Innovation and knowledge spillovers
are key to economic growth, and universities play a central role. In the U.S.,
academic institutions spent $48 billion on R&D, accounting for 56 percent
of basic research and 33 percent of total research in the U.S. (National
Science Board, 2008). Academic research takes two main forms: scientific
publications and, increasingly, patents. Promoting university innovation and its
diffusion is a major policy objective. This policy focus assumes that knowledge
spillovers are geographically localized. Thus it is important to understand how
geography, and the characteristics and policies of universities and states, limit
knowledge spillovers.
The importance of state borders: We focus on how state borders, and
distance, influence the diffusion of knowledge from private and public American
universities. While country borders typically signify zones with different
cultures, languages, and political institutions, American states are unlikely
to vary much on these dimensions. Moreover, separating state border effects
from pure distance effects is difficult. Nonetheless, because state borders are
not strongly associated with different linguistic, culture, or political
institutions, they provide a clean framework for investigating how local policy
influences knowledge spillovers.
We examine two ways that state
borders can affect university knowledge diffusion: local information, and policies
for commercializing university innovation. Local information is important when
dealing with tacit knowledge, which can be explained as inventors located
closer to the cited university having greater potential for learning than those
located further away, which in turn encourages development of local information
networks. The border effect should be stronger in states where inventors are
more likely to remain in the state when they move jobs, and when inventors are
more likely to have been educated at a local (in-state) university. State
policies can also influence the prevalence of such local information. For
example, “non-compete” labor laws make it more likely that inventors who shift
employers will leave the state.
The second method involves policies
that promote local commercial development of university innovations. This is
more likely to occur in states with a dense and vibrant community of scientists
and engineers. The state border is also likely to be more important for public
universities, which are often constrained or influenced by state government. For
example, public universities typically attach greater importance to promoting
local and regional development through their technology licensing policies
(Belenzon & Schankerman, 2009).
Our methods: We measure citations to
university-owned patents. Citations have been widely used to trace spillovers
from corporate R&D (Jaffe & Trajtenberg, 2002). However, citations to
university patents are an imperfect measure, and many scientific contributions
made by university faculty are never patented. We also examine the extent to
which corporate patents cite university scientific publications.
A citation indicates that the later
invention somehow builds on the earlier one, and that some knowledge transfer
has occurred. Jaffe, Trajtenberg, and Henderson (1993), for example, compare
the average distance of patents that cite another patent and a random control
group of patents, in the same field, that do not cite. They show that firms
located in the same city as the inventor are much more likely than others to
benefit from knowledge spillovers from that innovation.
Geography can be summarized as
identifying whether inventors are in same city, state, or country. Current studies,
which do not measure geographic distance, are unable to explore in detail how
distance affects citation rates. We address this gap by using the actual
distance between the locations of patent assignees (measured by Google Maps),
distinguishing between both the relationship between spillovers and geographic
distance, and the impact of state borders. We show that citations decline
sharply with distance up to about 150 miles, but are essentially constant
beyond that. This strongly suggests that direct personal interaction plays an
important role in knowledge flows.
Findings: Controlling for distance, we find that
inventors in the same state as the cited university are substantially more
likely to cite one of the university's patents than an inventor outside the
state. In contrast, we find that state borders have minimal, although varying,
impact on citations by patents to university scientific publications.
The impact of state borders on patent
citations differs widely across states. First, the border effect is larger in
states that do not have, or do not strongly enforce, “non-compete” labor laws.
These laws, which restrict employees from moving jobs to a competing firm
within the same state for some period of time, should reduce knowledge
spillovers and weaken the state border effect on citation behavior. Our
findings reinforce studies that show that non-compete laws increase
out-migration for job movers (Marx et al., 2007, 2010); we show that
non-compete statutes also affect the knowledge diffusion that labor mobility
generates.
Second, the border effect is stronger
in states with a higher fraction of inventors educated in state, more
scientists and engineers, and lower rates of interstate labor mobility for
scientists and engineers.
Third, the border effect is much
stronger for citations to patents from public universities. A substantial part
of this effect is associated with local development focus. This finding has a
potentially important policy implication. Belenzon and Schankerman (2009) show
that universities with strong local focus earn substantially less licensing
income from their inventions – but there may be offsetting benefits, such as
greater localization of knowledge spillovers. This is key to understanding whether
it makes economic sense for universities (or state governments) to promote
local development through local licensing. We find a genuine trade-off, which
policymakers should bear in mind.
Finally, we examine the differences
in knowledge spillovers across technology areas. In fields where information is
harder to transmit, direct social relationships are likely to play a larger
role, making knowledge spillovers more sensitive to geographic distance. We
find that localization occurs mostly in biotechnology, pharmaceuticals, and
chemicals, and much less so in electronics, information technology, and
telecommunications. This implies that some of the variation we observe in the
strength of the border effect across states may be due to differences in
technology specialization.