September 1998

THE WEALTH OF SHARED INFORMATION: SILICON VALLEY'S HIGH-VELOCITY LABOR MARKET, ENDOGENOUS ECONOMIC GROWTH, AND THE LAW OF TRADE SECRETS

By Alan Hyde(1)

The economic success of the computer industry in Santa Clara County ("Silicon Valley"), California may be what the whole world wants to emulate, yet remains imperfectly understood by economists and lawyers. This Article argues that the success of Silicon Valley rests importantly on three factors. First, its high-velocity labor market provides maximum incentives for employee production, and subsequent diffusion, of information. Second, this information is the crucial factor in economic growth, as best illustrated in the economic models of endogenous growth and increasing returns. Third, law had to facilitate this kind of economic growth, by eliminating constraints on employee mobility. Covenants not to compete were already unenforceable in California, and thereafter judges, juries, and lawyers worked to make trade secrets law largely ineffective as an impediment to employee mobility or information diffusion. Within the past two years, however, trial courts in Silicon Valley have increasingly issued injunctions against departing employees, in unreported decisions that this Article will criticize. However, by and large, Silicon Valley employees remain legally and practically free to join or start competing firms. This fact is possibly the most important factor in the Valley's growth.

These three factors: labor turnover, information diffusion, and favorable law-- may not be necessary nor sufficient to replicate Silicon Valley. Other factors have contributed to its growth, such as Californian optimism and Stanford University. And it may be possible to attain similar growth levels through different incentives to create and spread technical information, other than a high-velocity labor market. Nevertheless, this Article suggests strongly that the many cities around the world, that are nicknaming some geographic feature "Silicon", might be better advised to examine, and eliminate, impediments to employee mobility and to the creation and diffusion of technical information.

This conclusion is meant to be a direct challenge to the way American law normally thinks about trade secrets. Consider a problem that, with variations, will occupy much of our time. A scientist or engineer, employed by one company, wishes to work for a rival, or start his or her own firm. Perhaps he or she has been involuntarily terminated, or can see the writing on the wall. Now suppose Employer 1 asserts a property interest in some formula or piece of knowledge that the employee knows and might draw on in the new employment. The old employer's claim may be based on the common law of trade secrets, or perhaps on form contracts that the employee signed on employment or thereafter. May the old employer obtain an injunction against the employee's departure, or against any disclosure of trade secrets?

This is a complex legal question, as we shall see, but I can state simply the gist of my critique of current law. Nearly all American lawyers and courts would define this problem as one of balancing the employer's proprietary interest in its trade secrets, against the employee's interest in freedom of job mobility. The employee will be conceptualized as asserting only an individual interest, more or less compelling, in making a living. The public interest, if identified at all, will be limited either to an abstract interest in economic competition, or perhaps, if the case concerns doctors, dentists, or lawyers, in competition in a named professional job description. The formula or information will be assumed to have value only insofar as it is the property of the old employer, the new employer, or the employee. The law will be blind to the substantial social and private benefits that come when information diffuses rapidly through an economy. These benefits will become apparent when viewed through the lens of the best economic and sociological research.

As that work, however, is not well-known in the legal literature, this Article must take some time to summarize it. Part I reviews the work of AnnaLee Saxenian specifically on Silicon Valley, attributing its decisive triumph, over the Route 128 region of Massachusetts, to the Valley's organization into smaller, easily started firms, with significant interchange of employees and of information among employees. Part II is an introduction to recent work, primarily by Paul W. Romer, on the role of shared ("nonrivalrous" and "nonexcludable") information in economic growth, and the economic analysis of the production and diffusion of such information. Neither of these sections includes original research by me, though I believe I coined the phrase "high-velocity labor market" and know I am the first to bring these two literatures together. Taken together, they present a compelling picture of the sociology and economics of a high-growth region.

Part III examines the chief American legal impediments to the creation of such a high-growth region, specifically, two impediments to employee mobility: covenants not to compete with one's employer, and some interpretations of trade secrets law that prevent employees from taking new jobs or working in the areas they know best. Part IV examines whether aspects of California law might have contributed to Silicon Valley's happy confluence, of employee mobility and information production and diffusion, and concludes that, by effectively eliminating the two impediments to full employee mobility, it has. There are simple and complex elements to this legal story. In many areas of its achievement, California started with natural advantages, but went on to make the most of them. So it is with law. The "natural advantage" that facilitated Silicon Valley was the fact that covenants not to compete have been unenforceable in California since 1872. This helped Silicon Valley engineers skip to rival firms, or to their own start-ups in garages, while their counterparts on Route 128 pursued orderly career paths in the internal labor markets at Digital Equipment or Wang.

However, California's statute on covenants not to compete could not, by itself, have created Silicon Valley. Other states have identical statutes that courts have eviscerated. And other states employ the common or statutory law of trade secrets to permit employers to enjoin their employees from taking new positions or from taking information with them. Even California courts occasionally enjoin employees from disclosing information, or working in specific areas, on their new jobs.

The striking story here, though, is how seldom this occurs. Understanding this phenomenon will require that we examine, not only California's law "on the books", but the practical law "in action" of trade secrets in California. For while California could have tamed its trade secrets law by statute or court decision, it did not. Instead, local legal culture discourages employers from aggressive use of trade secrets laws against departing employees. Firms in Silicon Valley have learned--lawyers will tell them--that it is very difficult to enjoin employee disclosure of information or get damages from the second employer; that judges and juries in Silicon Valley often define trade secrets more narrowly than the relevant statute or appellate case law; that litigating against rivals or start-ups is unpopular in Valley culture; that firms that become plaintiffs suffer in loss of internal morale and in recruiting. The result of this complex story, which combines interviews in Silicon Valley with examination of California law, is that trade secrets law is largely, though not entirely, ineffective to impede employee mobility. The effect, I argue, has been a major contribution to the growth of Silicon Valley.

