1 / 39

The Influence of University Research on Industrial Innovation

The Influence of University Research on Industrial Innovation. Jinyoung Kim † Sangjoon John Lee †† Gerald Marschke ‡. † Department of Economics, Korea University †† Department of Economics, Alfred University ‡ Department of Economics, SUNY at Albany and IZA.

leal
Download Presentation

The Influence of University Research on Industrial Innovation

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript

  1. The Influence of University Research on Industrial Innovation Jinyoung Kim† Sangjoon John Lee†† Gerald Marschke‡ †Department of Economics, Korea University †† Department of Economics, Alfred University ‡Department of Economics, SUNY at Albany and IZA Prepared for the “Universities, Innovation, and Economic Growth,” Federal Reserve Bank of Cleveland, November 16-17, 2006.

  2. Objectives • We examine the influence of university research on industrial innovation through industry’s association with university-experienced scientists. • While the outputs of university research can be disseminated through the “traditional channels of open science” (paper publications, conferences, informal social networks)… • This paper focuses on the role of research personnel as a pathway for the diffusion of ideas from university to industry.

  3. Objectives Using U.S. patent data, we examine 1. Trends in university influence in the 1980s through the 1990s 2. Factors that influence a firm’s interaction with university research in the pharmaceutical and semiconductor industries during the 1990s

  4. Main Findings Economy-wide: Industry increased its employment of and/or collaboration with inventors with university research experience and with advanced degrees. Industry patents increasingly cited university patents. Pharmaceutical and Semiconductor Industries: Pharmaceutical industry made greater use of inventors with university background and cited university patents more often than semiconductor industry. Percentage of industry patents that involved inventors with university backgrounds and that cited university patents increased substantially in both industries.

  5. Main Findings Firm-level: Firms with large R&D operations in both industries and young firms in the pharmaceutical industry were more likely to interact with university researchers. Patents of firms that employ inventors with university patenting experience more likely to cite university patents as prior art. * More interaction does not necessarily mean more transfer of knowledge.

  6. Outline • Literature Review • Data Sources and Assembly Procedure • Empirical Findings • Conclusion

  7. Literature Review • Effect of university research on industry R&D and innovation important Jaffe (1989), Adams (1990), Mansfield (1991), Cohen, Nelson, Walsh (2002)

  8. Literature Review • Indirect/anecdotal evidence that transmission of knowledge via person-to-person contact important  Studies that find technological diffusion is geographical limited: Jaffe (1989), Jaffe, Trajtenberg, Henderson (1993), Zucker, Darby, Brewer (1998), Mowery, Ziedonis (2001)  Studies that show employment of or collaboration with university research personnel is important: Cohen, Nelson & Walsh (2002), Almeida & Kogut (1999), Zucker, Darby, & Armstrong (2001), Kim & Marschke (2005)

  9. Empirical work based on 3 data matching exercises • U.S. Patent Bibliographic data contain all patents in electronic form from 1975 • Patent information includes inventor and assignee names (but no unique identifier for either) • Name-matching exercise produced 2.3 million scientist panel that contains patent history and some educational information • Merged with firm-level information if assignee is public in pharmaceutical or semiconductor industry

  10. Data Sources

  11. Data Sources

  12. Data Assembly Procedure 1. Inventor name matching in Patent BIB data  The inventors listed is comprehensive and include only those inventors that make a creative contribution to the innovation underlying the patent  Method similar to Trajtenberg (2006)  Treat each entry in inventor name field as a unique inventor  With N names, generate N(N-1)/2 pairs (N = 5.1 million, N(N-1)/2 = 13 million)  For each pair, decide if the two names belong to the same inventor, based on the following criteria

  13. Data Assembly Procedure  Two names in a pair are matched if the SOUNDEX codes of their last names and their full first names are the same, their middle name initials are not different, and at least one of the following 3 conditions is met: (1) the full addresses are the same; (2) one name is an inventor of a patent that is cited by another patent whose inventors include the other name; or (3) the two names share the same co-inventor.

  14. Data Assembly Procedure  Two names in a pair are matched if the two names have the same full last and first names, and at least one of the following 2 conditions is met: (1) the two have the same zip code; or (2) they have the same full middle name.  Transitivity  2.3 million inventors out of 5.1 million names (45%)

  15. Data Assembly Procedure 2. Matching the Dissertation Abstract data to the inventors in the patent data  List of names matched to a unique inventor  Search those names in Dissertation data  When multiple names matched, randomize.  64,507 M.A. or Ph.D. degree holders out of 2.3 million inventors (3%)

  16. Data Assembly Procedure 3. Firm-assignee matching  Select firms in the Compact D/SEC data with SIC 2834 (pharmaceutical preparation) or 3674 (semiconductor and related devices)  Use subsidiary info in D/SEC data to accurately identify all subsidiaries of parent firms  Merge patent data with Compact D/SEC data  Use S&P data to track name changes  Merge info on firms’ founding years

  17. Data Assembly Procedure 4. Combining inventor and firm databases  Link inventor database to firm database to produce a data set on inventors and patents that include firm-level data  Add citation info to patents

  18. Empirical Findings Three Measures of University Influence • Whether an inventor with university patenting experience appears on industry patent • Whether an inventor with an advanced degree appears on industry patent • Whether an industry patent cites university patents

  19. Figure 1A. Percentage of industry patents that list at least one inventor named as inventor on a university patent in the previous 10 years (UNIV)

  20. Findings in Figure 1 • A steady increase in the interaction between university research and industry innovation • A big blip in 1995  Due to a change in patent law in 1995  Before 1995, successful applications received 17 years protection (from grant date) while applications filed after 1995 received a 20 year monopoly (from application date).  Patents applied for prior to 6/8/1995 but granted after 6/8/1995 received the longer of the periods of protection offered by the two schemes.  Industry patents relying on university research and thus basic research may have longer shelf life which created a strong incentive for firms to rush applications.

