Administrative burden and program access

Research on administrative burden demonstrates that program participation depends not only on eligibility rules but on the procedural demands through which eligibility is established and maintained. Herd and Moynihan (Reference Herd and Moynihan2018) conceptualize administrative burden as the costs individuals encounter in interactions with the state, distinguishing among learning costs, compliance costs, and psychological costs. This framework explains persistent gaps between eligibility and enrollment across means-tested programs (Moynihan, Herd, and Harvey Reference Moynihan, Herd and Harvey2015). Currie (Reference Currie, Auerbach, Card and Quigley2006) documents that take-up rates for means-tested programs fall substantially below 100% even among populations that clearly meet eligibility criteria and that administrative complexity is a primary driver of this shortfall.

In the Medicaid context, these costs have measurable effects on coverage outcomes. Take-up rates vary across states, ranging from approximately 50% among eligible parents in high-burden states to 96% in low-burden states. This variation reflects differences in enrollment procedures, renewal requirements, and verification systems rather than differences in statutory eligibility alone (Herd and Moynihan Reference Herd and Moynihan2020; Michener Reference Michener2018). Evidence from the post-pandemic Medicaid unwinding period reinforces this point. States that relied on automated data matching and ex parte renewals experienced substantially lower procedural disenrollment rates than states that depended on manual verification (Brooks et al. Reference Brooks, Alker, Green, Yafimenka and Lin2025; Tolbert, Rudowitz, and Drake Reference Tolbert, Rudowitz and Drake2023). The unwinding experience demonstrated at national scale that administrative design systematically structures who retains coverage and who does not.

The administrative burden framework is well established in identifying how procedural costs suppress participation (Burden et al. Reference Burden, Canon, Mayer and Moynihan2012; Herd and Moynihan Reference Herd and Moynihan2018; Moynihan, Herd, and Harvey Reference Moynihan, Herd and Harvey2015). Empirical applications have documented substantial effects on take-up across program domains (Heinrich Reference Heinrich2016), with the framework already extended specifically to Medicaid enrollment (Herd et al. Reference Herd, DeLeire, Harvey and Moynihan2013). Related friction concepts include sludge (Sunstein Reference Sunstein2019), and a recent review literature maps their conceptual relationships (Madsen, Mikkelsen, and Moynihan Reference Madsen, Mikkelsen and Moynihan2022). What this literature does not specify is how particular design features determine where in the enrollment process these costs bind. Nor does it identify when their cumulative effects become large enough to function as the dominant determinant of who a program actually serves.

Work requirements as continuous compliance regimes

Work requirements intensify administrative burden by transforming eligibility from a periodic determination into a continuous compliance process. Unlike standard Medicaid eligibility, which can often be verified through existing administrative data, work-conditioned eligibility requires ongoing documentation of qualifying activities. Each reporting cycle creates an independent point of potential exclusion even when underlying eligibility conditions remain unchanged (Sommers et al. Reference Sommers, Chen, Blendon, Orav and Epstein2020). The political roots of this design are well documented. Work conditions have historically functioned as mechanisms for distinguishing between “deserving” and “undeserving” recipients. Their structure often reflects assumptions about beneficiary behavior that interact with racial attitudes and public perceptions of program populations (Haeder and Moynihan Reference Haeder and Moynihan2023; Soss, Fording, and Schram Reference Soss, Fording and Schram2011).

The empirical record on Medicaid work requirements prior to H.R. 1 is concentrated in two state cases. Arkansas implemented work requirements under a Section 1115 waiver beginning in June 2018, requiring expansion adults to report 80 hours of monthly qualifying activities through an online portal. Within four months, approximately 17,000 individuals lost coverage (Sommers et al. Reference Sommers, Chen, Blendon, Orav and Epstein2020). One-third of the target population had never heard of the requirement, and 44% were unsure whether it applied to them (Herd and Moynihan Reference Herd and Moynihan2020). Georgia launched Pathways to Coverage in July 2023 under a Section 1115 waiver, adopting a fundamentally different architecture. Unlike Arkansas, Georgia imposed work requirements as a condition of initial eligibility rather than as a maintenance condition on an already-enrolled population. Thompson and Burke (Reference Thompson and Burke2007) provide historical context on the use of Medicaid demonstration waivers as instruments of executive federalism. This structural difference is consequential because it determines where in the enrollment lifecycle administrative burden exerts its strongest effect.

These two cases share a common outcome. In both states, procedural exclusion exceeded substantive screening as the primary driver of enrollment suppression. They differ, however, in a design feature that generates a testable expectation. Arkansas introduced burden at the maintenance stage, concentrating attrition among individuals who already held coverage. Georgia embedded burden at the point of entry, concentrating attrition among individuals who had not yet enrolled. The question that follows is whether this variation in design architecture produces predictable differences in where procedural attrition concentrates within the enrollment lifecycle.

From friction to rationing: A threshold argument

The administrative burden literature treats procedural costs as frictions that impede access without redefining how programs allocate coverage (Fox, Stazyk, and Feng Reference Fox, Stazyk and Feng2020; Herd and Moynihan Reference Herd and Moynihan2018). This assumption holds when burden operates at the margin, when program access is determined primarily by eligibility criteria and procedural costs modestly suppress participation among those who are eligible. The evidence from work-conditioned Medicaid programs suggests a different condition. In Arkansas, 95% of those who lost coverage met the work requirement or qualified for exemptions but failed to navigate the reporting process (Sommers et al. Reference Sommers, Chen, Blendon, Orav and Epstein2020). In Georgia, 54% of interested individuals never applied due to reporting requirements, 22% of denials were procedural, and 30% of disenrollments resulted from paperwork failures (Georgia Budget and Policy Institute 2025). At this scale, burden is not merely impeding access to a program that otherwise functions according to its eligibility criteria. It is determining who the program serves.

