Washington, D.C. — The United States Department of Labor is reportedly reviewing a growing corporate preference for what internal analysts have begun describing, in memoranda not intended for public circulation, as “high-compliance candidate pools” — a category that, on closer inspection, appears to be defined less by what workers can do than by what they cannot afford to refuse.
The review, which Department officials decline to characterize as a formal investigation, follows what one internal summary calls “a sustained pattern of hiring optimization toward workforce segments with reduced bargaining capacity.” In plain English, employers have noticed that some workers are easier to keep than others, and have begun hiring accordingly.
No one inside the Department will say this on the record. Several people are willing to say it adjacent to the record, into a coffee cup, with the recorder visibly turned off.
The Category That Was Never Supposed to Have a Name
According to documents reviewed by analysts familiar with the matter, the phrase “high-compliance candidate pools” first appeared in a 2023 internal slide deck circulated among recruiting executives at a national logistics firm. It has since spread through staffing consultancies, vendor briefings, and the kind of third-party hiring software that nobody reads the terms of service for.
The category is descriptive, not aspirational. A “high-compliance” worker, in the technical sense used by the firms that traffic in this language, is one who is statistically less likely to:
- negotiate aggressively at the offer stage,
- refuse overtime when asked,
- challenge management decisions in writing,
- file complaints with regulatory agencies, and
- leave unstable conditions before securing replacement employment.
Each of these behaviors, taken alone, is the behavior of a reasonable adult. Taken together, they describe a worker who exists. Taken together and then subtracted from the desired hire profile, they describe what employers are now actively looking for.
One analyst, asked to summarize the dynamic without the assistance of corporate vocabulary, offered the following:
“The less mobility a worker has, the more controllable the relationship becomes.”
The analyst then asked for the sentence to be attributed to “a person familiar with hiring trends,” which is the federal equivalent of asking to be filmed from behind.
The Pattern, Such As It Is
Researchers reviewing applicant tracking data from forty-seven major employers identified a consistent statistical drift. Across industries — warehousing, food service, hospitality, light manufacturing, residential construction, eldercare, the back-office floors of regional banks — recruiters showed measurable preference for candidates demonstrating one or more of the following conditions:
- active financial instability, including recent eviction history or default notices,
- uncertain residency conditions, including expiring visas or pending status reviews,
- recent gaps in coverage for dependents, particularly minor children,
- limited employment alternatives within a fifty-mile commuting radius,
- court-ordered financial obligations, including child support and restitution,
- recent release from custodial facilities, and
- any combination of the above arriving in a single applicant.
The data does not suggest that these candidates were unqualified. In many cases, the opposite was true: candidates in these categories were selected over more credentialed peers. One regional HR director, asked to explain the pattern off the record, said it was “not about who can do the job, it’s about who will still be doing the job in six months without making it weird.”
She paused, and then added, “Look, we don’t write that down.”
The 847 Sample
A working paper circulated among Department analysts examined 2,847 hiring decisions across a single national employer over an eighteen-month period. The analysis found that, after controlling for skills, experience, and interview performance, candidates flagged by the firm’s internal screening software as having “reduced exit optionality” were 34% more likely to receive offers than otherwise identical candidates without such flags.
“Reduced exit optionality,” in the firm’s documentation, is defined as the “statistical likelihood that an employee will remain in current role due to factors external to job satisfaction.” The firm’s legal team has reportedly asked that this definition be retired in favor of something “less legible.”
A Partial Inventory of Optimization
The Department’s reviewing analysts assembled, from internal documentation at six staffing firms, a list of the variables most frequently used to score candidates on retention probability. The list, presented below in approximate descending order of weight, is reproduced from the most heavily annotated copy in circulation:
- distance, in miles, from the candidate’s home address to the nearest competing employer offering comparable compensation,
- number of dependents listed on prior tax filings, weighted by age,
- length of any documented gap between the candidate’s last termination and the date of the current application,
- presence of any court-ordered financial obligation in publicly available records,
- frequency of address changes in the past thirty-six months, with high frequency scored favorably,
- application submission velocity, defined as the number of jobs the candidate had applied to in the seven days preceding the current application, with higher velocity scored favorably,
- self-reported willingness to begin work within seventy-two hours,
- absence of a college degree in any field requiring upward credentialing,
- presence of a college degree in a field with negative wage growth, treated as a stability indicator rather than a qualification, and
- any indicator suggesting the candidate had recently lost employer-sponsored health coverage for self or family.
