Pay to Leave: Stay-or-Pay Contracts, the New Regulatory Wave, and How Leaders Can Retain Talent Through Investment Rather Than Restraint
Download the article here:
Listen to a review of this article:
Abstract: Stay-or-pay contracts require workers to repay training costs, bonuses, or other sums if they leave before a set date. Once a niche practice, they have spread into nursing, trucking, aviation, retail, and other sectors, and regulators have responded. Connecticut extended its ban on employment promissory notes to all employers on October 1, 2026; New York enacted its Trapped at Work Act; and California, after passing a broad ban in 2025, delayed it to January 2027 while adding exceptions. This article examines stay-or-pay provisions, including training repayment agreement provisions, alongside the related evidence on noncompete agreements. It reviews economic theory on who should pay for training, empirical research on mobility restraints and wages, and documented organizational examples, from state enforcement actions against hospital and retail employers to investment-based education benefits. It then outlines five evidence-based responses for leaders who want to protect training investments without trapping workers, and three long-term capabilities for retention built on development rather than restraint.
Imagine starting a new job as a nurse, a truck driver, a pet groomer, or a regional airline pilot, and discovering that if you leave within two years you will owe your employer thousands of dollars. Not for anything you broke or took, but for the training the employer required you to complete in order to do the job. For many workers in the United States, this is not hypothetical. It is the logic of the stay-or-pay contract, and it has become one of the most contested employment practices of the decade.
The regulatory ground shifted again in the past few weeks. On September 30, 2026, California Governor Gavin Newsom signed Assembly Bill 1697, an urgency statute that took effect immediately. It postponed California's sweeping ban on stay-or-pay provisions, originally enacted through Assembly Bill 692 and designed to apply from January 1, 2026, so that it now applies to contracts entered into on or after January 1, 2027, while adding new exceptions for certain retention bonuses, paid time off advances, financial services inducements, and government grant-funded programs (Zadikany & Beglaryan, 2026). The next day, October 1, 2026, Connecticut's long-standing prohibition on employment promissory notes expanded from employers with 26 or more employees to all employers (Dorey et al., 2026; Seyfarth Shaw, 2026). New York's Trapped at Work Act, signed in December 2025 and amended in early 2026, is scheduled to take effect roughly a year after enactment, although commentators differ on the exact date (Vorys, 2026).
These state actions follow a turbulent period at the federal level. The Federal Trade Commission (FTC) adopted a near-total ban on noncompete agreements in April 2024, estimating that about 30 million workers, nearly one in five, were bound by them (Federal Trade Commission [FTC], 2024). A federal court set that rule aside in August 2024 before it took effect, and in September 2025 the FTC voted to abandon its appeals and pursue case-by-case enforcement instead (Jones et al., 2025; Maurer, 2024). The National Labor Relations Board's general counsel declared most stay-or-pay provisions presumptively unlawful in October 2024, a position that was rescinded in February 2025 (MacDonald & Carter, 2025). The result is a patchwork: federal pressure has eased, but state law is tightening, and state attorneys general and private plaintiffs have become active.
Why does this matter for leaders beyond legal compliance? Because stay-or-pay contracts reveal a deeper strategic choice. Organizations can try to keep people by making it costly to leave, or by making it worthwhile to stay. The first approach is cheaper in the short term and easier to implement. The second requires investment, management capability, and patience. The research reviewed here suggests that the second approach tends to produce better outcomes for workers and, in many cases, for organizations as well. This article explains what stay-or-pay provisions are, how prevalent they have become, what the evidence says about their effects, and how leaders can protect legitimate training investments while building retention on a sounder foundation.
The Stay-or-Pay Landscape
Defining Stay-or-Pay Provisions, TRAPs, and Related Restraints
A stay-or-pay provision is any contractual term that requires a worker to pay their employer, or a third party, if they leave employment before a specified time. The most common form is the training repayment agreement provision, often abbreviated as TRAP, which obliges a worker to repay the claimed cost of training if they separate early (Consumer Financial Protection Bureau [CFPB], 2023). Other variants include repayment of signing or retention bonuses, relocation assistance, or tuition support, and flat "quit fees" that are not tied to any specific cost. Connecticut's statute uses the term employment promissory note to describe an instrument requiring an employee to pay the employer if employment ends before a certain period, and prohibits requiring one as a condition of employment (Seyfarth Shaw, 2026).
Stay-or-pay provisions are closely related to, but legally distinct from, noncompete agreements, which restrict where a worker can work after leaving rather than imposing a payment. Both function as mobility restraints: they raise the cost of changing jobs. Advocates have argued that TRAPs can serve as substitutes for noncompetes in jurisdictions where noncompetes are restricted (Student Borrower Protection Center, 2022). Because the economic logic of the two practices is similar, and because the empirical research on noncompetes is far more developed, this article draws on both literatures, while being careful to note where findings apply to one but not necessarily the other.
The distinction between general and firm-specific training is central to the debate. In Becker's (1964) classic human capital framework, general training raises a worker's productivity at many employers, while specific training raises productivity only at the current employer. In competitive labor markets, Becker argued, workers would bear the cost of general training (because they capture its value) and firms would bear the cost of specific training. Training repayment agreements are often defended using this logic: if the training is transferable, the worker benefits and should share the cost. The problem, as regulators and researchers have noted, is that much of the training covered by TRAPs is in fact employer-specific onboarding, or is required to do the job at all, and the claimed costs are frequently not transparent (CFPB, 2023).
