V. Three Alternative Economic Analyses
As mentioned, I have never seen a trade secrets case, or any intellectual property dispute between an employer and an employee, that turns on economic analysis. Almost any economic analysis would be an improvement. For reasons I've suggested, Paul Romer's model, which carefully distinguishes between rivalrous and nonrivalrous information, is the best starting point. However, there are things to be learned from at least three alternative economic analyses, two of which have been applied to related legal problems--not trade secrets--in the academic literature. They are: the "property rights" school, applied to invention assignment agreements by Robert Merges; the "human capital" school, applied to covenants not to compete by Rubin and Shedd; and the somewhat psychological economic theories of monopoly in intellectual property, put forward by Fritz Machlup nearly half a century ago.
A. Property Rights and Contracting
Legal audiences have become familiar with the branch of economics that models the likelihood of contract under various assumptions concerning information and other transaction costs.(1) This is an important change from the law and economics approaches of a few years ago, that often assumed low transaction costs, and thus markets that would reach all Pareto-optimal trades until equilibrium was achieved. A particular kind of transaction cost that has received searching analysis is the allocation of property rights, associated with the work of Oliver Hart, mostly applied to problems of industrial organization and the nature and boundaries of firms.(2) This work is a much more elaborate and precise model of Ronald Coase's basic insight that firms (hierarchies) arise when individual contracts are too costly.(3) The Grossman-Hart-Moore approach might be described as a considerably more elaborate version of the difficulties of contracting. A particular insight of this school is that all contracts are incomplete, so the background law of control over assets, that is, property rights, will continue to be salient even when actual contracts are negotiated.
Several recent articles apply this approach to contracts between inventive employees and employing firms, asking what kinds of bargaining equilibria could result from various allocations of intellectual or other property. This work is a useful supplement to this Article, and I will review some of the competing models in this section.
The property rights approach is at best a supplement to this Article, because it begs the most important question that this Article has analyzed: whether there should be any property at all. Analyses in the Grossman-Hart-Moore approach start by assuming some kind of property right, whether tangible or intangible--an asset, a patent, the right to terminate a contract--and ask what kind of bargaining equilibrium would result from various allocations of that property right. This approach obviously translates best into legal applications for a class of problems where law, too, unambiguously creates that property right--a patent, for example. In such cases, it surely makes sense to ask what kind of bargains will result depending on whether patents are held by employees or employers. However, it is simply not possible within the property rights approach to ask the question of whether there ought to be property at all. Yet that is the precise legal question posed by cases that ask whether a particular chip specification or winemaking process is a "trade secret" or "general knowledge" in the "public domain."(4)
These bargaining models may also fail to address two other important aspects of the analysis of this Article: incentives; and welfare. As we shall see, they may assume away the difficult issue of the tradeoff between the comparative incentive effect of an intellectual property right in the hands of an employee or employer. They also assume the First Theorem of Welfare Economics, under which strong property rights and free contract will necessarily maximize public welfare. As we saw, recent models of endogenous growth challenge this assumption.(5)
Still, one welcomes any sort of economic analysis of these problems, and no doubt a convincing metatheory will some day combine an analysis of the question Property vs. Nonrivalrous Information, the question addressed in this Article, with a property rights analysis of the different question Whose Property?
The published bargaining models so far turn out to be, not surprisingly, highly
responsive to their underlying assumptions. One may choose between bargaining models
that assume efficient Pareto-optimal contracting between employers and employees;
bargains that assume imbalances in bargaining power that result in suboptimum
investment in production of ideas; or a model that assumes such high transaction costs
that efficiency demands employer ownership of all relevant intellectual property. These
different results stem largely from the simplifying assumptions in each model. I shall
summarize them in chronological order of their publication.
1. Pakes and Nitzan: Pareto-optimal negotiations at employee departure
Pakes and Nitzan attempt to model an optimum labor policy for hiring scientists for a project "when one takes explicit account of the fact that the scientist may be able to use the information acquired during the project in a rival enterprise."(6) They focus on negotiations as the employee departs, which they assume will reach a Pareto-optimal outcome, in which the employer will or will not make an offer to the employee that will induce him or her to stay. Under this assumption, the optimal hiring policy is thus one that permits the scientist to depart at any time, free to trade on information acquired on the last job.
