1. Scholarship and Inventive Activity in the University: Complements or Substitutes? Brent Goldfarb University of Maryland Gerald Marschke University at Albany, State University of New York Institute for the Study of Labor (IZA) Amy Smith University at Albany, State University of New York EPFL: September 30, 2006

2. Motivation Involvement of industry in University-based scientific research is on the rise: Have academics’ inventive and commercialization activities inhibited academic scholarship? (Guena & Nesta 2003; Mowery et al. 2001; Stokes 1997; Nelson & Romer 1996)

3. Important Question Historically, university research important to U.S. economic growth (e.g., Jaffe, 1989; Nelson and Rosenberg, 1993; Henderson, Jaffe et al., 1998). Between 1975-78, annual social rate of return to university research 28% (??) (Mansfield, 1991). Increase in inventive and commercialization activity in US universities (Henderson, Jaffe, et al 1998; Thursby and Kemp, 2001; Kenney, 1986; Argyres and Liebeskind, 1998; Owen-Smith and Powell, 2001).

4. Data 11 year panel (1990-2000) Stanford University research scientists (Biochemistry & Electrical Engineering) Stanford is the Canary in the Mineshaft

5. Findings Inventive activity increases publication quality (not quantity) in biochemistry. Weak relation in engineering.

6. Literature StudyFindingsDataTechnique + Solution to endogeneity problem Agrawal & Henderson (MS 2002 0 quantity + quality Panel: 236 MIT EE, CS & ME Depts None Breshchi et al. (2005) +Panel: 2957 Italian academics Matched comparison None Markeiwicz & DiMinin (2005); + quantity + quality Matched panel: 300 scientists Multiple disciplines 2SLS Instruments: University patents, department patents; Stephan et al. (2005) + quantity (field differences) Cross-section of Doctoral Recipients Instruments: lagged pubs Azoulay et al. (2006) + quantity 0 quality Panel of 3,862 life-scientists Lagged outputs as “selectors” into patenting regime.

7. Advances Instruments –Venture capital outlays in scientist’s area –Inventive activity of others in department Better Econometrics –Wooldridge quasi-differencing transformation to remove fixed effects Disclosures not patents Control for teaching

8. Data and methods Using scientist-level data we estimate: Invention endogenous Instruments: Effort demand shifter: affect inventor effort towards inventive and commercialization relative to publication output.

9. Count data specification Wooldridge quasi-differencing to remove fixed effects:

10. Data Tenure-track faculty in Stanford’s Electrical Engineering and Biochemistry Departments between 1990 and 2000. EE: 57 inventors, 197 inventions, 1362 pubs Biochemistry: 15 inventors, 52 inventions, 592 pubs Teaching histories Venture capital outlays by broad industrial category Sources: –ISI Web of Science –ISI’s Science Citation Index, weighted by journal’s citation index –Stanford University Records (OTL, Catalogues) –Venture Economics Database

11. Biochemistry: VariableMeanStd. Dev.MinMax Total Publications2.9613.567024 Weighted Publications5.4528.301058.667 Total Credits6.3286.894060 Total Inventions.428.95307 Total Licenses.141.50705 Note: 460 observations (scientist years) for 57 scientists

12. VariableMeanStd. Dev.MinMax Total Publications4.4184.213023 Weighted Publications34.17940.0950251.344 Total Credits5.1043.800017 Total Inventions.3881.00308 Total Licenses.321.96205 Note: 134 observations (scientist years) for 15 scientists Biochemistry:

13. All models include fixed effects, year dummies, tenure status, lagged weighted publications, and an intercept DV:Total PubsWeighted Pubs IVPoisson MLE GMMLS2SLS GMM Teaching2.72** (1.18).162 (.198) 2.73* (1.18) 2.13 (1.47) 2.50 ° (1.45) 1.91* (.96) Inventions.68* (.40).09 (.40).68 (.40) 1.76 (1.79).70* (.37) Licenses1.37 (4.51) ObsN = 366, n = 49 N = 346, n = 57 N = 403, n = 57 N = 346, n = 57 R 2 (LL)(-547).128.112.124 Sargan P- value.291.236 Electrical Engineering:

14. DV:Total PubsWeighted Pubs Variable Poission MLE LS2SLS GMM Teaching -1.59 (0.11) -8.42 (7.00) -0.52 (10.10) -9.07 (7.92) -10.12 ° (6.20) Inventions.03 (.04) 10.86 ** (3.12) 27.53 * (13.67) 13.58 *** (2.97) Licenses 15.51 ** (6.89) Obs N = 119, n = 15 N = 104, n = 15 N = 119, n = 15 R 2 (LL)(-198).00.03.01 Sargan P- value.94 All models include fixed effects, year dummies, tenure status, lagged weighted publications, and an intercept Biochemistry:

15. Results Each invention increases weighted pubs –40%-80% in biochemistry (mean = 34) –Not robust but: 10% in electrical engineering (mean = 5.2) Teaching complementary in EE only (maybe substitute in biochemistry?) Licenses (commercialization activities) complement publishing in biochemistry

16. Conclusions Commercialization concerns overblown Interpretation: Maybe just life sciences are more likely in Pasteur’s Quadrant and quality researchers excel in both.