Whiplash—Diverging Captive Insurance Rulings and Another IRS Victory

Editor's Note: This article, authored by Matthew Queen and published on behalf of the 831(b) Institute, reflects the Institute's advocacy efforts on behalf of small and midsized businesses that utilize 831(b) captives for risk management. It examines the potential implications of the CFM Insurance, Inc. v. Commissioner ruling in light of Swift v. Commissioner, including differing judicial interpretations of risk distribution and the "commonly accepted notion of insurance."

As legal interpretations and regulatory approaches continue to evolve, we encourage readers to stay informed on industry developments, consider multiple viewpoints, and consult with qualified professionals to gain a comprehensive understanding of the potential implications for captive insurance.

The latest Tax Court captive insurance holding, CFM Ins., Inc. v. Commissioner, T.C. Memo. 2025–83, explicitly contradicts the Fifth Circuit's holding in Swift v. Commissioner with regard to risk distribution, yet still hands a victory to the Internal Revenue Service (IRS).

This pair of opinions drives the concept of "risk distribution" into an increasingly complex and difficult-to-reconcile area. While the court delivered a surprising win to the taxpayer on the element of risk distribution, it spilled pages worth of ink on an extended analysis of whether the transaction constituted "insurance in the commonly accepted sense" in order to invalidate the 831(b) transaction.

This case focused on an 831(b) captive called "CFM," which provided a variety of coverages for a chain of grocery stores. The IRS challenged CFM's § 831(b) election, which allows small insurance companies to exclude premiums from taxable income, and the deductibility of premiums paid by the grocery stores.

Neither the Internal Revenue Code nor its regulations define "insurance," leaving federal courts to rely on common law. The foundational framework stems from the Supreme Court's 1941 decision in Helvering v. Le  Gierse, which established that insurance fundamentally involves a two-part test reviewing whether a transaction possesses the twin elements of risk shifting and risk distribution.

Over time, the Tax Court legislated from the bench and elevated dicta from Le Gierse to expand the inquiry into a four-part test, adding the prong of whether the transaction includes an insurable risk and the requirement to meet the "commonly accepted notions of insurance."

The Le Gierse holding, a life insurance case predating World War II, does not fully reflect the nature of property and casualty insurance. First and foremost, risk shifting is not an element of modern commercial property and casualty coverage. The very existence of large deductibles, coinsurance clauses, and self-insured retentions presupposes that risk retention is an integral part of the business, challenging the Le Gierse bedrock assumption that "risk shifting" is axiomatic to "insurance" for federal tax purposes.

Ironically, federal courts seem to agree. The analysis of risk shifting is dismissed so long as an 831(b) captive demonstrates that it possesses the minimum statutory capital required by law. Since funding the minimum statutory capital is a prerequisite to the issuance of a captive insurance license, this is a rule without consequence. Accordingly, all captive insurance companies possess risk-shifting. Since risk shifting is part of the core of the Le Gierse holding, the Le Gierse holding is substantially diminished in relevance over time. Further adherence to this case can distort what constitutes insurance for federal income tax purposes.

A novel argument in CFM Insurance involved the McCarran-Ferguson Act, which provides that the business of insurance is state law. CFM argued that the Act mandated federal courts to defer to Utah's determination that CFM was a valid insurance company. The court rejected this, distinguishing between "insurance" for tax purposes and "the business of insurance" regulated by states, asserting that federal tax consequences do not invalidate state law.

This portion of the CFM holding is, in the author's view, open to question. The IRS can significantly affect the business of insurance by shutting down captive insurance programs based on its interpretation of insurance. While the IRS does possess the power to prosecute abusive tax transactions, it has, in some cases, applied its authority to define insurance in ways that many in the industry view as problematic.

Regardless of the McCarran-Ferguson Act, the CFM court did find that the transaction had sufficient risk distribution. The court's conclusion heavily relied on each expert witness, including the IRS's witnesses, all agreeing that the total units of risk within the insurance transaction included the total of the customer transactions and products sold. However, the Tax Court realized that this conclusion is at odds with the Swift court's opinion and included language that this conclusion is unique in order to allow future cases to ignore this precedent on an as-needed basis.

There is no way to reconcile the holding on risk distribution from the CFM court with the Swift court's holding. The cases stand at odds with each other. Further, both cases reveal basic errors in mathematics.

Risk distribution is rooted in the statistical concept of the law of large numbers (LLN). The LLN states that as your sample pool increases, the mean (or average) of that pool gets closer to the total population mean. In the author's view, federal courts have misapplied certain statistical principles.

Page 36 of the opinion states as follows.

In all our previous microcaptive cases, we came to the same conclusion—there wasn't a large enough pool of unrelated risk for the policies issued to the related entities to satisfy the law of large numbers. Id. at 1213–14; see also Avrahami, 149 T.C. at 181–82 (seven types of policies to four entities insufficient); Syzygy, 117 T.C.M. (CCH) at 1169 (eight policies to one entity insufficient); Rsrv. Mech. Corp., 115 T.C.M. (CCH) at 1479–80 (eleven to thirteen policies for three 3 entities insufficient). [Emphasis added.]

Note again on pages 38 and 39 this language: "When asked about the specific policies that CFM issued, she couldn't answer even with a range of how many exposures would suffice to trigger the law of large numbers."

The highlighted portions of the CFM case reveal the issue. Federal courts, including the Fifth Circuit in the recent Swift decision, fail to grasp that the LLN is a continuous function. The LLN is not a "yes or no" sort of thing. It is always there. The LLN applies with a sample size of 1 unit or 1 billion units.

The only difference is that the mean of the group gets closer to the mean of the total population of the group as you add units of exposure to the sample. In other words, you can get closer to the true average of the population, but you'll never get there because you never have the full population in an insurance risk pool.

Then we see a reference on page 50 where the Tax Court mentions that the IRS's expert "testified that for certain policies as few as 30 exposure units would be sufficient to adequately distribute risk." This is noteworthy because the number 30 is a special number with regard to the central limit theorem (CLT).

The CLT provides an estimate of the variability of an average. So, if our LLN function yields an average expected loss cost of $X, then we can leverage the CLT to say that we expect the loss to be $X within some percentage up or down. Application of the CLT generally requires 30 units of exposure. This is not the same thing as the LLN but does broadly suggest the following: (1) risk distribution does not need thousands of units of risk, and (2) there is more than one way to achieve it, and assuming the LLN is the sole function by which to do so is a misinterpretation of actuarial science.

Despite the surprising win on risk distribution, CFM ultimately failed the "commonly accepted notion of insurance" prong, which doomed the transaction. This element involves a qualitative assessment of whether the company operates like a genuine insurance enterprise in its day-to-day activities. The Tax Court provided a laundry list of alleged managerial errors or purportedly slipshod work from the captive's operations as evidence that the transaction was not insurance in the commonly accepted sense. Ultimately, these criteria amount to a litany of perceived errors, both minor and somewhat notable, supporting what appears to be a predetermined conclusion. It is difficult to reconcile the court's analysis of the "commonly accepted sense" criteria with the facts presented. The definition of "insurance" should not hang in the balance of minor, inconsequential administrative details.

Although the case is another loss for the taxpayer, it is encouraging. The finding of risk distribution in this case, on the exact grounds rejected in the recently published Swift case, reveals continuing inconsistencies in how insurance is defined and governed by the courts. Moreover, the entire case line of "insurance in the commonly accepted sense" rests on an increasingly irrelevant life insurance case arising from another era. The courts remain divided on how to review captive insurance transactions. Irreconcilable court opinions generally result in a resolution, either via legislation or a higher court's intervention.