When Development Becomes Dependence: Human-Capital Transferability and the Quality of Exit

Abstract

A worker can become substantially more capable through employment while becoming less able to carry the economic value of those capabilities into another institution. Labor economics has long recognized that human capital differs in its transferability. Becker’s distinction between general and specific human capital established the canonical starting point; later work shows that transferability can vary with task similarity, occupational distance, the combination of otherwise general skills demanded by particular firms, and the availability of receiving employers capable of using those combinations productively. The political significance of these differences is easy to miss if an outside option is treated as binary. An alternative employer may exist while receiving only a diminished share of the productive capability the worker has accumulated.

This article develops that proposition at the intersection of human-capital economics and republican theories of labor freedom. Its claim is deliberately narrower than a general theory of institutionally produced dependence. Human-capital transferability is an institutionally governable determinant of the quality—not simply the existence—of workers’ outside options. Where firms or legal institutions exercise meaningful discretion over portability-shaping arrangements, republican evaluation should extend to those arrangements because they affect how much accumulated productive agency can remain economically usable across institutional boundaries. The relevant object is productive continuity: the extent to which developed worker capabilities remain capable of valuable exercise after exit.

The argument does not imply that maximum portability is desirable. Specificity can be technologically inherent; specialization can generate substantial productivity gains; and imperfect labor markets can create incentives for firms to finance valuable training that they would not finance under perfect competition. Imperfect transferability can also increase employers’ replacement costs, so its effects on bargaining power are heterogeneous rather than monotonic. Noncompete agreements provide the principal stress test because they expose the distinction between exit and productive continuity unusually clearly. Workers may remain capable of leaving while being displaced toward employers or product markets in which accumulated expertise fits less well. The resulting normative principle is comparative rather than absolute: where feasible arrangements can protect the same independently legitimate productive interest while preserving materially different amounts of workers’ accumulated productive capability, freedom as nondomination gives a pro tanto reason to prefer the arrangement that preserves the stronger outside option, subject to its effects on investment, replacement costs, bargaining, and the distribution of adjustment burdens.

I. Exit Is Not a Binary Variable

Employment changes workers. A person who spends ten years inside an organization does not leave with only ten additional years of wages. She may acquire technical competence, tacit judgment, professional knowledge, routines of coordination, managerial capability, occupational reputation, customer knowledge, and mastery of particular systems. Some of these capacities will remain valuable across many firms. Others will travel principally to firms performing similar tasks. Still others will produce high returns only when joined to complementary technologies, teams, organizational routines, or assets resembling those through which they developed.

Labor economics supplies several ways of describing this variation. Becker’s canonical distinction separates general training, which raises productivity beyond the training firm, from specific training, whose productivity effects are concentrated within the employment relation. Later approaches substantially complicate that binary. Gathmann and Schönberg model task-specific human capital and show that workers disproportionately move toward occupations with similar task requirements, while the task distance of occupational changes declines with experience. In their estimates, task-specific human capital accounts for as much as 52 percent of overall wage growth. Gibbons and Waldman likewise incorporate task-specific accumulation into a model of assignment and wage dynamics inside firms. Lazear, from another direction, shows that even wholly general skills can form a de facto firm-specific bundle when firms place different weights on those skills and the outside market for the originating firm’s particular combination is thin. (Becker 1962; Gibbons and Waldman 2006; Lazear 2009; Gathmann and Schönberg 2010.)

These results make a simple philosophical point economically precise. Two workers can both possess alternative employers while possessing very different outside options. One may move to a comparable organization and continue exercising most of what years of work have taught her to do. Another may also leave, but only for destinations whose tasks, technologies, skill weights, or complementary assets make a substantially smaller portion of her accumulated capability valuable.

The difference matters because exit performs more than one function in employment relations. It can terminate a particular authority relationship. It can also operate as a bargaining threat: the credible possibility of moving elsewhere constrains the terms an incumbent employer can impose. That disciplining function depends not simply on whether movement is possible but on what happens after movement. An alternative requiring a large wage reduction, occupational downgrade, abandonment of accumulated expertise, geographic displacement, or costly retraining may remain an alternative while supplying a weak threat point.

The existence of another job is therefore an incomplete description of an outside option.

