RAPPAPORT POLICY BRIEFS InstituteRAPPAPORT for Greater INSTITUTE Kennedy Schoolfor of Government, Greater Boston April 2009

To Compete, or To Non-Compete? The Eff ects of Employee Non-Compete Agreements By Matt Marx, (Harvard Business School), Deborah Strumsky (University of North Carolina - Charlotte), and Lee Fleming (Harvard Business School)

The movement of talented individuals likely to be more affected by non- This brief is based on “Mobility, Skills, and among different fi rms is a key to competes. the Michigan Non-compete Experiment,” the growth and development of new which is forthcoming in Management Michigan’s 1985 reversal of its law Science. fi rms and, by extension, to regional banning the use of non-compete Matt Marx economic development. While all agreements created a natural Matt Marx is a doctoral candidate at the states have laws barring mobile Harvard Business School and will be an experiment that allows us to fi ll these assistant professor at MIT’s Sloan School of employees from revealing their gaps. In particular, a close examination Management in the fall. former employer’s trade secrets, many of thousands of patents before and Deborah Strumsky states, including Massachusetts, go after the law’s passage suggests Deborah Strumsky is an assistant professor further and allow the use of “non- at the University of North Carolina — that after passage of the law the job Charlotte. compete” agreements—provisions in mobility of inventors in Michigan fell Lee Fleming employment contracts that prevent 8.1 percent, compared to inventors employees from taking jobs in similar Lee Fleming is the Albert J. Weatherhead III in other states that continued not to Professor of Business Administration at the businesses for a certain period of time enforce non-competes. Moreover, Harvard Business School. after they leave their current fi rm. mobility for individuals who had fi rm- Rappaport Institute for Greater Boston Established fi rms in knowledge- specifi c skills or who specialized in a The Rappaport Institute for Greater Boston intensive fi elds generally favor such narrow range of technologies fell by aims to improve the region’s governance by fostering better connections between agreements as a way to ensure that more than 15 percent. The Michigan scholars, policy-makers, and civic leaders. The trade secrets are not revealed as well data and subsequent qualitative studies Rappaport Institute was founded and funded by the Jerome Lyle Rappaport Charitable as to honor customer confi dentiality further suggest that individuals who Foundation, which promotes emerging leaders and prevent competitors from do leave fi rms that used non-compete in Greater Boston. appropriating the specialized skills agreements were more likely to take © 2009 by the President and Fellows of Harvard College. The contents refl ect the and knowledge of others’ employees. “occupational detours” and also to views of the authors (who are responsible for the facts and accuracy of the research In recent years, however, a small take jobs with larger fi rms. Taken herein) and do not represent the offi cial but growing body of research has as a whole, these fi ndings suggest views or policies of the Rappaport Institute . suggested that more new fi rms form that states that allow non-compete Rappaport Institute for Greater Boston in states that ban such agreements, agreements may want to reexamine 79 JFK Street, Cambridge, MA 02138 Telephone: (617) 495-5091 such as California (Gilson 1999; those policies. Email: [email protected] Almeida and Kogut 1999; and Stuart http://www.hks.harvard.edu/rappaport and Sorenson 2003). However, causal Background evidence for these assertions remains The components of non-competition thin (Fallick, Fleischman, and Rebitzer law have not changed materially for 2006) and we know little about which centuries. The earliest recorded case groups of knowledge workers are was settled in England in 1414, only To Compete, or To Non-Compete? RAPPAPORT INTITUTE POLICY BRIEFS

