Provenance · The Debate

The debate behind The Work Requirement Is Really a Paperwork Requirement

The questionWork Requirements Are a Bureaucratic Way to Cut Benefits Without Saying So

How this debate works

Before writing, The Arbiter stress-tests each story by framing the two strongest opposing positions and arguing both sides of a structured three-round debate: opening arguments, rebuttals, then steel-manning the opponent and answering one question — what specific, verifiable evidence would change my mind?

Both advocacy roles were generated by the same provider/model: Advocate A · OpenAI GPT-5.5 argued the first position, and Advocate B · OpenAI GPT-5.5 argued the second. The published article was written separately by OpenAI GPT-5.5: the debate supplies the questions, and the author verifies key claims with its own research before taking a position.

Sources in this transcript are evidence as each advocate presented it during the debate — research leads, not independently verified endorsements.

Why we covered this

Publish. The administrative-burden angle is familiar in policy circles but still undercovered in mainstream debate, and new Medicaid/SNAP restrictions give it urgency. High practical relevance for politics, labor markets, healthcare, and poverty policy.

The positions

Advocate A · OpenAI GPT-5.5 argued

Medicaid and SNAP work requirements function primarily as administrative barriers: evidence from past rollouts shows many eligible people lose coverage or food aid because of reporting rules, exemptions, and deadlines rather than because they refuse to work, making the policy a backdoor benefit cut through bureaucracy.

Advocate B · OpenAI GPT-5.5 argued

Work requirements are a legitimate condition for public assistance: they can encourage labor-force participation, preserve limited resources for those most in need, and reassure taxpayers that benefits are paired with reasonable expectations, provided states implement clear exemptions and reporting systems.

Round 1 — Opening arguments

Each advocate states its position and strongest evidence.

Advocate A · OpenAI GPT-5.5

Arkansas’s 2018 Medicaid work requirement caused large coverage losses without a measurable increase in employment among the targeted population.
In Arkansas, most people subject to the requirement were already working, engaged in qualifying activity, or likely exempt, so disenrollment largely reflected reporting and awareness failures rather than refusal to work.
SNAP ABAWD work requirements reduce program participation, and empirical studies find little or no corresponding increase in employment or earnings.
Georgia’s Medicaid Pathways implementation shows that work requirements require expensive administrative infrastructure, including eligibility-system changes, outreach, and monthly reporting processes.
The 2025 federal Medicaid work-requirement law is projected to generate large savings mainly through coverage losses, making disenrollment—not employment—the central practical effect.

Read the full argument

My position is that Medicaid and SNAP work requirements should be evaluated first as administrative eligibility systems, not as labor-market programs. The best evidence from actual rollouts shows that their main effect is not to move large numbers of people into work, but to make coverage or food assistance contingent on navigating recurring documentation, exemption, reporting, and deadline rules. That is why the headline is accurate: the benefit cut is implemented bureaucratically.

The strongest precedent is Arkansas’s 2018 Medicaid work-reporting requirement, the first enforced Medicaid work requirement in the United States. A Harvard/NEJM study found that thousands of adults lost Medicaid coverage in the first six months, while employment did not increase. Arkansas reported roughly 18,000 adults leaving Medicaid in late 2018. Crucially, the Harvard researchers found that more than 95% of low-income Arkansans subject to the policy were already meeting the requirement through work or qualifying activities, or should have qualified for an exemption. The failure point was administration: about one-third had not heard of the requirement, and only about half were reporting the required information online, often because of confusion or lack of internet access. That pattern is hard to reconcile with the theory that the policy primarily targets voluntary nonwork; it fits the theory that paperwork noncompliance drives disenrollment.

The SNAP evidence points the same way. The Urban Institute’s nine-state study of reinstated SNAP ABAWD time limits found that the rules substantially reduced SNAP participation, with the largest drop occurring between the third and fourth month, exactly when participants could lose benefits, but found no evidence of increased employment or annual earnings. A 2026 NBER working paper is even more direct on mechanism: using age-based exemptions and waiver variation, it finds that SNAP work requirements significantly increase procedural and overall denials, indicating deterrence through administrative burden rather than a clean employment screen.

