European Office

Patent family data and statistics at the European .

Peter Hingley; EPO, Walter Park; American University, USA

WIPO-OECD Workshop on Patent Statistics, 18 & 19 / 09 / 2003. EPO Patent families

1. Introduction

¾ 2. Construction of families

¾ 3. Sociology of families

¾ 4. Timeliness of families data

¾ 5. Statistical description of filing trends

¾ 6. Families as a basis for forecasting filings

¾ 7. Conclusions

WIPO-OECD Workshop on Patent Statistics, 18 & 19 / 09 / 2003. EPO Patent families 2. Construction of patent families

Priority forming DOCDB publications database First Filings Applications, Grants (with references to priorities)

Priority 1 Application A Priority 2 Application B Priority 3 Application C, Grant C

Rearrange as families (PRI file) : Family 1 Priority 1 Application A Status A Family 2 Priority 2 Application A, Application B Status A Family 3 Priority 3 Application C, Grant C Status G

Country & Bloc of Filing (EPC, Japan, US, Others), Applicant & Names, Priority date, Subsequent filings dates, model / Patent, Classification (IPC & NACE), PCT involved, EPO involved.

WIPO-OECD Workshop on Patent Statistics, 18 & 19 / 09 / 2003. EPO Patent families 3. Sociology of patent families

Families originating from first filings in EPC contracting states (including EPO first filings) in 1997.

Monolateral families ("First")

First: 127179 Out: 42700 T.P.F: 15539

Bilateral familes ("Out")

Trilateral patent families ("T.P.F.")

WIPO-OECD Workshop on Patent Statistics, 18 & 19 / 09 / 2003. EPO Patent families 3. Sociology of patent families

From EPC to any other Blocs (Priority Year 1997)

129179 EPC

US and/or and/or JAPAN

OTHERS

42700 (33%)

WIPO-OECD Workshop on Patent Statistics, 18 & 19 / 09 / 2003. EPO Patent families 3. Sociology of patent families

From EPC to Other Trilateral Blocs (Priority Year 1997)

129179 EPC

US and/or JAPAN

33880 (26%)

WIPO-OECD Workshop on Patent Statistics, 18 & 19 / 09 / 2003. EPO Patent families 3. Sociology of patent families

Trilateral patent families from EPC (Priority Year 1997)

129179 EPC

US and JAPAN

15339 (12%)

WIPO-OECD Workshop on Patent Statistics, 18 & 19 / 09 / 2003. EPO Patent families 3. Sociology of patent families

Trilateral patent families (Priority Year 1997)

129179 EPC 15339 (12%)

151652US23458 (16%) T.P.F. 20255 (6%) JAPAN 348329

2254 (2%) OTHERS 133254

WIPO-OECD Workshop on Patent Statistics, 18 & 19 / 09 / 2003. EPO Patent families 3. Sociology of patent families

Families subsequent filings to EPO (Priority Year 1997)

129179 48647 EPC EPO 32124 19979 36231 25143 151652US 3748 JAPAN 348329

5783 OTHERS 133254

WIPO-OECD Workshop on Patent Statistics, 18 & 19 / 09 / 2003. EPO Patent families 3. Sociology of patent families

Families from EPC using the PCT system (Priority Year 1997)

EPC 24186 (86%) 27999

US 12657 (45%) PCT 13281 (47%) JAPAN

19885 (71%) OTHERS

WIPO-OECD Workshop on Patent Statistics, 18 & 19 / 09 / 2003. EPO Patent families 3. Sociology of patent families

Consolidation of families - Minimum bound

Priority filings Subsequent filings x x x x x x Φ x x Θ x x x x x x x l = pΦ = yΘ = 10 p = 7 y = 8 Φ = average number of subsequent filings linked to a priority filing. Θ = average number of priority filings linked to a subsequent filing.

WIPO-OECD Workshop on Patent Statistics, 18 & 19 / 09 / 2003. EPO Patent families 3. Sociology of patent families

Consolidation of families

Minimum bound: cmin = p + y - l Important formula: l = pΦ = yΘ

Therefore: cmin = p - (Θ -1)y

Maximum bound is effectively p.

Preliminary comparison exercise by EPO / OECD on priorities in 1993. Consolidation reduced OECD Triadic patent families to 71% of the unconsolidated number. Minimum bound on EPO data was 67%.

System works for Bilateral families. Need to think carefully about the extension to Trilateral patent families. WIPO-OECD Workshop on Patent Statistics, 18 & 19 / 09 / 2003. EPO Patent families 4. Timeliness of data

Trends in Φ and Θ. EPC Contracting states priorities flowing to EPO (includes all first filings at EPO).

1.30

1.25

1.20

Theta 1.15 Phi Phi trend

1.10

1.05

1.00 1990 1992 1994 1996 1998 2000 2002

WIPO-OECD Workshop on Patent Statistics, 18 & 19 / 09 / 2003. EPO Patent families 4. Timeliness of patent family data

Trends in Φ and Θ. Japan priorities flowing to EPO.

