In the wake of three recent U.S. Tax Court rulings, taxpayers find themselves asking whether captive insurance companies are still a viable risk management vehicle for their business operations.
In each of these decisions, the court determined that the captive insurance companies at issue did not distribute risk and that the arrangement between the affected parties was not insurance in its commonly accepted sense. As such, the Internal Revenue Service continues to pursue captive insurance cases based upon the judicial standard of insurance, as set forth by the U.S. Supreme Court in Le Gierse and modified by the Tax Court in Sears, AMERCO, and Harper, which calls for the following “three prong test”:
- Whether the arrangement involves an insurance risk;
- Whether the arrangement provides both risk-shifting and risk distribution; and
- Whether the arrangement is recognized as insurance in its commonly accepted sense.
While it is imperative that a captive arrangement involve each of these elements, a bigger question centers around the IRS’s definition of what qualifies as the requisite risk distribution based on the judicial standard of insurance and the agency’s position in rulings. To facilitate a discussion around the question of risk distribution we can look at the Tax Court’s recent trilogy of cases that failed the requisite risk distribution test in Avrahami, Reserve Mechanical, and Syzygy; the Tax Court’s historical view of risk distribution; and the IRS’s position on risk distribution.
In Avrahami, the Tax Court’s initial focus was on risk distribution in determining that the arrangement did not constitute insurance for federal income tax purposes. The Tax Court’s decision was based on the determination that the captive did not distribute risk, as there were only 3-4 insured entities and the captive insured only a handful of risks. In the analysis, the Tax Court determined that the amounts from Pan America (the pooling entity) were not considered insurance for federal income tax purpose, as the pooling entity did not qualify as an insurance company under the “three prong test”.
The Tax Court in Reserve Mechanical concluded that the arrangement did not constitute insurance for federal income tax purposes. The Tax Court’s decision was based on the fact that it determined the captive lacked risk distribution, observing that there were only three insured entities and an insufficient number of risks (without specifying the number of risks) in contrast to thousands of risks in Securitas Holdings and Rent-A-Center. Again, the Tax Court challenged the insurance company status of the pooling entity, PoolRe, based on the judicial standard of insurance under the “three prong test”. The court also indicated that the circular flow of funds and the arrangement with PoolRe was not insurance in its commonly accepted sense.
In Syzygy, the Tax Court’s initial focus was on risk distribution in determining that the arrangement did not constitute insurance for federal income tax purposes. The Tax Court’s decision was based on the fact that it determined the captive did not distribute risk. In the analysis, the Tax Court concluded that the fronting carriers were not bona fide insurance companies. Accordingly, the reinsurance of those policies did not distribute risk and the reinsurance of those policies lacked the requisite risk distribution.
The Tax Court has emphasized that it is more important to determine the number of independent risk exposures rather than the number of insureds in each of the above cases; however, the Tax Court failed to indicate what would qualify as a sufficient number of risks and continued to focus on the number of insured entities. This leads one to question how the IRS determines what constitutes risk distribution, which is a key basis of determination in all three cases discussed above. Is risk distribution based on the number of insured entities or the number of insured events? To examine this further and better understand this key consideration in challenges of captive insurance companies’ viability as a risk management vehicle, we can review the court’s historical view with respect to what qualifies as risk distribution, along with the agency’s position on risk distribution.
The Court’s Historical View With Respect to Risk Distribution
While the courts have proffered a significant amount of guidance on the subject of risk shifting, they have not been as forthcoming on the subject of risk distribution. The opinion of the Supreme Court in Le Gierse briefly mentions risk distribution; however, subsequent rulings and court opinions further developed the concept of risk distribution.
Many captive cases have cited the U.S. Court of Appeals for the Second Circuit in Treganowan to explain and distinguish the risk distribution criteria.
“Risk shifting emphasizes the individual aspect of insurance: the effecting of a contract between the insurer and insured each of whom gamble on the time the latter will die. Risk distribution, on the other hand, emphasizes the broader, social aspect of insurance as a method of dispelling the danger of a potential loss by spreading its cost throughout a group. By diffusing the risks through a mass of separate risk shifting contracts, the insurer casts his lot with the law of averages. The process of risk distribution, therefore, is the very essence of insurance.” - Note, The New York Stock Exchange Gratuity Fund: Insurance That Isn’t Insurance, 59 Yale L. J. 780, 784.