Acemoglu and Pischke (1998, 1999) offered an important refinement. They showed that when labor markets are imperfect, for example because the current employer knows more about a worker's abilities than rival employers do, firms can have an incentive to pay for general training themselves. In this view, firm-sponsored general training is a normal feature of real labor markets rather than an anomaly that must be corrected by repayment clauses. Their model also implies two possible equilibria: one with high turnover and little training, and one with low turnover and substantial training. Which equilibrium a firm ends up in depends partly on choices it makes about how to treat its workers.
Prevalence, Drivers, and Distribution
Precise prevalence data on stay-or-pay provisions are limited, but the available evidence suggests they are far from rare. The CFPB reported that a survey by Cornell University's Survey Research Institute found nearly 10% of US workers were covered by a training repayment agreement in 2020 (CFPB, 2023). The Student Borrower Protection Center (2022) estimated that major employers using TRAPs operate in market segments employing more than one in three private-sector workers, though this figure describes the reach of those employers rather than the number of workers actually bound.
Noncompete agreements are better documented. In a nationally representative survey of 11,505 labor force participants, Starr, Prescott, and Bishara (2021) found that about 18% of workers were currently bound by a noncompete and that about 38% had signed one at some point. Noncompetes were common even among lower-wage workers and in states that do not enforce them. Few workers reported negotiating over them, and roughly a third received the noncompete only after accepting the job offer.
Stay-or-pay provisions are concentrated in particular sectors. The CFPB's 2023 review documented their use in nursing, trucking, retail, aviation, and other industries, along with reported amounts. Nurses reported owing between $5,000 and $50,000; commercial driver training was charged at more than $6,000 in cases where the carrier paid schools between $1,400 and $2,500; and a pet groomer reported owing $5,000 plus $500 for tools (CFPB, 2023). In nursing, the CFPB cited a National Nurses United survey showing that TRAP exposure was concentrated among early-career nurses: 44.8% of nurses with five or fewer years of experience and 45.3% of those with six to ten years reported exposure, compared with 24.3% of those with 11 to 20 years and 9.4% of those with more than 20 years (CFPB, 2023).
Several drivers explain the spread of these provisions. The first is labor scarcity in high-training occupations. Health care systems facing nursing shortages, carriers facing driver shortages, and regional airlines facing pilot shortages all invest in training new entrants and fear losing them to competitors who did not bear the cost. The second is the regulation of noncompetes: as states restricted noncompete agreements, particularly for lower-wage workers, repayment provisions offered an alternative way to raise the cost of leaving (Student Borrower Protection Center, 2022). The third is the shift of training costs onto workers more broadly, a pattern that critics describe as employer-driven debt (CFPB, 2023).
The distribution of the burden matters. Stay-or-pay provisions tend to fall hardest on early-career workers, lower-wage workers, and those with few alternatives, exactly the groups least able to absorb a sudden debt. The nursing data illustrate this clearly. And because many workers do not read or understand the provisions at the time of hiring, as Starr et al. (2021) documented for noncompetes, the restraint may only become visible when a worker tries to leave.
The regulatory response has unfolded unevenly, and its shape matters for leaders operating across jurisdictions. Connecticut has prohibited requiring employment promissory notes as a condition of employment since the 1980s, but until October 2026 the rule applied only to employers with 26 or more employees and offered no private damages remedy (Seyfarth Shaw, 2026). Colorado adopted limits on training repayment in 2022 as part of a broader noncompete reform, allowing repayment only for training distinct from normal on-the-job training and requiring the amount owed to decline over two years (Vickles & Klein, 2025). California enacted the most sweeping prohibition in 2025, voiding most contract terms that require a worker to pay a debt, fee, or penalty upon leaving and creating a private right of action with statutory damages of at least $5,000 per worker; AB 1697 then delayed its operation to contracts formed from January 2027 and broadened its exceptions (Zadikany & Beglaryan, 2026). New York's law bans employment promissory notes while permitting narrowly structured tuition and bonus repayment, enforced through Department of Labor penalties rather than private lawsuits (Vorys, 2026). At the federal level, the direction has been the opposite: the FTC's noncompete rule never took effect, and the NLRB general counsel's position that most stay-or-pay provisions violate federal labor law was withdrawn within months (Jones et al., 2025; MacDonald & Carter, 2025). The practical effect is that the rules a worker faces depend heavily on where they live and work.
Organizational and Individual Consequences
Organizational Performance Impacts
From an employer's perspective, stay-or-pay provisions appear to offer two benefits: they protect training investments and they reduce early turnover. Both claims deserve scrutiny.
On turnover, the evidence on mobility restraints suggests they do reduce movement, but at a cost. When Michigan inadvertently began enforcing noncompetes in 1985, Marx, Strumsky, and Fleming (2009) found that inventor mobility declined, with the largest effects among inventors with firm-specific skills and narrow specializations. Conversely, when Hawaii banned noncompetes for technology workers in 2015, Balasubramanian et al. (2022) found that worker mobility rose by about 11% and new-hire wages by about 4%. Restraints work in the narrow sense of keeping people in place. The question for leaders is whether keeping people in place is the same as keeping them engaged and productive.