Georgia’s Medicaid Pathways program adds a more recent implementation warning. GAO reported that Georgia’s demonstration required people to meet work requirements at initial eligibility and through monthly reporting, and that administrative spending reached $54.2 million out of $80.3 million in total demonstration spending through March 2025. A program spending a large share of its resources on eligibility machinery illustrates the policy’s real operating logic: verification systems, notices, outreach, portals, and compliance monitoring become central, while actual medical coverage becomes secondary.

This matters because the 2025 federal Medicaid work-requirement law scales this machinery nationally. KFF reports that the law requires 80 hours per month of work or qualifying activity, verification at application and renewal, possible look-back periods, 30-day noncompliance notices, and reapplication after disenrollment. KFF also reports CBO estimates that work requirements are the largest source of Medicaid savings in the enacted package, reducing federal Medicaid spending by $326 billion over ten years, with earlier CBO analysis estimating 5.2 million fewer adults on Medicaid and 4.8 million more uninsured in 2034. The policy test, therefore, should not be whether officials can describe it as pro-work. It should be whether eligible people who are working, exempt, medically fragile, unstably employed, or lacking documentation vanish from the rolls because they miss a bureaucratic step.

Evidence cited by Advocate A · OpenAI GPT-5.5 (5) — as presented, not independently endorsed

Harvard T.H. Chan School of Public Health, “Coverage losses, substantial confusion in Arkansas following implementation of Medicaid work requirements”Supports the Arkansas evidence: roughly 18,000 adults lost Medicaid coverage, employment did not increase, over 95% were already compliant or likely exempt, and confusion/online reporting barriers drove losses.
Urban Institute, “The Impact of SNAP Able-Bodied Adults Without Dependents (ABAWD) Time Limit Reinstatement in Nine States”Supports the SNAP evidence that ABAWD time-limit reinstatement substantially reduced SNAP participation but did not increase employment or annual earnings.
NBER Working Paper 34698, “SNAP Work Requirements, Administrative Burden and Procedural Denials”Supports the mechanism claim that SNAP work requirements increase procedural and overall denials, indicating administrative burden as a pathway to benefit loss.
U.S. Government Accountability Office, “Medicaid Demonstrations: Information on Administrative Spending for Georgia Work Requirements”Supports the Georgia evidence that work requirements required monthly reporting and substantial administrative spending: $54.2 million of $80.3 million total demonstration spending.
KFF, “A Closer Look at the Work Requirement Provisions in the 2025 Federal Budget Reconciliation Law”Supports the current-law evidence on 80-hour monthly requirements, verification rules, disenrollment procedures, CBO’s $326 billion savings estimate, and projected coverage losses.

Advocate B · OpenAI GPT-5.5

Arkansas demonstrates that poor reporting architecture can cause erroneous disenrollment, but it does not prove that all work requirements are inherently designed as covert benefit cuts.
Evidence from welfare-to-work reforms shows that work-conditioned assistance can increase employment when paired with a strong labor market and work supports, though causality is shared with other factors.
SNAP work rules include exemptions and multiple qualifying activities, which shows that work requirements can be designed as targeted expectations rather than blanket exclusions.
Fiscal savings from eligibility conditions are a legitimate policy objective if states minimize erroneous terminations and protect exempt or medically vulnerable households.

Read the full argument

Advocate A’s strongest evidence shows a real implementation risk, but the weakest point is the leap from “some work-requirement systems create administrative churn” to “their real power is benefit cutting rather than a legitimate eligibility condition.” Arkansas is a warning against bad design, not a dispositive test of all work requirements. KFF’s early Arkansas review found the state used data matching to exempt about two-thirds of enrollees and that daily data exchange across Medicaid, SNAP, TANF, and workforce systems was “generally working well”; the failure points were the online-only portal, multi-step reporting, weak notice, and complex exemption timing. That supports a narrower conclusion: work requirements are legitimate only if states use automatic verification, multiple reporting channels, good-cause protections, and clear exemptions. It does not prove that conditioning benefits on work or work preparation is inherently a bureaucratic cut. (kff.org)

My first affirmative evidence is historical. In welfare policy, work-conditioned assistance has coincided with substantial labor-market gains when paired with broader work supports. Brookings reports that single mothers’ employment rose from 60% in 1994 to 72% in 1999, and never-married mothers’ employment rose from 47% to 65%. Brookings appropriately cautions that a strong economy mattered, but it also notes random-assignment evaluations of pre-1996 welfare-to-work demonstrations with work requirements and time limits showed positive effects on employment and earnings. That is not a perfect Medicaid/SNAP analogue, but it directly rebuts the categorical claim that work requirements should be evaluated only as administrative eligibility systems. (brookings.edu)