This is the oldest article I have been able to find addressing the question of employer-employee disputes over the ownership of ideas. It long predates the Grossman-Hart-Moore approach and takes a cheerful approach to negotiations between the employer and the departing scientist. Today's models, as we shall see, are much likelier to hypothesize numerous inefficiencies in the contracting process. I think, however, that there is much realism in the Pakes-Nitzan analysis. A central assumption of their model is that Employer 1 should always be able to outbid either a second employer or a new start-up. Whatever information the scientist has, Employer 1 should be able to exploit more efficiently than a new employer or new enterprise: Employer 1 knows more about the information; the rival has start-up costs.(7) If, prior to the scientist's departure, Employer 1 is a monopolist over the information, its position is even stronger: competition will divert to consumers some of the gains that would otherwise accrue to Employer 1, the monopolist. Thus, to any employee threatening to defect, Employer 1 should be able to offer a bonus that would offset the amounts to made by moving to a rival or start-up. One form of such a bonus discussed by Pakes and Nitzan--frequently encountered in Silicon Valley, as they note--is to offer the favored employee stock options. Pakes and Nitzan conclude that there "will always exist a particularly simple labor contract consisting of a flat rate and a profit-sharing scheme which satisfies" the following conditions: maximize expected profits; ensure no rivals appear; and compensate the employee at no less than his alternative wage. It might therefore seem to follow that "provided all agents act optimally, we should not observe a scientist breaking away from an established firm to join or set up a rival."(8)
Why then do employees ever leave firms to go to rivals or start their own? Pakes and Nitzan are well aware that this occurs and, as this Article does, they focus on the semiconductor industry as one of rapid employee turnover. When employees leave, it must be because their employers didn't bid to keep them. This could be because it is easy for new firms to enter the market (low start-up costs, venture capital), so Employer 1's monopoly was ending anyway, or because the information is not that secret anyway (e.g. reverse engineering is feasible). Under these assumptions, Employer 1 will not offer the bonus and the scientist will depart. The optimum labor contract is thus one that permits such departures. "Provided the firm is free to choose among alternative labor contracts, it can provide an incentive structure which controls the mobility of the scientist--only inducing him to leave and set up a rival when it is in the firm's interest for him to do so."(9) This can be true simply because it was not worth it to Employer 1 to pay the bonus that would have kept the scientist. But it can also be true, as Pakes and Nitzan note, because "the establishment of a rival may increase the sum of the returns accruing to the two agents involved."(10) In Romer's and other models of increasing returns to endogenous technological change, precisely this characterizes Silicon Valley.
We will turn momentarily to more recent analyses with gloomier assumptions about the efficiency of negotiations between employers and departing employees. I must say that my sense, based on my interviews, is that Pakes' and Nitzan's picture is quite realistic. It is not a big deal to decide whether or not to hold an employee. Firms have a great deal of discretion in offering employment contracts; they are not simple price takers.(11) As shown in Part I, firms can decide at the internal hiring stage to structure long-term careers for employees (say, on Route 128), through internal labor markets, implicit contracts for lifetime employment, backloaded benefits. They can also offer such a contract at any later point in the employment relationship, for example to retain an employee who threatens to leave.
What would cause the contracting process, at the time of employee departure, to fail? The popularity in law schools of Oliver Williamson's work may have oversensitized us to problems of firm-specific investment, hold-ups, or end-game bargaining. In Silicon Valley's high-velocity labor market, rival firms and venture capitalists are available to provide accurate information about any employee's alternative earnings, as set by the market. If Employer 1 chooses not to outbid this value, it has probably concluded that what this employee knows is easily replaceable, or that the firm would itself benefit from increasing returns: for example, if its former employees move to other firms in the Valley and create increasing demand for Employer 1's products. In short, courts in trade secret cases should be alert for firms attempting to use the courts to impede employee mobility that the firm itself could easily have purchased from the employee, but chose not to.
I must point out three possible critiques of Pakes and Nitzan. First, they expressly do not model issues of incentives. By contrast, they assume "that the effort the scientist devotes to the project is independent of the terms of the contract."(12) We will have to look elsewhere for our really difficult trade-off: whether the gains from having employees able to market all they know outweigh potential losses if firms underinvest in production of information. Moreover, it seems likely to me that the incentives issue is related to the issue of whether Employer 1 can really always outbid Employer 2. In particular, I should think that one reason employees dream of their own firm is that they anticipate they would work harder for themselves than for another.(13) Second, Pakes and Nitzan seem to me to be guilty of Boyle's Paradox: information for them is simultaneously a commodity that Employer 1 monopolizes, and something the market evaluates efficiently (because it is known to all).(14) Third, they assume property rights in information that are weak enough so that it is feasible for an employee to depart a firm without disclosing the ideas in his head.(15) This may beg the question; it may also be an inaccurate statement of the law.(16)
1. Aghion and Tirole: incentives to create value; bargaining power
Aghion and Tirole model ownership of intellectual property as a question of an initial contract assigning property rights.(17) Unlike Pakes and Nitzan, and Anton and Yao, they do not focus on negotiations at the state at which the employee departs. Their model is precisely one of allocation of property rights as incentives to produce value.