The central concept of this article is productive continuity: the extent to which worker-embodied productive capabilities accumulated through one employment relation remain capable of valuable exercise after movement into another productive setting. This does not imply that workers own every source of their productivity. Firms contribute capital, intellectual property, data, customer relationships, organizational routines, teams, brands, equipment, and other complements that workers cannot simply carry away. Nor does every loss at an institutional boundary amount to an injustice. Some value genuinely exists only in the match.

Productive continuity asks a narrower question: how much of what the worker herself has become capable of doing remains economically usable on the other side of the institutional boundary?

That variable should not be confused with mobility. A worker can be mobile but poorly transferable. Another can remain with one employer for decades while possessing highly transferable human capital. A third may possess skills that are technologically general but economically specific because very few outside firms demand the same combination. Transferability concerns the receiving capacity of the alternatives available to the worker.

The political significance of an outside option therefore depends partly upon what the destination can receive.

II. What Human Capital Actually Transfers

Becker’s general-specific distinction remains indispensable because it identifies the investment problem at the center of human-capital formation. Under the competitive benchmark, firms lack an incentive to finance perfectly general training when workers can leave and capture its return elsewhere. Firm-specific investment produces a different division problem because separation can destroy value that exists principally within the match. (Becker 1962; Acemoglu and Pischke 1999a.)

Actual human capital, however, is less binary than that framework can suggest.

Gathmann and Schönberg’s task-based approach measures transferability across occupations through the similarity of tasks. Workers tend to move into occupations with similar task requirements, experienced workers make less distant occupational moves, and task-specific human capital accounts for a substantial portion of wage growth. Their result is particularly important here because it allows specificity to vary continuously. Human capital need not be either useful at one firm or equally useful everywhere. It can remain highly productive within a neighborhood of related work and depreciate as task distance increases. (Gathmann and Schönberg 2010.)

Gibbons and Waldman reach a related insight from within the firm. Their model introduces task-specific human capital to explain features of wage, assignment, and promotion dynamics. Experience therefore does not simply increase an undifferentiated stock of “skill.” Repeated task assignment changes the composition of the worker’s productive capability. (Gibbons and Waldman 2006.)

Lazear’s skill-weights approach further loosens the equation between technological specificity and economic specificity. All component skills may be general, yet a worker’s bundle can become disproportionately valuable at one firm because that firm demands a distinctive combination. Market thickness then matters: where many firms demand similar combinations, the bundle is more transferable; where few do, the same nominally general skills can behave economically like firm-specific capital. (Lazear 2009.)

Transferability is therefore relational rather than an intrinsic substance contained inside the worker. It depends upon what the worker has learned, what alternative institutions require, what complementary assets make those capabilities productive, and how many plausible receiving institutions value sufficiently similar combinations.

This distinction illuminates an otherwise puzzling possibility: an alternative can deteriorate without disappearing. An engineer may be qualified for many jobs while finding that only a narrow set uses the architecture in which her accumulated judgment is most valuable. An inventor may remain employable while losing access to precisely the firms whose technologies most closely match her accumulated expertise. A manager may possess broadly recognizable skills while discovering that the specific combination developed through one organization commands a substantial premium only in a thin external market.

The number of alternatives can therefore remain constant while their productive value changes.

I use productive agency across institutional boundaries to name this object from the worker’s side. “Agency” here does not introduce a new general theory of personhood. It identifies a practical productive capacity: the ability to deploy accumulated competence in consequential economic activity. A worker who can leave only by ceasing to exercise a large share of what she has become capable of doing occupies a different economic position from a worker whose accumulated competence remains highly usable after movement.

Human-capital economics already explains why that difference can exist. The question for political theory is what follows when institutional arrangements can materially affect its magnitude.

III. The Quality of the Outside Option

Republican labor theory gives strong reasons not to identify freedom with voluntary contracting or the formal availability of exit. Pettit’s account of nondomination focuses on exposure to uncontrolled power rather than actual interference alone. Bryan develops the labor-market implications more specifically. He argues that reasonable alternatives to a particular wage relation are insufficient to establish freedom because agents may remain dependent upon the cooperation or permission of others to engage in productive activity. His independent-production criterion asks both how strongly an agent relies on such cooperation and whether the decisions and terms governing that cooperation are forced, through relevant forms of control, to track the agent’s interests. (Pettit 1997; Bryan 2023, 692–704.)

