Benjamin Alarie is the Osler Chair in Business Law at the University of Toronto and the CEO of Blue J Legal Inc. Bettina Xue Griffin is a senior legal research associate at Blue J Legal.
In this article, Alarie and Griffin examine the Tax Court’s decision in Reserve Mechanical and the strength of its appeal on the issue of whether the taxpayer was exempt from tax as a valid insurance company under section 501(c)(15).
Copyright 2021 Benjamin Alarie and Bettina Xue Griffin.
All rights reserved.
The line between legitimate tax planning and unsuccessful tax avoidance is indistinct. Courts routinely grapple with whether a taxpayer has crossed the line from legitimate tax planning to unsuccessful tax avoidance. Legislated exemptions, such as the one provided under section 501(c)(15) for some insurance companies, are a public and open invitation to tax planning.
This kind of tax planning runs the risk, however, of being considered unsuccessful tax avoidance in circumstances in which the taxpayer has a weak or nonexistent business purpose for an insurance arrangement. In examining the bona fides of an insurance arrangement, courts will look behind the formal structure and scrutinize its reasonableness and business purpose to determine if the taxpayer has met the requirements for the insurance exemption.
It can be difficult to predict how courts will proceed when they engage in this examination, particularly because common law tests involve several criteria and ultimately rely on the exercise of judicial discretion. Still, it is now possible to use machine-learning models to gain insight into how courts have approached insurance arrangements based on the totality of the case law. Leveraging insights from a systematic analysis of earlier decisions in the form of a machine-learning model allows researchers to make accurate predictions about how future cases are likely to be resolved. Indeed, Blue J Tax achieves 94 percent accuracy in modeling how insurance arrangement cases will be resolved by the courts.
In this month’s installment of Blue J Predicts, our focus is on the taxpayer’s appeal of the Tax Court’s decision in Reserve Mechanical.1 We examine the strength of Reserve’s appeal on the issue of whether it was exempt from tax as a valid insurance company under section 501(c)(15). The Tenth Circuit heard oral arguments on May 4, and its decision is pending. Based on the facts found by the Tax Court, Blue J predicts with 77 percent confidence that Reserve’s appeal will be dismissed.2
The Tenth Circuit’s pending decision in Reserve Mechanical is expected to have implications for captive insurers and particularly for captive insurers that participate in risk redistribution through risk pools. There are (at least) three issues that the court is expected to address in its decision; we will focus on two of them. The first issue is how important the status of the reinsurer is (rather than the insurer itself) in determining whether the direct insurer should be considered a valid insurance company for tax purposes. The second issue concerns the weight to be placed on the use (or nonuse) of actuarial methods to determine the amount of the premiums to be paid to the captive insurer. The third issue concerns the weight to be placed on the relative importance of there being a circular flow of funds in the arrangement. Our analysis draws on the Blue J Tax machine-learning model of the insurance arrangement case law in addressing the second and third of these issues. We also examine how these two issues interact to make Reserve’s appeal unlikely to succeed.
A captive insurance company is generally a wholly owned subsidiary that is established to insure the parent and, in many cases, the parent’s other subsidiaries. Sometimes a company will create a captive insurer because it cannot obtain suitable insurance coverage elsewhere. In other situations, a company will be able to get suitable insurance from third parties but decides that a captive insurance arrangement is more attractive economically for tax or other reasons, such as inefficiencies in the insurance market caused by moral hazard or adverse selection. Regardless of the precise circumstances, under the right conditions and with good planning, taxpayers stand to obtain favorable tax treatment under section 501(c)(15). This provision creates a tax exemption for insurance companies if gross receipts for the tax year do not exceed $600,000, and more than 50 percent of gross receipts consist of premiums.
The terms “insurance” and “insurance company” are not expressly defined in the IRC, although for some purposes concerning life insurance, an insurance company is any company “more than half of the business of which during the taxable year is the issuing of insurance . . . or the reinsuring of risks underwritten by insurance companies.”3 Consequently, to characterize insurance activities and whether a particular entity constitutes an insurance company, recourse must generally be made to the common law. Common law has established four nonexclusive criteria to determine whether a particular arrangement constitutes insurance for federal income tax purposes: (1) the arrangement involves insurable risks; (2) the arrangement shifts the risk of loss to the insurer; (3) the insurer distributes the risk among its policyholders; and (4) the arrangement is insurance in the commonly accepted sense.4
Let’s now turn to the facts of the Reserve Mechanical case.
A. Reserve Was a Captive Insurer
Peak Casualty Holdings (Peak) was a Nevada limited liability company whose operations included cleaning used mining equipment. Peak was jointly owned by Norman Zumbaum and Cory Weikel. Zumbaum and Weikel engaged the services of Capstone Associated Services Ltd. (Capstone) to form and administer Reserve as a captive insurer to Peak. Capstone issued a feasibility study and prepared the documents to form Reserve as a corporation with a class B general insurance license issued by Anguilla (a territory of the United Kingdom). Reserve was wholly owned by Peak. Zumbaum and Weikel served as directors and executive officers for Reserve.