Job embeddedness research helps answer that question. Mitchell, Holtom, Lee, Sablynski, and Erez (2001) identified three dimensions that keep people in their jobs: links to colleagues and the community, fit between the person and the job and organization, and sacrifice, meaning what a person would give up by leaving. A stay-or-pay provision manufactures sacrifice artificially, without strengthening links or fit. A worker held in place only by a debt is embedded in a narrow and fragile sense. They may stay, but there is little reason to expect discretionary effort, and once the repayment period ends, the restraint disappears all at once.
There is also legal and reputational risk, which has grown considerably. HCA Healthcare's experience illustrates the stakes. Through its Specialty Training Apprenticeship for Registered Nurses (StaRN) residency program, newly hired nurses were required to complete training and to agree to repay part of its cost, reported to average about $10,000, if they left within two years; debts of nurses who left early were sent to collections (Attridge, 2025). In July 2025, the attorneys general of California, Colorado, and Nevada announced a settlement under which HCA agreed to stop enforcing and collecting on the provisions, pay restitution to nurses who had made payments, and pay penalties reported at about $2.9 million across the three states (Alvarez, 2025). HCA stated that most of its hospitals had dropped the agreements in early 2022 (Alvarez, 2025).
PetSmart faced similar scrutiny over training repayment provisions for pet groomers. A class action filed in 2022 alleged that groomers faced training debts exceeding $5,000 (Student Borrower Protection Center, 2022). In July 2025, Colorado's attorney general sued the company, alleging that recruits were promised free training and later charged up to $5,500, with debts pursued by collectors (Student Borrower Protection Center, 2025). The matter was reported to have settled in late 2025 for $225,000 and an end to enforcement of the provisions (Vickles & Klein, 2025).
In trucking, CRST Expedited faced a class action alleging that driver trainees accumulated debts for tuition and lodging and faced weekly deductions with interest and an additional $6,500 charge if their contracts ended within ten months; in December 2020, the parties reached a preliminary settlement of $12.5 million in a case centered on trainee wage claims (Fisher, 2020). In aviation, a putative class action filed in November 2025 alleges that Frontier Airlines required pilots who left or were terminated within two years to repay up to $59,190 for training (Towards Justice, 2025). These are allegations rather than findings, but they show that the litigation risk extends across industries.
Beyond legal exposure, stay-or-pay provisions can damage an employer's brand in labor markets where reputation travels quickly. Workers talk, particularly in occupational communities such as nursing and trucking, and media coverage of TRAP disputes has been extensive. An employer known for trapping workers may find it harder to attract the best candidates, raising recruitment costs that offset any savings from reduced turnover. There is also a subtler selection effect worth considering. Manchester (2012) found that education benefits retained workers largely by attracting people who were inclined to stay. By the same logic, though this has not been tested directly for TRAPs, a contract that penalizes leaving may be most acceptable to candidates who have the fewest alternatives, and least acceptable to the high performers with the most options. If so, stay-or-pay provisions could quietly shift the composition of an applicant pool in the wrong direction.
Individual Wellbeing and Stakeholder Impacts
For workers, the most direct effect of mobility restraints is on earnings. Lipsitz and Starr (2022) studied Oregon's 2008 ban on noncompetes for hourly workers and found that it raised average hourly wages by about 2% to 3%, along with increased job-to-job mobility. Johnson, Lavetti, and Lipsitz (2025), examining changes in noncompete enforceability across states over time, found that stronger enforceability reduced earnings and mobility, that effects spilled over to workers not bound by noncompetes, and that greater enforceability widened earnings gaps by gender and race. While these studies concern noncompetes rather than TRAPs specifically, they suggest that restraints on mobility tend to suppress the wage gains that come from being able to change jobs. Job changes are one of the main ways workers, especially early in their careers, achieve wage growth. A restraint that delays or prevents a move during those years may have effects that compound over a working life. The finding by Johnson et al. (2025) that enforceability widened gender and racial earnings gaps suggests that the burden of restraints is not evenly shared.
Financial strain is a second concern. The CFPB (2023) documented cases in which workers who left early were referred to debt collectors, and in which the claimed training costs appeared to exceed what the employer actually paid. For an early-career worker earning modest wages, a debt of several thousand dollars can be destabilizing. The CFPB also reported that, among registered nurses with TRAPs, about half knew they were taking on a debt obligation before accepting or continuing the job, which implies that the other half did not.
A third concern is voice. When a worker cannot afford to leave, they also lose the implicit leverage that the possibility of leaving provides. The California Nurses Association survey cited in reporting on the HCA settlement found that about 40% of new nurses nationwide had signed such provisions, and that about one-third said they felt restrained from raising concerns about unsafe staffing or working conditions (Attridge, 2025). In health care, aviation, and trucking, where safety concerns affect patients, passengers, and the public, the suppression of voice is not only a workforce issue but a stakeholder issue.