I have never seen a trade secrets case between an employer and employee that overtly invokes any economic analysis. My contention is that such analysis will often powerfully strengthen the case for the defendant, by emphasizing the economic growth associated with employee mobility and information diffusion. There are, of course, other possible economic analyses, not yet adopted by courts but encountered in the legal literature. Part V considers and criticizes two: the "property rights" approach that seeks to allocate property in information so as to facilitate private contract and lower transaction costs, and an approach that analyzes information as "human capital." Both approaches contain valuable insights, but replicate law's inattention to the social value in shared information, and ignore ways in which the contracting that they contemplate will produce suboptimum amounts of information. Part VI considers implications for future academic research.

I. Silicon Valley: The Wealth of Shared Information

Shared information leads to wealth and growth in an industrial district, in which firms exchange information that enables them to work with each other. Pioneering work was done on this concept by Michael Piore and Charles Sabel.(2) A vivid current example of rapid economic growth linked to rapid diffusion of information, largely through a high velocity labor market, is Silicon Valley, that is, Santa Clara County, California, as discussed by AnnaLee Saxenian, in her book, Regional Advantage: Culture and Competition in Silicon Valley and Route 128.(3) Saxenian identifies Silicon Valley's comparative advantage, over the computer industry around Massachusetts' Route 128, with Silicon Valley's organization into many small firms, as opposed to Route 128's domination by a few firms. Saxenian's book has come up in numerous conversations I have had in Silicon Valley, and her analysis seems to make sense to everyone with whom I spoke.(4) For my purposes it is not important whether or not Saxenian has discovered the key to Silicon Valley, so long as she has discovered an important factor that is not well understood by economists or lawyers, as this Article will show to be the case with modeling property in information.

Saxenian seeks to explain why Silicon Valley decisively passed Route 128 in the 1980's. Each region was a well-known center of the computer industry by the early 1970's. Each had strong ties to local universities, grew strong on defense contracting, but diversified into civilian production. Each employed workforces of roughly equal size in 1975. Yet between 1975 and 1990, California generated three times as many jobs as Massachusetts. In 1990, Silicon Valley-based producers exported electronics products worth more than twice as much as Route 128's. What accounts for the difference? In Saxenian's words:

Silicon Valley has a regional network-based industrial system that promotes collective learning and flexible adjustment among specialist producers of a complex of related technologies. The region's dense social networks and open labor markets encourage experimentation and entrepreneurship. Companies compete intensely while at the same time learning from one another about changing markets and technologies through informal communication and collaborative practices; and loosely linked team structures encourage horizontal communication among firm divisions and with outside suppliers and customers. The functional boundaries within firms are porous in a network system, as are the boundaries between firms themselves and between firms and local institutions such as trade associations and universities.

The Route 128 region, in contrast, is dominated by a small number of relatively integrated corporations. Its industrial system is based on independent firms that internalize a wide range of productive activities. Practices of secrecy and corporate loyalty govern relations between firms and their customers, suppliers, and competitors, reinforcing a regional culture that encourages stability and self-reliance. Corporate hierarchies ensure that authority remains centralized and information tends to flow vertically. The boundaries between and within firms and between firms and local institutions thus remain far more distinct in this independent firm-based system.(5)

Although Silicon Valley has many more firms, and many more small firms, these firms are linked by informal ties. High labor mobility among firms is central to the formation of these ties, dwarfing other sources (such as Stanford University School of Engineering). "During the 1970s, average annual employee turnover exceeded 35 percent in local electronics firms and was as high as 59 percent in small firms. It was rare for a technical professional in Silicon Valley to have a career in a single company. An anthropologist studying the career paths of the region's computer professionals concluded that job tenures in Silicon Valley averaged two years."(6) More recently, observers have commented on a pattern of even higher velocity employment in Silicon Valley, in which professionals never become "employees" at all, but move from job to job as consultants or free-lancers.(7)

Because professionals know and have worked with peers at other firms, there is a great deal of informal know-how sharing across firm boundaries. Economists increasingly recognize that in the real world, firms share significant information and knowledge across firm boundaries, and that willingness to do so can play a significant role in economic growth. The most important study is by Eric von Hippel.(8) Von Hippel studied "informal know-how training", engineers asking their counterparts at rival firms questions, that will be answered unless the answer is both secret and crucial to the firm. It turns out that firms, whose lawyers may posture publicly about the importance of secrecy, are in practice often willing sharers of information with direct rivals. The same phenomenon has been observed in biotechnological research.(9)

Saxenian's book includes a number of specific comparisons in which small, nimble Silicon Valley firms, working with subcontractors and customers, were able to steal a march on the few large firms dominating Route 128. Information turns out to be the key element in Silicon Valley's growth: technical information, how to solve problems; the specifications of new products coming onto the market, so that compatible products can be ready for them; convergence on superior methods or protocols, avoiding wasted effort. Of course, the excellent staff at Route 128 firms produce such information all the time. But only in the high-velocity labor market of Silicon Valley, where mobile employees carry this information among firms through formal and informal contacts, does information diffuse widely enough to realize its potential. [Consider inserting an example from Saxenian, e.g. new chip like 286 or 386 comes onto market and everybody in the Valley can have compatible products and programs reach market the same time].

Saxenian is an urban planner, not a lawyer, and her book is a challenge to lawyers' concept of information. Lawyers, as I shall show in more depth in Part III, might conceptualize much of that information, coursing through the Valley, as "property": the private property of the firm that employs the individuals who generate it. Law has many names for that kind of information that is the property of firms: trade secret, confidential information, proprietary information, an invention that has been assigned under an invention agreement. (Law also has names for the kind of information that everyone can make use of: we say it is the public domain, or general knowledge. However, as we shall see when we turn to law, knowledge about the capacity of another firm's forthcoming chip is not what lawyers usually call "general knowledge", unless the firm has published those very specifications). Saxenian's stories suggest that, as a general matter, if society wants the growth of Silicon Valley, its legal system probably wants to classify more information as in the public domain, and less as the private property of firms. I shall show in Part II that this is the precise conclusion of the economic analyses that try to model the problem. Part IV will show how the law in California has come to approximate this ideal.