Although Peak maintained insurance coverage with third-party commercial insurers for general liability, worker’s compensation, and other coverage, Reserve issued excess insurance coverage to Peak.
Capstone’s employees determined the premiums that Reserve would charge Peak for the policies using ratings based on Peak’s annual projected sales. For the relevant years (2008-2010), Reserve issued more than a dozen direct written insurance policies to Peak and two other relatively small and inactive companies related to Peak as the named insureds.
B. Reserve Joined a Risk Pool
Capstone made additional arrangements for Reserve, including facilitating Reserve’s participation in a risk pool with other Capstone insurance companies domiciled in Anguilla. This risk pool was operated by PoolRe Insurance Corp. (PoolRe).5 PoolRe was administered and managed by Capstone and was listed as the stop-loss insurer on each of Reserve’s policies. Thus, Reserve was the lead insurer for the policies, and PoolRe assumed an amount of excess risk. In return, PoolRe would receive a percentage of the premiums paid to Reserve, between 18.5 percent and 19.9 percent, depending on the tax year.
PoolRe entered into similar stop-loss endorsements with more than 50 other Capstone clients. PoolRe pooled the premiums that it was entitled to receive from these stop-loss endorsements and executed reinsurance agreements designed to redistribute those same premiums back to the Capstone entities, known as a quota share policy. Under their respective quota share policies, Reserve and each of the other Capstone entities agreed to assume coverage for a specified portion (quota share) of the risks that PoolRe had assumed (stop-loss pool). Reserve’s quota share was calculated so that Reserve was entitled to receive payments from PoolRe equal to the premiums that PoolRe was entitled to receive from Peak.
III. The Tax Court Decision
The Tax Court found that Reserve’s transactions with Peak were not insurance transactions and that Reserve was thus not entitled to a federal income tax exemption under section 501(c)(15). Specifically, Reserve failed to satisfy the third and fourth criteria of the common law test for insurance arrangements: Reserve did not distribute risk among its policyholders, and the arrangement was not insurance in the commonly accepted sense.
Past cases have examined two dimensions of insurance activity in assessing whether an insurer distributed risk: (1) the number of insureds; and (2) the total number of independent risk exposures. The Tax Court held that Reserve did not pool a sufficiently large number of unrelated risks through its direct insurance policies. While Reserve issued policies to three insureds, the primary insured was Peak. The operations of the other two related insureds were relatively insignificant. In Harper Group, the court found the existence of risk distribution had been satisfied when more than 30 percent of the captive’s gross premiums were derived from insuring unrelated parties.6 This appears to be the lowest judicially approved level of unrelated premiums for captive insurers to demonstrate risk distribution in the case law. However, according to the Tax Court, Reserve could not rely on the reinsurance and quota share policies with PoolRe to satisfy the risk distribution threshold because the court found the transactions with PoolRe were not bona fide and that PoolRe was not a bona fide insurance company. The court found that there was a circular flow of premiums between Peak, Reserve, and PoolRe.
In terms of the requirement that the transactions were insurance in the commonly accepted sense, the court gave significant weight to the fact that the premiums charged were unreasonably high and appeared to be priced in ways that were difficult to justify economically. For example, there was no explanation for why a policy for four years with greater coverage cost only approximately $15,000 more than a policy for one year with half the coverage. Occasionally, Peak spent more for one month of insurance coverage than for a year of coverage. The court noted that the initial feasibility study prepared by Capstone provided no information on the probability of a loss event. Also, the court found that only one claim was ever made under Reserve’s policies, and it was handled in an irregular manner. There was no investigation, no supporting documentation to substantiate the occurrence, and apparently no significant discussion regarding the amount of the claimed loss.
IV. Insights From Applying Machine Learning
In preparation for this analysis of the strength of Reserve’s appeal, we examined 100 cases involving insurance arrangements. We identified the key factual attributes that courts have examined in applying the legal test for insurance arrangements and then assembled a data set of these attributes that were present or not in each of these cases. Armed with this detailed data set of key factual attributes and the outcomes in the decided cases, we evaluated the facts of Reserve Mechanical and evaluated several scenarios for how the Tenth Circuit would be likely to decide the appeal. Given the findings of fact reached by the Tax Court, the Blue J Tax machine-learning model of insurance arrangement determinations predicts with 77 percent confidence that Reserve’s appeal to the Tenth Circuit will be dismissed.
As a legal matter, for Reserve to succeed on appeal, it must establish that (1) it did adequately redistribute risk and (2) that its transactions were insurance in the commonly accepted sense. It is unlikely that Reserve will satisfy either requirement.