1.70

1.60

1.50

1.40 Theta Phi 1.30 Phi trend

1.20

1.10

1.00 1990 1992 1994 1996 1998 2000 2002

WIPO-OECD Workshop on Patent Statistics, 18 & 19 / 09 / 2003. EPO Patent families 4. Timeliness of patent family data

Trends in Φ and Θ. U.S. priorities flowing to EPO.

1.60

1.50

1.40

Theta 1.30 Phi Phi trend

1.20

1.10

1.00 1990 1992 1994 1996 1998 2000 2002

WIPO-OECD Workshop on Patent Statistics, 18 & 19 / 09 / 2003. EPO Patent families 4. Timeliness of patent family data

Trends in Φ and Θ. Others priorities flowing to EPO.

3.00

2.80

2.60

2.40

2.20 Theta 2.00 Phi Phi trend 1.80

1.60

1.40

1.20

1.00 1990 1992 1994 1996 1998 2000 2002

WIPO-OECD Workshop on Patent Statistics, 18 & 19 / 09 / 2003. EPO Patent families 4. Timeliness of patent family data Fig. 1 a) Subsequent filings to EPO from various blocs of origin (y).

160000

140000

120000 All

100000

y 80000

60000 EPC 40000 US Japan 20000

Others 0 1993 1994 1995 1996 1997 1998 1999 2000 Year of priority forming first filing.

WIPO-OECD Workshop on Patent Statistics, 18 & 19 / 09 / 2003. EPO Patent families 4. Timeliness of patent family data Fig. 1 b) Numbers of families leading to subsequent filings at the EPO from various blocs of origin (p). Points are actual values from the PRI file, and lines are trends (see Annex A).

140000

120000 All 100000

80000 p 60000

EPC 40000 U.S. Japan 20000 Other 0 1993 1994 1995 1996 1997 1998 1999 2000 Year of priority forming first filing.

WIPO-OECD Workshop on Patent Statistics, 18 & 19 / 09 / 2003. EPO Patent families 5. Statistical description of filing trends

Origin All 1984 Origin All 1994

A H A H H A A A H H H A B First: 495539 B B First: 662257 G Out: 74784 B B Out: 108995 G G T.P.F.: 36567 B G T.P.F.: 49888 Origin All 1999 G G F C F E D F C C E D C F E D F C A E D C E D F H A ED H A Human necessities H A B Performing Operations B C Chemistry & Metallurgy First: 798634 B Out: 163648 B D Textiles & Paper T.P.F.: 31291 G E Fixed Constructions G F Mechanical Engineering G C G Physics F E D C C H Electricity F ED F ED WIPO-OECD Workshop on Patent Statistics, 18 & 19 / 09 / 2003. EPO Patent families 5. Statistical description of filing trends

Origin EPC Contracting States 1984 Origin EPC Contracting States 1994

H A H A H A H A H H A A G G G B G G G B First: 96716 B First: 107661 F Out: 24723 Out: 32634 B F T.P.F.: 14459 TPF: 10866 B B F Origin EPC 1999 F F E F E E D C E D C D D E C E C D D C H A C A Human necessities H A H A B Performing Operations

C Chemistry & Metallurgy G G First: 130999 B D Textiles & Paper G Out: 44628 T.P.F.: 5070 B E Fixed Constructions B F F Mechanical Engineering F F G Physics E D C E D H Electricity E C D C WIPO-OECD Workshop on Patent Statistics, 18 & 19 / 09 / 2003. EPO Patent families 5. Statistical description of filing trends

Origin Japan 1984 Origin Japan 1994

A A H A A H H A A B H H H B

B B B B First: 257983 First: 318261 Out: 25634 Out: 35250 T.P.F.: 14073 T.P.F.: 16442 C G G G Origin Japan 1999 G C C D C D D E F E C G E G F F F ED C A ED F ED A F H A Human necessities A B H B B Performing Operations H B C Chemistry & Metallurgy First: 356397 D Textiles & Paper Out: 45400 E Fixed Constructions T.P.F.: 13377 C G C F Mechanical Engineering C ED D G F E G Physics G F H Electricity ED F WIPO-OECD Workshop on Patent Statistics, 18 & 19 / 09 / 2003. EPO Patent families 5. Statistical description of filing trends

Origin U.S. 1984 Origin U.S. 1994

A H H A H A H A H A H A

B G First: 121733 G First: 48665 B G Out: 19734 Out: 34191 G B T.P.F.: 11103 B G T.P.F.: 17327 G B B F Origin U.S. 1999 F E F D C F ED C F ED C E A F D C ED C E C H D H A A Human necessities H A B Performing Operations B C Chemistry & Metallurgy First: 153350 B Out: 63762 D Textiles & Paper B G T.P.F.: 12003 E Fixed Constructions G F Mechanical Engineering F C G Physics G ED C C H Electricity F ED F ED WIPO-OECD Workshop on Patent Statistics, 18 & 19 / 09 / 2003. EPO Patent families 5. Statistical description of filing trends

