Correlation Does Not Imply Causation: The False Comparison of Silicon Valley and Boston’s Route 128

Source: Andrew Ryzhkov (Andrei Ryjkov)

It turns out that global climate change is caused by a decrease in the number of pirates. Don’t believe it? See the chart to the right.

This is an example of mistaking correlation for causation. Correlation does not imply causation.

A Developing Problem of Presumed Causation

The assumption that correlation implies causation arises often in the ongoing debate about the effects of noncompetes on the economy.

At the core of the debate is a common misconception that California’s ban on noncompetes is the reason that California surpassed Massachusetts (and all other states) in the high-tech industry. Even assuming there were any truth at all to that conclusion, the reality is likely that the ban is more correlative than causative. Otherwise, Oklahoma and North Dakota would be teeming with tech companies (as would the rest of California, outside of Silicon Valley). To state the obvious: they are not.

Nevertheless, this Silicon Valley/Massachusetts myth coincides with various mistaken interpretations of the conclusions to be drawn from preliminary research about the impacts of noncompetes. While indiscriminate general acceptance of these conclusions is harmless, the real problem lies with the fact that these assumptions are now creeping into legislative policymaking as the justification for new legislation. For example, see the Pennsylvania “Freedom to Work Act” bill (listing the supposed evils of noncompetes in support of the act) and Washington’s new noncompete law (noting, “[t]he legislature finds that workforce mobility is important to economic growth and development,” as the basis for banning certain noncompetes). And, now spurred on by a handful of anti-noncompete advocates, the FTC is getting on the bandwagon, ready to “try to . . . develop the evidence” to justify a change, i.e., searching for support for an ideological opposition to noncompetes. See FTC’s Simons Signals Potential Rulemaking on Noncompete Clauses.


The Silicon Valley/Massachusetts Miracle comparison seems to have its roots in AnnaLee Saxenian’s 1994 book, Regional Advantage: Culture and Competition in Silicon Valley and Route 128. (“Silicon Valley Versus Route 128: A look at how companies are shaped by the business and social cultures around them” provides an executive summary.)

Building on Dean Saxenian’s work, many scholars have taken their research into more fundamental, foundational directions, focusing on an empirical understanding the impacts of noncompete enforcement on employee mobility, innovation, entrepreneurship, the number and quality of startups, investment in employee development and training, and wages, among other things.

Much of this work has been performed by Professors Matt MarxEvan Starr, Norman BisharaLee FlemingOrly LobelJJ Prescott, and Deborah Strumsky, among others.

Over time, Professors Marx and Starr have emerged as two of the leading voices in this area. Instructively, each recognizes that noncompetes are not the sole cause of Silicon Valley’s preeminence, that much of the empirical research is still quite preliminary, and that noncompetes do provide some countervailing benefits. For example, as a recent teaser to one of Evan Starr’s studies explained,

“[E]mployees tethered to their firms by noncompete clauses and nontransferable skills receive more on-the-job training and often get hired with less experience.”

“Firms are more willing to invest in professional development when they don’t have to worry about their talent walking out the door,” Starr says. “Their focus shifts from buying talent to making talent.”

* * *

Training also leads to increased opportunities for promotions, which typically come with higher pay

And, as Professor Marx has observed, “If it were the case that workers made fully informed decisions about signing a non-compete and could negotiate higher compensation in exchange for doing so, these agreements could be valuable for both workers and firms.” (This would suggest, for example, that an approach focusing on restructuring how noncompetes are agreed to may be (part of) a thoughtful change in the law.)

Accordingly, recognizing the pros and cons of noncompetes, Professor Marx identified several ways to curb their abuses. He suggests:

Informing workers about non-competes in advance of hiring, requiring that employers provide legal consideration in exchange for signing a non-compete, and disallowing the practice of ex-post judicial modification of non-competes are all possible directions for reform.

Following extensive analysis, the Obama Administration convened a working group on noncompetes (in which Professors Marx and Starr and I, along with a handful of others, participated) and, based on the results, issued a Call to Action that did not call for banning noncompetes, but rather identified ways to reign in the abuses (including a ban on noncompetes for low-wage earners and healthcare workers and mandatory advance notice that a noncompete will be a requirement of the job).

The Hype

Despite the potentially wide-ranging conflicting impacts at stake, frequent headlines and Op-Eds decry the abuses, point to the (preliminary) research as presumed proof that noncompetes are bad, identify Silicon Valley’s ban on employee noncompetes as the purported panacea, and conclude with a call to dismantle hundreds of years of noncompete law.

Three recent examples of the gamut of headlines of articles and Op-Eds are Interns’ Job Prospects Constrained by Noncompete Agreements, Spread of noncompete agreements hurts workers and inhibits entrepreneurs, and Send Noncompete Agreements Back to the Middle Ages. The articles have become so ubiquitous that even the Onion has contributed its parodical take on the subject: Noncompete Clause In Lease Bars Tenants From Living Anywhere Else For 90 Days After Moving Out.