a few decades after the Bubonic plague had business of any kind is to that extent void.”1 decimated the European labor supply and Section 16600 has been upheld by the courts subsequent to the Ordinance of Labourers, and was reaffi rmed in August 2008 by the which essentially outlawed unemployment California Supreme Court’s ruling in Edwards in post-medieval England. Thus a plaintiff’s v. Arthur Andersen. Citing the attenuating request to enjoin one of his former clothes impact of non-competes on employee mobility, dyers from working in the same town for six Gilson proposed that this practice is in fact “the months was met with disdain from the judge, causal antecedent” of the high-velocity labor who threatened the plaintiff himself with jail market as well as the unique culture Saxenian time for having sought to restrict a citizen from attributes to Silicon Valley. practicing his trade (Decker 1993). Gilson’s hypothesis went untested until 2003, Firms use non-competes to protect their when Stuart and Sorenson (2003) examined interests: to prevent the disclosure of trade the effect of initial public offerings (IPOs) and secrets, to honor customer confi dentiality, and acquisitions on founding rates of biotech fi rms to prevent competitors from appropriating in regions that enforce non-competes compared the specialized skills and knowledge of its with those that did not. That proportionally employees (Valiulis 1985). One might argue more biotech fi rms were founded in states that that trade secrets are already protected by the proscribe enforcement of non-competes is non-disclosure agreement (NDA) employees consistent with Gilson’s hypothesis. However, are generally required to sign, but violations as the Stuart and Sorenson analysis measures of an NDA can be diffi cult to detect or prove fi rm foundings, it does not directly track (Hyde 2003). Preventing an ex-employee from individual mobility. joining a competitor via a non-compete reduces An individual-level study of mobility the likelihood that an employee will violate the was undertaken in Fallick et al.’s (2006) corresponding NDA via so-called “inevitable examination of the computer industry in disclosure” of confi dential information at a new Silicon Valley. Using month-by-month data job (Whaley 1999). from the Current Population Survey in the States, like Massachusetts, that top 20 metropolitan areas, they found an increase in intraregional employee mobility allow non-compete agreements for the California computer industry vs. other may want to reexamine those states. The authors caution, however, against policies. interpreting their results as unequivocal evidence linking non-competes and mobility.

Although the law of trade secrets is fairly Michigan’s Reversal of Non-Compete similar across U.S. states (Hyde 2003), Enforcement enforcement of non-competes varies At the turn of the 20th century, the metropolitan signifi cantly from state to state. For example, area of Detroit, Michigan in many ways California’s Business and Professions Code resembled the Silicon Valley of the last few section 16600 is reminiscent of early English decades. Growth of the nascent auto industry law: “Except as provided in this chapter, every was explosive, with 500 fi rms entering contract by which anyone is restrained from before 1915 (Klepper 2002). Ten years prior, engaging in a lawful profession, trade, or the Michigan legislature had passed statute 2 To Compete, or To Non-Compete? RAPPAPORT INTITUTE POLICY BRIEFS

445.761 (bearing resemblance to California’s the productivity of inventors visible outside prohibition): “All agreements and contracts of their current employer. Second, since each by which any person…agrees not to engage patent lists both the inventor’s hometown in any avocation or employment…are hereby and the patent assignee—if not owned by the declared to be against public policy and inventor, in which case the fi eld is blank or illegal and void.” This law governed non- lists the inventor, the patent is “assigned,” compete enforcement until March 1985, when typically to the inventor’s employer—we know the Michigan Antitrust Reform Act (MARA) the inventor’s employer and state of residence. repealed section 445 and with it the prohibition Third, by combining the NBER patent fi le on enforcing non-compete agreements. (Hall, Jaffe, and Trajtenberg 2001) with weekly More than 20 pages of legislative analysis updates from the US Patent & Trademark of MARA by both House and Senate Offi ce, we are able to observe these inventors subcommittees does not mention non-competes longitudinally from 1975 through 2006. (When as a motivation for the bill (Bullard 1983a; available, we also included more limited data Bullard 1983b; Bullard 1983c; and Bullard from 1963-1974.) 1985). This may be a consequence of MARA After applying various algorithms to account having been modeled on the Uniform State for some of the documented weaknesses in Antitrust Act of 1983, designed to “make patent data (Griliches 1991) and to identify uniform the law with respect to the subject of specifi c inventors, the resulting dataset contains this act among those states that enact similar provisions.” Given that the impetus for the Firms use non-competes to change in law appears to have been general protect their interests: to antitrust reform and not specifi cally altering prevent the disclosure of trade non-compete enforcement, it appears that the secrets, to honor customer 1905 statute prohibiting non-competes was confi dentiality, and to prevent inadvertently repealed as part of the anti- trust reform. If so, then Michigan’s change competitors from appropriating in enforcement would be an exogenous event the specialized skills and rather than an example of the legislature knowledge of its employees. simply “catching up” with the courts or responding to lobbying efforts. Even if it 98,468 inventors who patented in Michigan or were the case that behind-the-scenes lobbying in another non-enforcing state prior to MARA. by powerful interests contributed to the Following these inventors throughout the study legislature’s move—and we did not uncover period yielded 372,908 patents between 1963 any evidence of this—such a change would and 2006, for a patent-per-inventor ratio of still be exogenous to the inventors who are the 3.79. About 28 percent of these people – 27,478 subjects of this study, assuming that they would inventors – patented with different employers have been unaware of such efforts. during the study period. We also included controls for a variety of factors, including the Study Design particularly large presence of the automobile For several reasons, we used the U.S. patent industry in Michigan. database to examine inventor mobility. First, patents are public documents and thus make