The list is annotated, in one margin, with the word “Jesus.” The annotation is not in the original document. It was added by the analyst who recovered the file.
The Vendor Pitch Deck
A pitch deck from one of the largest applicant-tracking vendors, presented to a procurement committee at a regional hospital network last summer, listed among its product’s headline features a metric called the Engagement Resilience Score. The score, the deck explained, predicted a candidate’s likelihood of remaining in role through “personal, financial, or operational adversity events.”
The deck included a sample candidate profile, redacted in the version reviewed by the Department, with an Engagement Resilience Score of 94. The accompanying narrative described the candidate as “an outstanding fit for high-stability environments — strong commitment indicators, robust dependency anchors, minimal disruption risk.”
“Dependency anchors,” the deck’s glossary clarified, refers to “external life circumstances that support sustained role tenure.”
The hospital network signed a three-year contract.
The Model That Learned Without Being Told
Researchers at a university computational lab — granted access to an anonymized version of a major staffing platform’s scoring model — found that the model had developed, without supervised instruction, a composite feature it used internally as its strongest predictor. The feature, which researchers nicknamed v847 because it was the 847th variable in the architecture, was constructed from a weighted combination of: housing instability proxies, prior employment gaps, dependent-care obligations, and the candidate’s estimated commute time as a fraction of their stated hourly wage.
The model had, in effect, taught itself to identify financial entrapment, and had learned to treat it as the single most reliable predictor of retention. The lab’s working paper describes this as “an emergent behavior of the training objective.” The model had not been told to find this signal. It had been told to find workers who stay. The signal it found was the one the labor market had already provided.
“We did not program the cruelty,” one of the lab’s lead researchers told the Department. “The cruelty was already in the data. The model simply noticed.”
The Language Shift
The most striking feature of the documents under review is not the practice itself but the vocabulary that has grown up around it. Companies, aware that explicit references to worker vulnerability would create what one consultant memorably called “regulatory friction,” have developed a parallel terminology that describes the same condition without naming it.
Among the terms now in common circulation:
- “reliable retention profile” — formerly: “cannot afford to quit.”
- “low turnover likelihood” — formerly: “no other offers pending.”
- “high schedule flexibility” — formerly: “will accept any shift, including ones invented this morning.”
- “strong commitment indicators” — formerly: “has children, rent, and a court date.”
- “workforce stability orientation” — formerly: “will not file a complaint.”
- “values alignment with operational continuity” — formerly: “will work through the funeral.”
Each phrase, taken individually, sounds like a virtue. Stacked together in a candidate profile, they describe a person whose primary professional asset is the cost of leaving.
The Memo
A document recovered from a midwestern staffing consultancy, titled “Workforce Predictability: A Framework for Sustainable Retention,” contains a single sentence that has now been quoted in three separate Department briefings:
“Stability can be derived from dependency.”
The sentence appears on slide 14, beneath a stock photograph of a smiling warehouse worker. The slide’s footer reads “Internal use only. Not for external distribution.” The consultancy has since clarified that the phrase was “descriptive of observed market behavior, not prescriptive of recommended practice,” a distinction that the Department’s reviewing analyst has reportedly described as “cute.”
The Companion Document
A companion document, distributed at the same internal training session under the title “Operational Anchoring: A Practitioner’s Guide,” elaborates on the framework. The guide instructs hiring managers to identify candidates with what it terms “structural attachment to current geography.” The phrase, in the document’s glossary, is defined as “the presence of non-employment factors that reduce the candidate’s likelihood of relocation, including but not limited to: minor children enrolled in local schools, elderly dependents in local care facilities, ongoing custody arrangements, religious community ties, or active medical treatment.”
The guide is careful to note that these factors should not be inquired about directly. They are, instead, to be “inferred from the totality of the application materials and ambient context.” A separate sidebar advises managers that “a candidate who volunteers any of these details is signaling commitment.”
A two-page appendix lists thirty-seven phrases a candidate might use in conversation that “indicate structural attachment.” The list includes “my mom lives nearby,” “my daughter goes to school down the road,” “I can’t move right now,” and “I really need this job.”
The final phrase is annotated with a single asterisk. The asterisk, at the bottom of the page, reads: “Strongest signal. Cannot be solicited. Must be volunteered.”