The concept of rationing in public policy refers to mechanisms that allocate scarce resources among competing claimants, whether through explicit statutory thresholds or through implicit procedural mechanisms (Mechanic Reference Mechanic1997; Oberlander Reference Oberlander2003; Weimer and Vining Reference Weimer and Vining2017). Mechanic (Reference Mechanic1997) distinguishes between explicit rationing, in which authorities formally limit access through specified criteria, and implicit rationing, in which allocation occurs through queues, complexity, or other procedural mechanisms that constrain access without formal acknowledgment. Oberlander (Reference Oberlander2003) shows how political reluctance to articulate rationing decisions explicitly leaves implicit allocation to operate through administrative channels.

Administrative rationing as developed here occupies a specific position within this taxonomy. It is implicit rather than explicit, structural rather than discretionary, and it operates through program architecture rather than through caseworker decisions or queue length. One respect in which the concept departs from this taxonomy deserves acknowledgment. Rationing in its conventional sense presupposes a scarce resource divided among competing claimants. Coverage under eligibility-based Medicaid programs carries no enrollment cap, so no fixed supply of benefits is being divided. Administrative rationing borrows the allocative function of the traditional concept rather than its scarcity premise. Burden performs the work that scarcity performs in conventional rationing by determining which eligible claimants actually receive a benefit that is formally available to all who qualify. Formally stated, when administrative burden functions as the dominant determinant of program participation, it operates as a coverage allocation mechanism rather than as friction within an otherwise eligibility-driven process. Under this condition, statutory eligibility defines a potential population, but administrative design determines realized access. The framework is agnostic to policymaker intent. What matters is not whether burden was intended but whether design features structurally constrain participation more than eligibility criteria do. Table 1 summarizes how the three dominant frameworks each address a different dimension of burden but share a common limitation that administrative rationing addresses.

Table 1.

Theoretical frameworks for administrative burden in safety-net programs

Table 1. Long description

The table consists of five columns: Framework, Core claim, Mechanism, Explains, and Does not explain.

* Row 1: Herd and Moynihan (2018). Core claim: Burden is implementation friction. Mechanism: Learning, compliance, and psychological costs. Explains: Types of costs individuals face. Does not explain: How design parameters determine cost dominance or attrition distribution.

* Row 2: Soss, Fording, and Schram (2011). Core claim: Burden is a tool of disciplinary governance. Mechanism: Racialized paternalism in administration. Explains: Why burden falls disproportionately on marginalized populations. Does not explain: How program architecture generates attrition independent of caseworker discretion.

* Row 3: Lipsky (1980). Core claim: Burden is street-level rationing under resource constraints. Mechanism: Caseworker discretion allocates access. Explains: Individual-level variation based on frontline decisions. Does not explain: Design-level attrition patterns operating structurally without caseworker involvement.

* Row 4: Administrative Rationing (this article). Core claim: Burden is a structural coverage allocation mechanism. Mechanism: Design parameters like reporting frequency and verification methods produce attrition at specific stages. Explains: How specific design choices determine program participation. Does not explain: Whether policymakers intend burden to function as rationing.

The framework’s relationship to Lipsky’s (Reference Lipsky1981) account of street-level bureaucracy is central to the article’s theoretical contribution because both arguments identify administrative processes as performing allocative work that formal eligibility criteria do not. They locate that allocation in different places. Lipsky locates it in the discretionary judgments of caseworkers, whose individual decisions accumulate into patterns of access under conditions of resource scarcity. Administrative rationing locates it in the structural design of program architecture, where parameters such as reporting frequency, verification method, and point of imposition produce predictable attrition that operates independently of caseworker decisions. The implication is that even programs with minimal frontline discretion, including fully automated eligibility systems, can function as rationing mechanisms through their design architecture alone. Where Lipsky’s framework explains variation in access across caseworkers serving similar populations, administrative rationing explains variation in access across programs serving similar populations under different design configurations. Brodkin (Reference Brodkin2011) documents how new-managerialist reforms have progressively constrained caseworker discretion through performance metrics and standardized procedures, accelerating the shift in where allocation actually occurs from individual judgment to program architecture.

The structural mechanism through which administrative rationing operates does not produce uniform effects across populations. Prior research suggests that procedural exclusion is likely to fall more heavily on Black and Hispanic adults, rural residents with limited internet access, and individuals with disabilities (Haeder and Moynihan Reference Haeder and Moynihan2023; Soss, Fording, and Schram Reference Soss, Fording and Schram2011). Many in this last group qualify for exemptions but cannot navigate the documentation process required to claim them. The mechanism is partly cognitive. Scarcity, health problems, and constraints on executive functioning amplify the effects of any given level of administrative burden, so vulnerable populations bear disproportionate exclusion from the same design features (Christensen et al. Reference Christensen, Aarøe, Baekgaard, Herd and Moynihan2020). Comparative work across means-tested programs reinforces this point. Bureaucratic encounters in Medicaid impose substantially heavier burdens than encounters in WIC despite serving overlapping populations, with the difference attributable to program design rather than client characteristics (Barnes, Michener, and Rains Reference Barnes, Michener and Rains2023). Administrative rationing, therefore, raises a distributional question alongside the structural one. Whether burden operates as the dominant determinant of program participation, and how its filtering effects distribute across subpopulations, are theoretically related rather than separate questions. The same design features that produce procedural attrition in aggregate may generate patterned attrition across racial, geographic, and disability lines. Future empirical work will need to address both dimensions to characterize fully how design architecture allocates coverage. These structural and distributional dimensions also raise the empirical question of how the threshold between high burden and rationing can be identified.

The threshold dividing rationing from high-friction implementation is conceptual rather than empirically calibrated. Three approaches could operationalize it for case-level assessment. The first is the share of eligible non-participants whose non-participation is attributable to procedural failure rather than substantive ineligibility, with rationing identified where that share dominates eligible non-participation. The second is the responsiveness of enrollment to targeted burden reduction. Programs in which modest reforms produce proportional enrollment gains operate at the friction end of the continuum, while programs in which equivalent reforms produce only incremental gains are characterized by allocative architectures that distribute exclusion across multiple design features. Experimental work in adjacent program domains documents that targeted simplification of application materials can produce substantial take-up gains within months (Bhargava and Manoli Reference Bhargava and Manoli2015), providing a benchmark against which work-requirement program responses can be assessed. The third is the comparison of program participation rates across groups with substantively similar eligibility but differing exposure to procedural costs, which isolates the design-driven component of attrition. Each approach requires individual-level or longitudinal program data not yet available for work-conditioned Medicaid. The H.R. 1 implementation period will offer the first opportunity to apply these tests at scale.

