounds they achieve
… the more eye-catching their IPOs were
Lagged number of exits has no effect except for outdegree.
28
TAKE-AWAYS
First look into the importance of networks as a choice of organizational form in the VC industry
Shed light on industrial organization of the VC market
Ramifications for LPs choosing a VC fund
Deeper understanding of the possible drivers of VC cross-sectional performance
Raises interesting questions:
How do these networks arise?
What determines the choice of whether or not to network?
What are the costs?
29
…AND NEXT PAPER Large academic literatures on networks and collusion/competition and on market
entry
Look at whether macro-level networking in a VC market presents a barrier to new entry: It does!
Define markets by natural combination of state and industry
More networked VC markets experience less entry by outside VCs
The more networked a market, the less likely a potential entrant is to enter
But networking can also help a VC overcome this barrier to entry
Previous experience lead-managing deals in which an incumbent was an investor (in another market) not only mitigates the entry problem, but can actually overcome it
Previously investing along with an incumbent as a non-lead doesn’t have nearly as strong an effect
Not surprisingly, barriers to entry also affect pricing
Valuations are lower in more networked markets, and higher where entrants manage to get more market share
Deepens understanding of how VCs get benefits from being networked