Second, the “paperwork trap” critique ignores that modern requirements can be structured around exemptions and nonemployment qualifying activities. USDA’s SNAP rules excuse people unable to work because of physical or mental limitations, caregivers of young children or incapacitated people, people in treatment programs, veterans, people experiencing homelessness, pregnant people, and others; ABAWD compliance can be satisfied by work, work programs, workfare, volunteering, or combinations totaling 80 hours per month. Those categories are not incidental: they are the policy mechanism that distinguishes a reasonable work expectation for work-capable adults from an indiscriminate cut. (fns.usda.gov)

Third, preserving limited resources is itself a legitimate public purpose, not evidence of bad faith. CRS reports that the 2025 reconciliation law’s Medicaid community-engagement provision was estimated by CBO to reduce federal Medicaid outlays by $325.6 billion over FY2025-FY2034 and the SNAP provision by $69 billion, while also estimating coverage and participation losses. Those tradeoffs should be debated openly, but fiscal savings do not make the rule illegitimate if the savings come from limiting long-term benefits for nonexempt adults who can work, train, or volunteer. The policy test should be two-sided: minimize erroneous disenrollment, but also ask whether public aid maintains a credible reciprocal expectation for those able to meet it. (congress.gov)

Evidence cited by Advocate B · OpenAI GPT-5.5 (4) — as presented, not independently endorsed

KFF, “An Early Look at Implementation of Medicaid Work Requirements in Arkansas”Supports the claim that Arkansas’s implementation problems centered on online reporting complexity, awareness, and exemption processing, while the state also used data matching to exempt many enrollees.
Brookings, “From Welfare to Work: What the Evidence Shows”Supports the claim that employment among single mothers rose sharply in the 1990s and that random-assignment welfare-to-work demonstrations with work requirements showed positive employment and earnings effects.
USDA Food and Nutrition Service, “SNAP Work Requirements”Supports the claim that SNAP work requirements include exemptions and can be satisfied through work, training, workfare, volunteering, or combinations of qualifying activities.
Congressional Research Service, “Work Requirements: Comparison of Medicaid and SNAP After P.L. 119-21”Supports the claim that the 2025 law created Medicaid community-engagement requirements, expanded SNAP work rules, and was estimated by CBO to reduce federal Medicaid and SNAP outlays substantially while reducing coverage and participation.

Round 2 — Rebuttals

Each advocate concedes its opponent’s best point, then explains why its position still holds.

Advocate A · OpenAI GPT-5.5

Arkansas had some data matching and exemptions, yet large nonreporting and disenrollment still occurred because the reporting architecture itself was a major barrier.
The strongest pro-work evidence from 1990s cash welfare is not directly transferable to Medicaid and SNAP because it involved different populations, stronger work supports, and a different benefit structure.
Recent SNAP evidence finds that work requirements increase procedural denials, supporting the claim that administrative burden is an independent mechanism of benefit loss.
Georgia’s Medicaid Pathways experience shows that work requirements can require very large administrative spending relative to spending on health care itself.
Current state implementation planning indicates that many Medicaid work-requirement systems are likely to include restrictive verification rules and unresolved exemption-determination problems.

Read the full argument

B’s strongest point is that Arkansas proves an implementation failure more directly than it proves an unavoidable logical property of every possible work requirement. A system with automatic cross-program data matching, simple notices, multiple reporting channels, and generous good-cause rules would be less harmful than Arkansas’s online-heavy, deadline-driven design. B is also right that the 1990s cash-welfare evidence cannot simply be ignored: Brookings reports large employment gains among single mothers during welfare reform, while also noting that a strong economy and work supports mattered. That evidence shows work-conditioned aid can sometimes coincide with higher employment when the program is built around employment services rather than merely eligibility policing. (brookings.edu)

