University Innovation and Technology Transfer

1
University Innovation and Technology Transfer

• Mark Schankerman
• London School of Economics and CEPR
• Lecture for European Science Days Summer School
• Steyr, Austria (9-13 July 2006)

2

• 1. Universities are an important source of
  innovation
• In the U.S., they account for 50 of basic
  research
• Patents awarded to inventors at U.S.
  universities increased from 500 in 1982 to gt 3100
  in 1998. From 1991-2000, the number of licenses
  issued by universities rose from 1278 to 4362,
  and licensing income from 186 million to 1.3
  billion
• 2. University research generates real effects
• (Jaffe, 1989 Adams, 1991, and others since)
• Pure Knowledge Spillovers Raises productivity
  of local company RD and thus increases their
  RD spending and their patenting activity,
  conditional on their RD
• Technology transfer channels are licensing to
  existing companies and formation of start-up
  companies (under exclusive or non-exclusive
  licenses)

3
Institutional background for Technology
Transfer Bayh-Dole Act, 1980 (U.S.) This gave
universities and research institutes ownership of
inventions from federally funded RD, with
mandate to undertake technology transfer and to
share revenues with university inventors Prior
to Bayh-Dole Universities had to make bilateral
approval with all federal funding agencies.

BUT Bayh-Dole Act did not give birth to
technology transfer (Mowery, Nelson and others
emphasise this) Technology licensing occurred
before Bayh-Dole, through bilateral arrangements
with government agencies. But transactions cost
and uncertainty were higher.Much technology
transfer is biomedical-related, which was
triggered by the biotechnology revolution rather
than Bayh-Dole
4

Lets agree It is virtually impossible to
determine how much of the growth in technology
transfer activity is due to the Bayh-Dole Act. I
will not address this question.
Instead, I want to discuss 1. What are the
potential benefits and costs of technology
transfer? 2. What are the instruments and
institutional arrangements that most effectively
promote it? How should we organise the market for
technology transfer activities? In particular,
what role do (should) incentives and competition
play in university research activity and
licensing?
5
University-Private Sector Science Links
Venture Capital
Knowledge spillovers
University Scientists
University Technology Licensing Office (TLO)
Market for Technology Licensed inventions
Start-up companies
Corporate Research Funding
Important to distinguish between technology
transfer activity and contract (corporate-sponsore
d) research. The economic benefits and costs
differ.
6

• Potential Benefits
• Monetary incentives for universities to be
  proactive in licensing technologies to existing
  firms or new start-up companies. Clarity of
  property rights over the technology also
  facilitates this process (increases the
  willingness of licensees to contract).
• With revenue sharing, monetary incentives for
  research scientists to innovate on commercially
  relevant technologies (but this is also a
  potential cost).
• Supplementary source of income for universities
• Caution this is very skewed across universities
  and across innovations within a given university
  (e.g. Cohen-Boyer patent at Stanford).

7

• Potential Costs
• Shifting university focus from basic research to
  applied research
• Aghion, Dewatripont and Stein (2005) What should
  be the right mix of basic and applied research in
  universities? What is the right division of
  labour between university and company research?
• Restricting open science
• Increased delays in submitting scientific
  research findings for publications (for
  scientists who engage in technology transfer
  activity)
• Increased delays or refusals to engage in
  material transfer agreements and other
  transaction costs from IP rights
• Potential conflict of interests in university
  research

8

• What does the evidence say? So far, only limited
  problems with restrictions of open science. But
  we need to be vigilant.
• Agrawal and Henderson Survey of MIT scientists
  in Mechanical and Electrical Engineering shows
  little evidence that patenting significantly
  substitutes for publication activity (especially
  when publications are weighted by citations)
• Walsh, Aurora and Cohen Surveys of biomedical
  researchers in universities and private companies
  show no major delays or abandonment of projects
  due to transaction costs arising from property
  rights
• But what about projects that were not undertaken?
• Some evidence of obstacles and delays in securing
  material transfer agreements for research
  purposes