Finally, there are fairness concerns rooted in contract formation. Harris (2021) argued that training repayment agreements should be analyzed under the doctrine of unconscionability, given the imbalance in bargaining power and information between employers and entry-level workers, and proposed that worker training be designed and funded jointly by workers or their representatives, employers, and government. Whether or not courts adopt that framing, the argument captures why these provisions feel unfair to many workers: they are presented as standard terms, often with little explanation, at a moment when the worker has little ability to negotiate.
Evidence-Based Organizational Responses
Leaders face a real problem. Training is expensive, and losing newly trained workers to competitors is painful. But the evidence suggests that the most effective solutions lie in changing the economics of staying rather than the penalty for leaving. Table 1 summarizes five responses.
Table 1. Evidence-based responses to stay-or-pay risks
Response | Evidence Base | Key Practices | Illustrative Examples |
Audit and redesign repayment terms | CFPB (2023); Zadikany & Beglaryan (2026); Vorys (2026); Harris (2021) | Inventory all repayment clauses; limit to actual cost of transferable credentials; prorate; no repayment on termination without cause | HCA's 2025 multistate settlement; PetSmart's Colorado case |
Replace clawbacks with earn-in incentives | Mitchell et al. (2001); Zadikany & Beglaryan (2026) | Pay retention bonuses at the end of a period; vest over time; separate agreements | California's revised bonus exception; Frontier Airlines litigation as a contrast |
Prepay education without clawbacks | Manchester (2012); Benson et al. (2004); Acemoglu & Pischke (1998) | Upfront tuition; no tenure lock; transferable credentials; broad eligibility | Walmart Live Better U; Chipotle Cultivate Education; Starbucks College Achievement Plan |
Pair development with internal mobility | Kraimer et al. (2011); Benson et al. (2004); Hom et al. (2017) | Career paths after training; promotion on completion; manager sponsorship | Chipotle's crew-to-management pathway; Amazon Career Choice |
Disclose remaining restraints and limit noncompetes | Starr et al. (2021); Lipsitz & Starr (2022); Balasubramanian et al. (2022); FTC (2025) | Disclose at offer; plain language; drop noncompetes for most roles | FTC's 2025 Gateway Services order; CRST trucking settlement |
Audit and Redesign Repayment Terms
The first step is to know what an organization's contracts actually say. Many employers have repayment clauses scattered across offer letters, training agreements, bonus agreements, and policy handbooks, often drafted at different times by different teams. Regulators have focused on several recurring problems: claimed costs that exceed actual costs, repayment demands for training that is mandatory or employer-specific, lack of proration, collection actions against workers who were terminated, and opaque terms presented at hiring (CFPB, 2023). The emerging state laws codify many of these concerns. California's framework, as amended by AB 1697, permits repayment for transferable credentials only under defined conditions and allows certain retention bonus repayment only where the terms are in a separate agreement, the worker is told they may consult counsel and given time to do so, repayment is prorated and interest-free over no more than two years, and repayment applies only on resignation or termination for misconduct (Zadikany & Beglaryan, 2026). New York's amended law similarly permits repayment for transferable credential tuition only if the agreement is separate, the amount does not exceed cost, repayment is prorated, there is no acceleration, and repayment is not required after termination other than for misconduct (Vorys, 2026). Colorado has allowed recovery of training costs only for training distinct from normal on-the-job training, within two years of completion, and in amounts that decline over that period (Vickles & Klein, 2025).
Effective approaches:
Contract inventory: Catalogue every provision across the organization that requires a worker to pay anything on separation, including training, bonuses, relocation, tuition, and equipment.
Cost substantiation: Limit any repayment to the documented, actual cost of training, and never charge for onboarding, mandatory compliance training, or job-specific instruction.
Transferable credentials only: Reserve any repayment arrangement for training that yields a credential with recognized value at other employers.
Proration and caps: Reduce the amount owed steadily over a short period, without interest, and set a reasonable ceiling.
Termination protection: Do not seek repayment from workers who are laid off or dismissed for reasons other than misconduct.
Jurisdiction mapping: Track where workers are located and apply the most protective applicable standard as a baseline.
HCA Healthcare's experience shows how quickly exposure can accumulate. The StaRN program combined several features regulators have criticized: training required for the job, repayment tied to a two-year commitment, and collection actions against nurses who left (Attridge, 2025). The 2025 settlement required HCA to stop enforcing the provisions and to cancel outstanding debts in addition to paying penalties (Alvarez, 2025). For any organization with similar programs, the lesson is to review them before a regulator does. A practical way to begin is to ask three questions of every repayment clause: Is this training required to do the job? Does it produce a credential the worker could use elsewhere? Can we document what it actually cost us? A clause that fails the first test or the last two is a candidate for removal, not revision.
PetSmart's case adds a further dimension. The Colorado complaint alleged that recruits were told training would be free and later faced substantial repayment claims (Student Borrower Protection Center, 2025). Whatever the merits of individual allegations, the gap between how a program is described in recruitment and what the contract requires is a recurring source of legal and reputational risk.
A common pitfall: Organizations sometimes respond to new laws by updating contracts only in the affected state, leaving older provisions in place elsewhere. Given the speed at which states are acting, and the reputational cost of being identified as a TRAP user anywhere, a single enterprise standard is usually safer and simpler.