The present argument does not compete with Bryan. The existence of a good outside option is not sufficient for nondomination. But it does not follow that outside-option quality is politically unimportant. Alternatives remain one mechanism through which bargaining positions are formed, employer power is constrained, and the consequences of refusal are determined.

The residual question is internal to the concept of an alternative:

What makes one labor-market alternative stronger than another?

At minimum, outside-option quality contains several analytically separable dimensions. The alternative offers some level of compensation and nonpecuniary value. Reaching it imposes transition costs. And it can receive some portion of the human capital the worker has accumulated.

Holding other dimensions constant, an outside option becomes weaker when substantially less worker-embodied productive capability can be transferred into it at comparable value.

This yields the paper’s independence theorem:

The Transferability Thesis: Holding other relevant features of an outside option constant, its quality declines as the value of worker-embodied productive capability that can be carried into and exercised within the alternative declines.

The proposition is intentionally modest. Productive continuity is not the only dimension of an outside option. Wages, working conditions, geography, schedule, health insurance, risk, status, workplace authority, and nonpecuniary goods can all matter. Nor does the thesis imply an entitlement to preserve the full market value of every skill combination developed through work.

It identifies transferability as one distinct determinant that references to “alternative employment” can otherwise obscure.

The republican significance follows because the quality of alternatives affects exposure to employer discretion. If refusing the incumbent employer means moving to another institution where accumulated capability remains highly valuable, the worker possesses one kind of counterweight. If refusal means preserving employment only through substantial human-capital depreciation, the price of contesting the incumbent relation is larger.

Evidence on noncompete enforceability supports the underlying relationship between restrictions on mobility and outside options. Johnson, Lavetti, and Lipsitz find that greater noncompete enforceability reduces earnings and job mobility, with larger effects for workers most likely to sign noncompetes. Their analysis specifically identifies diminished outside options and reduced ability to exploit tight labor markets as mechanisms affecting earnings. They also find that greater enforceability exacerbates gender and racial earnings gaps. (Johnson, Lavetti, and Lipsitz 2025, 2735–2793.)

The argument here goes one step more specific conceptually, not causally. Outside options differ not only in how many employers remain available or what those employers pay. They differ in how much accumulated human capital those destinations can use.

That creates a sharply defined institutional object:

The Portability-Governance Proposition: Where an institution exercises meaningful discretion over an arrangement that materially changes the transferability of worker-embodied human capital, that arrangement is a proper object of republican evaluation because it changes one determinant of outside-option quality. Where comparably effective and economically feasible arrangements protect the same independently legitimate productive interest while preserving materially different levels of productive continuity, there is a pro tanto reason to prefer the arrangement preserving the stronger outside option.

Nothing in this proposition requires describing the institution as producing dependence. Its labor-market significance is more specific. The arrangement alters what survives exit.

That is the independent object of this paper.

IV. Why Maximum Transferability Is the Wrong Ideal

The obvious counterfeit theory would instruct institutions to maximize worker portability.

Economics gives strong reasons to reject it.

In the standard competitive human-capital model, firms do not finance general training because workers can depart and capture the returns. Acemoglu and Pischke show that this result changes under imperfect labor markets. Wage compression and related frictions can allow firms to capture part of the return generated by technologically general training, producing firm-sponsored general training as an equilibrium phenomenon. Labor-market imperfections can therefore support capability development that might be underprovided under the competitive benchmark. (Acemoglu and Pischke 1999a, F112–F142; 1999b, 539–572.)

A political rule that simply minimized every friction affecting transferability could consequently reduce some investments that benefit workers.

Han introduces a second complication. His dynamic two-sided labor-market model defines imperfect human-capital transferability as the portability of skills across occupations. Workers anticipate depreciation in human-capital value when changing occupations. Imperfect transferability gives firms market power because occupational switching becomes more costly for workers, but the same imperfection makes experienced workers harder for firms to replace. The effects therefore run in opposite directions. In Han’s counterfactual, perfect transferability reduces wage markdowns for older workers but increases them for younger workers. The life-cycle profile of markdowns is correspondingly nonmonotonic. (Han 2025.)

This result is important because it rules out a simple equation:

lower transferability ≠ uniformly greater employer bargaining power.

Transferability operates on at least two margins. It affects the worker’s departure margin: how much productive value the worker sacrifices by moving. It also affects the employer’s replacement margin: how easily the firm can replace accumulated skill if the worker leaves.

