First results - Club Gertech

A corrected count of priority filings
G. de Rassenfosse, H. Dernis, D. Guellec, L.
Picci, B. van Pottelsberghe
The Output of R&D activities: Harnessing the
Power of Patents Data
IPTS workshop, Seville, 14 – 15 May 2009
Outline
 Objectives
 Methodology
 First results
 Technical issues
G. de Rassenfosse, H. Dernis, D. Guellec, L. Picci, B. van Pottelsberghe, 2009
2
Objectives
Methodology
First results
Technical issues
Outline
 Objectives
 Methodology
 First results
 Technical issues
G. de Rassenfosse, H. Dernis, D. Guellec, L. Picci, B. van Pottelsberghe, 2009
3
Objectives
Methodology
First results
Technical issues
A new indicator of patented output
 Usefulness of monitoring patenting activity
 Create a new patent indicator based on priority
filings that best captures a country’s patented
output
 Two main characteristics
 A large geographical scope taken into account
 An improved methodology to recover missing data
G. de Rassenfosse, H. Dernis, D. Guellec, L. Picci, B. van Pottelsberghe, 2009
4
Objectives
Methodology
First results
Technical issues
Many patent indicators already available…
USPTO
EPO
Triadic
PF@NPO
PF_CORR
Type
Mostly SF
Mostly SF
Family
Only PF
Only PF
Home bias
Strong
Strong
Low
Strong
Medium
Pros
Unique std.
Valuable
Unique std.
Valuable
Unique std.
H. valuable
Close to inv Close to inv
Large cov.
Cons
Mix of PF
and SF
Mix of PF
and SF
Long delay
Many std.
Low quality
Many std.
Low quality
Propensity
Productivity
Source: de Rassenfosse and van Pottelsberghe (2009)
G. de Rassenfosse, H. Dernis, D. Guellec, L. Picci, B. van Pottelsberghe, 2009
5
Objectives
Methodology
First results
Technical issues
… so why come up with a new one?
USPTO
EPO
Triadic
PF@NPO
PF_CORR
Type
Mostly SF
Mostly SF
Family
Only PF
Only PF
Home bias
Strong
Strong
Low
Strong
Medium
Pros
Unique std.
Valuable
Unique std.
Valuable
Unique std.
H. valuable
Close to inv Close to inv
Compreh.
Large cov.
Cons
Mix of PF
and SF
Mix of PF
and SF
Long delay
Many std.
Low quality
Many std.
Low quality
Propensity
Productivity
Source: de Rassenfosse and van Pottelsberghe (2009)
G. de Rassenfosse, H. Dernis, D. Guellec, L. Picci, B. van Pottelsberghe, 2009
6
Objectives
Methodology
First results
Technical issues
Many reasons to build a new indicator
 Has a different meaning:
 Most global measure of patented inventive output
 Closest to the date of invention
 Very rich informational content
 Structure of family size
 Structure of offices of filings
G. de Rassenfosse, H. Dernis, D. Guellec, L. Picci, B. van Pottelsberghe, 2009
7
Objectives
Methodology
First results
Technical issues
Outline
 Objectives
 Methodology
 First results
 Technical issues
G. de Rassenfosse, H. Dernis, D. Guellec, L. Picci, B. van Pottelsberghe, 2009
8
Objectives
Methodology
First results
Technical issues
The methodology explained
 We want to compute the indicator for country i
based on, say, inventors.
 Let C be a set of patent offices (PO). It is
composed of the POs of EU27, EPO and USPTO.
 For every element c of C, compute the fractional
count of patents invented by inventors from i
 Problem: for some filings (frequently), the information
on inventors is missing in Patstat
 Solution:
 Browse all the second filings and look for the info
 If some filings remain unidentified, assign them to
inventors from the patent office’s country.
