Private Equity and Long Run Investments: The Case of Innovation Josh Lerner, Morten Sorensen, and Per Stromberg
Motivation We study changes in R&D and innovation for companies involved in buyout transactions. This helps us distinguish two opposing views of buyout transactions: Buyouts liberate firms from short-term agency problems arising in public firms (Jensen [1989]). Buyouts compromise long-term values to extract short-term rents (Shleifer, Summers [1988]). 2
Hypothesis 1: Private Equity investors as long-term investors Jensen [1989] The eclipse of the public corporation predicts that LBOs will become the dominant corporate organization form: Superior corporate governance. Concentrated ownership by active owners. Strong managerial incentives. Impact on innovation: Facilitate long-run investments that myopic (quarterto-quarter) public firms cannot. Avoid wasteful expenditures of others (Jensen [1993]) 3
I realize, gentlemen, that thirty million dollars is a lot of money to spend. However, it s not real money and, of course, it s not our money either. 4
Hypothesis 2: Private Equity investors as short-term investors Critics suggest different view (Shleifer, Summers [1988]): Investors compromise long-term value creation to enhance short-term performance. Renege on implicit and explicit obligations to employees and retirees. This helps investors flip offerings quickly, to pay large dividends: Boosts IRR and allow raising new funds sooner, enhancing fee income. Implication for innovation is a temptation to defer expenditures: These depress current accounting earnings with few immediate gains. 5
The Great Global Buyout Bubble, by Andrew Sorkin, New York Times 6
What we do We examine investment in innovation as one form of long-run investment. Investment in innovation presents an attractive testing ground: Costs must be written off immediately. Benefits may not be apparent for many years. But clearly important to long-term success. Systematic, well-understood measures available. We use patent data to overcome problems with gathering data for private firms. 7
Rationale for study Growth of private equity industry: Larger sample to work with. Investors now have more operational orientation but also more intense competition. Recently have observed greater representation of technology transactions. Patent data allow us to look beyond the publicto-private transactions: Private-to-private deals may have different features. 8
Relevant Literature (1/2) R&D and capital constraints: Greenwald, Salinger and Stiglitz [1991]: Case studies of auto and airline industries. Hall [1992]: 1247 R&D-performing manufacturing firms. Hao and Jaffe [1993]: Panel data on 81 firms in five high-tech industries. Himmelberg and Petersen [1994]: Panel data on 179 small high-tech firms. Conclusions: Internal finance availability important for R&D. Interpretation of pattern is challenging. 9
Relevant Literature (2/2) Hall [1992] considers public-to-private LBOs during 1980s: Notes these firms were doing little R&D before transaction. 4% of 1982 employment, but 1% of R&D. Concludes LBO wave unlikely to have much impact on innovation. Lichtenberg and Siegel [1990]: 43 whole-firm LBOs that filled out RD-1 survey R&D expenditures appear to increase after LBO on absolute and relative basis. More likely for R&D to increase for LBO firms than for matches. 10
Data: Transactions (1/4) Begin with CapitalIQ database: Transactions between 1/1/80 and 12/31/05. Focus on leveraged buyout investments. Classified in CIQ as Going Private, JV/LBO, LBO, Management Participated, or MBO. Closed and effective deals. Supplement with information from Dealogic, other CapitalIQ databases, Directory of Corporate Affiliations, Hoovers on financial characteristics, previous parents, exits. 11
Data: Match to patents (2/4) Match CIQ data to HBS patent database: Contains of all patents awarded through May 2007. Assignee names have been cleansed (relative to original USPTO data). Identify all LBO targets assigned at least one patent from three years prior to five years after buyout: Match on name and location. 12
Data: Divisional targets (3/4) If LBO target is a unit of a larger firm, patents likely to be assigned to parent. Identify all corporate parents in [-3,+5] window: CapitalIQ, Dealogic, DCA, Factiva, Google. Identify all patents assigned to parent with same inventor as LBO target. This method captures some, but not all, of the relevant parent patents. 13
