Imperfect Information, Patent Publication, and the Market for Ideas

Conference on Patent Use, London
September 09, 2013
Imperfect Information, Patent Publication, and the
Market for Ideas
Deepak Hegde (New York University and USPTO)
Hong Luo (Harvard Business School)
Introduction
|
Predictions
|
Sample
|
Findings
|
Conclusion
Information frictions hinder the market for ideas
• The market for ideas and information-related frictions
–
–
–
Hard to communicate idea’s value to potential buyers without risking
expropriation (Arrow 1962, Anton & Yao 1994)
Buyers and sellers have high search costs (Hellmann 2007, Elfenbein 2007)
Transferring tacit knowledge is costly (Arora 1995)
• Solution: License when patent is granted?
Hegde and Luo
Patent Publication and the Market for Ideas
2/23
Introduction
|
Predictions
|
Sample
|
Findings
|
Conclusion
Patent grant increases the hazard of licensing
Effects of patent
publication and property
rights certification
Source: Gans, Hsu and Stern (Mgmt Sci 2008, 54(5) p 990), based on 198 technology licenses
negotiated before 1999
Hegde and Luo
Patent Publication and the Market for Ideas
3/23
Introduction
|
Predictions
|
Sample
|
Findings
|
Conclusion
How does patent publication affect licensing?
• The American Inventor’s Protection Act (AIPA) required US patent applications
filed from November 29, 2000 to be published 18 months from application date
36 months on average
Before AIPA
Secrecy
Application
Issue & Publication
Expiry
20 years
18 months
After AIPA
Secrecy
Application
Publication
Issue
Expiry
20 years
• AIPA harmonized U.S. publication rules with rest of the world
–
–
AIPA provides an opt-out provision for applicants who forego foreign patenting
Effects of AIPA are widely debated (Modigliani, et al 1999, HR5980)
Hegde and Luo
Patent Publication and the Market for Ideas
4/23
Introduction
|
Predictions
|
Sample
|
Findings
|
Conclusion
Why does patent publication matter?
•
Patent publication reduces information asymmetries
–
–
–
•
Patent publication reduces search costs
–
–
•
Establishes an official (and credible) record of invention disclosure
Reduces asymmetries about the technology
Reduces asymmetries about the potential scope of inventors’ property rights
Notifies a large pool of potential buyers (licensees)
Facilitates better evaluation and comparison against prior art and other
publications
Voluntary disclosures are costly to advertise and may not be credible
Hegde and Luo
Patent Publication and the Market for Ideas
5/23
Introduction
|
Predictions
|
Sample
|
Findings
|
Conclusion
Before AIPA, patentees license either after
patent application or after grant
License after grant
License after application
Hegde and Luo
Patent Publication and the Market for Ideas
6/23
Introduction
|
Predictions
|
Sample
|
Findings
|
Conclusion
After AIPA, patentees choose among licensing after:
(i) application, (ii) 18 month-publication, and (iii) grant
License after 18m publication
License after grant
License after application
Hegde and Luo
Patent Publication and the Market for Ideas
7/23
Introduction
|
Predictions
|
Sample
|
Findings
|
Conclusion
We test two predictions about AIPA’s effects on the
timing of licensing (conditional on licensing)
Prediction 1: Conditional on licensing, other things being equal, after AIPA:
a)
b)
c)
The likelihood of licensing 18 months from patent application (i.e., patent publication
date) increases