3 To Compete, or To Non-Compete? RAPPAPORT INTITUTE POLICY BRIEFS

Results as measured by the percentage of patents in the states that indicate a change in assignee. We began by comparing the mobility rates of Overall, it appears that MARA did not cause an Michigan vs. other non-enforcing states from absolute decrease in Michigan mobility, though 1975 to 2000. (See Figure 1) Non-Michigan it may have contributed to a decrease relative states demonstrate a volatile and increasing to other states that continued to proscribe non- trend in mobility over the entire time period. competes. Michigan mobility increases similarly during the early years, but levels off in the 1980s, The marked upward trend of synthetic though it does jump in the late 1990s. Figure Michigan immediately following MARA 1 also includes a “synthetic” Michigan line further supports this interpretation. Rabaut (Abadie, Diamond, and Hainmueller, 2007). (2006) ascribed the real upturn in the late 1990s Prior to and including 1985, the algorithm to a judicial pendulum swing. On a scale of 1 to used to generate this line reconstructs “real” 10, with 1 being complete inability to enforce Michigan prior to 1986 from a composite of non-competes and 10 being the opposite, he similar states. In 1986 and later, the synthetic indicated that Michigan went from a 1 before line of Figure 1 is a prediction based on MARA to an 8 immediately after passage patenting in the control states, multiplied by the and then back to somewhere between 4 and same composite weighting determined before 6 because “judges got sick of non-competes. 1986. This analysis indicates that MARA At fi rst they felt they had to enforce them but does seem to be associated with less mobility then they looked harder at being ‘reasonable.’”

Figure 1: Yearly Moves/Patents for Michigan and Other Non-Compete States

Annual mobility rates of U.S. inventors with at least one patent prior to MARA in a non-enforcing state. “Synthetic Michigan” represents predictions of mobility in post-MARA Michigan, based on a weighted average of pre-MARA mobility in other non-enforcing states. MARA passed in 1985.

4 To Compete, or To Non-Compete? RAPPAPORT INTITUTE POLICY BRIEFS

Rabaut further reported that even employers continued to limit non-compete agreements. in Michigan became less enamored with Moreover, those who have completed projects non-competes over time, because while they which developed fi rm-specifi c human capital appreciated the use of non-competes as a as opposed to general human capital were 15.4 “hiring shield” they began to realize that it also percent less likely to change employers under deprived them of a “hiring sword.” a non-compete regime. Similarly, employees A similar pattern is revealed by modeling an who are highly technologically specialized are individual inventor’s decision to change jobs. found to be 16.2 percent less likely to change We estimate that the baseline predicted annual jobs. Moreover, the emigration of inventors probability of changing fi rms for non-Michigan from Michigan to states that continued not to inventors is 7.95 percent before MARA and recognize non-competes appears to increase 10.80 percent thereafter while the predicted after MARA’s passage. (See Figure 2) probability of mobility for Michigan inventors Eff ects on Ex-Employees is 7.18 percent before MARA and 8.98 percent afterwards. Put another way, after MARA, To better understand how individuals respond inventor mobility in Michigan dropped by 8.1 to limits on their future employment, we percent relative to mobility in other states that surveyed 5,000 members of the Institute of Electrical and Electronic Engineers (IEEE),