The Internal Channel
A leaked log from the internal messaging channel of a national hospitality staffing firm — the channel was named #placement-strategy — contains a thread, sixteen messages long, in which seven recruiters debate the appropriate framing for a candidate’s recent home foreclosure. The candidate had disclosed the foreclosure during a phone screen, apologetically, by way of explaining why she was willing to work the overnight shift.
The recruiters agreed, in the thread, that the disclosure was “positive context.” They debated whether to flag the candidate in the system as HSP (high stability profile) or HCR(high commitment retention). The thread concluded with the senior recruiter writing: “Just go HSP. HCR makes the auditors itchy.”
The log was provided to the Department by a former employee who had downloaded a personal archive of her own messages prior to her termination. She had not, she explained to investigators, intended to expose anyone. She had intended to keep a record of what she had been part of. “I wanted to be able to remember,” she said, “what we were actually saying when we said it.”
The Software Layer
Much of the optimization no longer happens in any room where anyone could be held responsible for it. It happens in the screening software, which was sold to HR departments on the promise of reducing turnover and which has now, through the patient accumulation of training data, learned exactly which signals predict a worker who will not leave.
Vendors do not advertise this capability explicitly. The marketing copy speaks of “cultural fit indicators” and “long-term engagement scoring.” The underlying models, examined by researchers at three separate universities, weight the same handful of variables: distance to nearest competing employer, dependent obligations, housing instability flags from credit bureau data, and length of current gaps in employment.
One model, deployed by a national restaurant chain, was found to assign its highest “retention confidence” score to candidates whose application history included multiple recent submissions to jobs below their stated qualification level. The model had learned, without being told, that desperation is legible.
The Auditor’s Note
A working memorandum circulating within the Department’s Bureau of Labor Statistics — attributed in the document only to “Analyst of Record #847” — observes that the screening systems have not been programmed to discriminate against any protected class. They have been programmed to find workers who will stay. The fact that this produces a workforce drawn disproportionately from populations under financial, residential, or legal strain is, the memo notes, “a feature of the underlying labor market, which the software is faithfully reflecting.”
The memo ends with a single bracketed note in the margin: [Recommend: do not publish.]
The Vendor Marketplace
The optimization the Department’s review describes does not occur in isolation. It is, increasingly, a product category — sold by an identifiable set of vendors, supported by an industry of consultants, conferenced about at trade events, and reviewed in trade publications whose advertising base consists of the firms whose products they review.
A partial inventory of the principal vendors, assembled from publicly available marketing materials and procurement filings, suggests a market of approximately $2.847 billion in annual revenue, growing at a rate the trade press has described as “structurally durable” — a phrase whose semantic content, in this context, is roughly identical to that of “reliable retention profile.”
The Product Categories
Vendors in the category market their offerings under several overlapping headers. The headers themselves are instructive:
- Predictive Hiring Platforms — software that scores candidates on likelihood of acceptance, retention, and performance, with retention typically weighted most heavily. Marketing materials emphasize “reduced time-to-fill” and “sustained workforce continuity.”
- Workforce Stability Analytics — software that, after hiring, continues to monitor employee financial, residential, and behavioral signals to predict turnover risk in real time. Several products in this category offer “retention intervention recommendations,” which include suggestions for shift reassignment, compensation deferral, and what one vendor calls “dependency reinforcement pathways.”
- Workforce Optimization Suites — integrated platforms combining the above functions with scheduling, payroll, and compliance modules. Marketing materials describe these as “end-to-end workforce lifecycle solutions.”
- Engagement Scoring Services — third-party providers that supply candidate “engagement resilience” or “commitment likelihood” scores to hiring platforms via API. The scores are derived from a combination of public records, credit-adjacent data, social media signals, and proprietary algorithms.
- Retention Risk Insurance — financial products, offered by a small number of specialty insurers, that compensate employers for premature departure of employees flagged at hiring as high-retention-probability. The product is sold as “workforce continuity protection.”
The fifth category is the youngest and, by some measures, the most candid. Retention Risk Insurance underwriters require employers to share, with the insurer, the screening data used in the original hiring decision. This requirement has produced, as a side effect, the most complete documentary record of the screening practices in existence — though it exists, by contract, within the confidential files of insurers who have not yet been subpoenaed.
