Open innovation and patent value in the US and Japan

Transcription

1 Do not quote or cite without permission. Some numbers are still preliminary. Open innovation and patent value in the US and Japan John P. Walsh* and Sadao Nagaoka** 15 November 2011 *Professor, Georgia Institute of Technology/Visiting Professor, Hitotsubashi University **Research Counselor, Research Institute of Economy, Trade and Industry and Professor, Institute of Innovation Research, Hitotsubashi University We would like to thank the research assistance by Naotoshi Tsukada, Wang Tingting, Hsin-I Huang, Taehyun Jung, YouNa Lee, and Yeonji No. We also thank RIETI and Kauffman/GRA for funding for this research. 1

2 Innovation and Collaboration Innovation is key to economic growth Innovation process becoming more complex Open innovation Markets for Innovation National Innovation Systems Inventors and their firms need to exploit information and opportunities outside the firm in order to combine one s own capabilities with resources from the external environment. While interdependent innovation seen as hallmark of the Japanese innovation system (Branstetter and Sakakibara, 1998), increasingly, the US system is also seen as moving toward an open innovation model (Chesbrough, 2003, Hicks and Narin, 2001) 2

3 Collaboration and Patent Value Collaboration plays an important role in helping a firm to exploit external capabilities and to facilitate access to new knowledge Bounded, local search limits ability to fully exploit potentially valuable information (Nelson and Winter, 1982, Cohen and Levinthal, 1990) Network ties [collaboration] can facilitate broader information flows and encourage innovation (Owen-Smith and Powell, 2004). But, innovation is a two-step process Invention: benefits from broader information access Commercialization: benefits from targeted information access 3

4 Research Questions How open is innovation? Differences across countries, sectors, firm size Japan >> US? Impact of collaboration on patent quality, commercialization rate Also, comparing survey and bibliometric indicators of collaboration 4

5 Inventor Surveys Pioneered by PATVAL, followed by RIETI, GT/RIETI Survey inventors (rather than R&D managers) Large sample, broad technology/industry coverage Comparative (JP-US, also Europe) 7

6 RIETI/GT Inventor Survey Sampling Frame: OECD Triadic Patent Families Advantage of using the TPF Reduce home country bias Focus on economically important patents (Random sampling would result in targeting most questionnaires to economically unimportant patents. Filing in multiple jurisdictions works as a threshold) Disadvantage of using the TPF Select subset of inventions, and even of patented inventions. Likely to be biased toward commercialized inventions Perhaps, bias against nonprofit, small, and/or independent inventors?-in fact, not much 8

7 Data Dual-response (post and web) mail survey Japan: Winter/Spring 07; US: Summer, 07 Japan: over 3600 responses (priority year 95-01) 20.6% response rate (27.1% adjusted for undelivered, ineligible, etc.) US: over 1,900 response ( 00-03) 24.1% response rate (31.8% adjusted) Very little difference between respondents and non-respondents in solo inventions, average number of inventors, use of university information or forward citations Data span NBER 2-digit technology classes Weight country averages to account for differences in technology mix (effects of weighting quite small) 9

8 Table 1. Basic Profile of Inventors, Japan, and US, triadic patents (NBER weighted) Japan US Sample size Academic Background University graduate (%) Doctorate (%) Demographics Female (%) Age (mean years, std. dev.) 39.1 (9.1) 47.2 (9.9) Organizational Affiliation Large firm (500+ employees)(%) Medium firm ( )(%) Small firm ( )(%) Very small firm (lt 100)(%) University (%) Other

9 Collaboration Types of Collaboration Co-assignment Prior work finds that such co-assignment is growing, but still quite rare (1.5%) in US (but more common in Europe and Japan) (Hicks and Narin. 2001; Hagedoorn, 2003) But, measure limited by available bibliometrics In Japan, need all co-assignees agreement to license; in US, each can license independently Cross-organizational co-inventors (Non-co-inventor) Collaboration partners 11

10 Table 2. Incidence of co-applications vs. that of research collaborations (%) Co-applications based on patent documents EU US Japan External co-invention Research collaborations, which do not involve coinventions

11 Figure 2. External Co-inventors, by organization type, US and Japan (NBER weights) 14

12 Figure 3. Any external co-inventor, by firm size, US and Japan 15

13 How Open is R&D? The level of solo invention is very similar in both countries (and highest among very small firms) as is mean number of inventors (just under 3) The level of co-inventions with external co-inventors are similar between Japan and US: slightly more than 10% of the inventions involve external inventors in both countries. SME co-invention higher in Japan However, only 1.8% of the US triadic patent have multiple assignees (about 10% in Japan, or 7% when we exclude affiliated firms) Any external cooperation (either co-invent, formal collaboration or informal collaboration) 26% in US, 29% in Japan 2+ partners: 10% in US; 5% in Japan In both countries, the bulk of these external co-inventors/collaborators are vertically related (either suppliers or users). Horizontal co-inventions/collaborations (collaborations among competitors) are much less common than vertical Collaboration most common in biotech, motors/generators and (in US) medical device 18