In Humana, the U.S. Court of Appeals for the Sixth Circuit ruled that there is no reason why risk distribution could not be present where a captive insures several separate corporations within an affiliated group and losses are spread among the separate corporate entities. In the case, Humana owned more than twenty separate affiliate hospitals and Health Care Indemnity, Inc. (HCI). The appellate court ruled that there was incidence of risk shifting and risk distribution in the insurance arrangement between the Humana affiliates and HCI. While the appellate court determined in this case that that risk distribution could exist in a brother-sister insurance arrangement, it failed to elaborate on the number of risks or insured parties it considered to be sufficient to constitute risk distribution.
In Kidde, the Court of Claims addressed risk distribution in more detail. Kidde Industries, Inc. (Kidde) was a decentralized conglomerate with approximately 15 divisions and 100 wholly owned subsidiaries. In addressing the issue of whether risk distribution occurred with respect to the transaction, the court distinguished risk distribution from risk shifting and suggested that risk distribution is less of a subjective observation and more of a mathematical exercise.
“Risk distribution addresses the risk that over a shorter period of time claims will vary from the average. Risk distribution occurs when particular risks are combined in a pool with other, independently insured risks. By increasing the total number of independent, randomly occurring risks that a corporation faces (i.e., by placing risks into a larger pool), the corporation benefits from the mathematical concept of the law of large numbers in that the ratio of actual to expected losses tends to approach one. In other words, through risk distribution, insurance companies gain greater confidence that for any particular short-term period; the total amount of claims paid will correlate with the expected cost of those claims and hence correlate with the total amount of premiums collected.”
It was held that Kidde Insurance Company (KIC) achieved risk distribution, not necessarily through the reinsurance of multiple unrelated insured entities, but through the total discrete risks that were insured.
On the issue of risk distribution, the Tax Court offered additional guidance in Gulf Oil. According to the court:
[R]isk distribution occurs when “there are sufficient unrelated risks in the pool for the law of large numbers to operate”. As the number of unrelated risks is increased, protection is improved against the chance that the severity and number of harmful events will be spread over time or in other ways in groupings disproportionate to the overall risk. That is, with an increasing number of ventures in a combined pool, the unusually favorable and unusually harmful experiences tend to stay more nearly in balance.”
The Tax Court also observed that “’unrelated’ risks need not be those of unrelated parties; a single insured can have sufficient unrelated risks to achieve adequate risk distribution.” Exploring this concept, the Tax Court noted that as a theoretical matter, risk distribution or pooling requires:
(i) Mass—Sufficient exposures so the law of large numbers takes effect,
(ii) Homogeneity—Sufficiently similar exposures so members of each risk group are exposed in approximately the same degree to a chance of loss, and
(iii) Independence—Separate and distinct exposure units, so that the loss does not affect all or a large segment of the risk group at the same time.
In Gulf Oil, the Tax Court and the Third Circuit held against the taxpayer based on insufficient risk shifting. However, the Third Circuit did not comment on the Tax Court’s statement that one insured could achieve risk distribution with multiple risks. Accordingly, the commentary is relevant to the issue of risk distribution.
In Malone & Hyde, the IRS asserted that the taxpayer had too few insured parties to demonstrate evidence of risk distribution. In the facts of the case, there were eight entities being insured (versus the 22-45 in Humana). The Tax Court stated in its decision that:
“[R]isk distribution is viewed as accomplishing one or more of the following: (1) spreading losses among insureds; (2) providing an insurance company with a pool of premiums from which to pay losses; and (3) serving as a prerequisite to the application of the law of large numbers.” (From Taylor, Taxing Captive Insurance: A New Solution for an Old Problem,” 42 Tax Lawyer 859,876 (1989).)
Citing their statement in Gulf Oil, the Tax Court reiterated:
“risk transfer and risk distribution occur only when there are sufficient unrelated risks in the pool for the law of large numbers to operate.”