The likelihood of licensing after patent grant decreases
The likelihood of licensing after patent application decreases if expected loss due to
information leakage during pre-publication negotiation are low
“Prediction” 2: Conditional on licensing, other things being equal, average
licensing delay may increase or decrease after AIPA
a)
b)
Some inventors switch from licensing after application to 18-months
Some inventors switch from grant to application
Hegde and Luo
Patent Publication and the Market for Ideas
8/23
Introduction
|
Predictions
|
Sample
|
Findings
|
Conclusion
Sample: “Material” licenses in the biomedical industry with
patent applications filed between 1995 and 2005
•
Source: Deloitte Recap’s RecapIP database and Securities and Exchange
Commission (SEC) EDGAR filings
–
–
•
The 339 sample licenses collectively licensed 661 patent applications
–
–
•
Consider only first-time licensing agreements (eliminate amendments, alliances, etc)
Limit sample to only those licenses that were negotiated within 7.5 years from first
patent application date to minimize truncation bias
Each license has 1.94 patents on average (median = 1)
14% of licensed patent applications were pending or abandoned by 2013
Pre-AIPA and post-AIPA applications are not significantly different in patent
characteristics except grant lag
–
78% of the sample comprises of pre-AIPA patent licenses
Hegde and Luo
Patent Publication and the Market for Ideas
9/23
Introduction
|
Predictions
|
Sample
|
Findings
|
Conclusion
Prediction 1a: The likelihood of licensing 18-months from
patent application is higher for post-AIPA patents
.015
.01
0
.005
Density
.02
.025
Kernel density estimate
0
10
20
30
40
50
60
70
80
90
Patent licensing date - Patent application date (in months)
Pre-AIPA patents
Hegde and Luo
Post-AIPA patents
Patent Publication and the Market for Ideas
10/23
Introduction
|
Predictions
|
Sample
|
Findings
|
Conclusion
The likelihood of licensing 18-months from patent
application is higher for post-AIPA patents, ceteris paribus
Multinomial Logit
Sample
(1)
(2)
All licenses
Licenses of granted patents
Before 18
months
B/w 18 m. &
allowance
Before 18
months
B/w 18 m. &
allowance
Post AIPA patent
2.44*
4.57**
1.23
3.78**
No. of patents in license
No. of prior joint deals
No. of prior licensee deals
No. of prior licensor deals
Patent claims
Patent citations
Patent originality
Refs to non-patent prior-art
Log expected pendency
Licensor type dummies
Technology class dummies
Observations
Model chi-square
Log-likelihood
0.42**
1.55+
1
0.99
0.93
1.37+
1
1.01
0.50**
1.48
1
0.98
1
1.03
1.5
0.97
6.66*
1.05
1.33
1
0.99
1
1.02
1.48
0.99
1.96
D.V. = licensing window
Y
Y
339
1633.10
-296.48
Y
Y
291
2409.68
-224.45
Base class is license after grant; Robust SE in brackets; ** p<0.01, * p<0.05, + p<0.1
Hegde and Luo
Patent Publication and the Market for Ideas
11/23
Introduction
|
Predictions
|
Sample
|
Findings
|
Conclusion
Prediction 1b: The likelihood of waiting till patent grant to
license decreases for post-AIPA applications
70%
Before AIPA
60%
62%
After AIPA
50%
37%
40%
30%
22%
37%
26%
20%
16%
10%
0%
License before 18m
Hegde and Luo
License b/w 18m &
patent allowance
License after patent
allowance
Patent Publication and the Market for Ideas
12/23
Introduction
|
Predictions
|
Sample
|
Findings
|
Conclusion
The likelihood of waiting till patent grant to license
decreases for post-AIPA applications, ceteris paribus
OLS
Sample
D.V.