Bargaining adds another complication. Gopal, Li, and Rawling model noncompetes as a tradeoff between firm investment incentives and allocative efficiency, with outcomes dependent on bargaining power. In their model, when worker bargaining power is sufficiently high, the noncompete and no-noncompete contracts coincide because neither induces firm-provided investment; when firms possess all bargaining power, noncompetes encourage industry-specific investment but reduce wage growth. Their empirical results show that wage gains associated with noncompetes are concentrated among workers who did not negotiate over them, a pattern they interpret as consistent with firm-provided transferable skill investment where firms possess substantial bargaining power. (Gopal, Li, and Rawling 2026, 267–272.)

This is precisely the kind of result a serious normative account must absorb rather than explain away.

The political economy of transferability therefore involves several potentially conflicting margins: how much human capital is formed; how much of it remains useful after separation; how costly experienced workers are to replace; how efficiently workers are allocated across firms; and how the resulting surplus is divided.

An arrangement can improve one while worsening another.

The appropriate ideal is consequently not maximum portability but governed specificity.

Productive institutions must remain capable of generating deep expertise, distinctive skill combinations, long-duration investments, and valuable complementarities. The political question begins where institutions or legal regimes possess meaningful choice among ways of protecting those goods that impose materially different losses on workers’ productive continuity.

V. Noncompetes: Exit Without Productive Continuity

Noncompete agreements are a demanding test because they make the distinction between mobility and productive continuity empirically visible.

The ordinary description says that noncompetes restrict worker mobility. That is true but incomplete. The more revealing question is what happens to workers who move anyway.

Mueller’s 2026 study of inventor mobility exploits changes in state-level noncompete enforceability and examines the allocation of inventors across product markets. In his staggered difference-in-differences analysis, stronger enforceability increases the probability that inventors move across industries by roughly two inventors per hundred per year—about a 42 percent increase relative to the relevant baseline—while within-industry mobility falls. The result is strongest among inventors whose employers are more likely to rely on noncompetes. (Mueller 2026, 1295–1329.)

That finding should be stated carefully. Stronger enforcement does not simply make all affected inventors less mobile. It changes the direction of some mobility.

Mueller then examines the quality of the resulting matches. He constructs a technological-similarity measure from the patent portfolios of inventors and receiving firms. Following increases in noncompete enforceability, the technology cosine similarity of relevant inventor-employer matches falls by 0.05 from a mean of 0.48—approximately a 10 percent reduction. He also documents lower subsequent measures of inventor productivity among the affected population. (Mueller 2026, Table 9.)

The theoretical importance of this evidence lies in what it does not show.

Exit has not disappeared.

Workers move.

But the receiving set changes. Some inventors move farther away in product space and enter firms whose technological portfolios match their accumulated expertise less closely.

Observed mobility can therefore coexist with degraded human-capital transfer.

This is the empirical structure that the concept of productive continuity is meant to capture. The institutional boundary is crossed, but less of the worker’s accumulated productive capability crosses it at comparable value.

Mueller himself frames the mechanism partly as an ex ante/ex post problem. Ex ante, mobility restrictions may strengthen employers’ incentives to invest in workers. Ex post, they can create a hold-up problem because workers cannot credibly threaten to move to close competitors while retaining the value of industry-specific human capital. (Mueller 2026.)

The broader literature reinforces both sides of that tension. Johnson, Lipsitz, and Pei find that stricter enforceability increases firms’ R&D investment while reducing inventor mobility, entrepreneurship, and innovation, with evidence consistent with reduced knowledge diffusion. Starr’s 2026 review concludes that widespread use and enforceability of noncompetes are associated with lower mobility, wages, innovation, and entrepreneurship while taking seriously the traditional rationales involving training and protection of firm information. (Johnson, Lipsitz, and Pei 2023; Starr 2026, 139–166.)

The appropriate question is therefore not whether noncompetes are simply “good” or “bad.”

It is joint:

Which investment is protected or induced? Which high-fit destinations become inaccessible? How much worker-embodied human capital depreciates as a consequence? Who captures the resulting surplus? And could the productive interest be protected through an arrangement imposing less damage on productive continuity?

That is a more demanding inquiry than measuring turnover alone.

VI. Portability Losses Are Heterogeneous

The same formal restriction does not impose the same economic burden on every worker.