Replace Clawbacks With Earn-In Incentives
If the goal is to encourage people to stay through a critical period, there is a simple alternative to clawing money back from those who leave: pay people more for staying. An earn-in retention incentive is paid at the end of a retention period, or vests in installments, so that nothing is ever owed back. The behavioral effect on retention can be similar, because leaving early means forgoing the payment, but the legal and relational character is entirely different. The worker experiences it as a reward rather than a debt.
In job embeddedness terms, an earn-in incentive still creates sacrifice, but it is sacrifice of a future gain rather than an imposed loss (Mitchell et al., 2001). California's revised framework recognizes the difference: it treats certain retention bonuses more favorably where workers can choose to defer receipt and where repayment terms are limited (Zadikany & Beglaryan, 2026). The safest design avoids repayment altogether. Earn-in designs also change the conversation between managers and employees. A manager discussing a retention bonus that will be paid next year is talking about a shared future; a manager reminding an employee of a debt they will owe if they leave is talking about a threat. The financial mechanics can be similar, but the relationship they create is not.
Effective approaches:
Pay at the end: Structure retention bonuses to be paid after the retention period rather than upfront with a clawback.
Graduated vesting: Pay in installments, for example quarterly or annually, so that the incentive to stay is continuous rather than a single cliff.
Separate and voluntary: If any repayment terms are used, keep them in a separate agreement that the worker can decline without losing the job.
Market-relevant amounts: Size incentives relative to what competitors offer, since the point is to make staying attractive, not merely leaving costly.
Combine with non-financial anchors: Pair incentives with mentoring, cohort experiences, and clear progression, which strengthen links and fit.
Regional aviation shows why the distinction matters. In 2023, Southern Airways Express was reported to be suing about 80 former pilots under training repayment agreements, seeking up to $20,000 each, while the airline's chief executive disputed the pilots' allegations and pointed to competition for pilots from higher-paying airlines (Schlappig, 2023). In 2025, a lawsuit against Frontier Airlines alleged repayment obligations of up to $59,190 for pilots leaving within two years (Towards Justice, 2025). Both cases concern allegations, but they illustrate how clawback structures can turn a retention tool into a source of conflict and litigation. In an industry where pilot shortages have been widely reported, an earn-in design would offer a retention incentive without creating a debt that workers must contest.
A common pitfall: Some organizations relabel a clawback as a "bonus" without changing its substance, for example by paying an upfront sum that must be repaid if the worker leaves. Regulators look at substance, and workers do too.
Prepay Education Without Clawbacks
The most counterintuitive finding in this literature is that education benefits can improve retention even when workers are free to leave with the credentials they gain. Manchester (2012), studying administrative data from a large employer, found that a tuition reimbursement program increased retention, and that most of the effect came from sorting: the program attracted workers who were more inclined to stay in the first place. Benson, Finegold, and Mohrman (2004) found that participation in tuition reimbursement reduced turnover while employees were taking courses; after employees earned graduate degrees, turnover rose, unless they were promoted. Taken together, these findings suggest that education benefits are a powerful recruitment and retention tool, provided they are paired with opportunities to use the new skills.
This is consistent with Acemoglu and Pischke's (1998) theory: in real labor markets, firms can benefit from paying for general training, especially when it allows them to identify and develop capable workers. The low-turnover, high-training equilibrium is achievable, but it requires the employer to make the initial investment.
Effective approaches:
Upfront payment: Pay tuition directly rather than reimbursing workers later, which removes the cash-flow barrier for lower-wage employees.
No tenure lock: Avoid requiring workers to repay tuition if they leave, relying on sorting and engagement rather than restraint.
Transferable credentials: Offer programs that lead to recognized degrees and certificates, which increase their attractiveness to candidates.
Broad eligibility: Make benefits available to frontline and part-time workers, not only salaried staff, after a short qualifying period.
Measure outcomes: Track participation, completion, retention, and internal promotion to evaluate return on investment.
Several large employers have built programs on this model. Walmart's Live Better U program, launched in 2018, covers tuition through an education partner; reporting on a Lumina Foundation study indicated that participants were substantially more likely to stay with the company than non-participants (Parisi, 2023). Chipotle's Cultivate Education program offers tuition assistance and debt-free degrees in selected fields; the company reported in 2021 that participating students had markedly higher retention and that crew members in the program were much more likely to move into management roles (Chipotle Mexican Grill, 2021). Starbucks's College Achievement Plan covers tuition for an online degree through Arizona State University; company leaders have reported more than 16,000 graduates and higher promotion rates among participants (Dungan, 2025). An earlier study conducted for the Lumina Foundation found that Cigna's education reimbursement program produced savings that exceeded its costs over three years, largely through improved retention (Taylor, 2016).
These figures are largely self-reported by the companies or by organizations that promote education benefits, and they cannot establish causation, particularly given the sorting effects that Manchester (2012) identified. But they are consistent with the academic evidence, and they show that large employers in retail, food service, and insurance have chosen to invest in transferable education without relying on repayment clauses. For smaller employers, the same logic can be applied at a smaller scale, for example by partnering with community colleges, apprenticeship programs, or industry training consortia that spread costs across employers and public funding sources. California's and New York's frameworks both make room for apprenticeships and government-supported training, which suggests that pooled models are likely to remain viable under the new rules (Vorys, 2026; Zadikany & Beglaryan, 2026).