Data: Trimming (4/4) For CapitalIQ buyouts from 1980 to 2005: 496 firms have successful patent applications filed from years -3 to +5 relative to buyout: Due to many recent buyouts Due to old economy nature of firms. 8,938 patents filed in this window, but >1/4 th assigned to Seagate: Second largest firm has 4%. We eliminate Seagate, leaving us with 6,398 awarded patents. 14
Investments and Exits by Year 80 70 60 50 40 30 20 10 0 Investments Exits 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 15
Patents and LBOs in sample 16
Patent Applications and Grants 1000 800 600 400 Applications Grants 200 0 1983 1986 1989 1992 1995 1998 2001 2004 2007 17
Industry composition 18
Number of patent applications, relative transaction 19
Methodology: three measures of innovation We measure three aspects of innovation: Patent quality: Economic impact. Basic/fundamental nature. Number of patent filings. Patent portfolio composition. 20
Patent quality measures Citation counts: Proxy for economic importance We use three year window to count citations Originality: One minus Herfindahl of classes of patents cited by patent. Generality: One minus Herfindahl of classes of patents citing patent. 21
Three-year citation counts for portfolio firm patents 22
Need for benchmark Citation rates change over time: Changing importance. Changing technology mixture. Changing propensity to cite overall. Use all patents in same USPTO technology class and grant year as control group: Compute baseline citation rates for these matching patents. Similar issues for originality, generality. 23
Average three-year citation counts for matching patents 24
Citations through 3 rd Year after Patent Grant For Patents Applied for in Years Relative to Buyout. Citations -3 to 0 +1 to +5 P-Value Unadjusted 1.99 2.49 [0.000] Relative 0.24 0.74 [0.000] 25
Multivariate statistical analysis We use Poisson and Negative Binomial specifications for citations. Consider both unadjusted and relative citation counts. Use individual year dummies as well as combined Post dummy. 26
Changes in citation count around buyout transactions 27
Fixed- and random-effects specifications 28
Three-year citation counts for portfolio firms (with fixed effects) 29
Other patent quality measures 30
Amount of patenting Truncation challenges: As-yet unissued patents. Assignment to corporate parents. Responses: Year and firm fixed effects. Limiting to observations before 1999. Limiting to firms with early and late patents. Still, this analysis is less conclusive 31
Number of patents granted 32
Composition of patent portfolios Now looking at firm level: Does the distribution of areas in which firms pursue innovation change? Where is increase in citations taking place? 33
Comparing citations in well and poorly populated classes 34
Comparing citations in growing and shrinking classes 35
Changes in citation count, controlling for patent class share 36
Summary of empirical evidence Patent quality appears to improve following buyouts: Not sacrificing originality or generality. Amount of patenting is probably not affected: More challenging analysis, due to data limitations. Composition of patent portfolio appears to focus on the more central technologies. 37
Concern #1 We may be double counting secondary buyouts: Same patents may appear both before and after a transaction. Repeat analysis, treating these patents separately: We count only first transaction. We delete these patents entirely. Makes little difference. 38
Concern #2 Are three years sufficient to capture patents citation counts? Citations are strongly serially correlated: Three-year citation count is a good proxy for total citation count. We also repeat analysis using 2 and 4 years windows: Find little or no difference to results. 39
Concern #3 Are differences due to investors cherry picking in divisional buyouts? Parent company may keep the best patents. We repeat analysis excluding divisional deals: Magnitude of difference in citations increases, still significant. Other results unchanged. 40
Concern #4 Results may be due to investors selecting targets with promising innovation portfolios: However, most targets are old economy firms where innovation is relatively small part of business potential. Pattern shows most of the improvement in years 2-3 following the transactions. Selection of targets unlikely to introduce large distortions. 41
Wrapping Up Innovation present a natural testing ground for understanding motivation of private equity firms. Our results are consistent with the positive view of private equity: We find an increase in innovation quality. No evidence of decline in fundamental nature of research. See a focusing of patent portfolios. Focusing of awards on high-impact areas. 42