A. No Redistribution of Risk
The reinsurance and quota share policies with PoolRe are the sole basis for Reserve to claim that there was an adequate distribution of risk. As Reserve conceded during oral argument before the Tenth Circuit, if the quota share policies are found to be invalid as insurance, then Reserve does not meet the 30 percent threshold.
This is a precarious situation because quota share risk pools like the one in Reserve Mechanical have been criticized by commentators as not sufficiently capable of redistributing risk to satisfy the common law test of insurance.7 Those pools can create a perfectly circular flow of funds that indicate the parties were not actually distributing risk but simply moving funds through intermediary accounts to obtain the tax exemption. These risk pools are by design not tailored to the underlying risk because the one-size-fits-all structure minimizes administrative and accounting costs. As Judge Kathleen Kerrigan stated:
The perfect matching of payments under the corresponding stop loss endorsements and quota share policies (from insureds to PoolRe, and from PoolRe to captives) indicates that the quota share arrangement was not the product of arm’s-length considerations. Peak’s risks that were insured through PoolRe were different from the risks that PoolRe ceded to Reserve under the quota share policies. The risks that PoolRe purported to assume under the stop loss endorsements related to various unrelated business activities and to policies covering various unrelated lines of insurance. Reserve has not shown that the risks were comparable in scale.8
As a legal matter, among the 100 insurance arrangement cases in our data set, there do not appear to be any in which the court explicitly analyzes how participation in a reinsurance or quota share risk pool can satisfy the common law requirement to distribute risk. Perhaps the Tenth Circuit decision will provide clarity on how an insurer could adequately redistribute risk through those pools. Given the remarks of Kerrigan and the lack of precedent for a successful risk redistribution through those pools with judicial approval and the prima facie circularity of the transactions that give the impression that these are not arm’s-length transactions, we are not optimistic that Reserve will succeed on this point.
B. Insufficient Insurance Indicia
The second hurdle that Reserve must overcome is proving that there are sufficient indications of insurance in the commonly accepted sense. One of these factors is whether the premiums were reasonable and the result of an arm’s-length transaction. The Tax Court found the premiums charged by Reserve to be unreasonably high. This factor will likely be the most difficult for Reserve to overcome as it undermines Reserve’s ability to satisfy that the transaction is insurance in the commonly accepted sense. The Tax Court found it significant that Peak paid insurance expenses of $95,828 to Reserve for 2007, and that amount increased to $412,089 for 2008 even though Peak and the two affiliates had no active business operations. These amounts were in addition to the premiums that Peak continued to pay for primary insurance.
Tax practitioners and insurance managers are also wondering how an insurer can demonstrate the reasonableness of its premiums and the importance of doing so in a transparent and replicable way. Relying on actuarial data and industry-accepted rates are obvious answers, but how well will those facts hold up under scrutiny by the courts?
Leveraging the Blue J machine-learning model, we uncovered the following insights on this topic:
Failing to use actuarial data or other objective methods to determine premiums is strongly correlated with an outcome of there being no insurance arrangement.
Blue J’s model shows, moreover, that a demonstration that actuarial methods were used is not necessarily sufficient for Reserve to succeed on appeal should the court also agree with the Tax Court that there was a circular flow of funds.
Blue J’s model also shows that when there is a circular flow of funds between the insured and insurer, the use of actuarial methods may be less meaningful as a factor than in situations in which there was no circular flow of funds.
C. Correlated With No Insurance
Blue J’s software allows users to find cases based on specific legal factors relevant for the analysis. Our analysis suggests that in the 100 cases examined, there is not one in which the parties failed to use actuarial data or other objective methods to determine premiums and the court found the transactions to be valid insurance for federal tax purposes. Although using actuarial data is not determinative of any outcome, the failure to use it seems to be strongly correlated with a finding of no insurance arrangement.
D. Actuarial Methods Not Enough
Reserve emphasized the use of objective premium pricing methods in its briefs and oral arguments. Although the absence of reliance on actuarial methods is strongly correlated with an outcome of there being no insurance arrangement, succeeding in establishing that actuarial methods were used to set premiums is a precarious basis for Reserve’s likelihood of succeeding in its appeal. Our analysis suggests that the outcome of the case is a coin toss (51 percent) if Reserve can establish that it used actuarial methods for setting its premiums.
We examined the facts found by the Tax Court and tested different hypothetical variant scenarios to measure the effect of the use of actuarial methods — and the absence of that use. We examined the interaction of the presence of actuarial methods with findings that there had been a circular flow of funds. Our findings are summarized in Table 1.