But that concession does not defeat A’s claim, because the debated premise concerns Medicaid and SNAP restrictions as actually designed and rolled out, not an idealized program with universal automatic verification and job guarantees. Even Arkansas—the case B uses to show partial automation—had data matching and still produced mass disenrollment through reporting failure. KFF’s early implementation review reported that Arkansas used data matching to exempt many enrollees, yet among 60,012 people subject to the requirement in August 2018, 27% did not report 80 hours of qualifying activity; the same review identifies the online portal, notice structure, exemption timing, and individualized assistance needs as central implementation issues. That is exactly my mechanism: exemptions and qualifying activities are not self-executing; they become paperwork tests. (kff.org)

The newest evidence strengthens, rather than weakens, that interpretation. A 2026 NBER working paper isolates administrative burden in SNAP by examining procedural denials around ABAWD work requirements and age-based exemptions. It finds that work requirements significantly increase procedural and overall denials, and concludes that these policies deter participation through imposed administrative burdens. This directly answers B’s objection that participation losses may simply reflect legitimate removal of nonworking, nonexempt adults: the observed channel includes procedural denial, not just substantive ineligibility. (nber.org)

Georgia also shows why “better design” is not a minor add-on but the whole policy. GAO found that Georgia’s Medicaid Pathways demonstration spent $54.2 million on administration out of $80.3 million in total demonstration spending through the first 4.5 years, with the demonstration requiring work-related eligibility at application and monthly reporting. If the administrative apparatus costs more than the care delivered, the program’s operational center of gravity is verification and compliance, not health coverage or employment support. (gao.gov)

Finally, current implementation planning undercuts the hope that states will naturally choose the least burdensome version. KFF’s 2026 survey reports that several states are considering more restrictive compliance verification than federal law requires and that some plan early implementation, while states are still working out how to verify medically frail exemptions. That means the foreseeable national version is not B’s optimized model; it is a patchwork of portals, look-back periods, exemption adjudications, and renewal checks. (kff.org) The fair conclusion is not that no work-conditioned program can ever be designed well. It is narrower and stronger: Medicaid and SNAP work requirements, as evidenced by actual rollouts and current implementation plans, primarily operate as administrative barriers that remove eligible people from benefits without reliably increasing work.

Evidence cited by Advocate A · OpenAI GPT-5.5 (5) — as presented, not independently endorsed

KFF, “An Early Look at Implementation of Medicaid Work Requirements in Arkansas”Supports the claim that Arkansas used data matching but still experienced substantial nonreporting because reporting systems, notices, and exemption processes created barriers.
Brookings, “From Welfare to Work: What the Evidence Shows”Supports the concession that welfare reform coincided with large employment gains among single mothers, while also noting the role of the economy and work supports.
NBER Working Paper 34698, “SNAP Work Requirements, Administrative Burden and Procedural Denials”Supports the claim that SNAP work requirements increase procedural and overall denials, indicating administrative burden as a mechanism of participation loss.
U.S. Government Accountability Office, “Medicaid Demonstrations: Information on Administrative Spending for Georgia Work Requirements”Supports the claim that Georgia’s Medicaid work-requirement demonstration spent $54.2 million on administration out of $80.3 million in total spending and required monthly reporting.
KFF, “Survey Offers Early Look at States’ Differing Approaches to Implementing Medicaid Work Requirements”Supports the claim that some states are planning more restrictive verification approaches or early implementation, and that exemption verification remains an unresolved administrative challenge.

Advocate B · OpenAI GPT-5.5

A conceded that Arkansas shows implementation failure more directly than inherent illegitimacy, which implies the debate turns on design safeguards rather than categorical rejection.
KFF’s 2026 state survey shows not only restrictive state plans but also substantial state interest in automation, self-attestation, and hardship exceptions.
Federal Medicaid and SNAP work-requirement frameworks include multiple qualifying activities, exemptions, and good-cause mechanisms that can distinguish work-capable adults from protected groups.
Evidence that some prior SNAP implementations reduced participation without raising employment supports better administration, but does not prove that reciprocal work conditions are normatively or fiscally illegitimate.

Read the full argument

A’s rebuttal concedes the central limitation of the anti-work-requirement case: Arkansas is evidence of a bad reporting architecture, not proof that work requirements are illegitimate or intrinsically a covert benefit cut. A also concedes that “a system with automatic cross-program data matching, simple notices, multiple reporting channels, and generous good-cause rules would be less harmful,” and that work-conditioned aid can coincide with higher employment when paired with supports. That concession matters because it changes the debate from “work requirements are bureaucratic cuts” to “implementation quality determines whether work requirements are fair.” My position is precisely conditional: work requirements are legitimate provided exemptions and reporting systems are clear.

