1) Venture capital-backed firms outperform similar non-VC firms in growth measures like sales, employment, and wages over 3-5 years. 2) When comparing different types of VC funds, foreign VC funds are associated with the best firm performance, while some government-supported funds perform similarly to private VC funds. However, LSVCC funds in Quebec are associated with relatively poorer performance. 3) Both private and BDC-supported VC funds in Canada perform on par with domestic private VC funds in metrics like R&D expenditures, employment, wages, and profits.

1. OECD Blue Sky III
Policy and program assessment
leveraging administrative data:
the case of venture capital in
Canada
September 2016
Ryan Kelly
Hankook Kim
The views and opinions expressed in this paper are those of
the authors alone and do not represent, in any way, the
views or opinions of the Department of Innovation, Science
and Economic Development or of the Government of
Canada

2. 1
Venture Capital (VC) is a financing instrument for potentially high-growth
innovative firms.
– Finances early-stage companies through concentrated equity stakes.
– Uses highly specialized knowledge of markets and technologies
– Relies on strict governance and due diligence
– Actively mentors portfolio firms to improve outcomes
VC is often thought to play a critical role in financing innovation and
commercialization.
– Financing innovation is subject to capital market failures (e.g., information
asymmetry)
– VCs use their expertise to mitigate information asymmetry and invest in innovative
firms that would be deemed too risky by traditional lenders
As part of its support for innovation, the Canadian government has
implemented a number of programs supporting VC.
– Forms of support vary: tax credits, government run funds, and fund-of-fund
investments.
– Opportunity to assess efficacy of different policy
Venture Capital and Gov’t support for Innovation

3. 2
1) VC-backed vs. Non-VC-backed comparison
– Uses matching estimator to compare VC-backed to otherwise similar non-
VC-backed
– Positive results associated with VC across several performance metrics
2) Private VC Funds vs. Government supported VC
Funds
– Given positive results from step 1, we focus on only VC-backed firms
– Uses multivariate regression to compare firm outcomes by type of VC fund
– Finds that some government supported VC funds perform on par with
domestic private VC funds
Research Strategy: two streams of separate, but complementary
analyses
Results are not causal. However, still relevant for policy as the
results identify if VC supports innovative firms with good
commercial prospects

4. 3
Official tax and administrative data sets compiled by Statistics
Canada.
– Business Register; corporate tax files; personal tax files
– The data contain a wide variety of firm-level information including sales,
revenues, employment, payrolls, and R&D expenditures
– Includes firms operating in Canada over 1999-2009
Thomson Financial data on VC fund investments linked to tax
data through business names and addresses:
– Data includes firms that received VC funding, the date of the first VC
financing, and the size of VC investment
– Following initial linkage, screens were applied to net out potential
erroneous linkages and make the data suitable for our analysis.
– 1,044 VC-backed firms 1999-2009 (1,545 in full 1990-2009 sample)
Data longitudinalized using a process called labor-tracking
– Methodology tracks individual employees (through Social Insurance
Number in tax filings) to identify firm relationships
Data

5. 4
1) VC-backed vs non-VC-backed

6. 5
Methodology: Propensity Score Matching
First: fit a logit model and calculate the probability of receiving a first round of VC
financing
where VCit is a dummy variable indicating whether a firm received its first round of
financing in time t, Xit is the full set of the desired covariates, αc is a constant, αt and αind
are time and industry-specific intercepts, and εit is the standard econometric error term
Second: define a set of potential matches for each treatment observation based
on its predicted probability of receiving VC, industry, and province.
Third: for each treatment observation, we select the firm among the potential
matches with the closest linear propensity score.
– Nearest neighbour without replacement
– More explicitly, Let p be the linear propensity score and I0 the set of non VC-backed
firms, the match for firm i, Mi, is defined as:
where
Pr( 1)it c t ind it itVC X         
  0min [ ] , | [ ]i i j i j
j
M abs p p j I abs p p     
0.25( )p 

7. 6
Treatment and Control Groups are similar across all relevant
covariates
Comparison between Matched Pairs—Propensity Score Matching
Covariate Mean VC
Mean Control
group
P-value of
Difference
Standardized
difference in
means
Ln Total Assets 14.485 14.453 0.735 0.020
Asinh Sales 11.618 11.835 0.539 -0.037
Ln Employment 2.717 2.765 0.544 -0.036
Ln Wages 10.683 10.674 0.787 0.016
Asinh Retained Earnings -8.089 -7.835 0.727 -0.021
Asinh Revenue 13.507 13.454 0.850 0.012
Asinh Net Income -7.572 -7.305 0.707 -0.023
Age 4.952 5.042 0.796 -0.016
Asinh R&D Expenditures 8.228 8.456 0.566 -0.035
Asinh Gross Profits 9.617 9.201 0.410 0.052
Linear Propensity Score -4.729 -4.754 0.860 0.011
Treatment
and control
groups have
no statistically
significant
differences
among any of
the matching
covariates
Standardized Difference in Means = 𝑋 𝑉𝐶 − 𝑋 𝐶𝑜𝑛𝑡𝑟𝑜𝑙 /𝑠 𝑉𝐶.
Asinh 𝑦𝑖 = log 𝑦𝑖 + 𝑦𝑖
2
+ 1
1/2
.
VC obs = 544; Control group obs = 544.
Comparison between Matched Pairs— IRAP
Count
VC Control Group
IRAP recipient in year of match 91 88
IRAP recipient during sample period 152 156
Treatment and control groups have
similar propensity to use a government
R&D support program (i.e., IRAP)