Saxenian's book is also an implicit challenge to the dominant trend in legal scholarship in employment law over the last twenty years, namely the construction and defense of legally enforceable contracts for long-term employment. The most sophisticated such work draws on economic theories of implicit contracts for employment.(10) I have always been a little uncomfortable with the claim that such contracts are really so universal or ought to be presumed legally enforceable, and have expressed my skepticism.(11) Speaking personally, reading Saxenian's book in 1994 crystallized for me what was missing from the debate on employment-at-will.(12) I have explored some employment law implications of labor markets like Silicon Valley's in other works.(13) This Article focuses on intellectual property disputes between employers and employees, and does not address issues of employment-at-will, though they lurk in the background. I will say here only that there are many ways to organize a labor market, and that courts should be alert to, and enforce, the actual employment contracts-in-fact that people have. Silicon Valley is not the only way to organize a labor market, and it is not for everybody. There is a definite trend for people to leave (say) computer programming as they age, and enter other labor markets.(14) In the 1950's and 60's, for historic reasons that I have explored elsewhere, firms really did make implied offers in fact of lifetime employment, so there is nothing wrong with courts enforcing such agreements. In Silicon Valley, by contrast, everyone understands that he or she is employed at will, if indeed "employed" at all (as opposed to consulting as a self-employed independent contractor). The economic discussion in Part II suggests that there is a certain type of economic growth that is hard to achieve without the high-velocity labor market of Silicon Valley, although it is possible that similar growth could be achieved through some combination of stable employment relations and other devices (lacking in Route 128) for diffusing information among firms. (This may have worked well in Japan for some decades). This discussion is quite speculative and will therefore be postponed to the Conclusion.

II. Some Economics of Nonrivalrous Information

Stated generally, the question of legal policy raised implicitly by Saxenian's and other work on Silicon Valley is whether, in general or specific, information should be private property or in the general domain. An economic analysis of the problem must be one in which this distinction is made, not assumed away.(15) It is most developed in the recent work on economic growth by Paul M. Romer. While his work has received a great deal of professional and journalistic attention, it has not yet been used by legal scholars. It is perhaps hard luck for Professor Romer that the introduction of his work to legal academia should come in an article dealing mainly with issues of employment and intellectual property law, to which Romer has not applied his own model. So I shall summarize Romer's work with more care than one might have devoted to work that is more familiar to legal readers.

Romer's model of economic growth specifically accounts for knowledge that is shared by all.(16) It addresses the following paradox. The following three premises all seem quite plausible. (1) Technological change lies at the heart of economic growth. (2) Technological change arises (at least in part) in response to market mechanisms. (3) At least some technology can be used over and over without extra costs: know-how, instructions, protocols.

If these three premises are accepted, there can be no equilibrium with price taking. The problem is the same one we have identified; many economists have spotted it too. Why should firms or individuals invest in the production of knowledge if they cannot exclude others from the final product? Given any set of factors of production, it would seem that there must always be one, investment in which would yield a greater return, than would the production of information that will be immediately available to rivals.

Romer reviews earlier attempts to deal with the paradox, which normally involve dropping one of the three premises. I will group these into five basic accounts. First, in some accounts, knowledge that can be used by anyone (which Romer terms "nonrivalrous") is treated as exogenous and publicly provided.(17) This drops premise (2), the role of market mechanisms in creating knowledge. While much knowledge is publicly-provided through schools, much of the information crucial to Silicon Valley is not of that type. As Saxenian shows, information can be created by firms for their own purposes, such as specifications of a forthcoming product, that is crucial for growth, because useful to other firms hoping to market compatible devices or programs.

Second, in other accounts, nonrivalrous, public information is the unintentional side effect of the production of conventional goods.(18) This also doesn't seem to capture the precise kind of information that I am interested in here, information that is not the side-effect, but the objective, of intentional investment in research and development. The "side effect" approach has other technical problems that Romer reviews but I shall not.

A third approach models nonrivalrous information as privately provided and compensated out of quasi-rents.(19) Romer discusses technical difficulties with this solution. For my purposes, I can observe (as Romer also does) that it doesn't address our problem. Rents or quasi rents will accrue to whichever aspects of production that we are treating as fixed, but the question will still remain when or whether those rents will compensate nonrivalrous information, or something else.

Fourth, human capital theorists sometimes assume away the distinction that Romer draws. They treat knowledge as "human capital" that, like other capital, is "rivalrous" (my use precludes your use).(20) This is just not true of the kind of knowledge we are concerned with here (I will return to this point in Part V.B, discussing rival law-and-economics accounts).

Finally, the only way all three premises can be maintained is to assume that the producer of the knowledge is some kind of monopolist.(21) Romer discusses some of the technical economic problems with all these assumptions. I will observe only that none well describes the kind of shared information that others have identified with the wealth of Silicon Valley, where knowledge is privately not publicly created, firms are rarely monopolies in a classic price-setting sense, and, given high job turnover, rivals cannot effectively be excluded from information.