Georgia’s October 2025 reforms provide one application of the second operationalization approach. The reforms reduced reporting frequency while holding eligibility criteria constant, generating a within-case comparison in which the only variable change was a single design parameter. If burden in Georgia operated as friction, the change should produce proportional enrollment gains. If burden operated as rationing, the change should produce only incremental gains because allocative function is distributed across multiple design features that the reform did not address. Three design parameters link the rationing framework to observable attrition patterns. Reporting frequency determines how many procedural checkpoints exist. Verification method determines whether the burden of proof rests on the individual or the state. Point of imposition determines where in the lifecycle the requirement binds.

The H1–H3 claims evaluated in this article are observable implications of the rationing claim. H1 holds that reporting frequency shapes procedural disenrollment by multiplying compliance checkpoints, since each additional checkpoint creates an opportunity for design architecture, rather than substantive eligibility, to determine participation. Empirical patterns consistent with H1 distinguish design-driven attrition from random implementation noise that would not concentrate by checkpoint frequency. H2 predicts that point of imposition determines where in the lifecycle attrition concentrates because design parameters channel exclusion to specific stages. Support for H2 distinguishes structural attrition produced by design from individual variation produced by caseworker discretion, which would not concentrate by lifecycle stage. H3 predicts that targeted burden reduction in Georgia produces incremental rather than proportional enrollment gains because allocative function is distributed across design parameters that a single reform leaves intact. Support for H3 distinguishes rationing from high-friction implementation, which would respond proportionally to reform. Each claim evaluates a distinct observable implication of the rationing claim, and evidence consistent with these claims would provide triangulating support for the framework, with H3 requiring post-reform data for confirmation.

Figure 1 presents the conceptual positioning of administrative rationing within the existing literature.

Figure 1.

Conceptual positioning of the administrative rationing framework within the administrative burden literature.

Figure 1. Long description

The flowchart is organized into five vertical levels connected by downward-pointing arrows.

Level 1: Three light gray boxes representing foundational literature. Left: Herd and Moynihan 2018, Burden as friction. Center: Soss, Fording and Schram 2011, Burden as governance. Right: Lipsky 1980/2010, Burden as discretion.

Level 2: A single gray box titled Shared consensus. It states that burden suppresses enrollment and that cost types have been identified.

Level 3: A red dashed box titled THEORETICAL GAP. It asks how design parameters produce predictable lifecycle-stage attrition in safety-net programs.

Level 4: A bold green box titled ADMINISTRATIVE RATIONING (this article). It defines this as when procedural exclusion rates exceed substantive ineligibility, where burden functions as a coverage allocation mechanism. Below this is a secondary green box titled Mechanism: Lifecycle attrition model, showing the formula: Reporting frequency times verification method times imposition point yields predictable attrition at specific enrollment stages.

Level 5: Three tan boxes representing applications. Left: Arkansas 2018-2019, imposed on enrolled population yielding maintenance attrition. Center: Georgia Pathways 2023-present, entry plus maintenance yielding full-lifecycle attrition. Right: H. R. 1 Mandate 2027-, 41 states, 6-mo. redetermination yielding predicted national-scale.

A horizontal double-headed arrow at the very bottom is labeled: Increasing design complexity and scale of implementation.

Case selection and comparative logic

This article employs a structured comparison of the only two state-level Medicaid work requirement programs operationalized in the United States prior to the federal mandate enacted in H.R. 1 (George and Bennett Reference George and Bennett2005). Arkansas and Georgia were selected not because they represent the full range of possible implementation designs but because they constitute the complete universe of operational cases in which actual enrollment, attrition, and expenditure data exist. Every other state that received waiver approval for work requirements either did not implement before court orders intervened or withdrew applications before enrollment began. This constraint on generalizability is acknowledged. The findings are treated as propositions to be tested as additional cases emerge during the H.R. 1 implementation period rather than as settled empirical conclusions.

The comparative value of these two cases lies in a specific design difference that generates testable predictions under the administrative rationing framework. Arkansas imposed work requirements only on individuals already enrolled in Medicaid through the ACA expansion. Georgia took a different approach. The Pathways to Coverage program imposed work requirements at the point of initial eligibility, requiring individuals to demonstrate compliance with the qualifying activities threshold before they could enroll, and also imposed continuing reporting obligations on those who did enroll. The structural difference is that Arkansas operates as a maintenance-only design while Georgia operates as an entry-and-maintenance design that adds an entry-stage burden Arkansas did not have. This difference determines where in the enrollment lifecycle administrative burden exerts its strongest additional effect. In a maintenance-only program, burden concentrates at reporting and redetermination because enrollees have already cleared the application process. In a program that adds entry-stage requirements, burden additionally concentrates at the pre-application and application stages because individuals must demonstrate compliance before entering the program. This structural variation provides the analytical leverage necessary to test H2, which predicts that programs imposing requirements at entry produce higher pre-application attrition than programs imposing requirements only at maintenance.

The two cases differ on more than the design variable of interest, and these differences require explicit attention. Arkansas imposed work requirements on adults aged 19–49 enrolled through the ACA expansion at incomes up to 138% of the federal poverty level. Georgia’s Pathways to Coverage targets adults at incomes up to 100% of the federal poverty level who are not otherwise eligible for traditional Medicaid, drawn from a non-expansion population that lacks ACA coverage entirely. The broader Medicaid contexts of the two states also differ in ways that extend beyond the expansion decision itself. Arkansas was among the earliest adopters of the ACA expansion, implementing it in January 2014 through a premium assistance model, so its work requirement was layered onto a mature expansion program with an established enrollee population (Sommers et al. Reference Sommers, Goldman, Blendon, Orav and Epstein2019). Georgia has declined expansion, maintains one of the highest uninsurance rates in the country, and has historically exhibited low take-up among its eligible populations (Tolbert et al. Reference Tolbert, Sammy, Clea and Anthony2026). A weak take-up environment could depress application completion in Georgia regardless of how the work requirement is structured. These population-level differences could in principle confound the design-architecture inference if attrition patterns reflect characteristics of who is targeted rather than how the requirement is imposed.