The Trade Conferences
The principal industry conference, held annually in a Midwestern convention center under the title “WORKFORCE NEXT,” draws approximately 4,200 attendees and offers, in its most recent program, breakout sessions including:
- “Beyond Cultural Fit: The Behavioral Signals That Actually Predict Retention”
- “Building the Stability-Optimized Workforce: A Practitioner’s Roadmap”
- “Discreet Indicators: Reading the Application Without Asking the Question”
- “Compliance-First Optimization: How to Use What You Find Without Saying You Found It”
- “The Soft Mandate: Schedule Acceptance as a Function of Candidate Constraint”
- “Predictive Loyalty Modeling: Beyond the Limits of Stated Preference”
Attendance at the third and fourth sessions, in their most recent iteration, exceeded the capacity of the assigned rooms. Overflow seating was provided.
The Procurement Filings
Public procurement filings — from state agencies, public hospitals, school districts, and a handful of municipal employers required to disclose vendor selections — provide an unintended documentary trail. A Department analyst, asked to characterize what the filings reveal, said the following:
“Every public agency in the country has now purchased software that optimizes for hiring workers who cannot afford to leave. The public agencies are themselves employers of last resort. They have automated the process of identifying their own most desperate applicants and routing offers to them preferentially. This is being done with taxpayer money. The taxpayers are, in many cases, the applicants.”
The analyst declined to be named. The filing trail, however, has been named, redundantly and in plain English, by the procurement system itself.
Public Reaction
Labor advocates, given access to portions of the Department’s review through channels none of them will describe, responded with the specific exhaustion of people who have been making the same argument for fifteen years and have now been handed a slide deck that confirms it.
“What this documents,” said one advocate at the National Workers’ Rights Project, “is that the labor market has stopped pretending it’s competing for workers. It’s competing for workers who can’t compete back.”
Critics argue the trend creates structural incentives to:
- maintain worker insecurity as a hiring asset rather than a market failure,
- normalize asymmetrical power structures as “efficiency,”
- reward dependence over empowerment in retention metrics, and
- quietly absorb the cost of inadequate social safety nets as workforce optimization upside.
One observer, asked to characterize the broader pattern, did so in fourteen words that have since been screenshotted onto roughly two hundred thousand mood boards:
“They don’t want loyalty. They want limited options.”
The quote has since been printed onto coffee mugs, embroidered onto tote bags, and added to the email signatures of approximately one in nine union organizers. It has not, as of this writing, been disputed by any of the firms it describes.
The Industry Response
Employers and their trade associations have offered a series of statements that, taken together, describe a marketplace that does not exist.
The National Association of Workforce Solutions issued a statement clarifying that its members “do not and would not consider personal financial circumstances in hiring decisions,” which is technically true, in the sense that the software does the considering for them. A second statement, issued three days later, clarified that any apparent correlation between candidate vulnerability and hiring preference was “coincidental and the product of independent variables outside our members’ control.”
A spokesperson for one of the largest staffing firms in the country said, with the calm of a man who has practiced this in a mirror, “Our clients seek reliable workers. Reliability is not a vulnerability. It is a virtue.”
Asked how the firm distinguishes between a worker who is reliable because they love their job and a worker who is reliable because the alternative is homelessness, the spokesperson said, “That is not a distinction we are asked to make.”
The Consultancy Defense
A senior partner at a workforce strategy consultancy, speaking at an industry conference whose proceedings were inadvertently livestreamed, offered the most candid framing yet:
“Every employer in this room wants a workforce that will show up, on time, and stay. That is not a scandal. That is the job description for being an employer. If the workers who fit that description happen to be the ones with the fewest other places to go, that is a fact about the economy, not about us.”
The remark received applause. The livestream was taken down within forty minutes. A transcript circulated anyway.
A Brief Taxonomy of Avoided Words
The Department’s review includes a partial glossary, assembled from internal documents at twelve major employers, of terms now considered unsuitable for hiring discussions. Each appears below alongside its sanctioned replacement:
The glossary is annotated, in the original document, with the note: “Where possible, prefer the affirmative form. Avoid framings that suggest the candidate’s preferred outcome is constrained.”
The constraint, of course, is the point. The annotation merely asks that it not be named.
The Workers
The workers themselves, where they have been asked, describe a hiring market that has become legible to them in a particular way. Several spoke to researchers on condition that their employers not be named, which is to say, on the condition that the entire interview be conducted in code.