They assume a research unit RU (who may be an employee) performing research for a customer C (who may be an employer). They also assume that the probability of discovery depends on the effort cost by the research unit, which cannot be specified ("noncontractible"), and the investment by the customer. Crucially, the "research unit" (employee) has no initial cash endowment and therefore cannot pay the customer anything; also, its income cannot be negative. These last two assumptions suggest to me that the "research unit" has been defined to be an employee.
Aghion and Tirole show that, under these assumptions, "whether C or RU should own the innovation hinges on two basic considerations: (a) the marginal efficiency of RU's effort compared with the marginal efficiency of C's investment; (b) the ex ante bargaining power of the two parties "(18) That is, if bargaining power were equal, the parties would efficiently contract to place the property right in order to provide incentive for the more marginally productive extra investment, be it RU's effort, or C's capital. (They analyze a number of common negotiated or implied contracts as split incentives for further effort and investment, such as trailer clauses and shop rights.) However, observable contracts may not enact this efficient result, for two reasons. First, since RU cannot compensate C for extra effort, C's will hold some property in ideas even where ownership by RU's would have resulted in higher total surplus. Second, unequal bargaining power may also result in suboptimum contracts for property rights. A recent empirical study of Aghion's and Tirole's thesis suggests that this last observation has serious observable consequences. Lerner and Merges studied the allocation of intellectual property rights in technology alliances between biotechnology firms, and found that bargaining power seemed to explain these allocations better than the incentive or wealth-creating effects stressed by Aghion and Tirole.(19)
I argued in Part IV that courts in trade secrets cases would do well to ask precisely the question that we can now call Aghion and Tirole's, namely, which allocation of property rights would provide the best incentives for further productive investments of effort and capital. This is of course an important question. It is not conceivable that parties would want to allocate property rights to diminish the surplus they will divide.
While Aghion and Tirole's analysis is most valuable, it is incomplete for at least four reasons that have already been mentioned. First, and most important, they assume the point that is at issue in this Article, namely, whether information is to be property at all, or rather nonrivalrous information diffusing or spilling over among firms. Theirs is a fair enough assumption in a typical patent case, where the issue is much likelier to be whether the employer or employee owns the patent. The question of this Article is, however, the unavoidable issue when the legal question is whether a given employee's knowledge is a "trade secret" (employer property) or "general knowledge" (nobody's property).
Second, the wealth of shared information informs even their set of problems, yet goes without express recognition by them. For, even in the class of genuine choices between employer property and employee property, we have seen that employee property is much less likely to remain secret long; it will move among firms with the employee and contribute to the kind of endogenous growth that we have discussed. This fact is not reflected in their model; I have been arguing throughout that it is the crucial fact distinguishing Silicon Valley from Route 128. Presumably their model can accommodate this by making some assumptions about the comparative advantage that RU has in creating value, given the tendency of information in RU's hands to spill over. However, we do not really have good estimates of this comparative advantage, so incorporating one in into the Aghion-Tirole model may just destroy its utility.
Third, they do not expressly analyze welfare effects of their model, perhaps because they assume, conventionally, that strong property rights and free contract will result in Pareto-optimality. Their own model, however, notes (but does not much develop) the role of unequal bargaining power in impeding efficiency. As important, I have argued throughout, is the tendency toward suboptimum levels of investment in research since firms cannot ever really monopolize the results, irrespective of their legal rights. In Romer's models, strong property rights and Pareto-optimal contracts between rightsholders will not yield socially optimum levels of investment in information.
Finally, if applied to legal problems, their analysis is a good starting point, but does not work well by itself and must be supplemented with assumptions that vary with the analyst. How is a court to determine whether the winery or the winemaker has, in Aghion and Tirole's phrase, "comparative advantage in creating value,"(20) so as to determine who owns the process of chardonnay reserve? As mentioned, their article makes some empirical assumptions that have already been questioned and must be regarded as quite tentative.(21)
If applied to legal problems, a time value would have to be added. For example, for a long time in its development of the popular 386 chip, Intel surely had "comparative advantage in creating value." But according to the California courts, at some time that advantage ended. After specifications were published in a trade magazine, did Chan, Hwang, and ULSI have as much or more advantage as Intel in using the specifications to design compatible math coprocessors?(22) When it became clear that Intel was not living up in good faith to its second source agreement with Advanced Micro Devices, so that AMD should have "reverse engineered" the 386 or been responsible for its failure to do so, does that mean that AMD now had "comparative advantage" over Intel?(23) When did this happen? How could we know?