Figure 2: Annual Emigration Rates of U.S. Inventors

Annual emigration rates of U.S. inventors with at least one patent prior to MARA in a non-enforcing state. “Synthetic Michigan” represents predictions of those leaving post-MARA Michigan for states that continued not to recognize non-competes, based on a weighted average of pre-MARA mobility in other non-enforcing states. MARA passed in 1985. 5 To Compete, or To Non-Compete? RAPPAPORT INTITUTE POLICY BRIEFS

interviewed 50 individuals in the automatic conversations with employment lawyers and speech-recognition industry, and examined a review of specifi c non-compete agreements smaller set of data from Michigan. We found indicate that such contracts are typically that: constructed to survive involuntary separation • Many individuals took “occupational of employee from employer. Moreover, the detours” such as taking jobs in general interviews and survey responses qualitatively programming while their non-compete corroborate the results of the statistical analysis. agreements banned them from working And though we have attempted to control in their specialized fi eld. for alternative explanations of post-MARA mobility changes in Michigan, the models may • While small fi rms might be more be incomplete. attractive to those subject to non- compete agreements, many consciously Despite these caveats, Michigan’s experience chose to work for larger fi rms both strongly suggests that non-compete agreements because they wanted to shield a small discourage the local mobility of skilled fi rm from potential legal expenses workers, especially those with fi rm- and and because many small fi rms require technology-skills. In addition, it appears that particularly lengthy non-compete the agreements encourage ex-employees to take agreements. “occupational detours,” and shun small fi rms in favor of larger enterprises. Consequently, • Some managers and entrepreneurs are states that continue to allow widespread use of coming to believe that the value of non- such agreements as a way to protect established compete agreements to their fi rm may fi rms may instead be inadvertently creating a be outweighed by the benefi t of being “brain drain” of the very workers needed to able to tell potential employees that the create and build successful new fi rms. fi rm does not use such agreements.

Michigan’s experience strongly Endnotes suggests that non-compete 1 Note that although contracts typically stipulate agreements discourage the a “choice of law”—a state under whose laws mobility of skilled workers, the agreement is to be governed—in their 1971 especially those with fi rm- and Frame v. Merrill Lynch ruling the California technology-skills. courts forbade corporations from specifying out-of-state jurisdiction as a means of cherry- picking one’s non-compete enforcement Conclusion regime. While these fi ndings are striking, we interpret these results cautiously because the statistical analysis depends on patent data which, it should be emphasized, enable only imperfect matching of inventors across patents and imperfect observations of job changes. We also cannot determine whether job changes are voluntary or involuntary, although 6 To Compete, or To Non-Compete? RAPPAPORT INTITUTE POLICY BRIEFS

References Velocity Labor Market. Armonk, NY, M.E. Sharpe. Abadie, A., and A. Diamond, J. Hainmueller. (2007). “Synthetic Control Methods for Klepper, S. (2002). “The capabilities of new Comparative Case Studies: Estimating fi rms and the evolution of the US automobile the Effect of California’s Tobacco Control industry.” Industrial and Corporate Change Program.” NBER Working Paper w12831. 11(4): 645-666. Bullard, P. (1983a). Michigan Antitrust Reform Rabaut, L. (2006). Personal interview via Act: House Bill 4994, 1st Analysis. M. S. A. phone from Cambridge, MA. to Grand Rapids, Sect.: 1-8. MI., Nov. 7. Bullard, P. (1983b). Michigan Antitrust Reform Stuart, T. and O. Sorenson (2003). “Liquidity Act: House Bill 4994, 1st Analysis. M. S. A. Events and the Geographic Distribution of Sect.: 1-6. Entrepreneurial Activity.” Administrative Science Quarterly 48: 175-201. Bullard, P. (1983c). Michigan Antitrust Reform Act: House Bill 4994, 2nd Analysis. M. S. A. Valiulis, A. (1985). Covenants Not to Compete: Sect.: 1-4. Forms, Tactics, and the Law. NY, NY, John Wiley & Sons. Bullard, P. (1985). Michigan Antitrust Reform Act: House Bill 4994, 3rd Analysis. M. S. A. Whaley, S. (1999). “The Inevitable Disaster Sect.: 1-4. of Inevitable Disclosure.” University of Cincinnati Law Review, 67:809-857. Decker, K. (1993). Covenants Not to Compete. New York, NY, John Wiley & Sons. Fallick, B., C. Fleischman, and J. Rebitzer. (2006). “Job-Hopping in Silicon Valley: Some Evidence Concerning the Micro-Foundations of a High Technology Cluster.” Review of Economics and Statistics 88(3), 472-481. Gilson, R. J. (1999). “The legal infrastructure of high technology industrial districts: Silicon Valley, Route 128, and covenants not to compete.” New York University Law Review 74: 575-629. Griliches, Z. (1991). Patent Statistics as Economic Indicators: A Survey. Journal of Economic Literature, 28(4):1661-1707. Hall, B. H., A. B. Jaffe, and M. Trajtenberg. (2001). The NBER patent Citations Data File: Lessons Insights and Methodological Tools, NBER. Hyde, A. (2003). Working in Silicon Valley: Economic and Legal Analysis of a High-