Additionally, the Tax Court stated that a single insured, under the right circumstances, could have sufficient unrelated risks to successfully achieve risk distribution. While the Tax Court did not opine as to a specific number of risks or insureds necessary to demonstrate risk distribution, they did express in the opinion that the eight subsidiaries “did sufficiently contribute to the pool of premiums”. Specifically, the insureds provided workers compensation insurance for 6,700 to 7,100 employees, automobile insurance for 1,800 automobiles, and general liability insurance for most of the taxpayer’s warehouses and retail stores.
In Harper the Tax Court found that a brother-sister affiliate insurer, successfully distributed risk where 29% of its business in the year at issue was from unrelated third-party business. Harper (who files a consolidated tax return) organized Rampart, an insurance company subsidiary, to provide marine legal liability insurance to Harper’s subsidiaries. Rampart also provided shipper’s interest insurance to customers using the shipping services of Harper’s subsidiaries. No other lines of business were insured by Rampart. Premium revenues from such customers accounted for 29%, 32%, and 33% of Rampart’s gross premium revenue in 1981, 1982 and 1983, respectively. The IRS disallowed deductions for insurance premiums that Harper’s domestic subsidiaries paid to Rampart during 1981, 1982, and 1983.
Rejecting the IRS’s view, the Tax Court in Harper concurred with the expert testimony of Dr. Neil A. Doherty, a PhD in Economics and a Professor of Insurance at the Wharton School of the University of Pennsylvania, who stated:
“premiums are fungible and indivisible, and once premiums are pooled together, the payor’s identity (i.e., the insured) is lost… . [T]he combination of the premiums … pay [the] total loss; and that’s the essence of the pool or risk distribution process.”
In this case, a relatively large number of unrelated insureds comprised approximately 30% of Rampart’s business thereby creating a sufficient pool of insureds to provide risk distribution. Consequently, the Tax Court held that Harper’s affiliated group was entitled to the claimed deductions for insurance premiums paid by its domestic subsidiaries to the affiliated insurer.
More recently in 2014, the Tax Court held in Rent-A-Center that the captive provided sufficient risk distribution. The captive’s insurance policies in question provided workers’ compensation for more than 14,000 employees, automobile liability for more than 7,000 vehicles, and general liability for more than 2,500 stores. Notably, the majority opinion in Rent-A-Center did not comment regarding how much risk was with any one insured; rather, the court focused on the sufficiency of the number of statistically independent risk exposures.
In Securitas, the Tax Court found that risk distribution existed regardless of the fact that substantially all of the risk resided with a single insured in one year (approximately 90%). The Tax Court’s focus was on whether the insurer (not the insured) distributed risk stating that,
“As a result of the large number of employees, offices, vehicles, and services provided by the U.S. and n-n-U.S. operating subsidiaries, SGRL was exposed to a large pool of statistically independent risk exposures. This does not change merely because multiple companies merge into one. The risks associated with those companies did not vanish once they fell under the same umbrella.”
The Tax Court emphasized that risk distribution is achieved by pooling a large enough number of unrelated risks. Similar to the conclusion reached in Rent-A-Center, the Tax Court’s analysis in Securitas differs from the IRS’s stated position in Situation 4 of Revenue Ruling 2005-40. In situation 4 the agency identified as relevant facts 12 brother/sister affiliates where no single insured represented more than 15% of the risk. These facts were not present in Securitas or relevant in Rent-A-Center.
The courts have taken a number of approaches in assessing whether risk distribution is present. The courts have heard insurance theorists conclude that a sufficient population of independent risks will allow for the law of large numbers to take effect as an appropriate criterion for determining the presence of risk distribution. The courts have also heard that the existence of multiple insureds/policyholders is required; however, the courts have not been definitive in expressing a single view in this area.
IRS’s Position on Risk Distribution
The IRS has touched on the concept of risk distribution in a number of rulings. Much like the courts, the IRS has not put forth a consistent position as to whether the number of insured entities or insured events is determinative in analyzing the risk distribution issue.