Romer's innovation is to separate the two kinds of knowledge that are often combined as "human capital": "rivalrous" knowledge where the use by one precludes use by another, typically because the knowledge is "embodied" in a particular person who can only be in one place at one time; and "nonrivalrous" knowledge that can be reused infinitely at no additional cost, such as programs, instructions, protocols, designs, or other know-how. In Romer's model, production has four basic inputs: capital, labor, rivalrous human capital, and nonrivalrous knowledge to which everyone has equal access. Capital is assumed to grow by the amount of foregone consumption, and labor and rivalrous human capital are assumed for simplicity to be fixed. (Rivalrous human capital obviously does grow as nonrivalrous knowledge grows, but for Romer the embodied kind of human capital has limits: individuals die and their embodied knowledge dies with them).(22)

Under these assumptions, Romer derives a solution in which nonrivalrous information can grow without bound and becomes the most important factor of production for growth (obviously, as in any economic model, this is more an assumption than a finding of the model; the interest is in the implications of such a model). The limitless growth of nonrivalrous information is crucial and what distinguishes it from other factors of production. Raw materials, water, workers' bodies can be overfished, used up, particularly if held in common without strong property rights. Information is not a common pond that can be overfished. It never wears out and private actors use it over and over again. Firms nevertheless invest in research, even though they will not be able to exclude rivals from the information eventually produced, because they can nonetheless charge more for their products than the costs of production. However, the equilibrium will be socially suboptimum: additional gains may be had if the government directly subsidizes research.(23) A companion paper reviews cross-country data on economic growth and finds that such growth is indeed better correlated with increased investment in human capital devoted to research, than with increased investment in physical capital.(24)

Romer's distinction between rivalrous and nonrivalrous goods has important implications for economic equilibrium and for public policy. Romer fully accepts the neoclassical analysis of rivalrous goods such as "land, fish, a worker's labor effort"--things that can be consumed by one user which then precludes consumption by a rival. He assumes that for such rivalrous goods, strong property rights and freedom of contract are unambiguously efficient. Pareto-optimal solutions require only the definition of strong property rights and freedom of contract (the First Law of Welfare Economics). However, the neoclassical approach does not work for markets in information. Nonrivalrous goods, that is, information: encoded broadcasts, designs, computer codes, basic research-- "make growth possible" but are not governed by an "invisible hand" equilibrium. "In a real economy, an inherent, unavoidable conflict exists between the incentives necessary to encourage the production of these goods and the incentives that lead to the optimal distribution of these goods, both to users and to the developers of other related nonrival goods. This means that private property rights and market exchange are not the perfect institutions for supporting growth. In fact, no simple description of the perfect institutional arrangement can exist. In any particular context, one must explicitly address the trade-offs both between the incentives for discovery and those for diffusion and between the limitations of market mechanisms and those of political mechanisms.(25) "To see why extremely strong property rights might be a problem, imagine that Bell Labs had been given a nonexpiring, ironclad patent on the discovery of the transistor. Or even worse, imagine that such a patent had gone to an organization such as IBM or General Motors. Think of how different the digital electronics and consumer electronics industries would be if every inventor who improved on the design of the transistor and every person who applied the transistor in a new setting had to negotiate with one of these large, bureaucratic organizations for permission to proceed."(26)

In other words, at least sometimes, socially optimal results in markets for information may require, according to Romer, restricted legal definitions of property (such as "trade secret"). I would add, in a conclusion that seems to me to follow necessarily from the last sentence but that Romer has not himself drawn, that this socially optimal result may also require legal restrictions on freedom of contract. It would do no good for law to adopt a narrow definition of patent or trade secret, in order to put more information into the public domain, if employers could, by boilerplate in form employment contracts (or even in negotiated contracts) get employees to agree to keep that very information secret. (Obviously it's a different matter if the legal resolution is that the information in question should not be in the public domain, or that society is indifferent so contract should prevail. All I'm saying is that if society has decided that the information should be in the public domain, law must address restrictions on freedom of contract, as well as legal definitions of patent or trade secret).

If we combine Romer's analyses of nonrivalrous information with Saxenian's and other empirical accounts of the actual process of innovation in Silicon Valley, we get an interesting and suggestive picture of the relationship between nonrivalrous information, labor mobility, and economic growth. (So far as Romer and I know, nobody yet has applied his work to labor markets).(27) "Industry lore from Silicon Valley holds that informal contacts and the significant degree of mobility among engineers and other skilled personnel within the U.S. semiconductor industry ensure that production experience is transferable to other national firms. In Japan, by contrast, the mobility of skilled labor between firms is more limited than in the United States. Yet national spillovers are thought to have arisen from government-sponsored cooperative research and other formal and informal ties between firms."(28) In this picture, US-style labor mobility is the precise mechanism by which specific technical information, produced by firms acting under market incentives, diffuses to contractors and rivals. This helps create new products and competition among similar products, lowering prices.

There are other ways in which letting information diffuse among firms can be socially optimum. The direct contributions to production, modeled by Romer, are surely the most important. There is another kind of growth, however, that comes from firms and consumers converging on particular technologies or protocols, even when the technology involved is not technologically superior, just because certain returns to scale increase when everyone is using similar technology. This phenomenon is discussed in the work on path dependence by Paul David and W. Brian Arthur, to which I can refer only briefly as it has become known in the legal literature.(29) Standard examples are the adoption of VHS videocassette technology or DOS operating systems over arguably technologically superior rivals. However, sometimes whole industries can spring up only when consumers can converge on some uniform technology even if not the best. The resulting development is path dependent. It could not have been predicted by neoclassical economic models and any resulting equilibrium is only probabilistic. Arthur, like Romer, makes the arresting claim that markets for information, unlike markets for goods, do not function like neoclassical markets. However, Arthur and Romer refer to different phenomena: for Romer, the tendency of market actors to produce research only in socially suboptimum quantity; for Arthur, the existence of multiple equilibria because of historically-contingent starting points. It is not necessary to choose between these accounts for present purposes. Both show legal and other decision-makers some potential benefits of weak property rights in information, leading to information diffusion, though as I say Romer's strikes me as the more important in understanding Silicon Valley.