3. Merges: sole ownership is less costly than shared ownership
a. Merges' argument
Merges, in a forthcoming article, makes a different implicit criticism of Aghion and Tirole: that shared ownership of intellectual property would create such insuperable obstacles to efficient contracting as to call rather for strong judicial default rules lodging intellectual property in employers.(24) Merges analyzes invention assignment agreements, like Aghion and Tirole's subject, an area in which there will be property, though it may be an open question whose. As such, there is no direct conflict between Merges' article and my own. One might read this Article on the question of whether there is to be property, and Merges' on the question of who ought to own such property as law recognizes. This would limit Merges' discussion to principles for construing agreements over recognized property, such as patents. However, if courts follow the recommendations of this Article, and narrow the legal definition of trade secret to the working definition in California courts (tangible positive information that is genuinely secret and would not have been created if it couldn't be secret), firms might be tempted to draft agreements purportedly making a broader class of information into company property. The construction of such agreements would implicate the issues raised in Merges' article.(25)
Merges directly applies the Grossman-Hart-Moore approach. Because contracts are costly to execute and invariably incomplete, there may be efficiency advantages in ownership (property) rather than contract. "[I]f the power of ownership were given to individual employees, it would be wasted in an orgy of expensive and ultimately self-defeating bargaining."(26) The firm arises in the first place because the knowledge of its respective employees are "strictly complementary [assets]: when they have economic value only in combination with other assets."(27) In such situations, "property rights theory tells us to be wary of divided ownership."(28) Additionally, if agents have property in their own inventions, they have incentives to ignore the firm's needs and devote on-the-job energy to their own; in such cases, firm ownership of inventions substitutes for a kind of impossible monitoring of employees.(29) While such employer ownership of intellectual property may appear to disadvantage employees, Merges argues that this is not so since employees retain an "exit option." The employee with a valuable invention, that has value independent of his employer's assets, is free to leave the firm without disclosing the idea.(30)
For independent consultants (as opposed to employees), Merges makes the contrary argument: they should normally own their own inventions, and any purported assignment to their customers should be narrowly construed, in favor of the consultants. There are several reasons for this. First, consultants' inventions are not complementary to firm assets, so there is no risk of hold-up. Second, letting consultants hold their own property rights "gives them the best incentive to perform." Third, property rights for consultants is a good "penalty default rule" in Ayres and Gertner's phrase, that is, it creates an incentive for the exceptional customer, commissioning research that will be complementary to firm assets, to disclose this fact to the consultant.(31)
b. Critique of Merges
Merges' discussion is subtle and valuable; I cannot discuss all its observations here. The value of unified ownership over complementary assets is an important one to which Merges rightly draws our attention. Courts should not lightly assume that parties have contractually allocated rights in information so as to diminish any economic value of that information. One wonders, however, whether this value is really the only important value that should shape the law; how important is the problem of high transaction costs due to divided property rights; and whether law is really prepared to enforce the property rights that Merges advises.
(1) The costs of firm ownership of intellectual property
This entire Article has been devoted to establishing some costs of corporate ownership of intellectual property. Corporate-owned intellectual property does not grow as rapidly as it could (Romer's point about the tendency to underinvestment), does not spill over rapidly (Saxenian's point about Route 128), does not provide as powerful incentive to human imagination as employee ownership (as Merges recognizes in his discussion of consultants), and dissipates rents in negotiations (this is Romer's point about how lucky we are that neither Bell Labs nor AT&T "owns" the transistor). Merges properly and correctly cautions courts not to presume that contracts have assigned property rights so as to minimize its value. By precisely the same token, courts will sometimes be compelled to construe contracts in order to provide for technology sharing, informal know-how trading, or nonrivalrous information in the public domain, and by the same logic. They will assume a Silicon Valley world, in which employers and employees implicitly agree to lodge property in information where it will increase each's wealth the fastest, that is, as nonrivalrous information.
Merges makes the common mistake that Romer criticizes: he recognizes correctly that information can be crucial to a firm's success, but assumes wrongly that this information can always be held as property that excludes other users, not as nonrivalrous information that can be used infinitely. Again, this is mainly because Merges is concerned with "inventions" that law has already determined will be patentable property. It's important that Merges' model be limited to this kind of excludable information, and not extended to matters that he doesn't address. Merges does not discuss any aspect of nonrivalrous information, none of the literature on increasing returns or endogenous growth, nor Saxenian's work on Silicon Valley.
(2) Is contracting for property a serious problem?