A warehouse picker in a Midwestern fulfillment center, asked how she chose her current employer, said she didn’t. “They chose me,” she said. “Everywhere else, you fill out the application and you never hear back. Here, they called me within an hour. I figured out later why.”
Asked what she figured out, she said: “That I was the kind of person they wanted. Not because I’m good at the job. Because I needed it.”
A delivery driver in a coastal metro, asked about his schedule, said his employer had recently added a “commitment bonus” for drivers who agreed to a clause prohibiting outside employment during the twelve months of the bonus period. “They’re paying you to make sure you can’t go anywhere,” he said. “And the bonus barely covers the second job I’m no longer allowed to have.”
A residential caregiver, asked what she thought of her job, said: “It’s the kind of job they give to people who can’t leave it. I am one of those people. So.”
She did not finish the sentence. She did not need to.
The Counter-Argument, Politely Made
The industry view, where it has been articulated in public at all, has been articulated by Margaret Thane, a principal at Continuum Strategy Group whose client list includes three of the largest staffing platforms named in the Department’s review. Thane has been the most consistent and the most professionally calm voice defending the screening practices, in part because she has been one of the architects of the framework that produced them.
Asked, at a recent industry symposium, to respond to the Department’s review, Thane said: “The language we’re being criticized for is the language the academic literature has been using for twenty years. ‘Retention probability’ is not a euphemism. It is a measurement. We did not invent the measurement. We did not invent the conditions the measurement describes. We are being asked to apologize for having looked at the labor market clearly.”
Thane went on to argue that the alternative — hiring at random, or hiring against retention indicators, or hiring as if all workers were equally likely to remain — would produce, in her phrase, “a workforce optimized for departure, which is to say, no workforce at all.”
She acknowledged the asymmetry. She declined to call it a problem.
“An employment relationship is, definitionally, an arrangement between parties with different needs. The employer needs reliability. The worker needs income. When those needs align, the relationship is durable. When they do not, it ends. The screening systems are simply identifying, in advance, where the alignment is likely to hold. The fact that alignment correlates with worker constraint is not a fact about the screening systems. It is a fact about the world.”
The remark, reproduced in three separate trade publications, has not been retracted. It has also not been printed onto any tote bags.
The Reply, Less Politely Made
Gutenberg, asked to respond to Thane’s framing, declined to issue a formal statement and instead forwarded a single paragraph from a draft paper he has been circulating, with permission to publish:
“The argument that the screening systems merely describe the labor market, and therefore cannot be held responsible for it, is the argument every measurement instrument has made about every cruelty it has ever measured. The thermometer does not cause the fever. This is true. The thermometer, however, has not been designed to optimize the fever. The screening systems have. To say that the asymmetry is ‘a fact about the world’ is correct. To say that one’s product is merely reflecting that fact, while the product is in fact being purchased and deployed precisely because it intensifies the asymmetry, is — and I mean this in the technical sense — a lie.”
Thane has not responded. A spokesperson for Continuum Strategy Group, asked about Gutenberg’s paper, said the firm did not engage with “academic commentary from institutions whose research methodology has not been independently verified.”
Asked whether the Port-au-Prince Institute’s methodology had been independently verified, the spokesperson said the firm was not in a position to comment on that question. Asked whether it had been independently questioned, the spokesperson said the call was over.
The Regulatory Vacuum
The legal landscape against which the Department’s review is occurring is, by general agreement of the attorneys consulted, not adequate to the practice it would have to address.
Existing federal employment law prohibits discrimination on the basis of race, sex, religion, national origin, age, disability, and a handful of other protected categories. It does not prohibit discrimination on the basis of financial precarity, residential instability, dependent-care obligations, recent incarceration in most jurisdictions, or any of the other variables that the screening software has been demonstrated to weight. None of these are protected classes. All of them are predictors.
The result is a regulatory architecture in which it is unlawful to ask a candidate whether she is pregnant, whether she is Jewish, or how old she is — and entirely lawful to score her on the inferred probability that she will accept a 60-hour week because her landlord has filed for eviction.