8. 7
VC-backed firms outperform non-VC-backed across a variety of
firm performance metrics
 , ,1
ln ln
n
i t y i ti
X X
y Yr Growth
n




Mean VC
Mean
Control
Group
Difference
P-value of
Difference
Sales
Growth after 1 year 48.7% 27.2% 21.6% 0.022
Growth after 3 years 100.1% 47.2% 53.0% 0.001
Growth after 5 years 137.4% 56.0% 81.4% 0.001
Employment
Growth after 1 year 34.4% 6.6% 27.7% 0.000
Growth after 3 years 41.5% -3.5% 45.0% 0.000
Growth after 5 years 50.6% 3.6% 47.0% 0.000
Gross Profit / Employment
Growth after 1 year 14.3% 26.7% -12.5% 0.183
Growth after 3 years 54.6% 48.1% 6.4% 0.668
Growth after 5 years 70.2% 71.3% -1.1% 0.958
Wages
Growth after 1 year 8.2% 4.6% 3.6% 0.061
Growth after 3 years 16.7% 11.8% 4.9% 0.106
Growth after 5 years 29.4% 19.3% 10.1% 0.029
R&D Expenditures
Growth after 1 year 25.3% 9.2% 16.1% 0.012
Growth after 3 years 24.9% 8.9% 16.0% 0.230
Growth after 5 years 48.9% 29.4% 19.5% 0.230
VC-backed firms grow
faster than non-VC-backed
No difference in
profitability
Wage growth suggests
higher value-added
employment
VC-backed firms increase
R&D following investment.
But, short-lived. Consistent
with an acceleration of the
commercialization of
innovation by the firm.

9. 8
2) Private VC vs Government
supported VC

10. 9
VC fund types in Canada
Foreign VC funds:
• Predominately high-performing US based VC funds
Private Independent Funds:
• Domestic VC funds without any formal association with larger financial institutions
such as banks or pension funds
Business Development Bank of Canada (BDC):
• The BDC is a crown corporation that invests federal funds directly in firms as well as
other VC funds.
Labour Sponsored Venture Capital Corporations (LSVCCs) based in Quebec:
• LSVCCs are retail VC funds supported through tax credits to investors; they often
have non-profit goals within their mandates (e.g., regional development).
LSVCCs based in the rest of Canada:
• Treated as separate category; industry observers suggest these funds maintain
portfolios more closely resembling VC investment than Quebec-based LSVCC funds
Other Government Funds:
• Predominately provincial government programs
PrivateFundTypesGovernmentsupported
FundTypes

11. 10
VC often co-invests with other VC.
– Better performance – Brander, Amit,
and Antweiler (2002).
BDC is the most likely to make joint
investments.
– More than 90% of BDC’s investments
are joint with other types of VC.
– Mandate to act as complementary
lender
LSVCCs in Quebec co-invest very little
with other VC.
– LSVCCs in Quebec tend to invest in
different industries and tends to invest
in later stages of the firm growth life
cycle than other VC
High degree of joint-investment
complicates straight comparisons
– Need to control in a multivariate
setting
Joint Investment by VC Types
–*Note: Based on 2,847 firms in the Thomson data that
have information on VC investor types. Joint investment in
this figure are not mutually exclusive. For example, the
investment made jointly by BDC and foreign funds could
have other participants.
0% 20% 40% 60% 80% 100%
LSVCC Quebec
LSVCC Rest of Canada
Other Gov't
Private Independent
Foreign
BDC
Joint-Investment with Other VC
Fund Types