Not a Circular Flow of Funds
Circular Flow of Funds
71 percent insurance arrangement
51 percent insurance arrangement
Absence of actuarial methods
38 percent insurance arrangement
23 percent insurance arrangement
We conclude that establishing that there had been the use of actuarial methods in Reserve’s case would be just enough to give Reserve a “more likely than not” chance of success in establishing that it was involved in an insurance arrangement. In an alternative hypothetical in which a circular flow of funds does not exist, success on this factor would put Reserve in a much stronger position.
E. Examining Interaction Effects
The Blue J Tax machine-learning model treats the various factors in a dynamic manner. The effect of any individual fact changes based on the existence or nonexistence of other facts.
We can learn some useful lessons from scrutinizing the results summarized in Table 1. For example, while the use of actuarial methods and industry-accepted rates is a key indicator of a valid insurance arrangement, its effect is weaker in some situations than in others. Our analysis has revealed that using actuarial data to determine premiums is slightly less significant in cases involving a circular flow of funds between the insured and the insurer (or in this case, between the insured, insurer, and reinsurer) than in the absence of a circular flow of funds. In other words, when the transactions are prima facie circular, the fact that the parties used objective data to determine the amount of premiums is given relatively less weight by the courts.
Referring to our earlier machine-learning predictions demonstrates this point. When there was a circular flow of funds, the likelihood of success in Reserve Mechanical dropped by 28 percent depending on whether actuarial methods were used (the change from 51 percent insurance arrangement outcome to 23 percent insurance arrangement outcome). The 28 percent represents the effect of the actuarial methods factor in this factual matrix.
When the facts change so that there is no circular flow of funds, the effect of the actuarial methods factor becomes more significant, at 33 percent (the change from 71 percent insurance arrangement outcome to 38 percent). One possible explanation for this difference is that using actuarial data to determine the amount of premiums is slightly discounted when other significant factors exist that cast doubt on the business purpose of the transactions.
This insight accords with legal intuition given that a circular flow of funds attracts greater scrutiny from the courts and the IRS. As Kerrigan explained in the decision, in cases involving captive insurers, the premiums charged were found to be reasonable “when it can be shown that the amounts agreed upon by the parties were the result of arm’s-length negotiations.”9 This requires looking beyond whether the premiums can be arrived at by actuarial means to consider whether the arrangement was driven by arm’s-length considerations.
Clearly, when there is a circular flow of funds between the insured, insurer, and reinsurer, there is a strong presumption against the existence of arm’s-length considerations. In Reserve’s case, this presumption is bolstered by the fact that all the insurers and reinsurers are clients of Capstone, and many of them are related in some way.
Blue J predicts with 77 percent confidence that Reserve’s appeal will be dismissed by the Tenth Circuit.
Reserve obtained leave from the court to file a supplemental brief after oral arguments concluded that may contain documentary evidence that could potentially be persuasive for the court. The Blue J prediction reported here is limited to the Tax Court’s findings of fact, the positions of the parties in their briefs before oral argument, and their oral arguments.
This article has explained how the Blue J Tax machine-learning model can identify novel patterns and insights on the likelihood of establishing an insurance arrangement using case law. Machine-learning algorithms are well equipped to identify the patterns from past decisions and identify the most salient considerations when courts are assessing whether a taxpayer has crossed that line from legitimate tax planning to unsuccessful tax avoidance.
1 Reserve Mechanical Corp. v. Commissioner, T.C. Memo. 2018-86.
2 To uncover the machine-learning insights detailed in this article, we assume for the purpose of analysis that Reserve will succeed on the issue of whether at least 30 percent of the premiums it received are from unrelated insureds. Additionally, we assume for the purpose of analysis that the Tenth Circuit will not determine the issue of whether Reserve is an insurance company solely on the bona fides of the reinsurer, PoolRe. As both these issues may be determinative of the outcome of whether Reserve is an insurance company, we assume for the sake of analysis that Reserve will satisfy these requirements in order to uncover the impact of other important factors for the purpose of analysis and to arrive at the 77 percent confidence prediction.
4 Harper Group v. Commissioner, 96 T.C. 45, 58 (1991), aff’d, 979 F.2d 1341 (9th Cir. 1992); Amerco v. Commissioner, 96 T.C. 18, 38 (1991), aff’d, 979 F.2d 162 (9th Cir. 1992).
5 PoolRe was initially domiciled in the British Virgin Islands before being redomiciled to Anguilla and obtaining its class B insurance license from Anguilla in April 2009.
6 Harper Group, 96 T.C. 45, aff’d, 979 F.2d 1341.
7 Jay Adkisson, ”Analysis of the IRS’s Big Win Against Risk-Pooled Small Captives in Reserve Mechanical,” Forbes, June 25, 2018.
8 Reserve Mechanical, T.C. Memo. 2018-86 at 41.
9 See also R.V.I. Guaranty Co. Ltd. v. Commissioner, 145 T.C. 209, 231-232 (2015); Harper Group, 96 T.C. at 60.