Origin Others 1984 Origin Others 1994

H A A H H A A H H A H G G A B G First: 92175 G Out: 4693 First: 114602 B T.P.F.: 525 B Out: 6920 F B F B T.P.F. 1660 G F B F Origin Others 1999 G E E C E C C E D D D C D C A F E D C H A F ED A Human necessities H H A B B Performing Operations B C Chemistry & Metallurgy G First: 157888 D Textiles & Paper Out: 9858 T.P.F.: 841 E Fixed Constructions B G F C C F Mechanical Engineering G E D C D G Physics E H Electricity F ED F WIPO-OECD Workshop on Patent Statistics, 18 & 19 / 09 / 2003. EPO Patent families 5. Statistical description of filing trends

Growth rates of IPC main Technical Areas from 1984 to 1994 EPC Japan

A; 10.5% H; 22.7% A; 35.1% H; 23.8% G; 14.9% B; 10.1% B; 9.6% F; 14.6%

C; 23.9% C; 9.4% D; 7.4% U.S. G; 21.7% D; 9.7% E; 1.9% E; 97.2% A Human necessities F; B Performing Operations H; 113.7% A; 106.4% C Chemistry & Metallurgy D Textiles & Paper E Fixed Constructions B; 73.4%

F Mechanical Engineering G; 127.3% G Physics H Electricity C; 73.4% F; 44.8% E; 43.1% D; 83.3% EPO Patent families 6. Families as a basis for forecasting filings

Econometric specification by Ordinary Least Squares (OLS)

Model fitted to subsequent filings at EPO, for data of 1985 and 1990. Forecasts made for subsequent filings at EPO for 1995 and 2000.

^

Log Y t = -21.04 + 0.449 log X1t + 0.669 log X2t + 0.895 log X3t + 0.772 EPO,

where ^

Y t = families with subsequent EPO filings per Source Country Worker at time t X1t = source country first filings per worker at time t X2t = source country real R&D stock per worker at time t X3t = source country real GDP per capita at time t EPO = 1 if the source country is an EPO state, and zero otherwise

All parameter estimates are significant at 1% level.

WIPO-OECD Workshop on Patent Statistics, 18 & 19 / 09 / 2003. EPO Patent families 6. Families as a basis for forecasting filings Econometric specification by Ordinary Least Squares (OLS) Families containing EPO Subsequent Filings in 1995: OLS forecasts vs. Actual filing numbers by country of first filing - natural logarithms.

12

Japan 10 U.S.A. 8 U.K.

6

4

Log # forecasted families 2

0 024681012 Phillippines -2 Log # Actual families

WIPO-OECD Workshop on Patent Statistics, 18 & 19 / 09 / 2003. EPO Patent families 6. Families as a basis for forecasting filings Econometric specification by Ordinary Least Squares (OLS)

Reasonable overall agreement between forecasted and actual families flowing to EPO in 1995.

BUT some of the largest countries are outliers:

Country Forecast Actual Deviation Japan 27 803 16 793 66% U.S.A. 20 236 24 990 -19% Germany 13 923 13 851 1% 1 926 2 056 -6% Total 80 914 75 732 7%

Generalised least squares (GLS) also used. GLS gives higher mean squares than OLS, but a better prediction of total filings.

WIPO-OECD Workshop on Patent Statistics, 18 & 19 / 09 / 2003. EPO Patent families 6. Families as a basis for forecasting filings Econometric specification by Ordinary Least Squares (OLS)

OLS and GLS both strongly overestimate families flowing to EPO in 2000.

This is due to the timeliness problem in the families (PRI) data:

Compare OLS forecasts for 2000 (p) with EPASYS data (y), using the formula y = (Φ/Θ) p

Country Forecast (p) Φtrend Θ Forecast (y) Actual Deviation Japan 34 456 1.093 1.478 25 481 18 385 39% U.S.A. 28 538 1.463 1.302 32 067 48 568 -34% Germany 17 820 1.265 1.123 20 073 22 405 -10% Italy 2 085 1.265 1.123 2 349 2 960 -21%

Should improve models – e.g. Data analysis over more years incorporating a time series approach. Modelling Euro-PCT filings distinctly from Euro-direct.

WIPO-OECD Workshop on Patent Statistics, 18 & 19 / 09 / 2003. EPO Patent families 7. Conclusions

1. “1 Priority = 1 Patent” is a useful definition.

2. Unconsolidated families describe the real patterns of world-wide patenting. Consolidated counts can be approximated statistically.

3. A large database with query access for researchers is a better idea than a prescribed set of filtered families by a single definition.

4. Families databases from patent publications are limited by the timeliness problem. Possible solutions include a) Updating counts by correlating to “live” patent office databases, b) Feeding families databases with unpublished applications.

5. Patent families give an added dimension to patent statistics (first filings patterns + transmission patterns to other offices).

6. Patent families should enable better forecasting models for international filings trends to be developed. And finally ... EPO Patent families 7. Conclusions

9. The following formula may be useful:-

l = pΦ = yΘ

Thank You.