Intellectual property will continue to be held outside of the firm, by consultants, other firms, and occasionally employees, no matter how many presumptions courts employ to assume firm ownership. Firms will and do contract for use of others' intellectual property. As mentioned above, some venture capitalists report that issues of ownership of intellectual property arise in every start-up that they've financed. I know of no empirical literature that suggests that contracting for rights in others' intellectual property is a serious problem. Merges' colorful invocation of "an orgy of expensive bargaining" responds more to the formal assumptions of the Grossman-Hart-Moore literature than to any known social problem.
For employee-held intellectual property (secrets, know-how, protocols, inventions, conceptions), the key negotiation, as Pakes and Nitzan confirm, is the exit negotiation. My Silicon Valley informants all confirm the importance of this negotiation, irrespective of anything the employee signed upon commencing employment. In Silicon Valley and elsewhere, the employee almost surely signed an agreement assigning inventions, possibly for some period after the end of employment, and agreeing to the secrecy of proprietary and confidential information.(32) Such agreements may strengthen the employer's position but do not eliminate the negotiation problem.
The employer will still have to evaluate what the employee has been working on, how valuable he is to the company, how damaging would be his working for a competitor. The employer will still have to decide whether to match or beat the employee's outside offer, sweeten it with stock options or some other compensation reflecting the new line of business that the employee may want to pursue. Such negotiations are as tricky and interesting as any other employment negotiations, but that's what business is all about, and Merges' proposals will not spare employers from them.
In that tiny subset of such negotiations that Merges wrongly assumes constitutes the whole world--the set in which the employee claims ownership of an asset that is complementary to firm assets in the sense that neither has value without the other--the firm should be able to put together the most attractive bid. Employees are not stupid. If their invention has more value when exploited by Hewlett-Packard than when exploited by the employee himself, the employee would be an idiot to leave Hewlett-Packard. In any case, the employment and venture capital markets will provide highly accurate assessments of the value of the employee's information in hands other than his employer's. Colorful talk of hold-ups simply begs the normative question of who is holding up whom: it assumes that the information is the employer's asset, when that is precisely what must be determined.
Most employee departures, following Pakes and Nitzan, do involve information that is an asset to the employee, but probably do not involve complementary assets. Merges' assumption that most employee inventions are complementary assets to firm assets is again more a formal assumption of the Grossman-Hart-Moore literature than an empirical finding about employee information.
In other words, the problems that Merges fears are not such big problems. They are faced hundreds of times a day in Silicon Valley's high-velocity labor market, when employees (or consultants) want to leave (or the firm wants to change their status) and negotiations ensue. I know of no empirical evidence that such negotiations are a particular social problem and my interviews in Silicon Valley have not turned up any such evidence. In short, I agree with Merges that the employee exit option is the key moment to examine, but I also believe that contracting at that stage is a theoretically efficient and socially common way of reaching optimal allocations of intellectual property, largely irrespective of any boilerplate signed earlier by the employee.(33)
(3) Are ideas in employees' heads property like any other property?
An important social and legal problem is the lengths to which law will go to enforce corporate ownership of information located in employee heads. Unlike the supposed orgy of bargaining, this is a real problem, as illustrated by the Evan Brown litigation, referred to above.(34) Merges, following Anton and Yao, asserts that an employee with a good idea not yet put into readable form is normally free to exit the company.(35) This is not always so.
Evan Brown is an engineer who has been ordered by a Texas court to disclose to his former employer algorithms that he has never written down anywhere. Brown claims to have developed these to solve problems, of intellectual interest to him, in converting older software into newer codes. These are continuing problems for anyone working with older software, but not something for which DSC Communications Corp., a communications company, markets programs. DSC claims, successfully so far, that the algorithms are their property under invention assignment and trade secret agreements that Brown signed on employment. The extensive press coverage seems mostly sympathetic to the employee.
Professor Merges does not discuss the case. I think that it flows directly from the kinds of "property rights" metaphors that he employs. As I said, I think that Merges is right to draw our attention to asset complementarity. And where an employee's invention is really valuable only as a complement to assets held by his employer--I express no opinion whether this is so of Evan Brown's algorithms--then it may be sound construction of their relevant contracts to suppose that the employer owns the invention. We could, in such a case, even describe the employee's invention as the "property" or an "asset" of the employer. However, this is a metaphor as well as a performative, and the metaphor must not obscure. An employee's invention can never be just the same kind of "asset" as a machine, factory, or notebook. Once the algorithm is disclosed, others cannot be excluded from using it, yet the production and diffusion of such algorithms and similar nonrivalrous information is, Romer shows, the most important factor in economic growth. This is why accounts of Brown's algorithms that characterize them as "property" or "assets" must be supplemented or corrected by accounts like Romer's, that carefully separate the wealth they represent by being shared, from the wealth they represent by being held secret.