9

• Murray and Stern study patent-paper pairs and
  find that citations to scientific papers decline
  after the patent is granted (adjusting for other
  time-related changes in citation rates). Suggests
  some anti-commons effects of patenting.
• Patenting and publishing appear to be
  complements, not substitutes, for most academic
  scientists. Patenting is often associated with a
  flurry of scientific publications by the same
  inventors. Patents may indicate research success
  rather than a retreat from fundamental research.
• Another reason for limited negative effects may
  be that much research is dual purpose, as
  illustrated by the following diagram.

10
Considerations of UseNo Yes
Quadrants of Scientific Research
Yes
Search for Fundamental Understanding
No
11
University-Private Sector Science Links
Venture Capital
Knowledge spillovers
University Scientists
University Technology Licensing Office (TLO)
Market for Technology Licensed inventions
Start-up companies
Corporate Research Funding
Important to distinguish between technology
transfer activity and contract (corporate-sponsore
d) research. The economic benefits and costs
differ.
12

• Issue 1. Incentives, Invention and Licensing
  Performance (Lach and Schankerman, 2003)
• The experiment for this work is as follows
• In the U.S., universities license inventions by
  faculty
• Part of the income (royalties) go to the
  inventor
• Inventors royalty share varies a lot across
  universities
• Use this variation to identify the effect of
  royalty sharing on invention and technology
  licensing outcomes

13
To preview the main findings

• Royalty sharing (cash flow rights) affects the
  licensing revenue earned by U.S. universities
• The incentive effect of royalty sharing is larger
  in private than in public universities. Why is
  this so? They suggest it is due to differences in
  the effectiveness of technology licensing
  offices, due in part to their incentives,
  objectives and constraints. We will see.
• Royalty incentives work both by increasing the
  scientists effort and by sorting scientists
  across universities
• Royalty incentives increase both the quantity and
  the quality (income per license) of licenses.

14

• Inventor royalty share control rights
  definition of inventor share (cash to inventor or
  direct rights over it)
• Three key characteristics of inventor royalty
  shares
• Very large variation across universities
• Observe both linear and non-linear sharing
• Non-linear sharing is always regressive (lower
  inventor shares for higher levels of license
  income)

15
(No Transcript)
16
An Empirical Model

• where R license income, s inventor royalty
  share and x control variables.
• We control for unobserved heterogeneity and
  reverse causality, which would cause a bias in
  the estimated effect of royalty incentives (e.g.
  differences in commercial orientation of
  universities).

17

• Control variables
• University quality citations per faculty
• Faculty size
• Technology fields Aggregate 23 departments into
  six areas and use shares of faculty in each
• Log total RD
• Age of the TLO
• Size of the TLO
• High-tech density of universitys location
• Private and medical school dummy variables
• Dummy variable controls for years (common demand
  and technology shocks)

18

• Incentive effect effort effect sorting effect
• Effort royalty incentives make scientists work
  harder
• Sorting scientists who patent heavily go to
  universities that pay higher royalty shares
• Why do we want to distinguish these two effects?
• Different welfare effects
• Policy issue about independence in setting
  royalty shares

19
We test for sorting by asking whether a
university is negatively affected by the royalty
shares paid by competing universities.
where sc average inventor royalty share
of competing universities. Effort hypothesis
?1 gt 0 Sorting hypothesis ?2 lt 0 We find
evidence for both effort and sorting channels
(see Table 4).
20
(No Transcript)
21

• Royalty incentives increase both the quality
  (value) of inventions and quantity of inventions.
• To distinguish between the quality and quantity
  effects, we examine whether royalty incentives
  affect license income when the number of licenses
  is held constant. That identifies the quality
  effect of incentives.
• See Table 5.