In Revenue Ruling 89-61, the IRS explained their understanding of the concept of risk distribution in depth as follows:
When additional statistically independent risk exposure units are insured, an insurance company’s potential total loss increases, as does the uncertainty of the amount of that loss. As the uncertainty regarding the company’s total loss increases, however, there is an increase in the predictability of the insurance company’s average loss (total loss divided by the number of exposure units). That is, by insuring a large number of statistically independent risk exposure units, a company takes advantage of the statistical phenomenon known as the law of large numbers. (When the sample number increases, the probability density function of the average loss tends to become more concentrated around the mean.) Due to this increase in predictability, there is a downward trend in the amount of capital that the company needs per risk unit to remain at a given level of solvency. In this sense, the additional insureds may make a company more efficient in the way its capital provides security, and thus may “help protect the company’s solvency.” Without an increase in capital, however, the increase in the predictability of the average loss is at the cost of an upward trend in the company’s risks of ruin (the probability that total losses may exceed total premiums and capital).
In late 2002, the IRS issued a triad of Revenue Rulings that provided guidance on situations in which arrangements between a corporation and a related insurance subsidiary will constitute insurance for federal income tax purposes.
Revenue Ruling 2002-89 addresses two scenarios in which a wholly owned captive insurance company provides insurance to its domestic parent corporation. In both situations, the parties conduct themselves consistently with the standards applicable to an insurance arrangement between unrelated parties for the insurance subsidiary to either directly insure or reinsure the parent’s professional liability risks.
In Situation 1, the insurance subsidiary earns 90% of its premiums from the parent company and 10% of the premiums from a third-party. The IRS held that this arrangement did not have the requisite risk shifting and distribution necessary for an insurance arrangement for federal income tax purposes. Historically, the courts have consistently held that insurance arrangements between a parent and subsidiary without sufficient third-party insureds are not valid insurance arrangements as the economic burden of loss is not passed from the insured to insurer. Thus, 10% of unrelated risks is not a sufficient volume to sufficiently dilute the impact of a potential loss to the parent.
In Situation 2, the insurance subsidiary earned less than 50% of its premiums from the parent and the remainder of its premiums from third-parties. The IRS held that the premiums of the parent and the third-party insureds were sufficiently pooled to demonstrate risk shifting and risk distribution is evidenced for the arrangement to constitute insurance for federal income tax purposes.
Revenue Ruling 2002-90 addresses a situation in which 12 geographically diverse subsidiaries with a significant volume of diversified, homogeneous risks had procured professional liability insurance from a sister insurance subsidiary. The insurance subsidiary did not insure any third-party risk. Each subsidiary had liability coverage for 5% to 15% of the total risks insured by the insurance subsidiary. The parties conducted themselves in a manner as unrelated parties and the parent, nor any other related party, guaranteed any of the risks insured by the captive.
Relying on precedent set by the Sixth Circuit in Humana and the Federal Claims Court in Kidde, the IRS held that there can be sufficient risk shifting and risk distribution in a brother-sister captive insurance arrangement where the insurance company accepts the risks of the insured and the risks can be sufficiently pooled so the losses relating to a certain subsidiary is substantially borne by the premiums paid by the other subsidiaries. A common parent does not preclude the validity of a brother-sister insurance arrangement.
The IRS formally opined on the validity of a group captive insurance arrangement in Revenue Ruling 2002-91. While addressing the creation of an insurance subsidiary by a small group of unrelated businesses in a concentrated industry, the agency ruled that such an arrangement might be an acceptable insurance arrangement for federal income taxation purposes. The IRS identified as a relevant fact that no member owned more than 15% of the group captive. The agency concluded premiums paid by the members would be deductible as ordinary and necessary business expenses and the group captive was treated as an insurance company for federal income tax purposes.
In Revenue Ruling 2005-40, the IRS examined whether insurance existed for federal income tax purposes under four fact patterns. In Situation 1, a domestic corporation, X, operated a large fleet of automotive vehicles representing a significant volume of independent, homogeneous risks, and insured these risks with Y, an unrelated domestic corporation. X was the only insured of Y. The agency stated that while this situation may constitute the shifting of risks from X to Y, the risks were not in turn distributed among other insureds (e.g., no risk distribution); therefore, the arrangement did not constitute insurance for federal income tax purposes.