12. 11
y is a performance measure.
– R&D expenditures, employment, wages, and gross profits.
VCF is a vector of identification variables for VC types
– BDC, LSVCCs based in Quebec, LSVCCs based in the rest of Canada, Other
Government, Foreign, Private Independent, Corporate and Institutional, and Other
– Minor VC types accounting for less than 10% of VC investments in Canada (i.e.,
Corporate and Institutional, and Other) are controlled for but the results are not
discussed.
Z is a vector of firm-level control variables.
– Years since the first VC financing, age and their squares.
– Growth stage dummies (seed, early, expansion and later) and a control for syndication.
– VC investment amount – a cumulative amount increasing with each successive round of
VC financing. Not distinguishable between different types of VC.
– Number of employees as measured by individual labour units (ILUs).
IND x TIME are cross-products of industry (NAICS 3-digit) and year.
– Control for industry-time-specific shocks.
– A lack of industry-specific deflators.
Empirical Model
𝑌𝑖𝑡 = 𝛼 +
𝑗=1
𝑚
𝛽 𝑚 𝑉𝐶𝐹𝑖𝑡 + 𝛾𝑍𝑖𝑡 + 𝛿𝐼𝑁𝐷 ∗ 𝑇𝐼𝑀𝐸𝑖𝑡 + 𝜀𝑖𝑡.

13. 12
BDC and LSVCCs in the rest of Canada perform on par with domestic private
VC funds
Other government supported fund types perform relatively worse; could
reflect non-profit objectives
Differences in firm performance across VC fund types
Ln R&D
Expenditures
Ln
Employment
Ln Wages
Ln Gross
Profits
Foreign 0.4544*** 0.5026*** 0.1633*** 0.4519**
(5.06) (5.04) (5.31) (2.37)
Private Independent 0.2897*** 0.0018 0.0873*** -0.2656
(3.65) (0.02) (3.08) (-1.56)
BDC 0.2209** -0.0594 0.0145 -0.3379*
(2.49) (-0.56) (0.46) (-1.68)
LSVCC Rest of Canada 0.1537 0.1381 0.0618** -0.0082
(1.64) (1.37) (2.18) (-0.05)
LSVCC Quebec -0.1807* 0.0605 -0.1283*** -0.246
(-1.68) (0.57) (-3.72) (-1.31)
Other Government 0.1343 -0.2145** -0.1027*** -0.8517***
(1.38) (-2.00) (-3.09) (-4.02)
Note: all specifications include control s for the amount invested, age, years since financing, stage of
investment, whether there has been joint investment, investment from other VC fund types, and industry year
cross-products. The t-statistics are in brackets. Standard errors are clustered at the firm level and are robust to
heteroskedasticity and autocorrelation. ***, ** and * indicate two-sided statistical significance at 1%, 5% and
10% levels, respectively.
Similar
performance
Relatively
poor
performance
Superior
performance

14. 13
Growth results show similar performance ranking
The growth results show a similar overall ranking across the performance
metrics, though differences are less pronounced and not consistently
significant
Differences may take time to accrue
Ln R&D
Expenditures
Ln
Employment
Ln Wages
Ln Gross
Profits
Foreign 0.0661 0.0787*** 0.0404*** 0.0825
(1.42) (2.73) (2.89) (1.38)
Private Independent 0.0583 0.0086 0.0159 0.0452
(1.53) (0.32) (1.30) (0.88)
BDC 0.0947** -0.0044 -0.0024 -0.0395
(2.00) (-0.13) (-0.16) (-0.68)
LSVCC Rest of Canada 0.0295 0.0379 0.0226* 0.0266
(0.63) (1.27) (1.83) -(0.48)
LSVCC Quebec -0.0542 -0.0171 -0.0391*** -0.0159
(-1.17) (-0.54) (-2.61) (-0.29)
Other Government 0.0038 -0.0369 -0.0373** -0.0955
(0.08) (-1.06) (-2.51) (-1.58)
Note: all specifications include a lagged dependent variable and control s for the amount invested, age, years since financing,
stage of investment, whether there has been joint investment, investment from other VC fund types, and industry year cross-
products. The t-statistics are in brackets. Standard errors are clustered at the firm level and are robust to heteroskedasticity and
autocorrelation. ***, ** and * indicate two-sided statistical significance at 1%, 5% and 10% levels, respectively.
Similar
performance
Relatively
poor
performance
Superior
performance

15. 14
VC-backed firms outperform non-VC-backed firms
– Outperform in terms of growth, even after controlling for initial innovativeness
– Higher wage growth over the long run (after 5 years)
– Initial surge in R&D expenditures consistent with acceleration of commercialization
Statistically significant differences in performance among firms backed by
different types of VC in Canada
– Foreign VC is associated with the best performance across all measures examined (i.e.,
R&D, employment, wages, and gross profits).
– Private VC performs on par with BDC and LSVCCs outside of Quebec.
– Quebec based LSVCCs and other government VC – mostly provincial programs – are
generally associated with relatively poor performance.
Importance of design of government support
– The literature highlights how non-profit mandates, e.g., regional development, may hinder
VC’s performance.
Interesting results but not a complete performance evaluation
– Only measures relative performance
– Does not capture indirect impacts of public support (e.g., BDC’s high intensity of
syndication with foreign and other VC)
– Cannot conclude regarding “good” or “bad” policy
Summary of Results