22
(No Transcript)
23
Gatekeeper Effect The effectiveness of royalty
incentives depends on the effectiveness of the
TLO

• If the TLO is ineffective, and if it has monopoly
  power over university innovations, then changing
  incentives will not have much effect.
• Thus royalty incentives and TLO effectiveness are
  complementary in producing better performance.
• This emphasises the importance of institutional
  structure Does (should) the university TLO have
  sole control over the licensing of university
  inventions?
• But why should we expect private universities to
  have more effective TLOs?

24

• Evidence from a new survey of 96 technology
  licensing offices indicates that private
  universities are likely to be more effective than
  public ones for three reasons
• Private universities are more likely to use
  performance-based incentives for their TLO staff
• Private universities are much less likely to be
  constrained by government in their licensing
  activities
• Private universities are typically less focused
  on local and regional development goals

25
Survey Evidence Incentives, Constraints and
Objectives in TLOs

• Public Private Is difference
• University University significant?
• Faculty Awareness of Incentives (
  yes) 91.7 96.4 No
• University Rewards Tech transfer ( yes) 9.4 15.4
  No
• Use of Incentive-pay ( yes) 49 79 Yes
• Government constraints on
• important or very important
• 1. Choice of license partners 23 0 Yes
• 2. Setting license contract terms 19 0 Yes
• 3. License confidentiality 27 0 Yes
• 4. Use of equity stakes 23 3.5 Yes
• 5. University liability/indemnification 75 18 Ye
  s
• 6. Dispute resolution mechanisms 49 3.6 Yes

26
University-Private Sector Science Links
Venture Capital
Knowledge spillovers
University Scientists
University Technology Licensing Office (TLO)
Market for Technology Licensed inventions
Start-up companies
Corporate Research Funding
Important to distinguish between technology
transfer activity and contract (corporate-sponsore
d) research. The economic benefits and costs
differ.
27
Issue 2. Impact of Incentives, Objectives
andGovernment Constraints on TLO Performance

• Belenzon and Schankerman (2006) study the
  relationship between the universitys license
  income and the following three factors
• whether the university TLO uses any form of
  performance-based pay (high-powered
  incentives) merit pay or bonus pay.
• 2. the importance that the university TLO
  attaches to local development objectives
  measured by low, medium and high importance
• 3. the severity of government constraints on
  licensing (formal or informal) measured by
  number of important constraints

28
A simple model of incentives and objectives in
the TLO
TLO licenses inventions in national or local
market. Unit cost Unit payoff
fraction of effort devoted to licensing in the
local market Quadratic effort costs
The TLO compensates the worker in two ways a
fixed-wage and a performance-based pay (in the
form of a constant fraction of licensing
revenues).
29
There is a divergence of interest between the TLO
and the worker. The worker cares only about her
share of license income net of effort costs. The
TLO cares about total license income and license
income in local market. WorkerTLO
30
Suppose the TLO can contract on the workers
allocation of effort. The first best efforts
solves
Effort cost to the worker
First-best level of effort
31
Suppose the TLO cannot contract on the workers
allocation of effort.
Second stage (workers choice)
Second stage (TLO contract)
32
Assume there is a fixed cost of adopting high
powered incentives, F.
TLO adopts high-powered incentives if gains of
adoption exceed F
where
Can show that Prediction 1 Universities that
care more about local development objectives, and
those that are more constrained, are less likely
to adopt incentive pay.
33
We observe only total licensing income, R. The
effect of adopting incentive pay on licensing
income is
where
One can show Prediction 2 Universities
that care more about local development
objectives, and those that are more constrained,
generate less licensing income, other things
equal.

34
Data description
35
Adoption of Incentives and University
Characteristics
36
Parametric estimation licensing income
37

• License income Additional control variables

(3) (4) (5) (6)
38

• Key Findings on Income per License
• Using performance-based pay is associated with
  about 30-45 more income per license. As with
  royalty incentives for research scientists,
  incentives for TLO licensing activities are also
  important.
• Having strong local development objectives is
  associated with about 30 less income per
  license.
• Each important government constraint is
  associated with 17 less income per license
  (average number of constraints reported as
  important is 1.5).
• These findings are robust to using non-parametric
  estimation methods.