A common pitfall: Organizations sometimes launch education benefits but make them hard to use, through complex reimbursement processes, narrow eligibility, or managers who discourage participation because of scheduling demands. Low utilization then becomes evidence that the program does not work.
Pair Development With Internal Mobility
Investing in people's skills raises an obvious risk: they may take those skills elsewhere. The research is clear that this risk depends heavily on what happens after training. Kraimer, Seibert, Wayne, Liden, and Bravo (2011) found that organizational support for development reduced turnover when employees perceived strong career opportunities within the organization, but was associated with higher turnover when perceived opportunities were low. Benson et al. (2004) similarly found that promotion after degree completion was the key factor in retaining newly credentialed employees. Hom, Lee, Shaw, and Hausknecht (2017), reviewing a century of turnover research, emphasized that understanding why people stay is as important as understanding why they leave.
The implication is direct. Training without a path forward can accelerate exits. Training paired with visible internal opportunities can strengthen retention. Recent data suggest this pairing is often missing: Workday's 2026 Global Workforce Report found that internal moves declined at 57% of employers and that promotion rates were essentially flat, while 79% of workers said they knew which skills they needed but only 66% said their employer helped them develop those skills (Workday, 2026).
Effective approaches:
Defined pathways: Map how training programs connect to specific next roles, with clear criteria and timelines.
Promotion on completion: Where feasible, link completion of significant training or credentials to eligibility for advancement or pay progression.
Internal first: Give trained employees early access to internal openings before external recruitment.
Manager sponsorship: Hold managers accountable for developing and moving talent, not hoarding it.
Visible success stories: Share examples of employees who advanced after development, making the pathway credible.
Chipotle's reported finding that crew members enrolled in its education program were far more likely to move into management illustrates the value of linking development to advancement (Chipotle Mexican Grill, 2021). Amazon's Career Choice program, which prepays tuition for eligible hourly employees after 90 days of employment, reported that more than 25,000 employees joined in a six-month period in 2022, as part of a broader commitment of $1.2 billion to upskill 300,000 employees (Amazon, 2022). Programs of this kind are most valuable when participants can see where the training leads, whether inside the company or beyond it. In health care, nurse residency programs offer a natural opportunity to apply this principle. A residency that leads to a defined first-year progression, specialty training opportunities, and recognized milestones gives new nurses reasons to stay that a repayment clause cannot provide, and it does so without the legal and reputational exposure that HCA encountered.
A common pitfall: Organizations invest in training but leave internal mobility processes unchanged, so that trained employees find it easier to get a promotion elsewhere than internally. Kraimer et al. (2011) suggest this combination is the worst of both worlds.
Disclose Remaining Restraints and Limit Noncompetes
Some restraints may remain justified, for example protecting genuine trade secrets or client relationships for senior employees. But the research suggests that how restraints are introduced matters greatly. Starr et al. (2021) found that many workers received noncompetes only after accepting job offers and that few negotiated them. Lipsitz and Starr (2022), Balasubramanian et al. (2022), and Johnson et al. (2025) together indicate that broad restraints on lower-wage and technical workers reduce wages and mobility. The FTC's 2025 action against Gateway Services, a pet cremation company, illustrates the current federal enforcement posture: the agency's proposed consent order would release nearly 1,800 workers from noncompete agreements that the company had required of almost all employees (FTC, 2025).
Effective approaches:
Early disclosure: Present any restraint, including repayment terms and noncompetes, in the job posting or offer, not after acceptance.
Plain language: Explain in simple terms what the restraint requires, how much could be owed, and under what circumstances.
Narrow scope: Limit noncompetes to the small number of roles where they protect genuine interests, and rely on confidentiality and trade secret protections for others.
Regular review: Reassess the need for each restraint periodically and remove those that are no longer justified.
Worker voice: Invite feedback from employees and their representatives on contract terms.
Gateway Services offers a cautionary example. According to the FTC, the company had required nearly all employees to sign one-year nationwide noncompete agreements from 2019 onward; the agency's order freed those workers and limited the company's use of non-solicitation provisions (FTC, 2025). The breadth of the restraint, applied to almost every role regardless of access to sensitive information, was central to the case.
The CRST trucking litigation illustrates the cumulative effect of multiple restraints. Driver trainees reportedly faced tuition and lodging debts, interest charges, and a substantial early departure fee (Fisher, 2020). When several restraints stack together, a worker's practical ability to leave can become very limited, which is precisely what regulators now scrutinize.
A common pitfall: Organizations frequently use the same contract template for all roles, applying restraints designed for senior executives to frontline workers who have no access to trade secrets. This both increases legal risk and signals distrust. Disclosure has value even where a restraint is lawful. A worker who understands at the outset exactly what they are agreeing to, and why, is less likely to experience the restraint as a betrayal later, and an employer that is willing to explain its terms plainly is more likely to have terms worth explaining.
Building Long-Term Retention Capability
The responses above help organizations move away from stay-or-pay practices. Building durable retention requires deeper capabilities.