The Hype in Context

It is a mistake to attempt to justify the anti-noncompete position with the overly simplistic Silicon Valley/Massachusetts dichotomy or preliminary research without understanding the error in the Silicon Valley-Massachusetts comparison, the limitations of the research, or the unintended consequences of the advocated-for reform.

The Silicon Valley/Massachusetts comparison started as little more than a thought experiment. As it has evolved, it has been supported by empirical research that is simply too nascent to be determinative at this point. But, more important, as the research continues to develop, it is starting to reveal nuances that alter the broader conclusions. For example, Evan Starr’s research suggests that, although noncompetes may inhibit the development of startups, the impact may be a positive one, skimming the cream from the top and cutting out the less likely to succeed startups.

Further, another recent study concluded, “employees . . . tended to be more productive, take fewer risks and align their behaviors with the goals of their employers” (at least in the mutual fund industry). Of course, as is the problem with this early stage of the research, another study, reaches the opposite conclusion: “In principle, companies might prefer a regime or a situation where they can keep their talent more easily. But at the same time, we find these results that suggest that their human capital becomes a little less productive in terms of innovation in this setting.”

So, while noncompetes may have played a part (perhaps some positive and some negative), there is much more to the story…

Silicon Valley is a Product of the Cluster Effect

A very simple (at least partial) explanation for Silicon Valley is that it had to happen somewhere, and California happened to be the right place at the right time.

Silicon Valley is an example of a recurring phenomenon known as a “business cluster” or “industry cluster.” “[A]s firms cluster around talent, and talent is in turn drawn to those firms, the result is a self-reinforcing trend toward ever-richer, ever-costlier metro areas that are economically dominant over the rest of the country.”

In that regard, Silicon Valley is where HP started and, importantly, had the foresight to foster a relationship with Stanford University, thereby sowing the seeds for the tech industry. Where HP faltered, however, was when, “[i]n 1976 an engineering intern at the company, Stephen G. Wozniak, built a prototype for the first personal computer (PC) and offered it to the company. Hewlett-Packard declined and gave Wozniak all rights to his idea; later he joined with Steven P. Jobs to create Apple Computer, Inc. (now Apple Inc.).” Of course, this then fed on itself.

And, while the PC industry started in Silicon Valley, Massachusetts companies (reacting as HP did) saw no future in them. Instead, companies like Digital Equipment CorporationWang Laboratories, and Prime Computer focused on minicomputers – an enormous mistake, perhaps second only to Dick Rowe’s decision to reject the Beatles.

In short, the time and environment were ripe for Silicon Valley to grow.

If Boston’s companies had Steve Wozniak and Steve Jobs, or had even just seen the potential of PCs, who knows how things would have turned out. To that point, as noted above, there have been many instances of business clusters over time.

Tin Pan Alley (1910)

For example, Tin Pan Alley was the Silicon Valley of its day for sheet music.

Starting around 1885, music publishing was a growing industry in several different cities, including Boston, Chicago, New Orleans, St. Louis, and New York. Yet, eventually, Tin Pan Alley in New York City managed to attract more and more artists and producers, ultimately becoming the epicenter of music publishing until the early to mid-1900s.

No one points to the shift from other cities to New York as the product of New York’s noncompete laws. Rather, it was the cluster effect, spurred on by increasing protection for an owner’s ability to control intellectual property (in that case, copyrights).

Interestingly, Massachusetts has its own high-tech industry cluster. Indeed, Silicon Valley’s draw notwithstanding, “Massachusetts boasts the nation’s best business environment for science and technology,” a ranking it has held since 2002. Indeed, when normalized for its size, Massachusetts actually has more venture capital investment than California.

And, more to the point, Massachusetts’s Kendall Square – “the most innovative square mile on the planet” – has become the Silicon Valley of biotech.

In short, the hype around noncompetes as the cause for the dominance of Silicon Valley over all others is misplaced, or at least, over emphasized. For example, a deeper inquiry into California law reveals that California’s employee noncompete ban notwithstanding, prior to 2008, California courts (the federal courts anyway) were willing to enforce restrictive covenants (like noncompetes) to a limited extent (using the so-called “narrow-restraint doctrine”). Moreover, companies used inter-company no-poach agreements (agreements not to solicit each other’s talent) as a self-help workaround until a 2009/2010 DOJ investigation and settlement.

Further, the focus on the difference between noncompete laws in California and other states ignores the trend in the courts to increasingly scrutinize noncompetes and require proof of a compelling threat to a former employer before they will enforce a noncompete.

Noncompetes are Certainly Sometimes Abused, But This is Nothing New, Nor is It the Cause of All of the Economy’s Woes 

The above is not intended to say that noncompetes have no downsides. They do. And significant changes to noncompete laws are warranted.

However, the scope of the abuses of noncompetes needs to be put in context, so policymakers can understand how widespread the problem actually is and how to properly tailor any legislation.

As a starting point, noncompetes are frequently abused. And, those abuses are primarily the use of noncompetes for low-wage and healthcare workers, the lack of advance notice given to employees that they will be required to sign a noncompete, and the use of overly restrictive agreements.