  21. Figures 1B, C. Percentage of industry patents that list at least one inventor named as inventor on a university patent in the previous 10 years (UNIV) * Firms may have been interacting with university research in early years at the same rate as in later years, but because universities infrequently patented before the 1980s, we do not detect it.

  22. Why the increase in UNIV? • Firms may have been interacting with researchers with university-research experience in early years at same rate as in later years, but because universities infrequently patented before the 1980s, we do not detect it. • Decomposition of UNIV increase shows • 2/3 of UNIV increase due to increase in availability of university-experienced researchers per patent • 1/3 due to increased likelihood that a university-experienced inventor used by industry

  23. Figure 2A. Percentage of industry patents that include at least one inventor with an advanced degree (ADVDEG)

  24. Figures 2B, C. Percentage of industry patents that include at least one inventor with an advanced degree (ADVDEG) * Is upward trend in percentage simply due to the fact that there are more inventors holding advanced degrees?

  25. Why the increase in ADVDEG? • Is upward trend in percentage of industrial patents with degreed inventors on them due simply to greater numbers of inventors holding advanced degrees? • Decomposition of ADVDEG increase shows • Little of increase due to increase in pool of degreed inventors • Most of increase due to increased likelihood that a degreed inventor used by industry

  26. Figure 3A. Percentage of industry patents that cite a university patent applied for within the previous 10 years (UCITE)

  27. Figures 3B, C. Percentage of industry patents that cite a university patent applied for within the previous 10 years (UCITE) * Is upward trend in percentage simply due to the fact that universities are patenting at higher rates?

  28. Why the increase in UCITE? • Is UCITE increasing because universities are patenting at higher rates or because university patents are more attractive to industry? • Decomposition of UCITE increase shows • Most of increase due to increase in number or university patents, not to increased likelihood that a university patent will be cited

  29. Firm-level Regressions • Dependent variables: the university influence measures (fraction of firm’s patents that list university-experienced inventors, inventors with advanced degrees, cite university-assigned patents) • Independent variables: number of inventors, number of employees, R&D expenditures per inventor, number of product lines, firm age • Separate regressions for semiconductor, pharma industries

  30. Main results • Size of R&D enterprise positively correlated with UNIV, ADVDEG, UCITE • Size of firm, scope, R&D expenditure per inventor not significantly correlated with dependent variable • In pharma firms, age negatively correlated with UNIV, ADVDEG, UCITE (significant in the case of UNIV) • For both types of firms, having university-experienced inventors increase likelihood that firm patents cite university patents

  31. Conclusion 1. Increasing use of inventors with past experience conducting university research and with advanced degrees economy-wide and in pharmaceutical and semiconductor industries. 2. Increase in citing of university patents economy-wide and in the pharmaceutical and semiconductor. 3. Use of university-experienced and degreed inventors higher in pharmaceutical industry. Citation rates also greater in pharmaceutical industry. 4. Firms with larger R&D enterprises disproportionately access university research. Younger pharmaceutical firms disproportionately access university-experienced inventors. 5. “Employing” university-experienced researchers is correlated with citing university patents.

  32. Does the presence of university-experienced researchers increase the firm’s ability to access university research? • Do firms with more university-experienced researchers produce more patents that cite university patents?

  33. Conclusion 1. Increasing use of inventors with past experience conducting university research and with advanced degrees economy-wide and in pharmaceutical and semiconductor industries. 2. Increase in citing of university patents economy-wide and in the pharmaceutical and semiconductor. 3. Use of university-experienced and degreed inventors higher in pharmaceutical industry. Citation rates also greater in pharmaceutical industry. 4. Firms with larger R&D enterprises disproportionately access university research. Younger pharmaceutical firms disproportionately access university-experienced inventors. 5. “Employing” university-experienced researchers is correlated with citing university patents.

  34. Number of patent applications

  35. Appendix A. SOUNDEX coding system A SOUNDEX code for a surname is an upper case letter followed by 6 digits. For example the SOUNDEX code for Kim is K500000, while that for Marschke is M620000. The first letter is always the first letter of the surname.

  36. The rules for generating a SOUNDEX code are: 1. Take the first letter of the surname and capitalize it. 2. Go through each of the following letters giving them numerical values from 1 to 6 if they are found in the Scoring Letter table (1 for B, F, P, V; 2 for C, G, J, K, Q, S, X, Z; 3 for D, T; 4 for L; 5 for M, N; 6 for R; 0 for Vowels, punctuation, H, W, Y). 3. Ignore any letter if it is not a scoring character. This means that all vowels as well as the letters h, y and w are ignored. 4. If the value of a scoring character is the same as the previous letter then ignore it. Thus if two ‘t’s come together in the middle of a name they are treated exactly the same as a single ‘t’ or a single ‘d’. If they are separated by another non-scoring character then the same score can follow in the final code. The name PETTIT is P330000. The second ‘T’ is ignored but the third one is not since a nonscoring ‘I’ intervenes.

  37. 5. Add the number onto the end of the SOUNDEX code if it is not to be ignored. 6. Keep working through the name until you have created a code of 6 characters maximum. 7. If you come to the end of the name before you reach 6 characters, pad out the end of the code with zeros. 8. Optionally you can ignore a possessive prefix such as ‘Von’ or ‘Des’. See "Using the Census SOUNDEX," General Information Leaflet 55 (Washington, DC: National Archives and Records Administration, 1995) for the detailed method.

More Related