G. de Rassenfosse, H. Dernis, D. Guellec, L. Picci, B. van Pottelsberghe, 2009
9
Objectives
Methodology
First results
Technical issues
A concrete example with Spanish inventors
EPO
USPTO
OEPM
Total PF from
Spanish
inventors
Missing at
OEPM
Other PO
G. de Rassenfosse, H. Dernis, D. Guellec, L. Picci, B. van Pottelsberghe, 2009
10
Objectives
Methodology
First results
Technical issues
A concrete example with Spanish inventors
EPO
162.85
USPTO
79.18
OEPM
Total PF from
Spanish
inventors
1711.07
Missing at
OEPM
49,85
Other PO
204.26
2207.21
G. de Rassenfosse, H. Dernis, D. Guellec, L. Picci, B. van Pottelsberghe, 2009
11
Objectives
Methodology
First results
Technical issues
A great amount of information recovered
 Two types of missing information:
 Inventor is known (e.g. has a name) but his country
code is missing: partially missing case
 Inventor is unknown: fully missing case
Results of the identification procedure (inventor criterion, year 2000)
NPO Total PF
AT
1307
BE
497
DE
44845
DK
1074
EP
11864
ES
1844
FI
2446
FR
12981
GB
23272
IT
8165
NL
2199
PL
2341
SE
4507
US
91624
Partially Fully Share missing
0
937
0,72
1
362
0,73
30 6312
0,14
4
801
0,75
0 3426
0,29
60
121
0,10
0
55
0,02
12513
465
1,00
143 17664
0,77
5719
184
0,72
3
131
0,06
0
99
0,04
1907 2452
0,97
0 23581
0,26
Share identified Remain missing
0,65
330
0,25
0,35
237
0,48
0,52
3063
0,07
0,94
47
0,04
0,89
366
0,03
0,62
69
0,04
0,69
17
0,01
0,59
5371
0,41
0,43 10062
0,43
0,43
3363
0,41
0,93
10
0,00
0,16
83
0,04
0,58
1835
0,41
0,37 14816
0,16
G. de Rassenfosse, H. Dernis, D. Guellec, L. Picci, B. van Pottelsberghe, 2009
12
Objectives
Methodology
First results
Technical issues
Results hold with another version and methodology
 Lucio launched computation in parallel:
 Different version of Patstat
 Slightly different identification rule
Comparison of patent count (year 1995)
AT
BE
BG
CY
CZ
DK
EE
FI
FR
DE
GDR
1504,18
1344,73
315,55
10,00
570,23
953,14
19,83
2104,22
11714,43
31476,44
Inventors
LP
Δ
1469,76
1337,22
317,63
10,00
565,25
906,78
19,83
2118,11
11531,07
31342,81
98%
99%
101%
100%
99%
95%
100%
101%
98%
100%
GDR
1338,2
1152,25
312,37
17
560,81
910,48
20,5
2160,76
11217,19
31239,83
Applicants
LP
Δ
1330,89
1137,52
312,75
5,00
558,28
910,07
19,50
2168,93
11263,04
31238,16
G. de Rassenfosse, H. Dernis, D. Guellec, L. Picci, B. van Pottelsberghe, 2009
99%
99%
100%
29%
100%
100%
95%
100%
100%
100%
13
Objectives
Methodology
First results
Technical issues
Outline
 Objectives
 Methodology
 First results
 Technical issues
G. de Rassenfosse, H. Dernis, D. Guellec, L. Picci, B. van Pottelsberghe, 2009
14
Objectives
Methodology
First results
Technical issues
EP and US filings may account for a large share of total
3
2,5
2
1,5
1
0,5
0
FI
DE
SE
GB
US
NL
DK
LU
AT
FR
IT
BE
IE
SI
HU
PL
CZ
ES
SK
LV
GR
BG
RO
LT
CY
M
EE
PT
Log of PF_CORR per capita
Count of priority patent applications (inventors), relative to capita. Year 2000.
EU27
EP
US
Missing
G. de Rassenfosse, H. Dernis, D. Guellec, L. Picci, B. van Pottelsberghe, 2009
15
Objectives
Methodology
First results
Technical issues
3
2,5
2
1,5
1
0,5
0
FI
DE
SE
GB
US
NL
DK
LU
AT
FR
IT
BE
IE
SI
HU
PL
CZ
ES
SK
LV
GR
BG
RO
LT
CY
M
EE
PT
Log of PF_CORR per capita
EP and US filings may account for a large share of total
EU27
EP
US
Missing
100%
80%
60%
40%
20%
0%
FI
DE
SE
GB
US
NL
DK
LU
AT
FR
IT
BE
IE
SI
HU
PL
CZ
ES
SK
LV
GR
BG
RO
LT
CY
M
EE
PT
Log of PF_CORR per capita
Composition of priority patent applications (inventors). Year 2000.
