Transcript Presentation slides

Indigenous innovation
and (lack of) intellectual property protection
IP Statistics for Decision Makers,
November 15–16, 2016
Sydney, Australia
Gaétan de Rassenfosse
Holder of the Chair of Innovation and IP Policy
Ecole polytechnique fédérale de Lausanne (EPFL), Switzerland
Joint with Emilio Raiteri, EPFL
@gderasse
Trade protectionism is a well-known concept, but
technology protectionism is much less discussed
Economic theory tells us that governments typically want:
Weakest possible
protection for foreign
firms in their domestic
markets to facilitate freeriding on foreign
technologies.
Strongest possible
protection in foreign
countries to maximize
returns to domestic firms.
2
The national treatment principle is enshrined in
international IP law treaties
“Each Member shall accord to the nationals of other
Members treatment no less favourable than that it
accords to its own nationals with regard to the
protection of intellectual property”
– key concept –
Already present in the
1883 Paris Convention
WTO
3
Yet, frequent evidence of discrimination against
foreigners in the judicial system
http://nyti.ms/1XZEsUf
http://reut.rs/296PQIU
Another source of discrimination may occur in the
examination process
“Industry representatives express mixed opinions on
whether there is anti-foreign bias in the issuance
or enforcement of patents in China. However, some
non-Chinese firms reportedly find it more difficult to
obtain patents in sectors that the Chinese
government considers of strategic importance.”
– it actually matters –
If foreign firms are wrongfully denied a patent,
local competitors may legally use, produce, and
sell their inventions in the home market.
5
The literature on the topic is thing
Webster et al. (2014) for a sample of about 50k U.S. patents in the early 1990s
filed both at the EPO and the JPO find that:
a. Japanese firms are more likely to have their applications issued at the JPO
b. European firms are more likely to have their applications issued at the EPO
Reasons for the effect are unclear, labeled “anti-foreign bias.”
But does this bias reflect discrimination? Or does it reflect greater familiarity of
local applicants with the own system? Perhaps greater incentives of local
applicants to push through a grant?
Liegsalz and Wagner (2013) focus on
the pendency of applications filed at
the SIPO between 1990 and 2002.
Local applicants receive faster patent
grants than foreign applicants.
6
We ask two questions
1. Is there evidence of anti-foreign bias in the issuance of
applications at SIPO?
2. If yes, does that matter? That is, does discrimination help the
development of indigenous innovation?
7
1. Is there evidence of anti-foreign bias in the
issuance of applications at SIPO?
Many factors affect the chance of being granted patent protection
Attorney effect
Firm effect
Invention quality
Country of residence ?
– should not matter –
…
Other effects
Technology effect
8
We need to establish a valid counterfactual outcome
Here, the counterfactual outcome measures what the grant
probability of an applications should be, if there were no
discrimination in the system.
1. Let’s compare apples to apples. Foreign filings at SIPO are
1
international filings so we restrict Chinese applications in
our sample to applications that have an international
equivalent. (Filters out many low-quality local
applications.)
2. We go one step further. Since all filings in our sample have
2
at least one international equivalent, we use information
on the grant outcome of these equivalents to establish the
counterfactual outcome.
9
Grant rate of ‘twin’ applications is the counterfactual
outcome
We call twin applications ‘one-to-one direct equivalents’
Claims only one patent
Is only claimed by one patent
These patents can
be seen as ‘twins’
USPTO
SIPO
We use information from the grant outcome at other offices
to predict what the grant outcome at SIPO should be.
10
We track the grant outcome of twins in seven offices
 Twins in: CA, EP, JP, KR,
RU, TW, US. Not always a
twin at each office.
 We obtain a baseline
probability of grant.
It informs us about the
likelihood of grant at
SIPO.
11
2. Does discrimination help the development of
indigenous innovation?
… and it was
This foreign patent
should be granted
We observe
some follow-on
innovation in China.
… but it was not
Do we observe
more follow-on
Innovation?
12
In a slightly more formal way
 Econometric specification for the first question:
(estimated as a LPM and logit).
We think of (β1 and) β3 as “abnormal” grant probability effect.
 Econometric specification for the second question:
(estimated as a probit, negative binomial and zero-inflated
negative binomial). A patent “should be granted” if PrHI = 1.
13
The construction of the database is a major
undertaking
500k applications, filed
at the SIPO in 2001–09
INPADOC legal status
google.com/patents
web crawler
SIPO website
web crawler
PATSTAT
 Foreign/local
applicant
 Grant outcome of CN
application
 Grant outcome(s) of
twin application(s)
 Name of attorney
agency
 Strategic area or not
 Forward cites
 etc.
Next challenge:
How to identify patent applications in strategic areas?
14
“The [Chinese Medium and Long Term Plan] is considered by many
international technology companies to be a blueprint for technology
theft on a scale the world has never seen before”
Report for the US Chamber of Commerce, McGregor (2010:26)
Image source: http://si.wsj.net
We estimate an invention “fixed effect” econometric
model of the probability of grant at the SIPO
We identify areas of “strategic importance” using the National
Medium and Long-Term Program for Science and Technology
Development 2006—2020.
– IPC –
Lower grant probability?
If yes: discrimination
Strategic
technologies
MLP
– IPC –
Non strategic
technologies
Result:
Same probability of grant,
controlling for 15 factors.
That is, β1=0.
16
Discrimination is highest in biotech, then for SEPs
Biotech
LPM
Logit
ICT
LPM
Logit
SEP
LPM
Logit
Energy
LPM
Logit
-20
-15
-10
-5
0
Percentage point change in grant probability
for applications by foreign vs. Chinese applicants
SEP: Standard-essential patents
17
Objection!
Maybe strategic patents by Chinese are indeed of higher quality
after all?
Well… don’t forget that we measure how the grant decision at
SIPO departs from the grant decision at other offices.
Why would the grant decision depart only in strategic areas?
We also show that the probability of grant of these Chinese
strategic applications at the USPTO is lower than the
probability of grant of JP and KR applications at the USPTO!
18
Objection #2!
Chinese applicants know their patent system better, so they
might simply be better at navigating its complexity.
Well… no reason to expect that such a “familiarity effect” would
be observed only in strategic technologies.
Note that a complex system is in itself a form of protectionism
(albeit of a weaker sort).
19
Objection #3!
Maybe Chinese applicants rely on better patent attorneys than
foreign applicants?
Well… again, no reason to expect that such an effect would be
observed only in strategic technologies.
Besides, we control for the “quality” of the patent attorney firm
by taking into account the average grant rate of applications
handled by the attorney firm.
 Account for factors such as the quality of the translation, how
well connected is the attorney firm with SIPO, etc.
20
Second question: we investigate the likelihood of
follow-on innovation, using forward citations