We now have two potential economic answers to the legal question of why it may be socially beneficial to let employees use, on their current job, information learned on a previous job. Let me present that question in a sharper form and apply the "endogenous growth" and "path dependence" literatures to it; I will also suggest some additional answers of my own, making no attempt to present these as formal economic models. The sharper form of the question comes from a conversation with Evan Chesler, Esq., a New York attorney who has represented employers in some important trade secret litigation, and I use it here with attribution to him. Suppose that Engineer was formerly employed by Employer 1 and now works at Employer 2. Employer 2 is experiencing a high failure rate in the manufacturing of widgets. A team of consulting engineers has identified five possible sources for the failure, and recommended the investment of $40 million to modify Production Process 1, a possible source of the problem. Suppose that Engineer, listening to this presentation, knows that Employer 1, during the time he worked there, invested $50 million in substantially the same production modifications now being proposed, only to learn that it was not possible to improve production quality in that way.

What is Engineer to do? If he says nothing, he violates his duties as an employee of Employer 2, since it is likely that they are about to pour $40 million down a rat hole. If he tells what he knows, he may violate his duty not to disclose the trade secrets and proprietary information of Employer 1. Employer 1 paid for, and considers that it owns, the "negative information" that Production Process 1 cannot be improved in a way that achieves better quality. Chesler argues that Employer 1 paid for that negative information. What possible social benefit is there, he asked me, in disclosing it?

The economics discussed so far provide two general answers. First (Romer), just this kind of information is crucial to high growth economies. Letting Engineer disclose it saves Employer 2 the waste and duplication of effort that strong trade secrets rights might entail. What Romer has written of free trade is equally true of free labor markets: "With free trade it is possible to put to better use human capital that might otherwise be used to redesign the wheel."(30) There is no social benefit in requiring multiple expenditure of resources to produce the same information (so long as the information will be produced anyway).(31) Second (Arthur), only convergence by manufacturers on the correct diagnosis of the problem will permit new firms to arise to solve it. There are social and private benefits to be gained from encouraging "informal know-how sharing" among firms competing in the market but producing similar products.(32) For example, suppose that the culprit in Chesler's hypothetical turns out to be, not the beguiling Production Process 1, but rather the unglamorous Production Process 4. The sooner that manufacturers achieve consensus that there must be changes in Process 4, is the sooner that consulting firms, engineers, industrial equipment manufacturers, and others will arise to meet the new market in Process 4 Improvement--the process which, by hypothesis, will actually yield productivity gains. In some cases, this efflorescence of problem solvers may require a genuine market for Process 4 improvement, not just one firm's knowledge that Process 4 is a problem, perhaps a problem without any solution apparent to it.

Let me suggest two additional reasons, that may or may not operate in any particular case, why there might be social advantage in letting Engineer tell Employer 2 about Employer 1's experience: reinforce the high-velocity labor market, and incentives to employees.

If Saxenian is correct, there are private and social gains to be had when employees can be induced to accept employment with very short tenure, knowing they will be on the job market again soon. So far, I have presented no reason to interfere with freedom of contract here, as an economist would see it.(33) The firm that wants to offer 1950s-style implicit contracts for lifetime employment, with promotions from an internal labor market and employees making orderly progression up defined promotion ladders, is free to offer that, although increasingly the courts will hold the firm legally to its promises. (In fact, some high technology firms have just such internal labor markets: firms around Boston, according to Saxenian, but, even in Silicon Valley, IBM and Hewlett-Packard are said to be firms that enable employees to make careers there and do less hiring from outside than do their competitors). The firm that wants to offer employees only a consultantship on a particular job is free to offer just that.

Some employees, without desirable skills, may have no choice but to accept short-term, unstable work relations. Talented engineers and professionals, however, have other options. Why do they ever agree to such short jobs? I don't know of systematic data, but when I interview people who work under these arrangements, they are confident they will find something else, they enjoy the excitement of Silicon Valley, and, like Saxenian, associate the growth of the Valley with these kind of employment relations. They expect, however, to be able to get another job on the basis of the experience they gained on the last one. I do not mean that they expect to be hired to obtain a known formula or algorithm or other secret of their former employer's, for no one I've talked to admits to being hired for that purpose. Rather, they expect they will be more valuable to the next employer, than they are to the current employer, because of the range of problems they have faced and seen solved.

Now while it might be in the interest of many or most Silicon Valley firms to get employees to work under these arrangements, a rogue firm might be tempted to free ride by offering only short-term work, yet still insist on its right to impede employee mobility to the next firm. If such restrictions are legally enforceable, managerial and professional employees, with the option to work in more stable arrangements will not accept high-velocity work--by hypothesis, a worse solution for Silicon Valley employers and for society. If, as I will show in Part IV, California legal culture does not let employers restrict employee mobility, this may reflect a tacit, and surely unenforceable, agreement among employers not to permit others to destroy the willingness of employees to accept short-term work.(34)

A final social reason to prohibit agreements that impede employee use of information learned on the last job might involve optimum incentives for the production of useful information. Again, this is not an all-or-nothing proposition. Many firms employ personnel under long-term contracts and engender astonishing amounts of useful information. However, in Silicon Valley, mobility, particularly mobility to start-up one's own company, is part of the motivational structure of employees. Employees dream of having the great idea (or associating with someone else who did) that venture capitalists will fund so that they can start their own firm. Perhaps a legal regime of employee mobility will produce more creative, engaged effort by employees than more traditional career reward structures. I know of nothing like data on this question. Universities appear to vary in their willingness to permit employees to hold intellectual property in their own names, yet I know of no systematic attempts to compare employee output at different universities along this dimension.(35)

By raising the question of incentives, however, haven't I undermined the entire theory? Suppose Silicon Valley's high-velocity labor market does give employees great incentives to create, secure that they can trade on these ideas at another employer, or with a venture capitalist funding their own start-up. Doesn't it deny those incentives to their employer? Don't employers have diminished incentives to invest in the production of information, to the extent that the information will be nonrivalrous? Neither Romer nor Saxenian tells us to be unconcerned with this question. Romer shows that firms will produce information even if they can't exclude others from its use, but he can't tell us for any particular firm or industry what mix of excludable information and nonexcludable information is optimum. Saxenian does suggest that Silicon Valley's impressive growth is linked to the proportion of actually-shared-information. However, it might turn out that entirely different factors explain the success of Silicon Valley. Perhaps Silicon Valley firms do underinvest in certain kinds of research, and will pay the price for that underinvestment at some future time.(36)