Three considerations limit this concern. First, the empirical pattern of interest is the location of attrition within the enrollment lifecycle, not its overall magnitude. Differences in income or labor force participation across the two populations would affect overall attrition rates but would not predict whether attrition concentrates at the maintenance stage or the pre-application stage. The lifecycle location is determined by where in the enrollment process the requirement is imposed. Second, the within-Georgia comparison generated by the October 2025 reforms holds the target population constant while varying a single design parameter, isolating the design-driven component of attrition. Third, the rationing framework concerns whether burden becomes the dominant determinant of participation among those who are substantively eligible. This question can be addressed within each case independently, regardless of how the eligibility threshold is set. The same logic extends to the broader program contexts. Arkansas’s mature expansion environment and Georgia’s high-uninsurance, low-takeup environment plausibly influence the magnitude of attrition in each case, but neither predicts the lifecycle stage at which attrition concentrates, which is the observable implication under evaluation. These considerations support a design-architecture interpretation of the observed lifecycle patterns, while acknowledging that population composition may also shape the magnitude of attrition and that the small case universe limits the generalizability of any single inference.

Table 2 presents the structural comparison across the two cases, the reformed Georgia design, and the H.R. 1 statutory framework. The design parameters vary along three dimensions that the administrative rationing framework identifies as determinative. The imposition point differs, with Arkansas applying requirements only at the maintenance stage while Georgia applies them at both entry and maintenance. Reporting frequency varies from monthly in both original designs to annual in Georgia’s reformed design, with H.R. 1 mandating a minimum six-month cycle. Verification method differs, with both state programs relying on individual self-report, while H.R. 1 permits retrospective administrative data matching. These differences predict where attrition will concentrate and at what rate.

Table 2.

Design parameter comparison across work requirement cases

Table 2. Long description

The table compares design parameters across four columns: Arkansas (2018–2019), Georgia original (2023–2025), Georgia reformed (Oct 2025-), and H. R. 1 mandate (2027–).

* Imposition point: Arkansas is Maintenance condition; all others are Entry and maintenance.

* Monthly hours: All four cases require 80 hours.

* Reporting frequency: Arkansas and Georgia original are Monthly; Georgia reformed is Annual at renewal; H. R. 1 mandate is Minimum every 6 months.

* Verification method: Arkansas is Online self-report only; Georgia original and reformed are Self-report (Gateway); H. R. 1 mandate is Self-report or admin data match.

* Redetermination: Arkansas, Georgia original, and Georgia reformed are Annual; H. R. 1 mandate is Every 6 months.

* Pre-application attrition: Arkansas is Not applicable (33 percent unaware, 44 percent unsure); Georgia original is 54 percent did not apply; Georgia reformed is To be determined; H. R. 1 mandate is Projected to exceed G A.

* Procedural attrition: Arkansas is 95 percent eligible or exempt; Georgia original is 22 percent denials and 30 percent disenrollments procedural; Georgia reformed is To be determined; H. R. 1 mandate is Unknown.

* Enrollment outcome: Arkansas lost 17,000 in 4 months; Georgia original had 8,077 after 2 years (7 percent of eligible); Georgia reformed is Projected 18,301; H. R. 1 mandate has 20 M plus affected.

* Status: Arkansas is Suspended by court; Georgia original is Operational; Georgia reformed is Extended to Dec 2026; H. R. 1 mandate is Required by Jan 2027.

Data sources

The analysis relies entirely on publicly available data. For Georgia, the primary source is the Georgia Budget and Policy Institute’s October 2025 report (Georgia Budget and Policy Institute 2025). The report draws on official program data released by the Georgia Department of Community Health, including monthly enrollment figures, denial and disenrollment counts disaggregated by procedural and substantive reasons, and program expenditure breakdowns. The denial and disenrollment percentages cited here (22% procedural denials, 30% procedural disenrollments) are derived from administrative counts published by the Department of Community Health. The pre-application attrition estimate (54%) draws on the same internal Department of Community Health monitoring data, obtained by GBPI through an Open Records Request. It tracks individuals through the application screen-flow. Of those who elected to be considered for Pathways and acknowledged the program requirements, 54% did not submit a complete application. The drop-off occurred because they failed to report a non-zero number of qualifying hours and activities or to request a reasonable modification. The figure is therefore best interpreted as an administrative measure of application-stage drop-off associated with the reporting requirement, rather than as a formal denial category. It provides stronger evidence than a purely speculative counterfactual estimate, but it remains less directly comparable to denial and disenrollment categories because the affected individuals did not complete applications. The Georgia Budget and Policy Institute is a state-level policy research organization with a stated equity orientation. The underlying program data it cites are publicly available and verifiable through Department of Community Health releases. Additional Georgia data come from the Department’s official program announcements (Georgia Department of Community Health 2025). For Arkansas, the primary sources are Sommers et al. (Reference Sommers, Goldman, Blendon, Orav and Epstein2019, Reference Sommers, Chen, Blendon, Orav and Epstein2020) and the Health Affairs analysis by Herd and Moynihan (Reference Herd and Moynihan2020). For H.R. 1, data come from Hinton, Diana, and Rudowitz (Reference Hinton, Diana and Rudowitz2025), the Center for Health Care Strategies (2025), and Schneider (Reference Schneider2026). Colorado’s publicly released fiscal impact estimates are used as an illustrative case of state-level implementation costs (Colorado Department of Health Care Policy and Financing 2026).