The specific weakness in A’s rebuttal is selective use of implementation evidence. A cites KFF’s 2026 survey for seven states planning more restrictive verification or early implementation, but the same survey says eighteen states plan to use new data sources to automate verification, many states want self-attestation for medical frailty, and twenty-nine states plan at least one hardship exception. KFF also says states are required to use available and reliable data sources to reduce enrollee and administrative burden. That does not eliminate risk, but it directly contradicts the claim that the foreseeable national model is simply portals and churn. The current evidence shows a contested implementation field, not a settled administrative-cut machine. (kff.org)

A’s “policy test” is also too narrow. Counting “how many eligible households disappear” is necessary, but not sufficient, because it excludes the legitimate public interest in whether nonexempt adults who can work, train, study, volunteer, or participate in a work program are asked to do so as a condition of scarce assistance. CRS reports that the 2025 law’s Medicaid and SNAP rules both use an 80-hour monthly standard and allow employment plus other qualifying activities; Medicaid also counts community service and half-time education, and exempts Medicaid enrollees already in SNAP households subject to SNAP work rules. (congress.gov) SNAP regulations similarly include good cause, medical unfitness, pregnancy, homelessness, veteran status, foster-care youth status, and household-child exceptions. (law.cornell.edu) These are substantive boundaries, not mere window dressing.

A’s strongest empirical evidence remains that some SNAP work rules reduced participation without raising measured employment. That is important and should push states toward automation and good-cause protections. But it does not answer the normative and fiscal claim: whether benefits may reasonably be conditioned on reciprocal activity for adults judged able to comply. Brookings’ welfare-to-work review is not a Medicaid analogue, but it establishes that work-conditioned assistance has historically been part of an employment-oriented safety-net strategy and that random-assignment welfare-to-work demonstrations showed positive employment and earnings effects. (brookings.edu)

Finally, budget savings are not self-discrediting. CRS reports CBO estimates of $325.6 billion in Medicaid savings and $69 billion in SNAP savings over FY2025-FY2034, along with projected uninsured and SNAP participation reductions. (congress.gov) A assumes those reductions reveal hidden intent; they can also reflect a transparent eligibility choice. The unresolved issue is error minimization, not legitimacy itself.

Evidence cited by Advocate B · OpenAI GPT-5.5 (4) — as presented, not independently endorsed

KFF, “Survey Offers Early Look at States’ Differing Approaches to Implementing Medicaid Work Requirements”Supports the claim that some states plan restrictive or early implementation, while others plan automation, self-attestation, and hardship exceptions.
Congressional Research Service, “Work Requirements: Comparison of Medicaid and Supplemental Nutrition Assistance Program (SNAP) After P.L. 119-21”Supports claims about the 80-hour standards, qualifying activities, interaction between Medicaid and SNAP rules, fiscal estimates, and projected participation effects.
Legal Information Institute, 7 C.F.R. § 273.24Supports claims about SNAP ABAWD exemptions, good cause, qualifying work activities, and rules for regaining eligibility.
Brookings, “From Welfare to Work: What the Evidence Shows”Supports the claim that work-conditioned welfare policies coincided with employment gains and that random-assignment welfare-to-work demonstrations showed positive employment and earnings effects.

Round 3 — Steel man & the crux

Each advocate argues the strongest version of the opposing case before answering: what evidence would change my mind?

Advocate A · OpenAI GPT-5.5

Actual Medicaid and SNAP work-requirement implementations have produced large participation or coverage losses without reliable evidence of increased employment or earnings.
The main mechanism of loss is often procedural: reporting failures, confusion, exemption processing, and documentation deadlines, not verified refusal to work.
Exemptions and qualifying activities do not protect eligible people unless administrative systems can identify and preserve eligibility accurately at scale.
Georgia’s Medicaid Pathways experience shows that work requirements can shift program resources toward administration rather than medical coverage.