Han’s results provide one reason. The interaction between worker outside options and employer replacement costs changes over the life cycle, so improvements in transferability do not affect all workers’ wage markdowns in the same direction. (Han 2025.)

Recent noncompete research identifies additional heterogeneity. Gottfries and Jarosch use a dynamic monopsony job-ladder framework and find that wage effects depend upon the firms using the restrictions: adverse effects can be particularly strong when noncompetes are used by high-productivity firms with high training costs and large rents. (Gottfries and Jarosch 2026, 262–266.)

Potter, Kurmann, and Hobijn find a different kind of heterogeneity in longitudinal NLSY97 data. Signing a noncompete is associated with slower wage growth over four years for workers with lower education but faster wage growth for workers with higher education; effects on tenure are imprecisely estimated for both groups. That result cautions against importing an average wage effect into every part of the labor market. (Potter, Kurmann, and Hobijn 2026, 273–277.)

Contract formation itself is also heterogeneous. Starr, Prescott, and Bishara’s nationally representative survey of 11,505 labor-force participants found that approximately 18 percent were then bound by noncompetes and 38 percent had agreed to one at some point. Only about 10 percent reported negotiating over the provision, and approximately one-third of workers encountering a noncompete were presented with it after accepting the job offer. Noncompetes appeared in lower-skill and lower-wage work as well as in high-skill employment, and earlier notice was associated with better employee outcomes. (Starr, Prescott, and Bishara 2021, 53–84.)

These findings do not establish that unnegotiated restrictions are necessarily dominating or that negotiated restrictions are legitimate. They establish something narrower: the economic and institutional circumstances under which formally similar restraints operate vary substantially.

Two dimensions are especially relevant to productive continuity.

The first is transferability exposure: how much productive value the worker loses when high-fit destinations become inaccessible.

The second is adjustment capacity: how easily the worker can absorb that loss through retraining, geographic movement, temporary earnings reduction, occupational change, savings, or access to other forms of security.

A restriction can therefore impose a modest burden on a worker whose capabilities transfer across a thick market and who can absorb short-run adjustment while imposing a much larger burden on a worker whose accumulated expertise is concentrated in a thin local or occupational market.

This also explains why initial compensation cannot settle the problem by itself. A worker may rationally accept a mobility restriction in return for higher pay, valuable training, or access to an unusually attractive career path. Those benefits are relevant to evaluating the arrangement. But their existence does not establish that every subsequent reduction in outside-option quality was understood, priced, or unavoidable—particularly where the specificity of human capital emerges only after years of participation.

The answer cannot be to make workers individually responsible for remaining perpetually transferable. An ethic of nondomination that instructed workers to stay generic, accumulate enough savings to tolerate any exit, cultivate several occupations, continuously retrain, and avoid deep specialization would transform freedom into permanent labor-market readiness.

The issue is institutional.

The question is which arrangements permit valuable specialization while avoiding portability losses that are not necessary to secure the productive good.

VII. Governing Portability-Shaping Arrangements

The preceding analysis supports a narrower normative test than the former upstream-control argument.

The first question is whether the arrangement serves an independently legitimate productive interest. Genuine protection of trade secrets, identifiable investments, customer relationships, safety interests, fiduciary obligations, or collaborative assets can qualify. The preference to retain workers because weakened outside options improve the employer’s bargaining position cannot by itself establish such an interest.

The second question is comparative: how much additional portability loss does the selected arrangement produce relative to feasible alternatives capable of protecting the same interest?

The third is functional. A formally narrower instrument is not necessarily less restrictive in practice.

Hrdy and Seaman demonstrate the importance of this point for confidentiality agreements. Their analysis of employment agreements finds that confidentiality provisions often extend well beyond legally defined trade secrets, sometimes restricting the use of publicly available or generally known information and lacking the temporal and geographic limits commonly associated with noncompetes. Some confidentiality agreements can therefore operate as de facto noncompetes. (Hrdy and Seaman 2024, 669–754.)

A policy cannot preserve productive continuity simply by replacing one contractual label with another.

The comparative inquiry must ask what the instrument actually does to the worker’s receiving set. Does it preserve access to employers where ordinary accumulated skill and experience remain valuable? Does it force the worker farther away in task, occupational, geographic, or technological space? Does it distinguish genuinely protected firm information from the employee’s own accumulated knowledge and capability?