Like most of the press coverage of this litigation, I find the spectacle of a court ordering a human to disclose his ideas to his former employer authoritarian and barbaric. That is not the point of this Article, however, which has tried as dispassionately as possible to examine the purely economic case for the expanded legal category of the public domain that is typical in the trial courts of Santa Clara County. Cases like Brown strike me as the inevitable consequence of loose talk about employee thoughts as employer "assets" or "property."
B. Human Capital analysis of employee information
[Rubin and Shedd(36) article in JLS would make covenants not to compete enforceable to the extent employer is regaining "its" investment in training. Critique. Since California grows without any covenants not to compete, they can't be necessary. Rubin and Shedd misuse human capital theory, under which it's not a meaningful question whether employers or employees "pay for" training. But like more standard human capital theory, they lump together as "human capital" what this Article shows must be kept secret: nonrivalrous information that contributes to economic growth when it spills over, and rivalrous information that is embodied in its employee-owner].
C. Machlup on the psychology of monopoly
I don't suppose anybody reads Fritz Machlup these days, or thinks about monopolies the way he did.(37) However, his analysis of monopolies, including monopolies of information, written almost half a century ago, makes more sense to me as an explanation of contemporary Silicon Valley practice than do most contemporary discussions.
In Machlup's models, the economic effect of a monopoly, specifically, the incentives it provides for further investment by the monopolist, varies under different psychological assumptions. He specifically applies the analysis to monopolies over information, such as patents.
The worst case for society is to grant a monopoly to a pessimist who is wrongly convinced that newcomers' competition will arise at any moment to destroy his business. (If the monopolist is correctly convinced of imminent competition, Machlup points out, the case is of no social significance since the monopoly ends). For such a pessimistic monopolist, "[i]nvestment in industrial research, development, and innovation will not appear promising in view of the supposedly imminent advent of competition. Inventions will be suppressed if the time for the amortization of the required new investments seems too short."
By contrast, Machlup hypothesized "the over-optimistic entrepreneur who underestimates the actual degree of pliopoly [ease of market entry] and overestimates the safe period. He need not be an actual monopolist, nor even imagine that he is one; it suffices that he believes it will take his competitors--imitators or makers of substitutes--longer than it actually does to start competing with him. This optimism is the best promoter of technical progress.[I]f imitation is rapid while the firms expect it to be slow, society will get the benefit of innovation as well as of rapid imitation.
"To buy innovation by paying with unnecessarily long delays of imitation is a poor bargain for society to make. Imitation always and necessarily lags behind innovation. It will be the best deal from the point of view of society if innovators optimistically overestimate this lag. If they expect the lag to be longer than it actually is, innovation will be enhanced and imitation will not be delayed. That it may create this socially wholesome illusion on the part of innovators is the strongest justification for a well-designed patent system."(38)
The structure of trade secrets in Silicon Valley bears out the truth of Machlup's mordant analysis a half-century on. It pays elaborate homage to trade secrets. Every firm uses trade secret agreements, that are exempt from California's statutory prohibition on covenants not to compete. It encourages firms to identify their wealth with their secrets, and offers to protect those monopolies. It thus encourages--very successfully--investment and innovation. Yet in fact a secret is only as secure as an employee is happy. Innovations thus diffuse rapidly and are easily imitated, lowering costs to consumers. What more could anyone want from a system of intellectual property?
There is a clash between recent developments in labor markets, and law's definition of trade secret. Law conceptualizes most employee information about how to do their jobs as the property of their employers. This makes a kind of sense when firms use internal labor markets, rarely hire from without, and offer valuable employees orderly lifetime careers inside the firm. It is incompatible, however, with high-velocity labor markets, such as Silicon Valley's, in which employees are expected to move frequently from firm to firm and cannot effectively be prevented from trading on the information they acquired at past employment.
It is certainly possible that such a high-velocity employment market will come to be associated with diminished employer investment in research, or training, or particular types of research or innovation. However, the preliminary soundings are that any such diminished employer investment is swamped by the growth associated with such a pattern of employee mobility and reduced intellectual property. Sources of such growth include incentives for employees and consultants to produce information; rapid diffusion of nonrivalrous information among competing firms through informal know-how trading, technology transfers, and, most importantly, employee mobility; industry convergence on diagnoses and protocols, avoiding duplicative wasted effort; and relatively low-transaction cost negotiations between employers and employees, against the background of a venture capital and employment market that provides accurate information of the market value of each employee's store of information.