We assume that a patent should be granted by SIPO if it was
granted at the USPTO and the predicted grant probability
score at SIPO is in the top quintile / decile (PrHI)

We estimate the likelihood of having at least one citation
(probit), as well as the expected number of citations (negative
binomial and zero-inflated negative binomial).

By “citation,” we mean citation at the SIPO by a Chinese
applicant.
21
Second question: we investigate the likelihood of
follow-on innovation, using forward citations
Top quintile
Top decile
… lots of control variables not reported…
Unduly refused applications are 3.4 percentage points more likely to
be built upon. They also receive 7–8 percent more citations (quintile).
Concluding remarks
We provide first-of-its-kind evidence that the patent system is
used to promote—and develop—indigenous innovation.
Patent system acts (is used?) as a barrier to entry in strategic
technological areas.
Discrimination against foreigners happens in apparent
violation of the national treatment principle.
Yet, no formal “proof” of intentional discrimination (e.g., result
could be driven by more experienced examiners beings
assigned to strategic technologies).
23
Thank you
@gderasse
[email protected]
Firm j
We also control for patent attorney effect
Patent i by firm j
with attorney a
We compute to grant rate of
applications by other firms to
measure attorney effect.
All other firms
attorney a
All patent applications
by attorney a
25
Yet another robustness test: Sensitivity to hidden bias


Remember our identifying assumption: we need an unbiased
counterfactual.
We test the sensitivity of our findings by injecting artificial
bias in the grant outcome of our twin applications (i.e., the
pseudo fixed effect ci).
26