14 Lessons for Bibliometrics US co-assignment significantly under-represents crossorganization collaboration Misses about 85% of cross-organization co-invention Misses almost 95% of cross-organization collaboration although reasonably good estimate in Japan (making cross-national comparisons vulnerable) Also university assigned patents under-estimate university contribution, especially in Japan Misses 36% of university-based patents in US Misses 83% of university-based patents in Japan Again, making cross-national comparisons vulnerable 19

15 Impact of Open Innovation on Patent Information Breadth and Quality Open innovation measures Breadth: Organizational heterogeneity of network (co-invention or other collaboration) Not the number of people, but what they are Jacques Mairresse Also university links Targeted: vertical links (customers, suppliers) more similarity in information source (though still more heterogeneous than single firm) Breadth of information flows as mediator? Use of information to suggest projects Other firms, public research organizations, published sources Breadth is number of dimensions with above median information use Control for project size, firm size, inventor human capital, technology 20

16 Measuring Patent Quality Inventor self-assessment of a) technical significance or b) economic value of invention compared to others in field during that year 4-point scale: top 10%, 25th-10th pctl, top half to 25th pctl, bottom half 11% in Japan; 13% in US rated in top 10% on economic value 16% in US rated in top 10% in technical significance Remember: triadic patents (expect mean value above average) Also correlated with size of project, commercialization, IPC classes and (in Japan) grant Forward citations also correlated with these value indicators, but weaker 21

17 Validating Measures of Value Self-reported quality have strong correlations with inventor months, commercialization and (to lesser extent) number of IPC classes. Forward citations (normalized) more weakly correlated. 22

18 23

19 Ordered logit regression of Technical Significance on R&D Cooperation, US Robust standard errors in parentheses ***p<.001 ** p<0.01, * p<0.05 p<.10 Includes controls for 1-digit NBER 25

20 Results Heterogeneous collaborations associated with broader information access Heterogeneous research teams produce higher quality patents (technical significance and economic value) Big effect is from 2+ alters in US; one alter in Japan (not shown), curvilinear effect n.s. [US] In US, robust to IV estimation (publication or patents/cap as instruments) University-industry collaborations also produce higher value patents Broader use of outside information accounts for some of this effect PRO info. especially important In US, info from other firms also important 28

21 Collaboration and Commercialization Moving from invention to innovation (commercialization) key process in generating value from inventive activity Commercialization: in-house, license or start-up About 60% of triadic patents in each country commercialized Control for inventor human capital, project size, upstream (basic and/or applied), filed year, firm size, sector, technical significance 30

22 Logistic regression of Commercialization on External R&D Cooperation, US (1) (2) (3) Cooperation heterogeneity * (0.0831) Vertical Cooperation ** (0.1562) * (0.1674) University Cooperation (0.2952) (0.3453) Info-firms *** (0.0222) Info-Pubs/patents ** (0.0235) Info-Univ./Gov. Lab ** (0.0415) Number of inventors (0.0346) (0.0343) (0.0352) PhD degree ** ** (0.1323) (0.1322) (0.1432) Inventor-months (0.0030) (0.0029) (0.0030) Big firm (> 500) * ** * (0.2530) (0.2558) (0.2698) Small firm (< 100) (0.3127) (0.3162) (0.3329) Self-Reported Technical Significance *** *** *** (0.0624) (0.0626) (0.0656) Upstream research *** *** *** (0.1766) (0.1762) (0.1858) Filed year (0.0619) (0.0622) (0.2440) Wald Chi-SQ (df) N

23 Results Patents based on vertical collaborations more likely to be commercialized (net of value) Collaborations with universities less likely to be commercialized (n.s. in US) In US, heterogeneous collaborations more likely to be commercialized (net of value), but not in Japan (and not if control for vertical collaboration) (For US), all these are robust to excluding value (For US) also robust to in-house use (not licensing effect) In both countries, info from other firms associated with greater chance of commercialization; information from publications and patents associated with less Use of university/govt lab sources associated with greater commercialization in Japan and less commercialization in the US Broad information does not effect commercialization 33

24 Conclusions Japan and US roughly equally open in their innovation processes Broad collaboration networks associated with higher quality patents Targeted (vertical) collaboration associated with higher rates of commercialization (net of value) Information flows explain some of these effects Broad information mediates heterogeneity-quality link Firm information flows (fine-grained matching) mediates verticalcommercialization link Consistent with evolutionary theory of firm and network theories of innovation 34

25 Questions? Comments? Suggestions? John P. WALSH Georgia Institute of Technology School of Public Policy 685 Cherry Street Atlanta, GA USA 36