In California, the legal system has facilitated this growth by making covenants not to compete unenforceable, and by adopting, at least until 1996, a narrow effective definition of "trade secret" that makes it very difficult for firms to prevent departing employees from starting or joining competitors. Other jurisdictions trying to encourage similar high growth districts should similarly make sure that employee mobility is not impeded, either by employment law or intellectual property law. Courts must be sensitive to the inhibiting effect that employer intellectual property rights can have on economic growth.
In particular, the following recommendations respecting employer-employee trade secrets disputes are preliminarily indicated. First, courts should require economic analysis of the likely effects of their decisions on incentives to produce information, and on the likelihood of the information diffusing. Second, courts must be sensitive to the real contract the parties have, and enforce that contract. An agreement to keep the employer's secrets may mean one thing when made by a new hire looking forward to an orderly lifetime career with that employer. It must mean something else when the employee and employer both know that the employee will be back on the job market when this project ends in six months. Third, courts should limit the definition of "trade secret" to tangible information, actually a secret in the industry, that would not have been created by the employer unless secrecy was reasonably contemplated. Doctrines like the "inevitable disclosure" of trade secrets must be rejected as antithetical to the public interest.
This article, so far as I know, is the first to examine the labor market implications of Paul Romer's work on endogenous growth, and the specific role that employee mobility can play in the process often blandly described as information "spillover" or "diffusion." Based particularly on AnnaLee Saxenian's comparison of Silicon Valley with Route 128, I have suggested that short job tenures and frequent mobility can contribute significantly to information diffusion and endogenous growth.
But is a Silicon Valley labor market necessary for rapid economic growth? Or could similar growth be achieved while maintaining more stable employment relations? There is much more to be understood about the relationship between the labor market and economic growth. Relevant is the strain of labor market scholarship on creation of new jobs, that analyzes the often disappointing results when countries shred their safety nets. Britain and Spain are examples of countries that attempted to use law to encourage short-term work and reduce unemployment benefits, yet achieved little in new job creation. It often turns out that the informal labor market presented very different barriers to job creation than did the formal market, and that laws on employment contracts were far from the most important barrier to job creation.(39)
It would be possible similarly to shred social safety nets, encourage short-term labor, yet not achieve the growth of Silicon Valley. It might also be possible to replicate Silicon Valley's growth with more stable employment contracts, but using other mechanisms that diffuse information among firms in the way that, in Silicon Valley, employee mobility does. Perhaps this is how Japan worked during its most recent growth years: stable employment relations, coupled with governmental and industry devices to spread information among firms.(40) In California, employee mobility may play the role that industrial planning did elsewhere--and play it better.
1. 111 A lucid introduction in Bernard Salanié, The Economics of Contract: A Primer (1997).
2. 112 Sanford Grossman & Oliver D. Hart, The Costs and Benefits of Ownership: A Theory of Vertical and Lateral Integration, 94 Q.J.Econ. 691 (1986); Oliver D. Hart & John Moore, Property Rights and the Nature of the Firm, 98 J.Pol.Econ. 1119 (1990); Oliver Hart, Firms, Contracts, and Financial Structure (1995).
3. 113 R.H. Coase, The Nature of the Firm, 4 Economica 386 (1937).
4. 114 The same problem arises with an economic analysis that assumes weak or unenforceable property rights and asks what kinds of bargains will result. James J. Anton & Dennis A. Yao, Start-ups, Spin-offs, and Internal Projects, 11 J.L.& Econ.Org. 362 (1995). I will refer to this interesting article, but it cannot address the question of strong property vs. no property.
5. 115 Romer, supra n.15.
6. 116 Ariél Pakes & Shmuel Nitzan, Optimum Contracts for Research Personnel, Research Employment, and the Establishment of "Rival" Enterprises, 1 J.Lab.Econ. 345 (1983).
7. 117 For further exploration of this point from a tax point of view, see Bankman, supra n.58. Since most new businesses lose money; continuing businesses have gains against which to offset these losses; and new ventures do not, it seems irrational for anyone ever to set up a new business. It would seem that an existing business could always outbid the investors in the new business.
8. 118 Pakes & Nitzan, supra n.116, at 353.
9. 119 Pakes & Nitzan, supra n.116, at 361.
10. 120 Pakes & Nitzan, supra n.116, at .
11. 121 See generally David Card & Alan B. Krueger, Myth and Measurement: The New Economics of the Minimum Wage 152-77, 369-86 (1995).
12. 122 Pakes & Nitzan, supra n.116, at 347.
13. 123 See generally Alan Hyde, In Defense of Employee Ownership, 67 Chi.-Kent L.Rev. 159 (1991).
14. 124 Boyle, supra n.14.
15. 125 Anton & Yao, supra n.114, make the same assumption, 11 J.L.& Econ.Org. at 369, analyzing the employee's choice whether or not to reveal a good idea to the employer.
