The decline of research in corporate research and development (R&D) Ashish Arora, Sharon Belenzon, Andrea Patacconi, Lia Sheer Along with: Honggi Lee, JK Suh Special acknowledgement: Wes Cohen 12/7/2018 1

Source: NSF S&E Indicators Business funded and performed research in the U.S., 1953-2015 Findings: Firms investing less in research even as overall R&D increases Absolute spending below 1999 in real terms Share of business in U.S. research is 30%, down from 45% (1990s) Share of research in business research is 20%, down from 30% (1990s) Implication: Less “R”, more “D” Source: NSF S&E Indicators

Trends in publishing by selected industries, U.S. firms 1980-2010 Scientific publication by U.S. corporations, 1980-2010 Publishing, patenting, and research: U.S. firms, 1980-2010 Trends in publishing by selected industries, U.S. firms 1980-2010 Publication Patent R&D Intensity (Right axis) “D” Patent Annual Publication R&D Intensity Annual Patent and Publication, per firm per year R&D Intensity Publications “R” Finding: Publication activity falls, while patenting increases, in a broad range of industries. Implication: Less R, more D Source: Arora, Belenzon, Patacconi, NBER 2015

Business R&D: Within industrial sector patterns Share of companies performing basic research out of all R&D performers, 1996-2006 Share of basic and applied research by industry sector 1996-2013 Share of research in business R&D Share of firms performing basic research Finding: The decline in corporate research is broad-based, present in a range of industrial sectors Implication: The causes “Less R, more D” are broad-based Source: NSF Science and Engineering Indicators, various years The figures present the share of basic and applied research in total domestic R&D paid and performed by companies. The data is bas on NSF’s Survey of Industrial Research and Development (SIRD) for 1996-2007 and the Business R&D and Innovation Survey (BRDIS) since 2008. Chem&Pharma- SIC 28 and NAICS 325; Electr&Semicond- SIC 357 & 36 and NAICS 334; Machinery- SIC 351–56 & 358–59 and NAICS 333.

Decline in private value of research Note: The estimates are from regressing stock market value against firm’s assets, R&D stock, publications stock, patents stock, and a complete set of dummies for year and industry. The sample consists of firms with at least one patent and one publication. The sample period is 1980–2006. M&A valuation Note: This figure presents estimates of elasticity of acquisition value with respect to publication and patent stocks by year cohorts for acquired firms. The estimates are from regressing stock market value against firm’s assets, sales, publications stock, patents stock, and complete set of dummies for year, industry codes, target country and acquisition year. The sample includes all SDC Platinum deals with complete information on target firm value, assets, and sales. The sample period is 1985–2007. Stock market valuation Finding: The value of publications has dropped and the value of patents has increased over time for both investors and managers Implication: Over time, firms and investors value “D” relative to “R” Source: Arora, Belenzon, Patacconi, NBER 2015

Finding: VC-backed startups also focus on D rather than R Publishing, patenting and investment, U.S. VC-backed firms, 1980-2006 Patents Publications VC investment (Right Axis) Patents, publications VC investment, total ($mm), Round A Patent, per firm per year Publication, per firm per year VC investment, per firm per year (R axis) Patents, publications, per firm VC investment, per firm ($mm), Round A Patents, publications and investment ($M), total Patents, publications and investment ($M), per firm “R” “D” Note: Figure A presents total annual flow of publications and patents over time for U.S. VC-funded patenting firms. Figure B presents annual publications and patents per firm over time for U.S. VC-funded patenting firms. The sample consists of U.S. VC-funded firms, first funded between 1980-2006. Publication and patents in any year are from any VC-funded firm receiving initial funding in the previous five years (excluding firms that have gone public). VC investment in a given year is total VC funding received by startups in the A round.VC data is from SDC Platinum VentureXpert database. Scientific publications include all articles published by our sample firms between 1980 and 2006, and covered in Web of Science "Science Citation Index" and "Conference Proceedings Citation Index-Science“. The analysis is based on a set of 5K firms, out of which 1.5k firms were identified as patenting firms (have at least one patent in the first five years after first investment) and are included in the analysis. An addition 15K firms are being processed for analysis. Finding: VC-backed startups also focus on D rather than R Implication: Startups will not pick up the slack Source: Arora et al. 2017. in progress

Using research in invention: Citations to scientific publications by corporate patents Average age of scientific publications cited in patents and average cites per patent, 1980-2006 Cited publication age (left axis) Citations to scientific literature as fraction of all citations Year Cites to scientific publications by corporate patents, 1986-2006 Share of citations to science Cited publication age Note: This figure presents average publication age of cited articles by patents (NPL) and share of citations to science per firm-year for our sample firms. Cited publication age is the difference between patent grant year and year of publication of the cited article. Scientific citation share is the ratio between patent citations to leading scientific journals and the total number of references the patent makes. Finding: Corporate patents cite science at higher rates over time; cited science is not older science Implication: R remain useful for D Finding: Corporate patents cite external science at higher rates over time Implication: Firms rely on externally funded R for D, including federally funded R Source: Arora, Belenzon, Patacconi, NBER 2015; Arora, Belenzon, Sheer, NBER 2017

Conclusions & implications Findings Corporations are withdrawing from research Startups will not fill the breach Research findings continue to be relevant for invention Corporations are using external research Interpretation: A division of innovative labor Reallocation of research from large corporate labs to more efficient and specialized research organizations (e.g., universities) Established firms source inventions from universities, often through start-ups Policy Implications Public funding for research is even more important for maintaining American competitiveness