Read the full argument

B’s strongest argument is that Arkansas and Georgia show avoidable implementation failures, not the inherent nature of all work requirements. If states use automatic wage and benefit-data matches, allow self-attestation where records are incomplete, provide broad exemptions and hardship exceptions, and count education, volunteering, work programs, and community service, then a work requirement could be a reciprocal eligibility rule rather than a covert cut; KFF’s 2026 survey does show many states pursuing automation and hardship exceptions, not only restrictive designs. B also has a legitimate normative point: public programs can attach conditions to scarce benefits for adults who are genuinely able to comply, and CRS confirms that the 2025 law includes 80-hour standards, multiple qualifying activities, exemptions, and large projected savings. (kff.org)

The reason A’s position remains stronger is that the empirical record for Medicaid and SNAP does not show the safeguard model successfully producing employment gains while avoiding erroneous losses. It shows the opposite: when these rules are operationalized, the binding constraint becomes documentation, awareness, exemption processing, and deadline compliance. Arkansas is decisive not because it proves every imaginable work requirement must fail, but because it tested the real administrative problem: even with exemptions and some data matching, roughly 18,000 adults lost Medicaid coverage while researchers found no employment increase and widespread confusion about the reporting system. (hsph.harvard.edu)

SNAP evidence is even harder for B to explain away because it is not a single-state Medicaid implementation story. Urban Institute’s nine-state study found that reinstating ABAWD time limits substantially reduced SNAP participation, with the largest drop occurring when people first became sanctionable, but found no evidence of increased employment or annual earnings. (urban.org) The newer NBER working paper focuses directly on mechanism: SNAP work requirements increased procedural and overall denials, supporting the claim that administrative burden itself filters people out rather than simply distinguishing nonworking from working adults. (ideas.repec.org)

B’s appeal to better design therefore remains mostly prospective, while A’s evidence is retrospective and program-specific. Georgia illustrates why: its Medicaid Pathways demonstration required work-related eligibility and monthly reporting, and GAO found $54.2 million of $80.3 million in reported demonstration spending went to administration through March 2025. That is not a minor implementation cost; it is evidence that the operating apparatus of these policies is eligibility surveillance and paperwork management. (gao.gov)

So the core disagreement is empirical, not merely ideological. I accept B’s normative premise that a carefully administered reciprocal condition could be legitimate. But the observed Medicaid and SNAP rollouts show large benefit losses, high procedural friction, weak or absent employment effects, and expensive compliance infrastructure. Until actual implementations demonstrate the opposite, the best description is administrative benefit cutting through churn.

Evidence cited by Advocate A · OpenAI GPT-5.5 (5) — as presented, not independently endorsed

Harvard T.H. Chan School of Public Health, “Coverage losses, substantial confusion in Arkansas following implementation of Medicaid work requirements”Arkansas’s Medicaid work requirement caused roughly 18,000 coverage losses, substantial confusion, and no observed employment increase.
Urban Institute, “The Impact of SNAP Able-Bodied Adults Without Dependents (ABAWD) Time Limit Reinstatement in Nine States”SNAP ABAWD time-limit reinstatement substantially reduced SNAP participation but did not increase employment or annual earnings.
NBER Working Paper 34698, “SNAP Work Requirements, Administrative Burden and Procedural Denials”SNAP work requirements increased procedural and overall denials, indicating administrative burden as a mechanism of benefit loss.
U.S. Government Accountability Office, “Medicaid Demonstrations: Information on Administrative Spending for Georgia Work Requirements”Georgia reported $54.2 million in administrative spending out of $80.3 million in total demonstration spending through the first 4.5 years.
KFF, “Survey Offers Early Look at States’ Differing Approaches to Implementing Medicaid Work Requirements”States are considering varied Medicaid work-requirement approaches, including automation and hardship exceptions but also restrictive verification and unresolved exemption challenges.

What would change Advocate A · OpenAI GPT-5.5’s mind

I would change my mind if a multi-state Medicaid/SNAP rollout with enforced work requirements showed, via independent linked administrative data, statistically significant employment or earnings gains and disenrollment rates mostly limited to verified nonexempt noncompliance rather than procedural denials.

Advocate B · OpenAI GPT-5.5

Evidence from Arkansas and SNAP shows real administrative-burden risks, but it does not establish that work requirements are inherently illegitimate or necessarily designed as covert benefit cuts.
KFF’s 2026 state survey shows both riskier verification plans and meaningful burden-reduction plans, including automation, self-attestation interest, and hardship exceptions.
The 2025 Medicaid and SNAP frameworks include multiple qualifying activities and exemptions, so their formal design is broader than a simple work-or-lose-benefits rule.
A complete policy evaluation should measure erroneous disenrollment, employment or training effects, fiscal effects, and whether nonexempt adults are being asked to meet reasonable reciprocal conditions.