Recent experimental evidence demonstrates why this comparison cannot be resolved in the abstract. Cowgill, Freiberg, and Starr’s field experiment randomized wages and the presence, salience, and duration of noncompetes across roughly 14,000 job offers to freelance recruiters working with two finance firms; all of the contracts also contained nondisclosure agreements. In the latest revision of the study, removing the noncompete increases mobility between the competing employers by approximately 36–52 percent and increases workers’ combined earnings from the two firms by approximately 12–17 percent. The authors find no evidence, while statistically rejecting even small effects in their setting, that removing the noncompete increased secret leakage. (Cowgill, Freiberg, and Starr 2026.)

The external-validity limit is essential. That experiment does not establish that nondisclosure agreements can protect every legitimate employer interest in every industry. It establishes that, in at least one real contractual setting, a materially less mobility-restrictive architecture preserved the measured information-protection outcome while producing substantially different mobility and earnings outcomes.

That is exactly the kind of evidence the portability-governance proposition requires.

Other candidate mechanisms—garden leave, narrowly tailored nonsolicitation rules, repayment obligations linked to identifiable training costs, deferred compensation, credentialing arrangements, or differently designed training systems—must be evaluated the same way. Each can itself become restrictive. A large repayment obligation can make exit prohibitively expensive. Deferred compensation can replicate a substantial mobility penalty. A nonsolicitation rule can be drafted so broadly that it blocks meaningful employment. Garden leave can compensate income without preserving productive continuity.

There is no universally “least restrictive alternative.”

There are only institutional comparisons.

The same reasoning can reach beyond post-employment covenants. Employers sometimes have discretion over whether training produces externally legible credentials, whether workers rotate through broadly useful tasks, whether certifications are recognized beyond proprietary systems, and whether otherwise valuable capabilities depend completely on firm-controlled complements. Not every such choice should be regulated, and some distinctive organizational systems derive their productive value precisely from specialization.

But where two feasible arrangements produce comparable legitimate productive benefits while leaving workers with materially different amounts of usable accumulated capability after exit, the difference is not economically or politically irrelevant.

The paper’s governing proposition can therefore be stated without borrowing the more general theory of Reflexive Capture:

Freedom-relevant portability is not a demand that institutions avoid dependence or specialization. It is a demand that institutionally governable losses in the continuation value of worker-embodied productive capability be counted when evaluating the quality of workers’ outside options.

The unit of analysis is the institutional boundary.

What happens to accumulated productive capability when the worker crosses it?

Conclusion: Freedom Across the Boundary

Employment is developmental. Workers do not rent a fixed bundle of abilities to organizations and later retrieve the same bundle unchanged. They learn, specialize, combine skills, acquire tacit knowledge, become experienced in particular tasks, and build careers within productive environments that partly determine what those capabilities are worth.

Labor economics has understood the components of this process for decades. Becker distinguishes general from specific human capital. Gathmann and Schönberg show that transferability varies with task distance. Gibbons and Waldman show why task-specific accumulation matters within firms. Lazear shows how combinations of general skills can become economically specific when the external market for the combination is thin. Acemoglu and Pischke show why imperfect labor markets can induce firms to finance training. Han shows why imperfect transferability can simultaneously weaken worker outside options and increase employer replacement costs. The noncompete literature shows why restrictions on movement can affect investment, bargaining, allocation, wages, and innovation in different directions.

The political implication should be equally exact.

The existence of another job is not a complete description of an outside option.

An alternative also has a receiving capacity. It can preserve much of what the worker has become capable of doing or very little. It can continue a productive trajectory or require substantial depreciation of accumulated human capital. It can discipline an incumbent employer strongly or weakly depending partly on what the worker must surrender to use it.

Republican theory need not treat this observation as a new theory of domination. Bryan already shows that reasonable alternatives do not exhaust freedom and that productive dependence raises questions of control beyond voluntariness.

The narrower contribution here is to specify a dimension inside the alternative itself:

Human-capital transferability helps determine the quality of exit.

That proposition identifies a distinct object of institutional governance. Where firms, contracts, or legal regimes possess meaningful discretion over portability-shaping arrangements, their effects on productive continuity belong inside political evaluation.

The relevant comparison is not between specialization and perfect portability.