What lawyers, judges deciding trade secrets cases, and other policy-makers would like from economics is some kind of model that trades off all the factors in this section. The model would measure the extent to which, for particular industries or kinds of information, making information publicly shared would contribute to Romeresque growth, convergence on helpful protocols, and incentives to employees, and weigh these factors against the possibility that making such information public would diminish firms' incentives to invest in its production. It is obvious that no such model exists. As Kenneth Arrow said to me in conversation, economics is nowhere near "even a sort-of-good model." This is no reason for lawyers to despair; it suggests the very case-by-case analysis. Lawyers for plaintiffs and defendants could incorporate the best economic analyses by addressing, in individual trade secrets cases, these crucial issues of public benefit in diffusion of information and incentives to employees, and weighing them against the question of incentives to the firm to produce information.

I will return to and refine this legal test, after surveying actual American law governing the allocation of intellectual property between employers and employees. It never (to date) incorporates economic analysis, and is very different from the analysis suggested so far.

1. *Professor of Law and Sidney Reitman Scholar, Rutgers. The State University of New Jersey School of Law, Newark, NJ 07102-3192, USA. Telephone: (973) 353-5463. Fax: (973) 353-1445. E-mail: hyde@andromeda.rutgers.edu.

This essay was preceded by years of intense schmoozing on the economics of human capital and employee information. Thanks, more than they can imagine, to Kenneth Arrrow, Ian Ayres, David Charny, Ron Gilson, Henry Hansmann, Al Klevorick, Mark Roe, Roberta Romano, Paul Romer, Chuck Sabel, Alan Schwartz. Comments on earlier drafts were made by Joseph Bankman, Robert Bone, Ronald Gilson, and Wendy Gordon. Some of my Silicon Valley interviewees must remain anonymous, but I must thank Dave Anderson, James Baer, David Berger, Steve Bochman, Cynthia Cannady, Jeff Chambers, Evan Chesler, Amy Christin, Mia Clark, Tim Clark, Peter Detkin, Paul De Stefano, Ian Feinberg, Boris Feldman, Paul Hall, Rob Kaplan, Mike Ladera, Mike Leventhal, Deborah Levinger, Janet Levinger, Art Lim, Thomas Nolan, Will Poole, Niels Reimers, Bertram Rowland, Deborah Satten, John Schoch, Gary Weiss. A Rutgers Faculty Colloquium, October 17, 1996, saw an earlier draft; thanks particularly to Bill Bratton, Sherry Colb, Jon Hyman, Howard Latin, John Leubsdorf, Jim Pope, Phil Shuchman, and David Troutt. A version was presented to the preconference Make or Buy: The Boundaries of the Firm, Columbia University, February 7, 1997 and then to the conference Corporate Governance Today, Columbia University, May 21-22, 1998; thanks to Mark Roe, David Charny, John Coffee, Ron Gilson, and Saul Levmore. Financial support was provided by the Sloan Foundation Program on Corporate Governance, Columbia School of Law. Above all, thanks to Paul Schachter, who brought the Abbruscato case into my employment law class and got me thinking about this problem.

2. 1 Michael J. Piore and Charles F. Sabel, The Second Industrial Divide: Possibilities for Prosperity (1984). See also Robert C. Allen, Collective Invention, 4 J.Econ.Behav.& Org. 1 (1983)(19th Century British iron and Steel).

3. 2 AnnaLee Saxenian, Regional Advantage: Culture and Competition in Silicon Valley and Route 128 (1994).

4. 3 See also Paul Milgrom & John Roberts, The Economics of Modern Manufacturing: Technology, Strategy, and Organization, 80 Am.Econ.Rev. 511 (1990).

5. 4 AnnaLee Saxenian, supra n.2, at 3-4.

6. Saxenian at 34-35, quoting American Electronics Association, Technical Employment Projections (1981)(statement of Pat Hill Hubbard). "[I]nter-firm worker mobility is a permanent feature of the labor market for semiconductor engineers, rather than an occasional symptom of job shopping during periods of rapid employment growth....Job changing is likely to be especially frequent in Silicon Valley. Within this industrial complex, small specialized semiconductor producers typically depend upon the external labor market (as opposed to the internal labor market of the firm) to meet their demands for skilled and experienced workers...." David P. Angel, The Labor Market for Engineers in the U.S. Semiconductor Industry, 65 Econ. Geog. 99, 103 (1989). In this nationwide survey of semiconductor engineers, half the job changes reported were moves in which both the old and new employers were located within Silicon Valley. 65 Econ. Geog. at 106.

7. There has been some journalistic notice of these labor markets, but little academic analysis yet. See, e.g., Rosabeth Moss Kanter, Nice Work if You Can Get It: The Software Industry as a Model for Tomorrow's Jobs, 23 Am.Prospect 52 (1995); Bernard Wysocki Jr., Flying Solo: High-Tech Nomads Write New Program For Future of Work, Wall Street Journal, August 19, 1996, p. A1, col. 6.

8. 7 The Sources of Innovation 76-92 (1988). The rapid growth of British iron and steel in the nineteenth century, discussed in Robert C. Allen, supra n.1, may, as von Hippel suggests (at 84), more accurately depict "informal know-how trading" than genuine "collective invention." See also Charles F. Sabel, Learning by Monitoring: The Institutions of Economic Development, in Handbook of Economic Sociology (Neil Smelser & Richard Swedberg eds. 1994); John Jewkes, David Sawers, & Richard Stillerman, The Sources of Invention (2d Ed 1969)(importance of mixing individual with corporate research). See also the discussion of informal know-how trading in winemaking in California, infra n.64.

