1528 23 DECEMBER 2016 • VOL 354 ISSUE 6319 sciencemag.org SCIENCE
INSIGHTS | POLICY FORUM
region or to the specific people running these
Although the temporal ordering of these
estimates suggests that KS leads to future VC
investment in the region, one might worry
that some unobserved time-varying factors
account for this effect. We therefore estimated the relation using instrumental variables (IVs) to try to eliminate the influence
of endogenous or simultaneous confounding
variables. Similar to a natural experiment, an
instrument uses variation in the odds of being treated that appear random with respect
to the outcome of interest. By predicting
treatment on the basis of some third factor,
unrelated to the outcome, and estimating
the effects using these predicted values instead of the actual values, an IV isolates the
effects associated with this exogenous source
of treatment, which then has a causal interpretation, just as when one can assign the
To calculate our instruments, we used KS
campaigns in categories distant from the in-
dustries in which VCs invest (i.e., the KS and
VC categories had almost no common words
in their descriptions). We used the number
of KS campaigns in distant categories, such
as art and dance, to predict the number of
KS campaigns in the categories that closely
matched VC industries. We then used those
predicted values—instead of the observed KS
campaigns in matched categories—to esti-
mate the effect of KS on VC.
We instrumented three different measures
of KS activity using two different instruments: (i) the total number of KS campaigns
(successful and unsuccessful) that closely
matched VC, using the total number of KS
campaigns that were distant from VC; (ii)
the number of matched KS campaigns that
were successful, and (iii) the number of successful matched KS campaigns in the technology category (five other nontechnology
KS categories also matched to VC industries). Both (ii) and (iii) were instrumented
using the number of successful distant KS
campaigns. We estimated the effects for each
year to explore whether the importance of
KS in attracting future VC investments has
been changing over time.
The results of these IV regressions are
reported in the SM. Although the 2009 estimates for the overall number of campaigns
and successful campaigns do not differ significantly from zero, by 2010, a 1% increase in
the number of successful matched KS campaigns in 1 year predicted a more than 0.10%
increase in the number of VC investments in
the same year. VC funding can quickly follow
a successful KS campaign, as successful campaigns attract the attention of investors and
as entrepreneurs tout their campaigns when
pitching investors. A 1% increase in the number of successful KS campaigns in the technology category in 2009 predicted a 0.36%
increase in the number of VC investments in
the county that year.
Two notable patterns appear. First, the
results become stronger as the KS measure
becomes more closely restricted to projects
of possible interest to VC investors. By comparing the predicted values using each of
these models, one can estimate the proportion of the effect coming from each group of
projects: Successful matched KS campaigns
appear to account for all of the effect of the
overall number (successful and unsuccessful) of matched KS projects, and projects
in the technology category within the set
of matched KS projects appear responsible
for most of the effect (89.4%) of successful
matched KS campaigns.
Second, these regressions suggest that
the importance of successful KS to VC investments in a region has been rising over
time. In 2015, for example, a 1% increase in
successful KS campaigns corresponded to
a more than 0.35% increase in VC investments in the same year; a 1% increase in
successful KS campaigns in the technology
category predicted a more than 1% increase
in VC investments in the same year. As CF
has gained legitimacy and entrepreneurs
have learned how to use CF platforms, it
may increasingly become a complementary
source of funds to VC.
The results suggest an interesting and important relation between CF and VC. If these
trends continue, the rise of CF may not only
fund more innovation in a more diverse set of
places but also expand access to VC and other
forms of finance in these same regions. j
REFERENCES AND NOTES
1. A. L. Zacharakis, G. D. Meyer, J. Bus. Venturing 13, 57
2. C. I. Rider, Admin. Sci. Q. 57, 453 (2012).
3. O.Sorenson, T.E.Stuart, Am. J. Sociol. 106, 1546(2001).
4. S.Shane, T.E.Stuart, Manag.Sci. 48,154(2002).
5. R. Kitchens, P. D. Torrence, Econ. Dev. J. 11, 42 (2010).
6. J. Guzman, S. Stern,Science 347, 606 (2015).
7. P. R. Krugman, Geography and Trade (MIT Press,
Cambridge, MA, 1991).
We thank the Ewing Marion Kauffman Foundation, the Yale
School of Management, and the Coleman Fung Institute of
Engineering Leadership for financial support. This research
has received funding from the NSF (grant no. 1536022). Any
opinions, findings, conclusions, or recommendations expressed
in this material do not necessarily reflect the views of any of
these supporting organizations.
0.5 0.9 1.1 2 1060 KS
Crowdfunding and venture capital at work
Distributions at county-level of matched Kickstarter (KS) campaigns, venture capital (VC) investments,
and the ratio of the amount of KS to VC funding, 2009–2015. Increasing blue to red indicates a higher
ratio of KS to VC funding.