becomes negligible. An uninformed regression will
not take this into account, and hence will tend to
bias the estimate toward zero. An informed regression,
however, will overweight the information
associated with firms with positive values (i.e., the
firms that received funding). Applying this technique,
the high-quality firms will more strongly
influence the estimates—in essence, magnifying
the information of the rare events.
Ideally, we would use population-level information
about the propensity of firms to receive VC
funding. Unfortunately, the propensity of technology
firms to receive VC funding is not known generally,
and hence we are unable to correct for this
problem (the rare-events methodology that we use
assumes that the sample is representative). However,
following the logic behind the rare-events
bias, we can state with certainty that this problem
will attenuate our estimates. In this regard, our
hypothesis testing exercise is particularly conservative.
The rare-events logit procedure also allows
us to explore the sensitivity of the results to different
assumptions about the baseline probability
of technology firms to successfully obtain venture
funding. The results are not sensitive to these
assumptions.
DATA
To explore the relationship between the content
of business planning documents and the observed
outcome of the proposed venture, we exploit a
sample of funding requests submitted to a single
VC firm based in the Northeast United States
from April 1999 to February 2002. The requests
in the sample were received during the peak of
the dot-com bubble and its immediate aftermath.
The VC firm that received the requests partnered
with a major Internet portal; over 89 percent of the
requests proposed to create dot-com firms (defined
with reference to the taxonomy of Internet business
models proposed in Afuah and Tucci (2003)).
The potential implications of sampling during this
period are addressed more fully in the discussion
below.
The sample consists of 1,063 requests for first
round VC funding for which the focal venture capitalist
(FVC) maintained paper records.11 These
becomes negligible. An uninformed regression will
not take this into account, and hence will tend to
bias the estimate toward zero. An informed regression,
however, will overweight the information
associated with firms with positive values (i.e., the
firms that received funding). Applying this technique,
the high-quality firms will more strongly
influence the estimates—in essence, magnifying
the information of the rare events.
Ideally, we would use population-level information
about the propensity of firms to receive VC
funding. Unfortunately, the propensity of technology
firms to receive VC funding is not known generally,
and hence we are unable to correct for this
problem (the rare-events methodology that we use
assumes that the sample is representative). However,
following the logic behind the rare-events
bias, we can state with certainty that this problem
will attenuate our estimates. In this regard, our
hypothesis testing exercise is particularly conservative.
The rare-events logit procedure also allows
us to explore the sensitivity of the results to different
assumptions about the baseline probability
of technology firms to successfully obtain venture
funding. The results are not sensitive to these
assumptions.
DATA
To explore the relationship between the content
of business planning documents and the observed
outcome of the proposed venture, we exploit a
sample of funding requests submitted to a single
VC firm based in the Northeast United States
from April 1999 to February 2002. The requests
in the sample were received during the peak of
the dot-com bubble and its immediate aftermath.
The VC firm that received the requests partnered
with a major Internet portal; over 89 percent of the
requests proposed to create dot-com firms (defined
with reference to the taxonomy of Internet business
models proposed in Afuah and Tucci (2003)).
The potential implications of sampling during this
period are addressed more fully in the discussion
below.
The sample consists of 1,063 requests for first
round VC funding for which the focal venture capitalist
(FVC) maintained paper records.11 These
การแปล กรุณารอสักครู่..