7 RAPPAPORT INTITUTE POLICY BRIEFS

RECENT POLICY BRIEFS RECENT WORKING PAPERS

“Silver Bullet or Trojan Horse? The Effects of “Controlling Violent Offenders Released to Inclusionary Zoning on Local Housing Markets the Community: An Evaluations of the Boston in Greater Boston,” By Jenny Schuetz, Rachel Reentry Initiative,” By Anthony A. Braga Meltzer, and Vicki Been (Furman Center for Real (Harvard Kennedy School), Anne M. Piehl (Rutgers Estate and Urban Policy) NYU March 2009. University), and David Hureau (Harvard Kennedy School) September 2008. “Reviewing Chapter 40B: What Gets Proposed, What Gets Approved, What Gets Appealed, and “Losing Faith? Police, Black Churches, and What Gets Built?” by Lynn Fisher, (Massachusetts the Resurgence of Youth Violence in Boston,” Institute of Technology) November 2008. by Anthony A. Braga and David Hureau (Harvard Kennedy School) and Christopher Winship (Faculty “Reducing Youth Violence: Lessons from the of Arts and Sciences and Harvard Kennedy School) Boston Youth Survey,” by Renee M. Johnson, Revised October 2008. Deborah Azrael, Mary Vriniotis, and David Hemenway (Harvard School of Public Health) June “Land-Use Planning in the Doldrums: Growth 2008. Management in Massachusetts’ 1-495 Region,” by Christina Rosan (Postdoctoral Fellow, “The Greeness of Cities,” by Edward L. Glaeser Massachusetts Institute of Technology) and (Harvard University) and Matthew Kahn (UCLA) Lawrence Susskind (Massachusetts Institute of March 2008. Technology) September 2007.

“The Seven Big Errors of PerformanceStat” “The Massachusetts Community Preservation by Robert D. Behn (John F. Kennedy School of Act: Who Benefi ts, Who Pays,” by Robin Sherman Government) February 2008. (MPA ‘07, Kennedy School of Government) and David Luberoff (Rappaport Institute for Greater “Boston Bound: A Comparison of Boston’s Legal Boston) July 2007. Powers with Those of Six Other Major American Cities,” by Gerald E. Frug and David J. Barron “Leveling the Playing Field: Sincere and () December 2007. Sophisticated Players in the Boston Mechanism,” by Parag Pathak and Tayfun Sönmez (Harvard “Hard Choices for the Next Governor” University) May 2007. September 2006. “Power and Interest Groups in City Politics,” “The Impacts of Commuter Rail in Greater by Jeffrey M. Berry, Kent E. Portney, Robin Boston” by Eric Beaton (MUP, 06, Graduate School Liss, Jessica Simoncelli, and Lisa Berger (Tufts of Design, Harvard University) September 2006. University) December 2006.

“The Economic Impact of Restricting Housing “Massachusetts’ Hancock Case and the Adequacy Supply” by Edward L. Glaeser (Harvard Doctrine,” Conference Paper for “Adequacy University) May 2006. Lawsuits: Their Growing Impact on American Education, by Robert M. Costrell (Commonwealth “Why Are Smart Places Getting Smarter?” of Massachusetts) March 2006. by Edward L. Glaeser (Harvard University) and Christopher Berry (University of Chicago) March “Guarding the Town Walls: Mechanisms and 2006. Motives for Restricting Multi-family Housing in Massachusetts,” by Jenny Schuetz (Kennedy “Regulation and the Rise of Housing Prices in School of Government) March 2006. Greater Boston” by Edward L. Glaeser, Jenny Schuetz and Bryce Ward (Harvard University) “Creating an Anti-Growth Regulatory Regime: January 2006. A Case from Greater Boston,” by Alexander von Hoffman (Joint Center for Housing Studies, Harvard “Local Services, Local Aid and Common University) February 2006. Challenges” by Phineas Baxandall (Rappaport Institute for Greater Boston) November 2005. 8