16. 126 See Brown v. DSC Communications Corp., 1998 WL 2366 (Tex.App. 1998)(dismissing interlocutory appeal of injunction ordering a former employee to disclose algorithms, existing only in his own head, to his former employer). I will discuss this case further in section V.A.3., in connection with Robert Merges' recent work. In the mountain of press coverage of this case, see also John S. Pratt and Peter Dosik, Whose Idea Is It? Company Sues Ex-Employee, 20 Nat.L.J. (Oct. 20, 1997); Geanne Rosenberg, An Idea Not Yet Born, But a Custody Fight, N.Y. Times, Sep. 8, 1997, at D3, col.1.
17. 127 Philippe Aghion & Jean Tirole, The Management of Innovation, 109 Q.J. Econ. 1185 (1994).
18. 128 Aghion & Tirole, supra n.127, at 1190.
19. 129 Josh Lerner & Robert P. Merges, The Control of Strategic Alliances: An Empirical Analysis of Biotechnology Collaborations, presented to Columbia University Center for Law and Economic Studies, December 2, 1996. [now National Bureau of Economic Research Working Paper, July 1997].
20. 130 Aghion & Tirole, supra n.127, at
21. 131 Lerner & Merges, supra n.129.
22. Supra nn.80-88.
23. 133 Supra n.93.
24. 134 Robert P. Merges, Property Rights Theory and Employee Inventions, Calif.L.Rev. (forthcoming). I used the April 26, 1998 draft, included in the collection Corporate Governance Today, published by the Project on Corporate Governance, Columbia University Center for Law and Economic Studies.
25. 135 As mentioned above, supra n.75 [State Farm case], it is apparently an open question in California whether courts will enforce an agreement that purports to make a trade secret out of something that would not have been held to be a trade secret by the courts.
26. 136 Merges, supra n.134, at 70 [Columbia volume].
27. 137 Id. 74. Note the assumption, criticized by Romer and throughout this Article, that knowledge is a rivalrous asset.
28. 138 Id. 78.
29. 139 Id. 81-84. This point draws on the familiar multitask principal-agent literature. See, e.g., Bengt Holmstrom & Paul Milgrom, Multitask Principal-Agent Analysis: Incentive Contracts, Asset Ownership and Job Design, 7 J.L.Econ & Org. 24 (1991).
30. 140 Merges, supra n.134, at 95-100.
31. 141 Merges, supra n.134, at 88-89. "Penalty default rules" are named and explained in Ian Ayres and Robert Gertner, Filling Gaps in Incomplete Contracts: An Economic Theory of Default Rules, 99 Yale L.J. 87 (1989).
32. 142 See supra nn.76,79 [Zachariades litigation; Neumyer survey; interviews].
33. 143 It is interesting to speculate why contract, which may have been overrated by an earlier generation of law-and-economics scholars, is so underrated today. It seems like only yesterday that freedom to bargain over strong property rights was supposed to be cheap, efficient, and the sure road to social Pareto-optimality. What happened to turn this rosy model into today's "orgy of expensive bargaining"? Today's contracts scholars, particularly in the incomplete contracts paradigm, stress all the things that keep a contract from happening (mistrust, information disparities, uncertainty about present or future states of the world) or from being effective even if negotiated (contractual incompleteness, defection, mistrust, let alone spite, envy, and ressentiment). I think of this whole school as "Dostoevskian Law and Economics." In Grossman-Hart-Moore models, the alternative to this distasteful bargaining is the smooth, orderly, authoritarian world of integrated ownership and hierarchy, and perhaps this picture is as appealing to some scholars as was the invisible hand of bargain to their predecessors. There are, however, reasons why it is not always good for Xerox to own Apple, for National Semiconductor to own Intel, for Stanford University to own Hewlett-Packard. There are also reasons why it is not good for any of those employers to own all the relevant information as corporate assets. There are finally reasons that often elude the Grossman-Hart-Moore analysis why it is not a good thing for any of these employers to own their employees.
34. 144 Supra n.126.
35. 145 Supra n.134, at 95-96.
36. Paul H. Rubin & Peter Shedd, Human Capital and Covenants Not to Compete, 10 J.Leg.Stud. 93 (1981).
37. 147 But see Margreth Barrett, Intellectual Property 284 (1995), excerpting another of Machlup's studies.
38. 148 Fritz Machlup, The Economics of Sellers' Competition: Model Analysis of Sellers' Conduct 54-56 (1952).
39. 149 Social Protection versus Economic Flexibility: Is There a Trade-off? (Rebecca M. Blank ed. 1994).
40. 150 Supra n.27.