9. 8 Walter W. Powell, Inter-Organizational Collaboration in the Biotechnology Industry, 152 J.Inst.& Theoret. Econ. (1996), and the interesting comment by Henry Hansmann, What Determines Firm Boundaries in Biotech?, 152 J.Inst.& Theoret.Econ. (1996).

10. 9 Paul C. Weiler, Governing the Workplace (1990); Stewart J. Schwab, Life-Cycle Justice: Accommodating Just Cause and Employment at Will, 92 Mich.L.Rev. 8 (1993); Katherine Van Wezel Stone,. The casebook, Steven L.Willborn, Stewart Schwab, & John F. Burton, Jr., Employment Law: Cases and Materials (2d ed.1998) is largely built around the assumption of such implicit contracts and their implications in various unexpected corners of employment law.

11. Alan Hyde, Book Review (of Paul C. Weiler, Governing the Workplace), Columb.L.Rev. (proposals for universal unjust discharge protection are very expensive and would only result if society were more outraged at unfair discharge than it appears to be); Alan Hyde, In Defense of Employee Ownership, 67 Chi.-Kent L.Rev. 159, 183-84 (1991)(longterm employment contracts are associated with high monitoring costs, particularly where trust has evaporated).

12. 11 Thinking about high technology has helped others see virtues in employment-at-will. Samuel Issacharoff, Contracting for Employment: The Limited Return of the Common Law, 74 Tex.L.Rev. 1783, 1792 (1996).

13. Alan Hyde, Employment Law After the Death of Employment, 1 U.Pa.J.Emp.&Lab.L. 99 (1998); Alan Hyde, Employee Organization in High Velocity Labor Markets, 50 Proc.NYU Conf. on Labor (forthcoming, 1997).

14. 13 According to a survey conducted by the National Science Foundation and the Census Bureau, six years after finishing college, 57 percent of computer science graduates are working as programmers; at 15 years the figure drops to 34 percent, and at 20 years--when most are still only in their early 40's--it is down to 19 percent. In contrast, the figures for civil engineering are 61 percent, 52 percent and 52 percent. ... A programmer who becomes, say, an insurance agent after failing to find programming work counts in the statistics as an employed insurance seller, not an unemployed software worker." Norman Matloff, Now Hiring! If You're Young, New York Times, January 26, 1998, at A19, col. 1.

15. James Boyle, with characteristic verve, has pointed out how economists often simplify matters by treating information simultaneously as "perfect": "free, complete, instantaneous, and universally available" and as a commodity ("costly, partial, and deliberately restricted in its availability"). He modestly names this "Boyle's Paradox." Shamans, Software, and Spleens: Law and the Construction of the Information Society 35 (1996). While it is true that some economic models do make the simplifying assumption that information is either property or cost-free, I don't think that the entire economics profession has been as heedless of the issue as one might gather from Boyle's nevertheless valuable critique.

16. 15 Paul M. Romer, The Origins of Endogenous Growth, 8 J.Econ.Persp. 3 (1994); Paul M. Romer, Endogenous Technological Change, 98 J.Polit.Econ. S71 (1990). "Shared" information is my term to refer to the class of information that Romer calls, with more precision, "nonrivalrous" and "nonexcludable." "Rivalry is a purely technological attribute. A purely rival good has the property that its use by one firm or person precludes its use by another; a purely nonrival good has the property that its use by one firm or person in no way limits its use by another. Excludability is a function of both the technology and the legal system. A good is excludable if the owner can prevent others from using it. A good such as the code for a computer program can be made excludable by means of a legal system that prohibits copying or by means of encryption and copy protection schemes." Id. S73-S74. Most information is "nonrivalrous" in Romer's terminology. The exception is information that is lodged in a human body. "Romer's ability to solve quadratic equations" or "Hyde's knowledge of employment law" are rivalrous goods since each of us can only be in one place at one time. "Knowing how to solve quadratic equations" or "knowledge of employment law" are nonrivalrous goods (also nonexcludable).

In this Article, I do not want to lean too heavily on Romer's distinction between the dimensions of rivalry and excludability, since my precise inquiry (unlike Romer's) is to the legal system. An employee's knowledge of his former employer's operations is sure to be "partially excludable"; our precise inquiry is to what that means, and that will involve examination, not just of formal law on the books, but practical reality. I will use the phrase "shared information" to mean information that, as a practical matter, is available to other firms in an industry. As we shall see, information may be "shared" in my sense despite contrary legal restrictions "on the books" that might make it "excludable" in Romer's terminology. I realize that in Romer's terminology, "shared information" is a pleonasm since all information, other than that strictly embodied in an individual, is "nonrivalrous": if it can be converted to a bit stream, its use by one does not preclude use by another.

17. Robert M. Solow, A Contribution to the Theory of Economic Growth, 70 Q.J.Econ. 65 (1956); Karl Shell, Toward a Theory of Inventive Activity and Capital Accumulation, 56 Am.Econ.Rev.Papers & Proc. 62 (1966); Karl Shell, A Model of Inventive Activity and Capital Accumulation, in Essays on the Theory of Optimal Economic Growth (Karl Shell ed. 1967).

18. 17 Kenneth J. Arrow, The Economic Implications of Learning by Doing, 29 Rev.Econ.Stud. 155 (1962); Robert E. Lucas, Jr., On the Mechanics of Economic Development, 22 J. Monet. Econ. 3 (1988).

19. 18 Zvi Griliches, Issues in Assessing the Contribution of Research and Development to Productivity Growth, 10 Bell J.Econ. 92 (1979); Karl Shell, Inventive Activity, Industrial Organization and Economic Growth, in Models of Economic Growth (James A. Mirrlees & Nicholas H. Stern eds. 1973); Paul M. Romer, Increasing Returns and Long-Run Growth, 94 J.Polit.Econ. 1002 (1986).

20. Gary S. Becker, Kevin M. Murphy, & Robert Tamura, Human Capital, Fertility, and Economic Growth, 98 J.Polit.Econ. S (1990).