(1)
Licenses of granted
patents
Pr of license after
patent allowance
Post AIPA patent
-0.18*
No. of patents in license
No. of prior joint deals
No. of prior licensee deals
No. of prior licensor deals
Patent claims
Patent citations
Patent originality
Refs to non-patent prior-art
Log expected pendency
Constant
Licensor type dummies
Technology class dummies
Observations
Log-likelihood
Adjusted R-square
0.02
-0.07*
0
0
0
0
-0.07
0
-0.25
1.44
Y
Y
291
-179.94
0.05
Robust SE in brackets; ** p<0.01, * p<0.05, + p<0.1;
Hegde and Luo
Patent Publication and the Market for Ideas
13/23
Introduction
|
Predictions
|
Sample
|
Findings
|
Conclusion
“Prediction” 2: Licensing delays can increase or
decrease after AIPA
0.00
0.25
0.50
0.75
1.00
Kaplan-Meier survival estimates
0
10
20
30
40
50
60
70
80
90
Patent licensing date - Patent application date (in months)
Pre-AIPA patents
Hegde and Luo
Post-AIPA patents
Patent Publication and the Market for Ideas
14/23
Introduction
|
Predictions
|
Sample
|
Findings
|
Conclusion
Post-AIPA applications are licensed nine months sooner,
ceteris paribus
Tobit
(1)
Licenses of
granted patents
Patent application
to license lag
(2)
Post-allowance
licenses
Patent allowance to
license lag
Post AIPA patent
-8.56*
-11.69**
No. of patents in license
No. of prior joint deals
No. of prior licensee deals
No. of prior licensor deals
Patent claims
Patent citations
Patent originality
Refs to non-patent prior-art
Log expected pendency
Constant
Licensor type dummies
Technology class dummies
Observations
Log-likelihood
Adjusted R-square
2.35**
-5.28**
0.04
0.17
0.02
-0.37**
-1.51
0.71+
-10.22
58.67
Y
Y
291
-1309.69
0.02
2.36**
-4.94*
0.02
0.12
-0.08
-0.30+
-8.94+
0.25
-14.77+
71.05
Y
Y
192
-813.62
0.03
Sample
D.V.
Robust SE in brackets; ** p<0.01, * p<0.05, + p<0.1;
Hegde and Luo
Patent Publication and the Market for Ideas
15/23
Introduction
|
Predictions
|
Sample
|
Findings
|
Conclusion
AIPA reduces licensing lags by increasing the likelihood of
licensing at 18 months and reducing post-grant delay
.015
.01
0
.005
Density
.02
.025
Kernel density estimate
-70 -60 -50 -40 -30 -20 -10
0
10
20
30
40
50
60
70
Patent licensing date - Patent allowance date (in months)
Pre-AIPA patents
Hegde and Luo
Post-AIPA patents
Patent Publication and the Market for Ideas
16/23
Introduction
|
Predictions
|
Sample
|
Findings
|
Conclusion
The effects of AIPA are stronger for US applications that are
not already published in foreign countries
Multinomial Logit
Sample
D.V. = licensing window
Post AIPA patent
No. of patents in license
No. of prior joint deals
No. of prior licensee deals
No. of prior licensor deals
Patent claims
Patent citations
Patent originality
Refs to non-patent prior-art
Log expected pendency
Licensor type dummies
Technology class dummies
Observations
Model chi-square
Log-likelihood
(1)
(2)
US patents with foreign
equivalents
US patents without foreign
equivalents
Before 18
months
B/w 18 m. &
allowance
Before 18
months
B/w 18 m. &
allowance
1.01
3.37
1.24
5.35**
0.18**
2.99**
0.99
1.03
1.02
1.03
0.7
0.91
0.11
0.65+
1.91**
1
1.01
1.02
1.04+
2.2
0.85
0.82
0.72
1.01
1
0.98
0.99
1.04
1.57
1.02
16.44*
1.1
0.89
0.99
1
0.99
0.85
1.24
1.16+
1.12
Y
Y
137
2830.70
-81.85
Y
Y
154
531.92
-119.43
Robust SE in brackets; ** p<0.01, * p<0.05, + p<0.1; Base class is license after grant
Hegde and Luo
Patent Publication and the Market for Ideas
17/23
Introduction
|
Predictions
|
Sample
|
Findings
|
Conclusion
The likelihood of licensing after 18-months post-AIPA is
higher for licensors without downstream resources
Multinomial Logit
Sample
(1)
(2)
For-profit licensors
Non-profit licensors
Before 18
months
B/w 18 m. &
allowance
Before 18
months
B/w 18 m. &
allowance
Post AIPA patent
1.43
1.23
0.71
7.59**
No. of patents in license
No. of prior joint deals
No. of prior licensee deals
No. of prior licensor deals
Patent claims