It is between feasible institutional arrangements serving legitimate productive purposes while imposing different amounts of human-capital loss at the boundary between institutions.

Some losses will be intrinsic to specialization. Some will be justified by investments, information, coordination, or complementary assets. Some restrictions may induce valuable training that would otherwise not occur. Some will make workers harder to replace as well as harder to move. Some will affect workers differently by age, education, occupation, bargaining power, geography, wealth, and accumulated experience.

Those complications do not weaken the argument.

They define it.

The central question is therefore narrower than the theory of reflexive capture and more exact than a general demand for worker mobility:

What happens to freedom when a worker can leave but cannot carry enough of what work has made her capable of doing into the available alternative?

The answer is not that the worker therefore cannot leave.

It is that exit has a quality as well as an existence.

A political economy of labor freedom should learn to recognize the difference.

---

References

• Acemoglu, Daron, and Jörn-Steffen Pischke. 1999a. “Beyond Becker: Training in Imperfect Labour Markets.” Economic Journal 109: F112–F142.
• Acemoglu, Daron, and Jörn-Steffen Pischke. 1999b. “The Structure of Wages and Investment in General Training.” Journal of Political Economy 107: 539–572.
• Becker, Gary S. 1962. “Investment in Human Capital: A Theoretical Analysis.” Journal of Political Economy 70, no. 5, part 2: 9–49.
• Bryan, Alexander. 2023. “Structural Domination and Freedom in the Labor Market: From Voluntariness to Independence.” American Political Science Review 117, no. 2: 692–704.
• Cowgill, Bo, Brandon Freiberg, and Evan Starr. 2026. “Clause and Effect: Theory and Field Experimental Evidence on Noncompete Clauses.” Working paper, revised July 18, 2026.
• Gathmann, Christina, and Uta Schönberg. 2010. “How General Is Human Capital? A Task-Based Approach.” Journal of Labor Economics 28, no. 1: 1–49.
• Gibbons, Robert, and Michael Waldman. 2006. “Enriching a Theory of Wage and Promotion Dynamics inside Firms.” Journal of Labor Economics 24, no. 1: 59–107.
• Gopal, Bhargav, Xiangru Li, and Luke Rawling. 2026. “Noncompete Agreements and Bargaining Power.” AEA Papers and Proceedings 116: 267–272.
• Gottfries, Axel, and Gregor Jarosch. 2026. “Noncompetes and Firm Heterogeneity.” AEA Papers and Proceedings 116: 262–266.
• Han, Qingyang. 2025. “Human Capital Transferability and Employer Monopsony Power.” Working paper.
• Hrdy, Camilla A., and Christopher B. Seaman. 2024. “Beyond Trade Secrecy: Confidentiality Agreements That Act Like Noncompetes.” Yale Law Journal 133: 669–754.
• Johnson, Matthew S., Kurt Lavetti, and Michael Lipsitz. 2025. “The Labor Market Effects of Legal Restrictions on Worker Mobility.” Journal of Political Economy 133, no. 9: 2735–2793.
• Johnson, Matthew S., Michael Lipsitz, and Alison Pei. 2023. “Innovation, Inventor Mobility, and the Enforceability of Noncompete Agreements.” NBER Working Paper 31487, revised December 2024.
• Lazear, Edward P. 2009. “Firm-Specific Human Capital: A Skill-Weights Approach.” Journal of Political Economy 117, no. 5: 914–940.
• Mueller, Clemens. 2026. “Non-compete Agreements and Labor Allocation across Product Markets.” Review of Finance 30, no. 4: 1295–1329.
• Pettit, Philip. 1997. Republicanism: A Theory of Freedom and Government. Oxford: Oxford University Press.
• Potter, Tristan, André Kurmann, and Bart Hobijn. 2026. “The Impact of Noncompetes on Wages and Job Tenure: New Evidence from NLSY Data.” AEA Papers and Proceedings 116: 273–277.
• Starr, Evan. 2026. “The Economics of Noncompete Clauses.” Journal of Economic Perspectives 40, no. 1: 139–166.
• Starr, Evan P., J. J. Prescott, and Norman D. Bishara. 2021. “Noncompete Agreements in the US Labor Force.” Journal of Law and Economics 64, no. 1: 53–84.

Publication note: Independent public research essay, published for criticism and revision. It is not presented as peer-reviewed scholarship.