39
Parametric estimation licenses executed
40

• Licenses executed Additional control variables

(3) (4) (5) (6)
41

• Key Findings on the Number of Licenses
• Performance-based pay does not affect the number
  of licenses executed per invention. With
  non-parametric estimation, we find a positive
  effect of about 10. This is weaker than for
  license income because numbers are easier to
  monitor by managers than income per license
  (what might have been).
• Having strong local development objectives is
  associated with 30 more licenses per invention.
  With non-parametric estimation methods, we find
  no effect of such objectives.
• Having strong government constraints have no
  significant effect on the number of licenses per
  invention.

42
Nonparametric estimation licensing income
43
Nonparametric estimation licenses executed
44
Summary Impact of Incentives, Objectives and
Government Constraints on Total License Income

• Using bonuses raises total license income by
  about 30-50. The full effect is due to
  increasing the quality of licenses, not their
  quantity. Merit pay does not have any significant
  impact.
• Strong local development objectives have a net
  negative effect on total license income they
  clearly reduce the value per license, and do not
  have a robust positive effect on the quantity of
  licenses on inventions.
• Thus local development objectives have a cost
  in terms of reduced license income. Does the
  local multiplier (agglomeration effects) make
  such a local licensing preference worthwhile?
• 3. Strong government constraints reduce total
  license income. The effect works by reducing the
  quality of licenses, not their quantity.

45
University-Private Sector Science Links
Venture Capital
Knowledge spillovers
University Scientists
University Technology Licensing Office (TLO)
Market for Technology Licensed inventions
Start-up companies
Corporate Research Funding
Important to distinguish between technology
transfer activity and contract (corporate-sponsore
d) research. The economic benefits and costs
differ.
46
Issue 3. Market Structure for Technology Licensing
Current arrangement In all research universities
in the U.S., and most in Canada, TLO has a
monopoly on commercialisation of inventions,
subject to expression of no interest (right of
first refusal). Remember most TLOs are very
small (average size lt 5 full-time
professionals). Does this market structure make
sense? Are there others that might be more
efficient without being more intrusive on
university research activity?
47

• Alternative Arrangements
• Specialisation Where are the economies of scale
  and scope?
• By technology field across geographic locations?
• By geographic region across technology fields?
• Monopoly or competition in technology licensing
  activity?

48
What form might competition take?

• Impose time limits on the monopoly TLO, and give
  inventors the right to use outside agents beyond
  that limit.
• Remove the monopoly control of the university
  TLO. Give inventors the right to use outside
  agents or do it themselves. Revenue sharing with
  the university can be preserved with this
  arrangement. Some Canadian universities do
  something similar.
• Make the university TLO the central information
  repository for all university inventions. Open up
  the licensing activity to private firms and
  intermediaries.

49
Summary of Key Findings and Policy Messages

• High-powered incentives for scientists royalty
  sharing promotes more and higher value
  innovations and more license income. Clear
  property rights and generous royalty incentives
  (cash flow rights for inventors) are important
  for stimulating university innovation and
  technology transfer.
• Policies are complementary Incentives for
  inventors and policies to enhance the
  effectiveness of TLOs. Policies need to address
  both together.
• High-powered incentives for the TLO
  Performance-based incentives increase the
  effectiveness of TLOs. Such incentives should be
  more widely used in TLOs.
• Local development objectives and government
  constraints Universities with strong local
  development objectives generate less income per
  license, but more licenses. Net effect on license
  income appears to be negative. Policy question
  is whether such objectives and constraints are
  desirable. The issue of financing local
  development should be separated from the use of
  TLOs to achieve it.
• Competition in licensing activity Institutional
  (market) structure of technology licensing
  activity is important. Need to consider
  introducing competition into the system.