Level Funded, Fully Insured, Self-Funded: Three Architectures, Not Three Products

Industry conversations place level funded on a spectrum between fully insured and self-funded, as if these were product tiers differentiated by complexity and risk tolerance. A broker might say level funded is “like fully insured but with upside,” or “self-funded with training wheels.” These framings are wrong in a way that produces real confusion about what level funded can and cannot do. Fully insured, self-funded, and level funded are not three products on a continuum. They are three architectures with different risk ownership structures, different regulatory treatment, and different capital requirements. The failure to understand the architectural distinction leads to purchasing decisions made on the wrong criteria.

The Fully Insured Architecture

#

In a fully insured arrangement, the employer purchases a group health insurance policy from a licensed carrier. The carrier owns 100 percent of the claims risk. The employer pays premium. The premium belongs to the carrier whether claims are high or low. This is not a nuance. It is the defining structural feature: the money moves one direction and does not come back.

The carrier holds statutory reserves and surplus against future claims obligations. These are carrier assets, maintained under state insurance department solvency requirements, and they are not employer assets under any legal theory. When a fully insured group has a favorable claims year, the carrier retains the difference between premium collected and claims paid. That difference funds carrier reserves, carrier operations, and carrier profit. The employer’s only recourse for favorable claims experience is the possibility of a lower renewal premium, which the carrier is not obligated to provide.

The employer receives almost no claims data. State insurance regulations govern what carriers must and may disclose, and the standard practice in the small group fully insured market is to provide aggregate utilization summaries at the carrier’s discretion. An employer in the fully insured market typically cannot obtain claims data at a level of detail sufficient to evaluate plan design, compare provider costs, or make informed decisions about coverage changes. The data asymmetry is structural. It follows from the fact that the carrier, not the employer, owns the risk and therefore controls the information.

Regulatory treatment reinforces the ownership structure. Fully insured plans are regulated by state insurance departments. State mandated benefit laws apply, requiring coverage of specific conditions, treatments, providers, and services that vary by state and can number in the dozens. State premium taxes apply, generally ranging from approximately 1.75 to 4 percent of premium depending on the state, with 2.5 percent the most common rate. State rate review processes apply, meaning the carrier must file rates with state regulators and justify increases. The National Association of Insurance Commissioners collects data on carrier medical loss ratios in the small group market, and the ACA’s medical loss ratio requirements (80 percent for small group) constrain but do not eliminate carrier profit on fully insured products.

The Traditional Self-Funded Architecture

#

In a traditional self-funded arrangement, the employer assumes the financial risk for providing health care benefits. There is no carrier in the insurance sense. The employer funds claims directly from operating capital or a dedicated health benefit fund. Claims are paid as they are incurred, and the employer absorbs all variance.

The employer contracts with a TPA or establishes an Administrative Services Only arrangement with a carrier. In an ASO arrangement, the carrier provides its network, claims processing systems, and administrative infrastructure but does not bear insurance risk. The distinction between ASO and fully insured is the location of risk: in ASO, the carrier runs the operation, but the employer owns the claims. In fully insured, the carrier runs the operation and owns the claims.

Stop loss is purchased separately in the traditional self-funded model, if it is purchased at all. Large self-funded employers with thousands of employees may forego stop loss entirely because their covered population is large enough that claims variance is statistically manageable. The law of large numbers works in their favor. For a group of 5,000 employees, the probability that actual claims deviate significantly from expected claims in any given year is low enough that the employer can budget for health care costs with reasonable confidence. For a group of 25 employees, the same statistical protection does not exist. One catastrophic claim can consume the entire annual health care budget.

This statistical reality is why traditional self-funding was historically limited to employers with 100 or more employees. The capital requirement for absorbing claims variance at small group sizes is disproportionate to the potential savings. A 25-person employer that self-funds without stop loss needs to hold cash reserves large enough to cover a worst-case claims year, and for a small group, the worst case is significantly worse, relative to the expected case, than it is for a large group. The variance problem is the structural barrier that kept small employers in the fully insured market for decades.

Self-funded ERISA plans are subject to federal regulation under the Department of Labor and are exempt from state insurance regulation. The ERISA preemption that creates this exemption is the same preemption that applies to level funded plans structured as self-funded, and LFP-01.03 examines it in detail. The practical consequences are significant: state mandated benefits do not apply, state premium taxes do not apply, and the employer operates under a single federal regulatory framework regardless of how many states its employees work in.

The Level Funded Architecture

#

Level funded borrows from both architectures and adds a structural element that neither possesses on its own.