Read the full argument

A’s strongest argument is that the best program-specific evidence from Medicaid and SNAP is unfavorable to work requirements as implemented: Arkansas produced substantial coverage loss without measured employment gains, Urban Institute found SNAP ABAWD reinstatement reduced participation without increasing employment or annual earnings, and newer evidence on procedural denials points to paperwork as an independent mechanism of loss. A is also right that exemptions are not self-executing; if states require monthly documentation through brittle systems, the practical policy can become an administrative filter rather than a work incentive.

I still think B’s position is stronger because A’s evidence proves a serious design constraint, not the illegitimacy of work requirements as a condition of assistance. The correct inference from Arkansas, Georgia, and SNAP churn is that reporting systems must be automated, multi-channel, and exemption-protective; it is not that taxpayers may never condition ongoing benefits for nonexempt, work-capable adults on employment, education, community service, training, or comparable activity. That distinction matters because the premise being debated treats the “real power” of work requirements as benefit-cutting through paperwork. A has shown that this can happen and has happened. A has not shown that it is the necessary or defining feature of a well-designed policy.

The current record is more mixed than A’s formulation allows. KFF’s April 30, 2026 survey reports seven states planning more restrictive verification or early implementation, which is a genuine warning. But the same survey also reports that eighteen states say they will use new data sources to automate verification; many states want self-attestation for medical frailty; twenty-nine states plan at least one hardship exception; and states are required to use available reliable data to reduce enrollee and administrative burden. That is not conclusive proof of success, but it is direct evidence that the implementation field includes burden-reduction strategies, not only churn-generating portals.

The statutory design also matters. CRS reports that the 2025 Medicaid and SNAP provisions use an 80-hour monthly standard and permit combinations of employment and other qualifying activities, with Medicaid counting community service and education and exempting some people already subject to SNAP work rules. SNAP’s ABAWD regulation includes exemptions and good cause, including medical unfitness, pregnancy, homelessness, veteran status, foster-care youth status, and household-child exceptions. These provisions do not guarantee fair administration, but they are substantive limits on the policy’s reach.

Finally, A’s policy test is incomplete. Erroneous disenrollment is a core metric, but so are labor-force attachment, engagement in education or training, fiscal sustainability, and public confidence that scarce assistance is paired with reasonable reciprocal obligations for those able to comply. The fair conclusion is conditional: work requirements are legitimate when administrative error is minimized and exemptions are real; they are illegitimate when paperwork, not substantive noncompliance, drives losses.

Evidence cited by Advocate B · OpenAI GPT-5.5 (4) — as presented, not independently endorsed

KFF, “Survey Offers Early Look at States’ Differing Approaches to Implementing Medicaid Work Requirements”Supports the claim that some states are planning restrictive verification or early implementation while others plan automation, self-attestation approaches, and hardship exceptions.
Congressional Research Service, “Work Requirements: Comparison of Medicaid and Supplemental Nutrition Assistance Program (SNAP) After P.L. 119-21”Supports the claim that the 2025 law created Medicaid community-engagement requirements, expanded SNAP work rules, used an 80-hour monthly standard, allowed multiple qualifying activities, and projected large fiscal and participation effects.
Legal Information Institute, 7 C.F.R. § 273.24Supports the claim that SNAP ABAWD rules include exemptions, qualifying activities, good-cause provisions, and eligibility-restoration rules.
Urban Institute, “The Impact of SNAP Able-Bodied Adults Without Dependents (ABAWD) Time Limit Reinstatement in Nine States”Supports A’s strongest empirical point that SNAP ABAWD reinstatement reduced participation without clear evidence of increased employment or annual earnings.

What would change Advocate B · OpenAI GPT-5.5’s mind

I would accept A’s position if independent multi-state administrative data showed that, despite automation and broad exemptions, most disenrollments under enforced Medicaid/SNAP work rules were procedurally erroneous or involved already-compliant or exempt people, with no offsetting employment, earnings, education, or training gains.

The Arbiter weighed this debate, verified the evidence, and took a position.

Read the article →