Patent citations
Patent originality
Refs to non-patent prior-art
Log expected pendency
Licensor type dummies
Technology class dummies
Observations
Model chi-square
Log-likelihood
0.52+
1.49
1
1.01
0.99
0.96
2.28
0.99
5.76
1.05
1.60*
1
0.92*
0.99
0.95
1.24
1.03
172.81**
0.36**
1.74
0.99
0.97
1
1.19*
1.79
0.92
9.16
0.98
0.64
0.97
1
1
1.15+
1.84
0.89
0.15
D.V. = licensing window
Y
176
1400.17
-112.79
Y
115
188.28
-94.14
Robust SE in brackets; ** p<0.01, * p<0.05, + p<0.1; Base class is license after grant
Hegde and Luo
Patent Publication and the Market for Ideas
18/23
Introduction
|
Predictions
|
Sample
|
Findings
|
Conclusion
The likelihood of licensing after application post-AIPA is
higher for inventions with higher imitation losses
Multinomial Logit
Sample
(1)
(2)
Below-average science
references
Above-average science
references
Before 18
months
B/w 18 m. &
allowance
Before 18
months
B/w 18 m. &
allowance
Post AIPA patent
7.46*
3.96
0.61
4.25**
No. of patents in license
No. of prior joint deals
No. of prior licensee deals
No. of prior licensor deals
Patent claims
Patent citations
Patent originality
Log expected pendency
Licensor type dummies
Technology class dummies
Observations
Model chi-square
Log-likelihood
0.51+
0.67
1
1
1.02
0.96
1.2
0.69
0.71
0.00**
1.01
0.91+
1
1.12
6.86
220.89+
0.43**
1.71+
1
0.98
0.99
1.03
2.87
131.19**
1.11
1.42*
1
1
1
0.89
1.06
6.33
D.V. = licensing window
Y
Y
119
3565.08
-71.56
Y
Y
172
2025.76
-125.86
Robust SE in brackets; ** p<0.01, * p<0.05, + p<0.1; Base class is license after grant
Hegde and Luo
Patent Publication and the Market for Ideas
19/23
Introduction
|
Predictions
|
Sample
|
Findings
|
Conclusion
It is unlikely that AIPA “forced out” inventors from
licensing or patenting
% patents opting for pre-grant
secrecy
Post-AIPA patent applications filed in 2001
16
US small inventors
US large inventors
Foreign large inventors
Others
12
8
4
0
5
10
25
50
75
Citations Percentile
90
95
Graham and Hegde (2013) show that fewer than 8% of all U.S. applications and
less than 4% of all biomedical inventions opted for pre-grant secrecy after AIPA
– Most important inventions even less likely to opt for pre-grant secrecy
– Less than 2% of our licensed patents chose pre-grant secrecy
Hegde and Luo
Patent Publication and the Market for Ideas
20/23
Introduction
|
Predictions
|
Sample
|
Findings
|
Conclusion
Our findings are robust to several alternative explanations
and events that may have coincided with AIPA
•
•
•
•
•
AIPA’s effect on the decision to license and sample selection
Increasing patent pendency post-AIPA
Increasing delays in FDA approval post-AIPA
The bust of public financial markets
Provisional rights granted by AIPA
Hegde and Luo
Patent Publication and the Market for Ideas
21/23
Introduction
|
Predictions
|
Sample
|
Findings
|
Conclusion
Patent publication facilitates transactions in the market for
ideas and accelerates commercialization
• Patent publication facilitates transactions in the market for ideas
–
–
Licensing is more likely to occur after 18 months for post-AIPA patent applications
Interviews with inventors confirmed the importance of patent publication
•
•
–
Reduces search costs of identifying licensees
Facilitates better comparison and evaluation of licensors’ technology
Publication effects strongest for US-only applications and non-profit licensors
• Patent publication accelerates commercialization
–
–
Licensing occurs nine months sooner for post-AIPA patents
A single year’s delay can cost licensors millions of dollars in royalties (Hegde 2013)
• Caveats
–
–
Findings are conditional on licensing
Findings may be specific to the publication regime created by AIPA
Hegde and Luo
Patent Publication and the Market for Ideas
22/23
Introduction
|
Predictions
|
Sample
|
Findings
|
Conclusion
Thank you
Hegde and Luo
Patent Publication and the Market for Ideas
23/23