From fully insured, level funded borrows a single monthly payment that provides budget predictability. The “level” in level funded is this payment: a fixed monthly amount, set by underwriting, that the employer pays on the same schedule and through the same administrative process as a fully insured premium. Level funded also borrows the bundled product experience. The employer does not separately procure stop loss, TPA administration, and network access. These are packaged into a single product offering by a TPA partnered with a stop loss carrier, or by a carrier offering a level funded product line.

From self-funded, level funded borrows employer ownership of the claims fund, ERISA preemption and the regulatory treatment that follows from it, access to claims data, and plan design flexibility that allows the employer to customize benefits within ERISA and federal regulatory requirements rather than accepting the state-mandated benefit framework that applies to fully insured plans.

What level funded adds is the bundle itself and the accessibility it creates. The claims fund, stop loss, and administration are packaged into a single product by a single entity. The aggregate stop loss, included by design rather than purchased separately, defines the employer’s maximum annual liability. The combination of bundling and a defined liability cap makes the self-funded architecture viable for employers with five to fifty employees who could not manage the components separately and who lack the capital reserves to absorb claims variance without aggregate protection.

Level funded is not a compromise between fully insured and self-funded. It is the self-funded architecture made accessible to small groups through bundling and mandatory stop loss packaging. The distinction matters because a compromise implies a midpoint, something that is partly one thing and partly another. Level funded is structurally self-funded. The employer owns the claims fund. The plan operates under ERISA. The claims data belongs to the plan. The regulatory treatment is federal, not state. The only thing level funded shares with fully insured is the payment mechanism: a fixed monthly amount that looks and feels like a premium. Everything underneath that payment is self-funded architecture.

Why the Distinction Is Structural

#

The consequences of misunderstanding the architecture are practical and financial.

Surplus. A fully insured employer has no surplus claim. The premium is gone. A level funded employer may have a surplus claim, depending on contract terms. A self-funded employer owns all surplus by definition. An employer who does not understand the architecture may not realize they are entitled to surplus return in a level funded plan, or may not scrutinize the contract terms that govern whether and how surplus is returned. The reconciliation mechanics in LFP-01.05 examine the variation in surplus treatment across the market.

Regulation. An employer who treats level funded as “basically fully insured” may not understand that state mandated benefit laws do not apply to their plan. They may include benefits they are not required to include, or they may fail to realize they have the flexibility to design a plan that fits their workforce rather than accepting a standardized state-mandated benefit set. In the other direction, the same employer may not understand that they have accepted fiduciary responsibility under ERISA, with legal obligations of loyalty and prudence that fully insured employers do not bear. Fiduciary exposure is a real liability that most small employers sponsoring level funded plans do not know they carry.

Data. The claims data a level funded employer receives is not a feature added to an insurance product. It is a structural consequence of the architecture. The employer owns the plan. The plan generates claims. The claims generate data. The data belongs to the plan. In fully insured, the carrier owns the plan and the data. The difference is not about carrier generosity or product features. It is about who owns the risk, and therefore who owns the information that risk generates. That data enables plan design changes: adjusting deductibles, adding or removing covered services, evaluating whether a high-cost specialty drug should be covered under the plan or managed through a carve-out. It enables vendor evaluation: comparing network discount performance across TPAs, assessing PBM pass-through rates, measuring utilization management effectiveness. None of this is possible in fully insured, where the employer receives aggregate summaries at the carrier’s discretion and has no mechanism to compare what they are paying against what they could be paying under a different arrangement.

Capital structure. The three architectures impose different capital requirements on the employer, and these requirements constrain which employers can operate under which architecture. Fully insured requires no employer capital reserve. The premium is the employer’s entire financial obligation. Traditional self-funded requires capital reserves sufficient to absorb claims variance, which for a large employer might mean a dedicated health benefit fund of several million dollars. Level funded requires no capital reserve beyond the monthly payment, because the aggregate stop loss defines the employer’s maximum liability. This is the structural innovation that opened self-funding to small groups: not the self-funded architecture itself, which has existed for decades, but the aggregate stop loss packaging that eliminates the capital reserve requirement. The Kaiser Family Foundation’s survey data shows that approximately 63 percent of all covered workers are in self-funded plans, with the rate reaching 79 percent at firms with 200 or more employees and approximately 20 percent at firms with three to 199 employees. The level funded segment is growing within that smaller-firm category, with 36 percent of covered workers at small firms enrolled in level funded plans as of 2024. The disparity in self-funding rates by employer size reflects the capital barrier that level funded is designed to lower.

The practical consequence of treating these architectures as products is that purchasing decisions get made on the wrong questions. The product frame asks: which is cheaper, which is simpler, which has the best benefits. The architecture frame asks: who owns the risk, who owns the data, what happens to the surplus, what regulatory framework governs the plan, and what obligations has the employer accepted. An employer who answers the product questions without answering the architecture questions does not understand what they bought.