Analytical approach

The analysis proceeds in three stages. The first examines each case independently, mapping enrollment outcomes onto the three burden categories identified by Herd and Moynihan (Reference Herd and Moynihan2018) and identifying the lifecycle stage at which procedural attrition concentrates. The second compares the two cases along the design dimensions specified in the observable implications, particularly reporting frequency, verification method, and point of requirement imposition. The third applies the comparative findings to the H.R. 1 statutory framework, modeling the projected attrition implications of specific design choices that states will face during implementation.

The analysis employs structured, focused comparison in the tradition established by George and Bennett (Reference George and Bennett2005). The goal is to identify design features that covary with outcomes across cases and to develop propositions that can be tested as additional cases emerge. This approach is well-suited to a context in which the universe of operational cases is small but the theoretical questions are well-defined. The paper is a theory-building study that uses empirical cases to develop and evaluate a theoretical extension rather than to estimate average treatment effects. The structured comparison allows for process tracing across each case to identify the mechanisms through which specific design features generate specific attrition patterns. The H.R. 1 projections in section “H.R. 1 and projected national implications” extend the comparative logic by applying the observed design-outcome relationships from the two precedent cases to the design parameters established in the federal statute. These projections are anchored in the CBO’s combined baseline estimate of six million coverage losses attributable to work requirements and six-month redetermination provisions and are structured as conditional predictions tied to specific design scenarios rather than as unconditional forecasts (Congressional Budget Office 2025).

Program design and implementation

Arkansas became the first state to implement Medicaid work requirements when the Arkansas Works program began enforcing community engagement requirements in June 2018. The program applied to adults aged 19–49 enrolled in Medicaid through the ACA expansion. Enrollees were required to report 80 hours per month of qualifying activities, including employment, education, job training, job search, or community service. Reporting was conducted exclusively through an online portal, with no telephone or in-person reporting options available during the initial implementation period.

Several features of the Arkansas design are important for understanding the attrition patterns that followed. First, the program imposed requirements on a population that had already been enrolled in Medicaid and was receiving benefits. These were not new applicants navigating an unfamiliar system. They were existing beneficiaries who had previously satisfied all eligibility criteria and were now required to meet an additional condition to retain coverage they already held. Second, the reporting mechanism was entirely digital, which imposed a compliance cost barrier that compounded the learning costs facing enrollees who did not know a new requirement had been imposed (Hill and Burroughs Reference Hill and Burroughs2019). Enrollees who lacked reliable internet access or familiarity with the state’s online system had no alternative pathway to demonstrate compliance. Exemptions existed but required navigating the same online system, compounding compliance costs for enrollees who might otherwise have qualified for relief from the reporting obligation.

Attrition outcomes and burden distribution

Within four months, approximately 17,000 individuals lost Medicaid coverage for failing to report qualifying activities (Sommers et al. Reference Sommers, Chen, Blendon, Orav and Epstein2020). The critical finding from subsequent analysis was that the overwhelming majority of those who lost coverage were not removed because they were actually failing to work. Sommers et al. (Reference Sommers, Chen, Blendon, Orav and Epstein2020) found that approximately 95% of those who lost coverage either worked enough hours to satisfy the requirement or should have qualified for one of the program’s exemptions. The coverage losses were driven not by a failure to meet the substantive condition but by a failure to navigate the reporting process through which compliance was documented. This finding provides strong evidence for the proposition that procedural attrition can dominate substantive screening in work-conditioned programs.

The distribution of awareness failures further illuminates the burden structure. Survey data revealed that one-third of the target population had never heard of the new work requirement, despite the state’s outreach efforts. An additional 44% were unsure whether the requirement applied to them (Herd and Moynihan Reference Herd and Moynihan2020). These figures indicate that learning costs alone were sufficient to place a majority of the affected population at risk of coverage loss before compliance costs or psychological costs even entered the picture. The online-only reporting portal amplified this effect by creating a single, digitally gated compliance pathway that excluded enrollees without reliable internet access, a population that overlaps substantially with those the program was designed to serve (Hill and Burroughs Reference Hill and Burroughs2019).

Because Arkansas imposed requirements on an already-enrolled population, the attrition concentrated at the maintenance stage of the enrollment lifecycle. The pre-application stage was not affected because the affected individuals had already applied for and received Medicaid before the work requirement existed. This pattern is consistent with H2’s prediction that the point of requirement imposition determines where in the lifecycle burden concentrates. In the Arkansas design, the maintenance stage bore the full weight of the new administrative obligation, and the reporting checkpoint became the primary mechanism through which eligible individuals were removed from the program.

The Arkansas experience was terminated prematurely by judicial intervention. A federal district court vacated the waiver approval in March 2019, finding that the Secretary of Health and Human Services had failed to adequately consider whether the program would promote the objectives of the Medicaid Act. The Eighth Circuit Court of Appeals upheld this ruling. The four-month operational period was sufficient to demonstrate the magnitude and character of procedural attrition but insufficient to reveal whether attrition rates would stabilize, increase, or decrease as enrollees became more familiar with the reporting system.

Program design and the entry-condition architecture

Georgia’s Pathways to Coverage program differs from the Arkansas model in a structural respect that has significant implications for where administrative burden concentrates. Pathways impose work requirements at the point of initial eligibility and also as a continuing maintenance condition. Individuals cannot enroll without first demonstrating 80 hours per month of qualifying activities, and once enrolled they must continue to document compliance to retain coverage. This means the administrative burden of documenting compliance is present from the first interaction a prospective enrollee has with the program, before any coverage begins (Georgia Department of Community Health 2025).

The program was approved by CMS in October 2020 and launched in July 2023 after an extended implementation period during which the state invested in upgrades to its eligibility and enrollment technology systems. Pathways cover adults aged 19–64 with household incomes at or below 100% of the federal poverty level who are not otherwise eligible for traditional Medicaid. Georgia chose this design as an alternative to the ACA’s standard Medicaid expansion, which would have extended coverage to all adults below 138% of the federal poverty level without work requirements. The state’s decision created a program with a narrower income threshold and behavioral conditions that operate simultaneously as eligibility criteria and as administrative requirements (Michener Reference Michener2018).