EU27
EP
US
Missing
G. de Rassenfosse, H. Dernis, D. Guellec, L. Picci, B. van Pottelsberghe, 2009
16
Objectives
Methodology
First results
Technical issues
The ranking may differ substantially for some countries
Rank of various patent counts (inventors), relative to capita. Year 2000.
TRIADIC
FI
GR
NL
FR
DE
ES
PL
PT
GB
US
Rank correlation
Normalised variance
3
18
4
8
1
16
21
20
11
5
0,58
EPO
USPTO
1
18
4
9
2
15
21
19
11
10
0,97
0,41
3
18
6
11
4
15
21
20
8
1
0,97
0,72
PF_CORR Δ
1
21
6
10
2
18
16
28
4
5
0,89
0,50
2
4
2
3
3
3
5
9
7
9
 Lowest rank correlation
 Variance similar to that of triadic patents
G. de Rassenfosse, H. Dernis, D. Guellec, L. Picci, B. van Pottelsberghe, 2009
17
Objectives
Methodology
First results
Technical issues
CY, MT, BE, NL do not rely much on their patent system
100%
90%
80%
70%
60%
50%
40%
30%
20%
10%
0%
CY
MT
BE
NL
LU
IE
AT
EE
DK
PT
ES
FI
SI
IT
SE
SK
FR
LV
CZ
LT
DE
HU
GB
GR
BG
US
PL
RO
Share of PF filed abroad
Most international inventors (year 2000).
G. de Rassenfosse, H. Dernis, D. Guellec, L. Picci, B. van Pottelsberghe, 2009
18
Objectives
Methodology
First results
Technical issues
CZ, EE, LU, CY are ‘heavily’ used by foreign inventors
100%
90%
80%
70%
60%
50%
40%
30%
20%
10%
0%
CY
MT
BE
NL
LU
IE
AT
EE
DK
PT
ES
FI
SI
IT
SE
SK
FR
LV
CZ
LT
DE
HU
GB
GR
BG
US
PL
RO
Share of PF filed abroad
Most international inventors (year 2000).
100%
90%
80%
70%
60%
50%
40%
30%
20%
10%
0%
CZ
EE
LU
CY
SK
US
GB
LV
HU
LT
BG
NL
DK
AT
DE
PL
IE
SE
ES
PT
BE
FI
SI
FR
IT
RO
GR
MT
Share of PF from national inventors
Most international offices (year 2000).
G. de Rassenfosse, H. Dernis, D. Guellec, L. Picci, B. van Pottelsberghe, 2009
19
Objectives
Methodology
First results
Technical issues
The EPO is gaining ground
Distribution of Dutch priority filings over time (inventor criterion)
100%
80%
60%
40%
20%
0%
1991
1992
1993
1994
NPO
1995
EU26
1996
EPO
1997
1998
1999
2000
USPTO
Distribution of Belgian priority filings over time (inventor criterion)
100%
80%
60%
40%
20%
0%
1991
1992
1993
1994
NPO
1995
EU26
1996
EPO
1997
1998
1999
2000
USPTO
G. de Rassenfosse, H. Dernis, D. Guellec, L. Picci, B. van Pottelsberghe, 2009
20
Objectives
Methodology
First results
Technical issues
Outline
 Objectives
 Methodology
 First results
 Technical issues
G. de Rassenfosse, H. Dernis, D. Guellec, L. Picci, B. van Pottelsberghe, 2009
21
Objectives
Methodology
First results
Technical issues
Technical issues
 The results presented are based on queries
where:
 appln_kind = ‘A’
 NOT IN tls216_appln_contn
 Can we improve the baseline query? E.g.:
 appln_kind != ‘U’
 appln_kind = ‘A’ OR appln_kind = ‘W’
 Include continuals
 Other ideas welcome
G. de Rassenfosse, H. Dernis, D. Guellec, L. Picci, B. van Pottelsberghe, 2009
22
Objectives
Methodology
First results
Technical issues
An (unstable) software available
G. de Rassenfosse, H. Dernis, D. Guellec, L. Picci, B. van Pottelsberghe, 2009
23