Embeddedness Through Investment Rather Than Sacrifice
The job embeddedness framework suggests that people stay when they are connected to others, when they fit their work and organization, and when they would give up something valuable by leaving (Mitchell et al., 2001). Stay-or-pay provisions attempt to work only on the third dimension, and only through an imposed cost. Durable retention comes from building all three dimensions through investment: mentoring and cohort-based learning that create links, job design and career development that improve fit, and growing pay, skills, and opportunity that make staying genuinely valuable. Kraimer et al. (2011) identified supportive relationships, including with supervisors and mentors, as antecedents of perceived organizational support for development. Organizations that build embeddedness through investment create a form of retention that does not expire when a repayment period ends. Practically, this means treating the first two years of employment, the period most stay-or-pay provisions target, as an investment window rather than a lock-in period. Structured onboarding, assigned mentors, cohort-based training, regular development conversations, and early stretch assignments all build the connections and fit that keep people engaged beyond any contractual horizon. Hom et al. (2017) noted the growing attention in turnover research to why people stay, and the answer is rarely that they cannot afford to leave.
A Low-Turnover, High-Training Equilibrium
Acemoglu and Pischke's (1998) model suggests that organizations can settle into one of two patterns: high turnover with little training, or low turnover with substantial training. Stay-or-pay provisions try to achieve low turnover without changing the underlying pattern, holding workers in place rather than making the organization the best place to use their skills. The alternative is to become an employer where training leads to growth, where skills are used and rewarded, and where workers choose to stay because their prospects are better inside than outside. This requires sustained investment and leadership attention, but it shifts the organization into the more favorable equilibrium. The benefit compounds over time: an employer known for developing people attracts candidates who value development, reinforcing the sorting effects that Manchester (2012) documented.
Restraint-Light, Compliance-Ready Contract Design
The legal landscape will keep changing. Connecticut, California, New York, and Colorado have all acted, other states have considered or adopted restrictions, and state attorneys general and private plaintiffs have demonstrated willingness to bring cases (Vickles & Klein, 2025). Organizations need a standing capability to monitor developments, review contract templates regularly, and apply consistent standards across jurisdictions. The most efficient approach is to adopt a restraint-light default: minimal repayment obligations, limited noncompetes, clear disclosure, and strong protection for workers who are terminated. An organization that already meets the most protective standard does not need to scramble each time a new law passes. This capability should sit jointly with human resources, legal, and business leaders, because contract terms are not only legal documents but also expressions of how the organization views its people. A periodic review that asks whether each restraint is still necessary, still lawful, and still consistent with the organization's values can prevent the gradual accumulation of terms that no one would design deliberately today. Equally important is a governance norm that any new restraint must be justified by a specific, documented business interest and reviewed periodically.
Conclusion
Stay-or-pay contracts promise employers a simple answer to a hard problem: how to protect investments in people who are free to leave. The evidence suggests the answer is less effective and more costly than it appears. Mobility restraints tend to reduce wages and mobility, they can suppress worker voice in safety-critical industries, and they now carry significant legal and reputational risk as states tighten the rules. At the same time, the evidence on education benefits, development, and internal mobility shows that organizations can protect their investments by making staying worthwhile.
The regulatory wave of 2025 and 2026 is best understood not as a compliance burden but as a prompt to rethink retention strategy. Leaders who replace restraint with investment are likely to find that they retain people who want to be there, which is the kind of retention that matters most.
Practical takeaways for leaders:
Know your contracts: Inventory every provision that requires workers to pay on separation, and review it against the most protective applicable law.
Never charge for the job itself: Do not seek repayment for onboarding, mandatory, or employer-specific training.
Prefer earn-in to clawback: Reward staying with incentives paid over time rather than penalizing leaving with debts.
Invest upfront in transferable education: Prepay tuition without tenure locks and let sorting and engagement do the retention work.
Connect training to careers: Pair every major development investment with a visible internal pathway.
Disclose and narrow: Present any remaining restraint early and plainly, and reserve noncompetes for roles that genuinely require them.
Build embeddedness: Strengthen links and fit, not just the cost of leaving.
Research Infographic
References
Acemoglu, D., & Pischke, J.-S. (1998). Why do firms train? Theory and evidence. The Quarterly Journal of Economics, 113(1), 79–119.
Acemoglu, D., & Pischke, J.-S. (1999). Beyond Becker: Training in imperfect labour markets. The Economic Journal, 109(453), F112–F142.
Alvarez, K. (2025, August 12). HCA to pay $2.9 million to settle allegations of illegal TRAP agreements. Staffing Industry Analysts.
Amazon. (2022, June 28). Over 25,000 employees joined Amazon's Career Choice in the last 6 months.
Attridge, M. (2025, July 24). Nurses win $1.5 million settlement over hospital training debt scheme. Courthouse News Service.
Balasubramanian, N., Chang, J. W., Sakakibara, M., Sivadasan, J., & Starr, E. (2022). Locked in? The enforceability of covenants not to compete and the careers of high-tech workers. Journal of Human Resources, 57(S), S349–S396.
Becker, G. S. (1964). Human capital: A theoretical and empirical analysis, with special reference to education. National Bureau of Economic Research.