In Situation 2, the facts were the same as in Situation 1, except Y also insured the risks of Z, a domestic corporation unrelated to Y or X. Z constituted 10% of the risk insured by Y, and X constituted 90% of the risks insured by Y. Again, the IRS found that while risk shifting may exist in Situation 2, the requisite risk distribution did not exist to constitute insurance. The agency stated that even with two unrelated insureds, there was an insufficient pool of other premiums to distribute the risk among.
In Situation 3, a domestic corporation, X, conducted business through 12 limited liability companies, of which it was the single member. The LLC’s were disregarded entities and entered into insurance arrangements with Y, an unrelated party. None of the LLC’s accounted for less than 5% or more than 15% of the total risk assumed by Y under the agreements. The IRS stated that if an entity is disregarded, its activities are treated in the same manner as a sole proprietorship, branch, or division of the owner. Therefore, applying that rule to Situation 3, Y had entered into an insurance arrangement only with X. As a result, the arrangement between X and Y did not constitute insurance for federal income tax purposes.
In Situation 4, the facts are the same as in Situation 3, except that each of the 12 LLC’s elected to be classified as an association for federal income tax purposes. The arrangement between Y and each LLC then constituted the shifting of a risk of loss from each LLC to Y. The risks of the LLC’s were distributed among various other LLC’s. Therefore, the arrangement constituted insurance for federal income tax purposes. Additionally, because the arrangement with the 12 LLC’s represented Y’s only business, and those arrangements were insurance contracts for federal income tax purposes, Y was an insurance company within the meaning of IRC Section 831(c) and Section 816(a), and the 12 LLC’s were entitled to deduct the amounts paid as insurance premiums under IRC Section 162.
In Revenue Ruling 2008-8, the IRS examined whether insurance existed for federal income tax purposes in a Protected Cell Company (PCC). In Situation 1, a domestic corporation, X, insured its professional liability risks with Cell X, an unrelated domestic corporation. X was the only insured of Cell X. The agency stated that in the absence of unrelated risk, Cell X lacks both risk shifting and risk distribution; therefore, the arrangement did not constitute insurance for federal income tax purposes.
In Situation 2, the facts were the same as in Situation 1, except Y, a domestic corporation, owned 12 domestic subsidiaries and insured the professional liability risks of each subsidiary with Cell Y. Each subsidiary was geographically dispersed and operated on an independent basis. The facts indicated the 12 subsidiaries had a sufficient volume of independent, homogenous risks, and none of the subsidiaries had less than 5% or more than 15% of the risk. The IRS stated that there was a sufficient pool of premiums to distribute the risk and the risk was shifted to Cell Y; therefore, the arrangement constituted insurance for federal income tax purposes.
What These Positions Tell Us About Risk Distribution
The presence of the risk distribution component of an insurance arrangement is both a qualitative and quantitative determination. The exposures shifted to an insurer must be separate and diverse yet must also be voluminous enough for the law of large numbers to take effect. All but one of the aforementioned captive insurer cases involve multiple, albeit related insureds. While the courts have indicated that a single insured could conceivably contribute enough diversified risk to create risk distribution, they have yet to rule specifically on this point. As so clearly discussed by the Court of Claims in Kidde, risk distribution requires application of the law of large numbers and multiple potential risks of loss at the insured level.
While the assertions of the courts in Humana, HCA, and Kidde demonstrate incidences of distribution, they do not discuss the considerations of risk distribution with a single insured. They have even found that risk distribution occurred, in part, based on the number of insureds. Perhaps a fallacy of reasoning in the aforementioned cases is the potential misapplication of the law of large numbers. While the law of large numbers is not a legal statute but a scientific theorem, it is fairly obvious that the composition of the risks, rather than the number of contributors to the pool of risks, provides a proper basis for the evaluation of risk distribution. The law of large numbers provides that as the independent units of exposure increase, the likelihood that the actual percentage of units that suffer the loss will not differ by more than a specified amount from the estimated probability that each unit will suffer a loss. This is consistent with the analysis of the law of large numbers and risk distribution by Samuelson in his essay Risk and Uncertainty: A Fallacy of Large Numbers (Scientia, vol. 57 (April–May 1963), pp. 1–6.)