21. 20 Romer cites Joseph A. Schumpeter, Capitalism, Socialism, and Democracy (1942). Later in this Article I will discuss work of Fritz Machlup in this tradition, infra n.139.

22. 21 What the published article calls "rivalrous" and "nonrivalrous" knowledge had earlier been termed by Romer "embodied" and "disembodied" knowledge. I have argued that a robust line between the embodied and disembodied is unlikely to yield an equilibrium solution. Alan Hyde, Bodies of Law (1997)(inconsistent social constructions of the body more important than biology in social thought).

23. 22 Romer, supra n.15, S93-98; Paul M. Romer, Implementing a National Technology Strategy with Self-Organizing Industry Investment Boards, 1993 Brookings Papers on Economic Activity: Microeconomics 345, 353-362.

24. Paul M. Romer, Capital, Labor, and Productivity, 1990 Brookings Papers on Economic Acitivity: Microeconomics 337, 354-66; Romer, supra n.15 [Origins].

25. 24 Romer, supra n.22 [1993 Brookings], 354-55.

26. Romer, supra n.22, at 358.

27. 26 Personal conversation, July 9, 1998.

28. 27 Douglas A. Irwin & Peter J. Klenow, Learning-by-Doing Spillovers in the Semiconductor Industry, 102 J.Polit.Econ 1200, 1205 (1994). Irwin and Klenow's data may debunk this somewhat, purporting to show that learning spills over just as much between firms in different countries as between firms within a given country. However, as they note, their results "are not informative about the transmission mechanisms of spillovers," id. 1218. It may well be that in the United States generally and California specifically, given the structure of employment and trade secrets laws, the labor market is a feasible, indeed excellent, mechanism of information spillover. Other countries, with different laws and traditions of labor mobility or entrepreneurial culture, might have to rely on different institutions of information diffusion, which might be more or less effective, generally or as to specific types of information.

29. 28 W. Brian Arthur, Increasing Returns and Path Dependence in the Economy (1994); Paul David, Clio and the Economics of QWERTY, 75 Am.Econ.Rev.Papers & Proc. 332 (1985); Mark J. Roe, Chaos and Evolution in Law and Economics, 109 Harv.L.Rev. 641 (1996).

30. 29 Romer, supra n.23[Brookings 1990), at 351.

31. 30 Steven N.S. Cheung, Property Rights in Trade Secrets, 20 Econ.Inq. 40, 47 (1982), discusses the following "dissipation of rents" with trade secrets: industrial espionage, imitation, potential litigation, unnecessarily delayed research.

32. 31 Eric von Hippel, supra n.7.

33. 32 I am well aware of the extensive repertoire of moral and legal arguments that would deny the term "freedom of contract" to relations between employers and employees. This entire Article brackets these out in order to ask the strictly economic question of whether, in general, rights to information should rest in employers, or employees (and through them, the public). I will similarly not invoke any literature arguing for a moral right in employees to own their thoughts or creations. See, e.g., Steven Cherensky, A Penny for Their Thoughts: Employee-Inventors, Preinvention Assignment Agreements, Property, and Personhood, 81 Calif.L.Rev. 597 (1993). For a critique of the metaphor of ownership of one's own body, see Hyde, supra n.21, at .

34. 33 "For [managerial] employees, their knowledge of the industry, indeed their detailed knowledge of their ex-employer, may be the only marketable attribute they have. Stripped of the ability to use that knowledge, these employees may have little to offer beyond the knowledge of a freshly-minted (and far cheaper) M.B.A. graduate." Edmund W. Kitch, The Expansion of Trade Secrecy Protection and the Mobility of Management Employees: A New Problem for the Law, 47 S.Car.L.Rev. 659 (1996).

35. 34 Pat K. Chew, Faculty-Generated Inventions: Who Owns the Golden Egg?, 1992 Wis.L.Rev. 259, 282-85; cf. Rochelle Cooper Dreyfuss, The Creative Employee and the Copyright Act of 1976, 54 U.Chi.L.Rev. 590 (1987)(arguing that professors, not universities, should own the rights in their work; no economic analysis). Until recently, American universities could have provided another model, in addition to Silicon Valley, of a market in which intellectual property effectively vested in the public, via employees, or in the employees themselves, but not in employers. Some versions of this policy resulted in enormous benefit for both the university and the employees or students, for example, when the university was Stanford and the individuals were Bill Hewlett and David Packard. Stanford engineering Professor (later Dean) Frederick Terman "wooed them back [from graduate work at MIT and management training at General Electric, respectively] with Stanford fellowships and part-time jobs." The understanding from the beginning was that they would form a business in a garage, producing products to which Stanford would not assert any property interest, and this proved a remunerative decision not only for them but for Stanford. David Jacobson, Founding Fathers, 26 Stanford No. 4, July/August 1998, at 59, 60.

However, recently universities have become more assertive of their property rights as employers, as Chew shows. See also the much-publicized tale of Petr Taborsky, the graduate student who spent a year in prison, including eight weeks on a chain gang, for "stealing" his own discovery from a university laboratory after the university, and the corporate sponsor of the research, claimed the patent. Man Imprisoned in Dispute Over Invention Completes Sentence, New York Times, April 3, 1997, at A22, col. 2; Mireya Navarro, Dispute Turns a Researcher Into an Inmate, New York Times, June 9, 1996, at A22, col. 1; William Booth, From University Lab to the Chain Gang: Student "Inventor" Is Imprisoned in Intellectual Property Dispute, Washington Post, June 7, 1996, at A1.

36. 35 Employers in high-velocity labor markets invest less in training than do similar employers that implicitly offer long-term employment in internal labor markets. Peter Cappelli, Rethinking Employment, 33 Brit.J.Indus.Rel. 563, 576 (1995)(reviewing literature); Paul Osterman, Skill, Training, and Work Organization in American Establishments, 32 Indus.Rel. 125 (1995).