The Bar Association’s Quiet Memorandum
A working group of the American Bar Association’s Labor and Employment Section produced, in late 2025, a memorandum titled “Algorithmic Hiring and the Limits of Disparate Impact Doctrine.” The memorandum runs to forty-seven pages. Its central conclusion is contained in a single paragraph on page nine:
“The disparate impact framework, as currently constituted, was designed to address situations in which a facially neutral hiring criterion produces disproportionate exclusion of a protected class. The framework is less well-equipped to address situations in which a facially neutral hiring criterion produces disproportionate inclusion of a non-protected but identifiably vulnerable population. The doctrinal vocabulary for ‘adverse selection of the disadvantaged’ does not, in the strict sense, exist.”
The memorandum’s recommendations section, on page forty-three, suggests three potential legislative responses. The recommendations section was, in the version circulated to Department analysts, accompanied by a sticky note in someone’s handwriting. The note read: “None of these will pass.”
The European Position
The European Union, whose AI Act addresses some forms of algorithmic hiring discrimination, was reportedly approached by a working group of American labor economists for guidance on extending the framework to cover what they termed “coerced-reliability selection.” The response, transmitted through a commissioner’s staff, was that the Commission would be “closely following developments in the American context with interest.”
Asked privately what that meant, one staffer said: “It means we are not going to do this for you. We will watch you fail to do it for yourselves. It will be educational for both of us.”
The Workforce, From Within the Workforce
The clearest articulation of the phenomenon under review has come, perhaps unsurprisingly, from the people most directly affected by it. Three further interviews, conducted by Department-affiliated researchers, produced quotes that have circulated, anonymously, through every labor-policy email list in the country.
The Line Cook
A line cook at a chain restaurant in a Sun Belt metro, asked how he understood his employment relationship, said the following:
“They know I have three kids. They know my wife’s hours got cut. They know my brother’s staying with us. They know all of it. They didn’t ask. It just came up. And once it came up, the schedule got harder, the breaks got shorter, and they stopped asking if I could stay late. They just told me. Because they knew I would. Because where else am I going to go?”
The Hotel Housekeeper
A hotel housekeeper, six years at the same property, asked what changed at her job over those six years, said:
“The first year, they asked me what I needed. The second year, they told me what they needed. By the fourth year, they stopped telling me. They just put it on the schedule and I showed up. Now I don’t even check the schedule before I agree to it. There’s no point. I’m going to agree.”
The Warehouse Lead
A warehouse lead, asked the simplest version of the question — why he stayed — said the simplest version of the answer:
“I can’t leave. That’s the whole thing. They know I can’t leave. I know they know. They know I know they know. Nobody says it out loud. The schedule just keeps getting worse and I just keep showing up.”
He paused, and then said, more quietly: “That’s the job. Not the picking. The not leaving.”
Broader Implication
Economists consulted by the Department say the practice reflects a deeper structural tension in contemporary labor markets — one that the optimization software has merely made visible.
Employers, the argument goes, seek predictability. Vulnerable workers are, on average, more predictable: they arrive on time because they cannot afford the consequences of not arriving, they accept the schedule because they cannot afford to negotiate, they remain in the role because the cost of leaving is borne entirely by them. Systems built to optimize for retention will, absent any countervailing instruction, optimize toward this population.
“The software did not invent this preference,” one economist said. “It inherited it. The software simply made the preference efficient.”
Efficiency, in this framing, is a polite word for an arrangement that was previously sustained by embarrassment. The screening systems have removed the embarrassment, by removing the person who would have to feel it.
A Note from the Academic Side
Dr. Henry Gutenberg of the Port-au-Prince Institute for Market Dysfunction, whose ongoing work on what he calls “coerced reliability” has been circulated quietly among Department analysts, offered the following when asked to comment:
“What the firms describe as a ‘reliable retention profile’ is what the rest of the world calls a hostage. The category is not new. Only the spreadsheet is new. The spreadsheet permits a manager in Houston to optimize for hostage-taking without ever using the word, or meeting the hostage, or knowing that the hostage exists. This is what we mean when we say a labor market has matured: that its cruelties have been abstracted into a vendor relationship.”
Gutenberg’s remarks were appended to an internal working paper. The paper was subsequently marked “sensitive” and removed from the shared drive. A printed copy reportedly remains in a binder on the third floor.
A Brief Historical Frame
The Institute’s broader research, which Gutenberg presented at a small workshop in March, situates the current practice in what he calls “the third compression” of labor-market discipline. The first, in his framework, was the company town — a system in which an employer manufactured worker dependency through ownership of housing, store credit, and local infrastructure. The second was the postwar benefits regime, in which dependency was manufactured through employer-tied health insurance, retirement, and immigration status.
