Knowledge spillovers play a critical role in economic
growth. The basic argument is that change is often achieved by recombining
existing ideas in novel and unexpected ways. However, because knowledge
frequently lies outside the boundaries of the inventing firm, it is important
to understand how outside knowledge feeds and interacts with internal R&D.
I propose a new construct of knowledge recombination in the context of
sequential innovation, conceptualizing new combinations of knowledge as the
ability of a firm to combine its past inventions with follow-up external
developments. Differences in the ability of firms to creatively reuse their own
knowledge with external ideas explain a large part of the cross-firm
differences in the value of intangible assets.
Using patent and citation data, I propose a new way of
measuring the re-absorption of “spilled” knowledge by inventing firms. I
distinguish between two types of citations from outside inventors: internal and
external. Internal citations are made by patents connected to the firm. An
example would be Microsoft citing a 2010 Intel patent that cites a 2008
Microsoft patent. In this example, Intel's follow-up development of the
original Microsoft patent is internalized by Microsoft in its later invention.
External citations, by contrast, are those that do not feed back into the
original inventor's research program in any subsequent period.
On the one hand, spillovers encourage innovation by
providing the basis for future research. On the other hand, firms don’t want to
spawn research that may ultimately undermine their market position or render
their own research obsolete. The prospect of future research undermines the
process of knowledge creation in the first place. This effect can be so large
as to create a “no-growth” trap where the incentives of early inventors to
invest in R&D completely disappear. However, little work has been done to
examine whether this negative effect of spillovers may be mitigated. If the
inventing firm could reabsorb its spilled knowledge in a future period together
with subsequent developments made by other inventors, it might be able to
escape the no-growth trap and sustain its long-term returns.
Previous research mostly focused on how private profits
from innovation can be shared contractually between early inventors and their
followers. The consensus is that explicit agreements between the first and
second innovator can be beneficial, because they compensate the early innovator
for their losses due to shorter patent life. However, designing and enforcing
contracts that would allow enough transfer of profits and eliminate the
question of incentives is arguably a very difficult task. Coordination costs,
overlapping patent claims (“thickets”) and hold-up problems are some of those issues.
This difficulty emphasizes the need to deepen our understanding of how private
rents can be captured by rent-seeking enterprises, especially when inventions
are highly dependent on one another. By combining past firm-specific knowledge
with outside follow-up developments, the firm can potentially not only mitigate
loss of profits by offering new products, but also to capitalize on R&D
complementarities between its past research efforts and relevant outside
developments.
An example that illustrates the relationship between
spillovers and private rents to innovation is the Computed Tomography (CT)
scanner invention. The CT technology was invented and patented by EMI. While
being extremely successful and profitable for EMI, it also inspired hundreds of
subsequent inventions by firms such as Pfiser, Syntex, Picker, and General
Electric. A few years later these firms dominated the CT scanner market,
forcing EMI to drop out.
The spillovers generated by the original EMI invention
were mostly external to EMI, meaning that those spillovers were not useful to
EMI. While EMI was responsible for the creation of one of the most influential
medical inventions of the 20th century, it lacked the capability needed follow
up its invention with outside discoveries and remain at the technology
frontier. This brings us to our central question: do firms whose inventions
result primarily in external knowledge spillovers have a lower stock market
value than firms whose inventions result in more internal knowledge spillover?
In economics, the number of times a patent is cited
provides useful information about patent’s importance. Citations are also used
to measure knowledge spillovers. Both patent quality and knowledge spillover
aspects of citations can help determine the inventing firm’s profits, as well
as shed light on the relationship between innovation and firm value.
I use information on patent citations to learn about
subsequent development of knowledge. Patent citations are a legal instrument
used to determine the scope of patent protection. The inventor identifies prior
projects that are technically related to the patent application. If patent B
cites patent A, that indicates that patent A contains a piece of knowledge on
which patent B builds. (Clearly, this piece of knowledge is excluded from the
scope of patent B.) I use citations to construct sequences of patents that cite
each other. I analyze each sequence to classify patents along the sequence as
internal, external, or self-citations.
Using data on about 1,200 American patenting firms
between 1981 and 1997, I find a strong, positive relationship between measures
of knowledge recombination and stock market value. Specifically, I find a
positive relationship between market value and internal citations, and a negative
relationship between market value and external citations. Increasing the share
of internal citations by the sample average and assuming that the change comes
from having fewer external citations is associated with an increase in $70
million in market value. Moving from the 10th to the 90th percentile in share
of internal citations is associated with an increase of about 10% of firms'
average value.
As one would expect, I find a stronger relationship
between value and citations in more cumulative, rather than discrete,
technology areas (for example, telecommunications versus drugs).
Internalization is less likely to be effective in industries where inventions
are more dependent on one another. In such industries reapplying knowledge has
a stronger effect on value.