This dual function requires explicit acknowledgment because it complicates the procedural-versus-substantive distinction on which the administrative rationing framework depends. An individual who works 80 hours per month but fails to document those hours through the Gateway portal does not fail to meet the program’s substantive intent. That individual fails to satisfy the administrative process through which the substantive condition is verified. In a conventional Medicaid program, the eligibility criterion (income below a threshold) can be verified through existing administrative data without the applicant actively documenting compliance. In a work-conditioned program, the eligibility criterion (80 hours of qualifying activities) cannot be separated from the documentation process through which it is demonstrated.

The rationing claim holds in this context because even among individuals who do work 80 hours, the documentation requirement independently excludes a measurable share. The 54% pre-application attrition and the 22% procedural denial rates in Georgia may include individuals who satisfied the underlying work-condition standard but were unable to complete the documentation process. The framework, therefore, identifies rationing not by the mere existence of eligibility criteria but by evidence that procedural documentation requirements exclude substantively eligible or potentially eligible individuals at scale.

Enrollment outcomes across the lifecycle

After two years of operation through June 2025, pathways enrolled approximately 8,077 individuals, no more than 7% of the estimated eligible population of uninsured, low-income adults from working households (Georgia Budget and Policy Institute 2025). Through June 2025, the program cost approximately $110 million, of which less than one-third funded actual health care benefits. The majority of expenditures supported eligibility system upgrades, outreach, and administrative overhead (Georgia Budget and Policy Institute 2025).

The program data reveal three distinct attrition points that correspond to the three stages of the enrollment lifecycle. At the pre-application stage, approximately 54% of interested Georgians did not complete an application after encountering the qualifying-activities reporting requirement. These were individuals who were aware of the program and expressed interest but did not submit applications because they could not document or demonstrate compliance with the 80-hour requirement at the point of entry. At the application stage, approximately 22% of denials resulted from procedural reasons, including missing documents and incomplete verification steps, rather than from a determination of substantive ineligibility. At the maintenance stage, approximately 30% of disenrollments were attributed to paperwork failures, including missed reporting deadlines and incomplete documentation submissions, rather than to a change in the enrollee’s underlying eligibility status (Georgia Budget and Policy Institute 2025).

These figures are consistent with H1 and H2 simultaneously. The concentration of attrition at the pre-application stage (54%), which accounts for the largest single share of enrollment suppression, is consistent with H2’s prediction that programs imposing requirements at entry produce higher pre-application attrition than maintenance-only programs. The presence of substantial procedural exclusion at the maintenance stage (30% of disenrollments) under monthly reporting is consistent with H1’s prediction. H1 predicts that high-frequency reporting drives procedural disenrollment by multiplying checkpoints at which design architecture, rather than substantive eligibility, determines outcomes.

The October 2025 reforms and a prospective test of H3

In October 2025, Georgia received federal approval to extend Pathways through December 2026 and to reduce reporting frequency from monthly to annual at renewal, add caregiving of a child under six as a qualifying activity, and establish retroactive coverage from the first day of the application month (Georgia Department of Community Health 2025).

These reforms create the empirical setting for a prospective test of H3, which predicts that targeted burden reduction in Georgia produces incremental rather than proportional enrollment gains. The reforms did not expand who is eligible. They changed how eligibility is documented and verified. The state projects enrollment will reach 18,301 by October 2026, a more than twofold increase (Georgia Budget and Policy Institute 2025). If this materializes, it would provide preliminary directional support for H3 because the reforms primarily altered administrative and documentation conditions while leaving the income threshold and 80-hour requirement intact. This projection has not yet been validated by actual enrollment figures, and the strength of the H3 finding depends on post-reform data that will become available during the program’s extended operational period.

At the same time, the projected 18,301 would still represent a fraction of the eligible population. Even under the reformed design, the vast majority of individuals who meet the program’s substantive eligibility criteria would remain unenrolled. This outcome is more consistent with the rationing interpretation than the friction interpretation advanced in this article. A friction model would predict that targeted burden reduction should produce proportional enrollment gains because the burden is a discrete obstacle that, once removed, allows the underlying eligibility criteria to determine participation. A rationing model predicts only incremental gains because the allocative function of burden is distributed across the program’s entire administrative architecture and cannot be eliminated by modifying a single design feature. Reducing reporting frequency alleviates one compliance cost but does not address the learning costs and psychological costs operating at earlier lifecycle stages. Georgia’s projected enrollment trajectory under the reformed design is more consistent with the rationing prediction than with the friction prediction, although definitive assessment will require actual post-reform enrollment data.

Design architecture and attrition concentration

The Arkansas and Georgia cases, when placed side by side, reveal a pattern that is difficult to explain through eligibility criteria alone but becomes coherent when analyzed through the lens of program design architecture. Both programs required 80 hours per month of qualifying activities. Both targeted low-income adults in similar age ranges. Both operated under Section 1115 waivers with CMS oversight. Yet, the most analytically important difference was where attrition concentrated within the enrollment lifecycle. This difference corresponds directly to the structural distinction between maintenance-condition and entry-condition designs.

In Arkansas, attrition concentrated at the reporting checkpoint because the requirement was layered onto an already-enrolled population (Sommers et al. Reference Sommers, Chen, Blendon, Orav and Epstein2020). In Georgia, attrition concentrated at the pre-application stage because the requirement was embedded at the point of entry (Georgia Budget and Policy Institute 2025). The dominant source of enrollment suppression in Georgia was front-loaded in a way that the Arkansas design, by its structure, could not produce.