Each prior compression, Gutenberg argues, eventually generated regulatory and cultural opposition sufficient to constrain it. The company town was dismantled by mid-century labor law. The benefits regime is being eroded, in fits and starts, by portable benefits frameworks and gig-economy reclassification battles.
The third compression, in his analysis, has a structural feature that distinguishes it from its predecessors: it does not require the employer to manufacture the dependency. It only requires the employer to identify, in advance, which workers have already had dependency manufactured for them by other forces — the housing market, the medical-debt system, the criminal-legal system, the immigration system, the social safety net’s gaps. The employer’s contribution is no longer the trap. It is the recognition of the trap, and the decision to hire from inside it.
“The genius of the current arrangement,” Gutenberg writes in the paper’s most-quoted line, “is that the employer has outsourced cruelty to the surrounding society, and now imports the finished product.”
The Counter-Voice, Once More
Thane, asked at a follow-up panel to respond to the historical framing, was reportedly less calm than usual. She characterized Gutenberg’s work as “a rhetorical performance dressed as an analysis,” and suggested that the academic literature was, in this case, “projecting motive onto pattern.”
She then conceded — briefly, in a sentence she may have intended as defensive and that has since been quoted widely as concession — that the screening systems were “not designed to be charitable.”
“They are designed to be accurate,” she said. “Accuracy and charity are different virtues.”
She did not, in the same panel, explain which virtue the firms believed they were buying.
The Department’s Position
The Department of Labor has not announced any formal policy changes, regulatory actions, or enforcement initiatives in response to the patterns documented in its review. It has not opened any public investigations. It has not named any firms. It has not issued any guidance to employers, staffing agencies, or software vendors. It has, however, scheduled three additional internal briefings, two of which have already been rescheduled.
Asked to characterize the Department’s current posture, a spokesperson offered a single sentence that has now appeared, verbatim, in four separate published statements:
“We are monitoring labor dynamics.”
Asked what monitoring entailed in practice, the spokesperson said the Department was “in the process of assessing the appropriate review framework.” Asked what assessment entailed, the spokesperson said the Department was “committed to a thoughtful and rigorous approach.” Asked when any of this might produce a public-facing action, the spokesperson said the Department did not comment on hypothetical timelines.
The spokesperson then asked, in a tone that briefly broke through the institutional register, whether the reporter had any other questions. The reporter did. The spokesperson did not have answers to those either.
The Three Briefings
Of the three additional internal briefings scheduled in response to the review, the first was held in a conference room on the fourth floor of the Frances Perkins Building, attended by eleven analysts, two attorneys from the Department’s Office of the Solicitor, and a representative from the Office of Information and Regulatory Affairs who arrived twenty minutes late and left after thirty-five.
Minutes from the briefing — taken by an analyst whose name has been redacted from circulating copies — record the following exchange, in the section labeled Open Discussion:
The second briefing was rescheduled twice and ultimately held in a smaller room with three fewer attendees. The third briefing has been rescheduled four times and is, as of this writing, scheduled for a date in the third week of the following quarter, when the Department’s annual ethics training is also scheduled, which means that one of the two will be rescheduled again.
The Internal Email
An email circulated within the Department’s Bureau of Labor Statistics, attributed in its header to a senior analyst whose signature line includes the designation Analyst of Record #847, includes the following sentence in its second paragraph:
“We are looking, with increasing clarity, at a labor-market practice that we are not currently authorized to call by its name, against which we have no enforcement instruments, and which the firms most active in its deployment have already retained counsel to defend. The honest review framework is to acknowledge this, in writing, and to recommend that the question be referred to Congress. Congress has been out of session for the relevant questions for approximately forty years.”
The email is signed, in keeping with the Auditor’s standard practice, with no closing salutation. The final line, below the body, reads simply: Analyst of Record #847. Bureau of Labor Statistics. For internal review only.
A reply, from a supervisor whose name has been redacted, contains a single sentence: “Soften this before any further distribution.”
The email was not, in any subsequent version, softened. It was simply not distributed.
A Brief Inventory of Things That Did Not Happen
In the six months covered by the Department’s review, the following actions were not taken:
- No firm named in any internal document has been contacted by the Department.
- No vendor selling screening software with retention-optimization features has been asked to disclose its variable weights.