While noncompetes certainly seem to be more widely used than in the past, many have seized on the perception that employers are increasingly using noncompetes for lower level employees, and have correlated that with slow wage growth since the Great Recession, blaming the latter on the former.  For example, see Resistance to Noncompete Agreements Is a Win for Workers. However, we don’t actually know if either of those assertions is true.

As to whether the use of noncompetes has been increasing, it’s hard to say; there simply are no longitudinal studies (of which I am aware) that have found that the use of noncompetes has risen over the years. We know only that, as Professor Starr explained, “roughly 18 percent of the U.S. workforce [was] bound by a non-compete [in 2014]. Among low-skill workers, . . . without a college degree, it’s about 15 percent.”

But, we also know is that the use of noncompetes dates back at least to medieval times, when master craftsmen tried to restrain their apprentices from using the skills the masters taught them. And, a century ago, noncompetes were already being used for low wage workers, including for example, to prevent a laundry route delivery driver from servicing the same route for a new employer.

As to the effects on wages, we don’t know whether there is something about the way noncompetes have been used recently that has stifled wage growth. However, slow wage growth has apparently been a persistent problem for at least the last 50 years – not just since the Great Recession or concomitant supposed increase in use and abuse of noncompetes.

The bottom line is that we do not know for sure how noncompete use has changed over years, and we cannot pronounce noncompetes to be the cause of slow wage growth.

The Troubling Cognitive Dissonance on the Issue

There is a different but related substantial drag on the economy:  The misappropriation of trade secrets – one of the key things that noncompetes are designed to prevent.

The Center for Responsible Enterprise and Trade ( and PricewaterhouseCoopers estimate that the cost of trade secret misappropriation is between one and three percent of U.S. GDP, possibly costing U.S. companies as much as $480 billion per year. Indeed, the threat to the economy and the innovation reflected in our trade secrets is so great that it led to the passage of the Defend Trade Secrets Act of 2016 (establishing a federal private right of action for trade secret misappropriation).

Misappropriation of trade secrets is directly related to the enforcement of noncompetes because the biggest threat to trade secrets is employees, especially as they move from one job to another. Indeed, one study suggests that 85 percent of trade secret thefts are committed by either an employee or a party to a contract. And, another reveals that as many as 72 percent of employees take – and are willing to use – their employer’s trade secrets when they leave their jobs. Similarly, [t]he U.S. Chamber of Commerce estimates that 75% of employees steal from the workplace and that most do so repeatedly.”

Accordingly, with so much focus on the risks posed to companies’ trade secrets by the movement of employees, there is a bit of cognitive dissonance in the rush to ban noncompetes, which are oftentimes the best tool to prevent that very risk.

Be Careful What You Wish For

With many states vying to change their noncompete laws in a vain hope of developing the next Silicon Valley, lawmakers need to understand the potential unintended consequences of significant changes (beyond measures designed to address noncompete abuses), including, for example, significantly increasing the likelihood that trade secrets will be unlawfully taken to a competitor and the volume of more-costly trade secrets litigation.

Such laws would also harm positive aspects of noncompetes, including, in addition to the benefits identified above, that employees “who learn of their noncompete before accepting their job, . . . have 9.7% . . . higher earnings, are 4.3 percentage points more likely to have information shared with them (a 7.8% increase relative to the sample average), are 5.5 percentage points more likely to have received training in their last year (an 11% increase), and are 4.5 percentage points more likely to be satisfied in their job (a 6.6% increase) relative to those employees without noncompetes.”

So, the mad push – based on early stage empirical evidence and faulty assumptions – to change noncompete laws to match California is, in the end, not only unnecessary, it may be counterproductive.

The Takeaways

I end where I began:  The proper regulation of employee noncompetes is an extremely complicated issue with no silver bullet. If policymakers believe that the prevailing market forces are insufficient to adequately limit the abusive use of noncompetes, legislative action is certainly an appropriate response. But, absent making the affirmative decision to base legislative policy on a visceral antipathy toward noncompetes, policymakers should engage in a critical analysis and avoid the temptation to rush to judgment. In particular, they need to be aware of – and at least consider, if not carefully weigh – the unintended adverse consequences of a policy that would ban noncompetes wholesale. The research into those considerations is, however, even more nascent than the research into the theorized adverse impacts of noncompetes.

In light of this, lawmakers should tread cautiously down any path that leads to the elimination of noncompete agreements. Indeed, the Obama Administration in its Call to Action and every state changing its noncompete laws has so far recognized this as well – and all have taken a tempered, considered approach.

Hopefully, that trend will continue.


This post was originally published on Law360 under the title, “Misconceptions In The Debate About Noncompetes.”

Thank you to Steve Riden for helping me edit it down from a tome (there’s a lot to cover)! And thank you to Erika Hahn for hours of help getting it finalized for Law360 and here.

Please note that the most current version of both the 50 State Noncompete Chart and 50 state and federal survey chart of trade secret laws can always be found on my firm’s resources page.