This pattern is consistent with H2, which predicts that programs imposing requirements at entry produce higher pre-application attrition than programs imposing requirements only at maintenance. The point at which a work requirement is imposed in the enrollment lifecycle predicts where procedural attrition concentrates. Maintenance-only designs produce attrition at reporting checkpoints among individuals who have already gained coverage. Designs that add entry-stage requirements produce additional attrition at the gateway to the program itself, preventing individuals from ever entering the system. The practical consequence of this difference is significant. In the Arkansas model, individuals experienced coverage and then lost it. In the Georgia model, individuals never received coverage in the first place. Both outcomes represent procedural exclusion of eligible individuals, but the Georgia design makes that exclusion less visible because the excluded population never appears in enrollment data. They are filtered out before they become enrollees, and their absence from the program registers as low take-up rather than as coverage loss (Béland, Rocco, and Waddan Reference Béland, Rocco and Waddan2016).

Design variable hierarchy

The three design variables specified in the observable implications can now be compared across the two cases to assess how each plausibly contributes to procedural attrition. Both programs required monthly reporting in their original implementations, imposing 12 annual compliance checkpoints at which documentation failure could trigger coverage loss. The similarity in reporting frequency means that this variable cannot explain the difference in attrition location. It can, however, explain the overall magnitude of procedural attrition in both programs. Georgia’s October 2025 reform, which reduced reporting from monthly to annual, provides the empirical setting for examining H1, which holds that reporting frequency shapes procedural disenrollment by multiplying compliance checkpoints. State projections of enrollment increases provide preliminary directional support consistent with the mechanism, with confirmatory evaluation requiring actual post-reform enrollment data not yet available.

Both programs relied on individual self-report rather than retrospective administrative data matching, placing the compliance burden entirely on the individual. Neither employed payroll records or educational enrollment databases for verification. The reliance on self-reporting amplified procedural attrition at every lifecycle stage (Soss and Keiser Reference Soss and Keiser2006). H.R. 1 permits retrospective verification using administrative data, creating statutory authority for states to shift this burden to the state and potentially reducing compliance costs at the maintenance stage.

The point of requirement imposition is the design variable that most clearly differentiates the two cases and produces the most pronounced difference in attrition patterns. These findings suggest a hierarchy among the three design variables. Point of imposition determines where attrition concentrates. Reporting frequency determines how much attrition occurs at each checkpoint. Verification method determines how the compliance burden is distributed between the individual and the state. All three variables contribute to total procedural exclusion, but they operate through different mechanisms and at different stages of the enrollment lifecycle. The consistent finding across both cases is that substantive eligibility criteria define a potential population while administrative requirements determine which substantively eligible individuals actually receive benefits. This is the empirical foundation for the rationing claim advanced in section “From friction to rationing: A threshold argument,” where administrative burden becomes the principal determinant of program participation.

The federal design framework

H.R. 1 transforms the policy question from whether individual states should pursue work requirements to how 41 states will simultaneously build the administrative infrastructure to enforce a federal mandate. The law establishes a set of mandatory design parameters while leaving some implementation details to state discretion and to the interim final rule that the Secretary of Health and Human Services must issue by June 1, 2026.

The mandatory parameters include several features that, based on the Arkansas and Georgia evidence, can be expected to generate substantial procedural attrition. The law requires that states verify work requirement compliance at application and at renewal, with renewals for expansion adults occurring every six months rather than annually (Hinton, Diana, and Rudowitz Reference Hinton, Diana and Rudowitz2025). This six-month redetermination cycle creates two annual compliance checkpoints at which enrollees must demonstrate that they have met the 80-hour monthly threshold. States may choose to verify compliance more frequently, which would increase the number of procedural checkpoints further.

The law also specifies a look-back provision that requires states to review whether an applicant met the work requirement in at least one of the one to three months prior to application (Center for Health Care Strategies 2025). This provision means that prospective enrollees must have been completing and tracking qualifying activities before they apply. This presupposes awareness of the program and its rules at a point when the individual has had no formal interaction with the Medicaid system. For individuals who learn about the program and wish to apply immediately, the look-back provision creates a mandatory waiting period during which they must accumulate a documented compliance record before they can receive coverage. This feature replicates and potentially amplifies the pre-application learning costs that suppressed enrollment in Georgia’s entry-condition design.

The consequences of non-compliance under H.R. 1 are more severe than under either state waiver program. Individuals who lose Medicaid coverage due to work requirement non-compliance are barred from receiving premium tax credits to purchase coverage through the ACA Marketplaces (Hinton, Diana, and Rudowitz Reference Hinton, Diana and Rudowitz2025). This provision eliminates the most accessible alternative coverage pathway for individuals who experience procedural disenrollment, converting an administrative failure into a complete loss of subsidized health insurance. Neither the Arkansas nor the Georgia program included an equivalent penalty, and the addition of this provision can be expected to increase the psychological costs borne by enrollees navigating the compliance process.

The implementation timeline is compressed in ways that compound the risk of administrative failure at the state level. States must begin outreach between June and August 2026, with full implementation by January 2027. Congress appropriated $200 million for implementation, but Colorado alone projects costs of $45.8 million for fiscal year 2026–2027 (Colorado Department of Health Care Policy and Financing 2026), suggesting that the federal appropriation will cover a small fraction of aggregate state-level costs. States that have never operated work requirement programs will need to build verification portals, develop exemption determination processes, integrate data systems, and train eligibility workers, all within approximately 18 months of the law’s enactment. Many already face application processing backlogs and long call center wait times under existing Medicaid caseloads (Brooks et al. Reference Brooks, Alker, Green, Yafimenka and Lin2025).

The policy diffusion literature suggests that when states face compressed timelines to adopt complex policy designs, implementation quality varies substantially based on existing administrative capacity and prior experience with similar programs (Pacheco Reference Pacheco2012; Shipan and Volden Reference Shipan and Volden2008). States that expanded Medicaid early and invested in robust eligibility systems may be better positioned to implement work requirements without generating excessive procedural attrition. States that expanded later or historically underinvested in administrative infrastructure may face greater implementation risk (Grumbach Reference Grumbach2022; Michener Reference Michener2018).