Benson, G. S., Finegold, D., & Mohrman, S. A. (2004). You paid for the skills, now keep them: Tuition reimbursement and voluntary turnover. Academy of Management Journal, 47(3), 315–331.
Chipotle Mexican Grill. (2021, April 8). Chipotle to offer free degrees in agriculture, culinary and hospitality to employees [Press release].
Consumer Financial Protection Bureau. (2023). Issue spotlight: Consumer risks posed by employer-driven debt.
Dorey, D. R., Laing, L., Lawes, B., & Wieselthier, S. (2026, September 30). Employer guide for workplace laws taking effect in October 2026. SHRM.
Dungan, A. (2025, June 11). How Starbucks boosts retention and engagement with upfront L&D aid. HR Grapevine.
Federal Trade Commission. (2024, April 23). FTC announces rule banning noncompetes [Press release].
Federal Trade Commission. (2025, September 4). FTC takes action to protect workers from noncompete agreements [Press release].
Fisher, T. (2020, December 23). CRST reaches $12.5M settlement in trainee wage lawsuit. Land Line.
Harris, J. F. (2021). Unconscionability in contracting for worker training. Alabama Law Review, 72.
Hom, P. W., Lee, T. W., Shaw, J. D., & Hausknecht, J. P. (2017). One hundred years of employee turnover theory and research. Journal of Applied Psychology, 102(3), 530–545.
Johnson, M. S., Lavetti, K., & Lipsitz, M. (2025). The labor market effects of legal restrictions on worker mobility. Journal of Political Economy, 133(9), 2735–2793.
Jones, M. P., Mee, S. E., Fischer, D. L., Bramble, J. P., & McGreevy, M. (2025, September 10). FTC rule abandoned: FTC to pursue individual enforcement actions. Morgan Lewis.
Kraimer, M. L., Seibert, S. E., Wayne, S. J., Liden, R. C., & Bravo, J. (2011). Antecedents and outcomes of organizational support for development: The critical role of career opportunities. Journal of Applied Psychology, 96(3), 485–500.
Lipsitz, M., & Starr, E. (2022). Low-wage workers and the enforceability of noncompete agreements. Management Science, 68(1), 143–170.
MacDonald, A. T., & Carter, E. (2025, February 14). Acting NLRB general counsel rescinds controversial memoranda. Littler Mendelson.
Manchester, C. F. (2012). General human capital and employee mobility: How tuition reimbursement increases retention through sorting and participation. ILR Review, 65(4), 951–974.
Marx, M., Strumsky, D., & Fleming, L. (2009). Mobility, skills, and the Michigan non-compete experiment. Management Science, 55(6), 875–889.
Maurer, R. (2024, August 20). Texas judge strikes down FTC noncompete ban. SHRM.
Mitchell, T. R., Holtom, B. C., Lee, T. W., Sablynski, C. J., & Erez, M. (2001). Why people stay: Using job embeddedness to predict voluntary turnover. Academy of Management Journal, 44(6), 1102–1121.
Parisi, T. (2023, June 6). Walmart's Live Better U program turns 5, expands to Canada. HR Brew.
Schlappig, B. (2023, October 9). Southern Airways Express suing pilots who quit, and it's messy. One Mile at a Time.
Seyfarth Shaw. (2026, May 14). Connecticut adopts broad pay-or-stay prohibition.
Starr, E. P., Prescott, J. J., & Bishara, N. D. (2021). Noncompete agreements in the US labor force. The Journal of Law and Economics, 64(1), 53–84.
Student Borrower Protection Center. (2022). Trapped at work: How big business uses student debt to restrict worker mobility.
Student Borrower Protection Center. (2025). Colorado attorney general sues PetSmart to halt predatory training debt scheme.
Taylor, T. (2016, July 19). New study supports a strong ROI for companies that invest in employee tuition. HR Dive.
Towards Justice. (2025, November 7). Towards Justice sues Frontier Airlines over training repayment agreement provision that traps pilots into employer-driven debt [Press release].
Vickles, H. F., & Klein, H. F. (2025, December 3). Stay or pay? States are saying "nay": What employers should know about emerging restrictions on training repayment agreements. Venable LLP.
Vorys. (2026, March 2). Payback policies pivot, part II: New York's Trapped at Work Act amendments signed into law.
Workday. (2026, October 5). Workday global workforce report: AI is rewriting jobs more than it's cutting them [Press release].
Zadikany, R., & Beglaryan, S. (2026, October 2). California delays "stay or pay" law to January 1, 2027 and expands key exceptions: An update on AB 1697. Mayer Brown.
Jonathan H. Westover, PhD, Chief Research Officer (Nexus Institute for Work and AI); Co-Founder & Chief Workforce and Learning Officer (Future State University); Founder & CEO (Human Capital Innovations); Professor of Organizational Leadership & Change (UVU). Read Jonathan Westover's executive profile here.
Suggested Citation: Westover, J. H. (2026). Pay to Leave: Stay-or-Pay Contracts, the New Regulatory Wave, and How Leaders Can Retain Talent Through Investment Rather Than Restraint. Human Capital Leadership Review, 40(2). doi.org/10.70175/hclreview.2020.40.2.3