26 Appendix 37

27 Sample characteristics Sample composition of firms by size is similar to patent propensity-adjusted composition of population 100% 80% 60% 61.9% 67.4% 84.5% 81.1% 82.5% 40% 11.5% 9.9% 20% 0% 26.5% 22.7% 7.5% 7.7% 7.0% 8.1% 11.2% 10.5% Sales and Receipts, U.S. Census 2002 Innovativeness adjusted sales (Acs & Audretsch 1988) Patent propensity adjusted sales (Scherer 1983) GT/RIETI Unweighted GT/RIETI Weighted Very small (emp<100) M edium Large (emp³ 500) 38

28 Instrumental Variable Probit (top 10% technical significance), US Cooperation heterogeneity Number of inventors PhD degree Inventor-months Big firm (> 500ee) Small firm (< 100ee) (1) Probit 0.09 (0.05) 0.03 (0.02) 0.14 (0.09) 0.01*** (0.00) (0.18) 0.58** (0.20) (2) IV Probit 1.16*** (.05) 0.02 (0.02) 0.20*** (0.06) (0.00) 0.07 (0.12) 0.07 (0.17) Wald Chi-SQ (df) N Robust standard errors in parentheses ***p<.001 ** p<0.01, * p<0.05 p<.10 Includes controls for 1-digit NBER Hausman test rejects null that Heterogeneity is exogenous. Collaboration instrumented by inventor publishing in last 3 years 39

29 Table 1. Composition of the sample NBER Class Sub-Category Name Japan US 11 Agriculture, Food, Textiles 1.6% 0.4% 12 Coating 2.2% 1.8% 13 Gas 1.2% 0.6% 14 Organic Compounds 3.3% 3.2% 15 Resins 3.4% 4.4% 19 Miscellaneous-chemical 5.7% 12.1% 21 Communications 4.4% 7.8% 22 Computer Hardware 1.6% 2.2% 77 Computer Software 3.4% 4.8% 23 Computer Peripherals 2.1% 1.7% 24 Information Storage 3.3% 2.2% 31 Drugs 3.5% 5.3% 32 Surgery & Medical Instruments 2.3% 6.4% 33 Biotechnology 2.4% 2.1% 39 Miscellaneous-Drug&Med 1.2% 1.5% 41 Electrical Devices 2.5% 2.3% 42 Electrical Lighting 2.6% 1.7% 43 Measuring & Testing 3.0% 3.2% 44 Nuclears & X-rays 2.0% 2.0% 45 Power Systems 4.7% 4.6% 46 Semiconductor Devices 3.5% 2.9% 49 Miscellaneous-Elec. 3.3% 1.7% 51 Materials Processing & Handling 2.8% 2.9% 52 Metal Working 3.7% 2.6% 53 Motors, Engines & Parts 4.0% 2.9% 54 Optics 2.7% 2.2% 55 Transportation 2.0% 1.7% 59 Miscellaneous Mechanical 3.4% 2.2% 61 Agriculture, Husbandry, Food 2.4% 0.7% 63 Apparel & Textile 2.1% 0.5% 64 Earth Working & Wells 0.5% 0.3% 65 Furniture, House Fixtures 1.4% 0.5% 66 Heating 2.3% 0.4% 67 Pipes & Joints 1.8% 0.7% 68 Receptacles 1.5% 1.0% 69 Miscellaneous-Others 6.2% 6.4% 40

30 Triadic patents in the total picture Japan 8% in terms of the share of granted patents by Japanese applicants US 23% of US patent grants (any country of origin) are triadic [Jensen, et al., 2005] Level of commercialization of triadic and non-triadic patents in Japan according to the RIETI Survey Triadic non-triadic The share of the patents used internally by a firm (%) The share of the patents licensed by a firm (%) 56% 41% 23% 14% 41

31 Appendix Table Appendix.2 Bibliographic and other indicators by self-reported economic value of patent (Japan) Forward citation claims scope in IPC JP grant Bottom half Top half, but not top 25% Top25%, but not top 10% Top 10% (USA) Forward Citations Number of claims Bottom half Number of IPC codes (scope) Top half, but not top 25% Top 25%, but not top 10% Top 10%

32 Any formal/informal collaboration by NBER 43

33 Figure 1. Commercialization of the inventions JAPAN US Any commercialization Use by the applicant/owner for its product or production Pure inhouse Licensed Is cross-license includded Use for starting a new company Note. pure in-house= used by the applicant/owner only for its internal use (neither license nor the use through a startup) 44

34 Factor analysis of information sources (US) Rotated factor loadings (pattern matrix) Variable PRO Public Spill Pubs Patents Univ Govt Customers Suppliers Competitors Variance explained 14.3% 10.7% 10.4% (blanks represent abs(loading)<.4) 45