Projected attrition under three design scenarios

The comparative evidence from Arkansas and Georgia, combined with the mandatory parameters established by H.R. 1, allows for a structured projection of how attrition patterns may develop across the 41 expansion states. Three design scenarios are modeled below, based on the discretionary choices available to states within the H.R. 1 framework. These projections are not statistical forecasts. They are structured inferences drawn from the observed relationship between design features and attrition outcomes in the two operational precedents. Table 3 summarizes the three scenarios across the design parameters identified in the administrative rationing framework, linking each configuration to its projected attrition pattern. Each scenario is assessed relative to the CBO’s combined baseline estimate of six million coverage losses from work requirements and six-month redetermination provisions.

Table 3.

Projected attrition patterns under three H.R. 1 design scenarios

Table 3. Long description

The table consists of 11 rows of design parameters compared across three scenarios.

* Verification frequency: Scenario 1 is Maximum permitted; Scenarios 2 and 3 are Six-month minimum.

* Look-back period: Scenario 1 is Three months; Scenarios 2 and 3 are One month.

* Verification method: Scenario 1 is Self-report only; Scenario 2 is Retrospective admin data match; Scenario 3 is Admin data match plus pre-populated.

* Reporting channels: Scenario 1 is Online portal only; Scenario 2 is Online portal; Scenario 3 is Online, telephone, and in-person.

* Exemption determination: Scenarios 1 and 2 use Manual review; Scenario 3 uses Automated cross-referencing.

* Outreach approach: Scenarios 1 and 2 use Informational campaign; Scenario 3 uses Direct application assistance.

* Projected pre-application attrition: Scenario 1 is High (replicates Georgia pattern); Scenario 2 is Moderate (reduced learning costs); Scenario 3 is Low (minimized but not eliminated).

* Projected maintenance attrition: Scenario 1 is High (replicates Arkansas pattern); Scenario 2 is Reduced (burden shifted to state); Scenario 3 is Minimal (automated verification).

* Coverage losses relative to C B O baseline: Scenario 1 Substantially exceeds six million; Scenario 2 Approximates or falls below six million; Scenario 3 is Well below six million.

* Procedural floor: Scenario 1 is No mitigation; Scenario 2 is Partial mitigation; Scenario 3 is Maximum mitigation within statutory constraints.

* Position on friction-rationing continuum: Scenario 1 is Rationing (burden as principal determinant); Scenario 2 is Threshold case (residual entry-stage burden may function as partial rationing); Scenario 3 is Friction (burden as marginal cost within eligibility-driven access).

Note. The CBO baseline of six million reflects combined coverage losses from work requirements and six-month redetermination provisions (Congressional Budget Office 2025; Schneider Reference Schneider2026).

Scenario 1 represents a high-burden design in which a state adopts the maximum verification frequency permitted under the statute, requires self-reported documentation at each compliance checkpoint, and implements a three-month look-back period at application. This scenario replicates the core design features that produced 95% procedural attrition in Arkansas and 54% pre-application deterrence in Georgia. Under this design, the projected attrition pattern would include substantial pre-application deterrence driven by the look-back provision and learning costs, high procedural denial rates at application, and recurring procedural disenrollment at each six-month redetermination. This scenario sits clearly within the rationing range of the framework. Procedural exclusion among substantively eligible individuals dominates participation outcomes, and burden functions as the dominant determinant of who receives coverage rather than as friction within an otherwise eligibility-driven process.

The Marketplace exclusion penalty would further suppress re-enrollment among individuals who experience procedural loss, creating a compounding effect over successive compliance cycles. If attrition patterns resembling Georgia’s 93% enrollment suppression rate were to emerge across a significant portion of the 20 million expansion adults, actual coverage losses could exceed the CBO’s combined 6 million estimate. The CBO projection accounts for behavioral responses and exemptions but may underestimate the compounding effect of procedural failure across multiple compliance cycles, particularly when the Marketplace exclusion eliminates the primary alternative coverage pathway for individuals who experience procedural disenrollment.

Scenario 2 represents a moderate-burden design in which a state adopts the minimum six-month verification frequency, implements retrospective administrative data matching using payroll records and program enrollment data to verify compliance wherever possible, and uses a one-month look-back period. This scenario incorporates the burden-reduction lessons from Georgia’s October 2025 reforms while operating within the H.R. 1 mandatory framework. Retrospective verification shifts the compliance burden from the individual to the state, reducing procedural attrition at the maintenance stage. Coverage losses under this scenario would likely approximate or fall below the CBO’s combined baseline of six million for work requirements and six-month redeterminations. This scenario sits at the threshold between rationing and friction. Burden reduction at the maintenance stage moves the design partway toward eligibility-driven access, but the entry-stage requirements and look-back provision retained under the federal mandate continue to operate as structural filters. Whether this configuration crosses below the rationing threshold depends on how strongly the residual entry-stage burden suppresses participation among the substantively eligible. The H.R. 1 implementation period will provide the evidence needed to classify this scenario more precisely.

Scenario 3 represents a low-burden design in which a state maximizes the use of automated exemption determination and cross-references existing data systems to pre-populate compliance verification. The state also provides multiple reporting channels including telephone and in-person options alongside online portals and implements robust outreach that includes direct application assistance. This scenario pushes the boundaries of state discretion under H.R. 1 and may require favorable interpretation of the forthcoming interim final rule. Procedural attrition would be minimized at each lifecycle stage. This scenario sits within the friction range of the framework. Substantive eligibility criteria remain the primary determinant of who participates, with administrative burden operating as a residual cost that modestly suppresses access at the margins. Even under the most generous interpretation of state discretion, however, the structural features of the federal mandate impose a procedural floor below which burden cannot be reduced through state-level design choices alone. These features include the six-month redetermination cycle, the look-back provision, and the Marketplace exclusion penalty.

The variation across these three scenarios illustrates the central finding of this article. The same statutory eligibility criteria, applied through different administrative architectures, will produce dramatically different enrollment outcomes across states, with some configurations operating as rationing and others as friction. The design choices that states make during the implementation period will function as de facto coverage determinants, shaping who receives Medicaid and who does not in ways that are not specified in the statute (Volden Reference Volden2006).