Whereas Gulf Oil explores and Malone & Hyde comments on the potential for a captive insurance company with one insured to experience risk distribution, this concept has yet to be applied. It is significant to note that the IRS has not commented on the number of insureds or the number of risks that must be present to constitute a valid insurance arrangement. Rulings, such as Revenue Ruling 2002-90, Revenue Ruling 2005-40 (Situation 4), and Revenue Ruling 2008-8 (Situation 2) have utilized multiple insureds and adopted the balance sheet and net equity approach. These rulings utilized the “three prong test” to determine the validity of such arrangement and failed to issue authority requiring a minimum number of insureds or insured risks for sufficient risk distribution.
As a familial relationship, even one of parent-subsidiary is not a perfunctory preclusion to an insurance arrangement for federal income tax purposes; the IRS is also emphasizing the validity of the Moline Properties doctrine, which states that affiliated corporations should be treated as separate taxable entities.
Risk distribution in an insurance arrangement is achieved not as a function of the number of contributors of risks but a function of a combination of the volume and independency of risks. Risk distribution will occur within a population of risks to which the law of large numbers can be adequately applied. When a captive insurer with one insured can demonstrate that a sufficient volume of premiums have been accumulated against a separate and diverse group of exposures, the courts and IRS indicate that risk distribution is present. Provided all other indications of a valid insurance arrangement are present, such as whether an insurable risk is present, the existence of risk shifting, and an arrangement of insurance in its commonly accepted form, insurance premiums paid to the related entity could be deductible as ordinary business expenses under tax code Section 162.
What’s Next for the Taxpayer?
As discussed above, there is no definitive authority discussing what constitutes risk distribution for federal income tax purposes. In determining the presence of risk distribution, the courts have considered its existence under two fact patterns: (1) where there existed an adequate number of insured entities, and (2) where there existed an adequate number of insured risks.
The IRS’s early ruling position when discussing the issue of risk distribution has been consistent with an analysis of the number and physical diversity of the insured risks. (See PLR 200121019 and Revenue Rulings 89-61 and 60-275.) In more recent Revenue Rulings, the agency identified as relevant facts 12 brother/sister affiliates where no single insured represented more than 15% of the risk and the IRS found risk distribution to exist. (See Revenue Rulings 2000-90 and 2005-40.)
In light of the recent strikes against the taxpayers in Avrahami, Reserve Mechanical, and Syzrgy, which failed the requisite risk distribution test, taxpayers are left questioning the IRS’s position with respect to risk distribution. Based on the court’s historical view of risk distribution and the agency’s stated position on risk, is it the number of insured entities or the number of statistically independent risks? Additionally, if it is more important to determine the number of independent risk exposures rather than the number of insureds, exactly how many risk exposures are required for adequate risk distribution?
Without a conclusive definition of risk distribution, there is no absolute way to conclude on what ultimately qualifies as the requisite risk distribution without further guidance from the IRS. However, there is an opportunity for taxpayers to present a captive insurance case where all other components of a valid insurance arrangement are present, and the captive (with only a few affiliated insureds) either has a sufficient number of independent risk exposures on its own or utilizes a pooling entity for purposes of achieving risk distribution. Under such fact patterns, the IRS will be obligated to opine specifically with respect to the issue of risk distribution for a captive with a few insureds with no pooling entity or a captive that utilizes a pooling entity to achieve risk distribution and comment further on the issue of whether it is more important to have a sufficient number of independent risk exposures or a specific number of insured entities.
This column does not necessarily reflect the opinion of The Bureau of National Affairs, Inc. or its owners.
Laurie Bizzell is a Senior Tax Manager in the Insurance, International and Private Equity Tax Practices at Bennett Thrasher LLP where she provides tax consulting and compliance services to U.S. and foreign alternative investment funds, captive insurance companies and structuring U.S. and foreign companies in various tax jurisdictions.She may be reached at (678) 218-1312 or Laurie.Bizzell@btcpa.net.