The Worldwide Count of Priority Patents: A New Indicator of Inventive Activity

Transcription

1 The Worldwide Count of Priority Patents: A New Indicator of Inventive Activity Gaétan de Rassenfosse University of Melbourne (MIAESR and IPRIA), Australia. Joint with H. Dernis (OECD), D. Guellec (OECD), L. Picci (IPTS & U Bologna), B. van Pottelsberghe (ULB) PATSTAT User Meeting, OECD (Paris), November

2 A comprehensive patent count Develop the most basic patent count: worldwide count of priority filings by inventor (or applicant) country of residence Complete matrix of all patent counts. Easy to select the subset of triadic patents, transnational patents, Belgian inventors produce ~ 1,000 patents/y patents with family size > 1, etc. 350 Belgian patent office EPO Rest of Europe USPTO Rest of the world 2

3 Many ypotential applications Caveat: combine patents from different jurisdictions, so must be used with caution. We propose p ways to deal with this specificity. Can be tailored to specific research questions. Already used in many academic studies and policy reports, including: Turlea et al. (2009, 2010); Danguy, de Rassenfosse and van Pottelsberghe (2009); Picci (2010), Nepelski and De Prato (2011); Thomson (2012); Picci and Savorelli (2012); De Prato and Nepelski (2012); Danguy (2012); de Rassenfosse, Thomson and Webster (2012); de Rassenfosse, Schoen and Wastyn (2012) 3

4 Confronted with one major technical issue Information on inventor country of residence is missing in many instances (58% of all priority documents filed in ) Proportion of priority filings with inventor country of residence, selected patent offices Solution: exploit family linkages to recover missing information Data for years , PATSTAT APRIL CN US CA DE BE GB SE IL AU FR BR Methodology Accuracy Getting the code 4

5 Browsing 7 sources of information 1. Priority document itself 2. > > Direct equivalents* 3. > Other second filings* * Earliest selected if many documents 4. > Applicant information from priority document 5. > Applicant information direct equivalent* 6. > Applicant information other second filings* 7. > Patent office of reference Optiona al Rationale for sources 4 6: information is usually highly correlated Rationale for soure 7: if information is missing in all the other sources,,probability is high that there is no second filings. The patent is probably a local one 5

6 PATSTAT DB Priority filings Select a PF Is the inf ormation on inventor(s) available? N Y Is the inf ormation on inventor(s) available? Select the earliest SF Y N N SF of Save info Go to next PF Source ref: 1 Save info Go to next PF Source ref: 2 or 3 Are all the SF browsed for inventors? Y Is the information on applicant(s) available? N Y Save info Is the inf ormation on applicant(s) available? Select the earliest SF Y N N SF of Go to next PF Source ref: 4 Save info Are all the SF browsed f or applicants? Y Assign country of patent office Save info Go to next PF Go to next PF Source ref: 5 or 6 Source ref: 7 6

7 Sources 1, 2 and 7 are most often used Large heterogeneity in source importance Share of information recovered, by source of information at selected patent offices AT AU BR FR IL US Mean 0% 20% 40% 60% 80% 100% Data for years , PATSTAT APRIL

8 Validity tests suggest high accuracy Three validity tests to assess the accuracy of the methodology: Comparing with external data source (FR) Bypassing source 1 and comparing replenished data with true data (CA, DE, US) Same as above, but for patent offices with large number of singleton patents ts (CN, KR, RU) Overview of the accuracy of the information recovered for priority patents filed at the French patent office, 3k randomly-drawn patents Total number of priority applications compared 3,000 Total number of applications for which country of inventor was correctly identified 2,921 Of which French inventors correctly identified 2,817 Of which foreign inventors correctly identified 104 Applications for which country of inventor was wrongly assigned 79 97% accuracy 8

9 High accuracy of information recovered Good result at the French patent office may be driven by low international profile of patents. Assigning g all patents to French inventors leads to a 95% accuracy. Overview of the accuracy of the information recovered for priority patents filed at the CA, DE and US patent offices, 3k randomly-drawn patents Canada Germany US Total number of priority applications tested 3,000 3,000 3,000 Total number of applications for which country of inventor was correctly 2,900 2,902 2,656 assigned Of which national inventors correctly identified 2,319 2,740 2,299 Of which h foreign inventors correctly identified d Applications for which country of inventor was wrongly assigned Distribution of allocation error (%): But the application was allocated to: The correct country is: Canada FOR Germany FOR US FOR Canada/Germany/US d/ / FOR % 9 9

10 High accuracy of information recovered Our methodology might not work well for patent offices with a high number of singleton patents (little information from sources 2 3 and 5 6) Overview of the accuracy of the information recovered for priority patents filed at the CN, KR and RU patent offices, 3k randomly-drawn patents China Korea Russia Total number of priority applications tested 3,000 3,000 3,000 Total number of applications for which country of inventor was correctly assigned 2,976 2,966 2,943 Of which national inventors correctly identified 2,901 2,926 2,879 Of which foreign inventors correctly identified Applications for which hcountry of finventor was wrongly assigned Distribution of allocation error (%): But the patent was allocated to: The correct country is: China FOR Korea FOR Russia FOR China/Korea/Russia FOR

11 The paper comes with the MySQL source code MySQL script freely available. Written to be easy to use: simply run it in any SQL console Available on the data section of my website: Can be reworked so that it can be used to improve the data for all records in tls206_person 11