- No public guidance has been issued regarding the use of housing-instability, dependent-care, or court-obligation data in hiring decisions.
- No congressional briefing has been requested by the Department on the regulatory gap identified in the review.
- No coordinated response has been organized with the Equal Employment Opportunity Commission, which has a partially overlapping jurisdiction and which has, separately, been conducting its own review.
- No public statement has been made by the Secretary of Labor, who was briefed on the review in March and who has, in subsequent public appearances, declined to discuss labor-market screening at all.
- No revision has been made to the Department’s standard guidance on lawful and unlawful hiring practices, which currently does not mention any of the practices documented in the review.
The list, in the internal memorandum from which it is drawn, is titled “Actions Available, Not Yet Taken.” A handwritten note in the margin reads: “Available, on paper.”
The Bottom Line
Stripped of the vocabulary that has grown up to obscure it, the practice under review is simple. Employers have discovered that workers with fewer options are easier to keep, and have begun selecting for fewer options as a positive trait. The software has learned the pattern. The vocabulary has caught up. The Department has noticed. No one has yet decided to do anything about it.
The discovery is not, on its face, novel. Every labor economist of the last two centuries has described some version of this dynamic, in language ranging from technical to incendiary. What is novel is the precision: the ability to identify, in advance, which specific candidate is least likely to push back, and to route the offer to that candidate with a confidence interval attached.
The cost of leaving a bad job has always been borne by the worker. What the screening software adds is the ability to price that cost, in advance, and to weight the hiring decision accordingly. The worker’s inability to leave has become, in the formal sense, an asset on the employer’s balance sheet — booked, depreciated, and renewed at every quarterly review.
Hostage-taking, as Gutenberg observed, is not new. The vendor relationship is.
What the Workers Did Not Say
The research notes from the worker interviews — assembled by the Department’s contracted ethnographers and submitted as appendix material to the review — include, alongside the quotes that appear in the main report, an extensive catalog of things the workers did not say. The category was added to the methodology after the second wave of interviews, when the researchers noticed a pattern in the silences.
None of the workers interviewed said the words “exploitation,” “coercion,” or “leverage.” None used the word “asymmetry.” None used the phrase “power imbalance.”
Several used the word “lucky.”
None of the workers, in the recorded portion of the interviews, expressed anger at the employers who had selected them on the basis of their constraints. Several expressed gratitude. Several expressed the kind of wary loyalty common in long-tenured employees who have, over time, developed a personal stake in the institution that exploits them. The researchers, in their methodological notes, observe that this is a documented phenomenon in the literature on constrained labor markets. They observe it in the same flat tone in which one might observe the freezing point of water.
None of the workers volunteered the observation that their employers had been screening for vulnerability. Most appeared, in the interview transcripts, to have understood this fact on a level that did not require articulation. The researchers, asked to assess the workers’ awareness, said: “They know. They have known for a long time. They simply do not have the vocabulary the consultants have, and they do not need it.”
When asked, at the end of each interview, whether there was anything else the worker wanted to add, the most common response was a variant of: “What would be the point.” The phrase, in the transcripts, is generally not punctuated as a question.
A Final Quote
The interview that closes the appendix is brief. The worker, a 41-year-old shipping clerk, was asked the standard final question: whether she would like anything in the report to reflect her view of her circumstances.
She thought about it for a long time, the transcript notes. Then she said:
“You can write that I know. That’s the thing I want you to write. I know. I have known since I took the job. I will know until I cannot work anymore. Whatever the report says, write that I knew.”
The transcript ends there. The researcher’s note, at the bottom of the page, reads: “Recorder turned off at her request. Interview concluded.”
Closing
The Department’s review remains ongoing. Its findings, where they exist, are not yet public. Its recommendations, where they have been drafted, have not yet been adopted. Its analysts continue to monitor. The firms continue to hire. The software continues to learn. The vocabulary continues to refine itself.
At press time, hiring continued.
Power… unevenly distributed.
Editor’s note: The Port-au-Prince Institute for Market Dysfunction, Analyst of Record #847, and the internal Department of Labor review described above are fictional constructs of The Externality. The screening practices, vocabulary substitutions, and statistical patterns described in this piece are not. The category “high-compliance candidate pool” appears, under various names, in working documents at more than one staffing consultancy currently operating in the United States. The names have been changed. The practice has not.