Captives: The Harper Test

Does the Arrangement Involve Insurance Risk?

The earlier captive insurance cases involved standard insurance coverages, such as commercial auto, property, general liability, workers compensation, and product liability. There was no need for any of these decisions to define "insurance risk." But in R.V.I. Guar. Co. v. Commissioner, 145 T.C. 209 (2015), the Treasury argued the company's primary insurance policy wasn't insurance, making necessary a discussion of this factor.

Petitioner R.V.I. Guaranty Co., which took its name from the acronym for residual value insurance (RVI), indemnified the insured if the value of an asset dropped below a certain dollar figure. The court provided the following explanation:

Assume that an automobile with an initial purchase price of $20,000 is leased for 3 years and that its expected residual value upon lease termination is $10,000. RVI might insure that automobile for 90 percent of the expected residual value, yielding an insured value of $9,000. If, at lease termination, the automobile had an actual residual value of $8,500, the RVI policy would indemnify the lessor for $500, assuming the lessor satisfied all terms and conditions of coverage. The lessor would bear the $1,000 initial layer of loss.

Source: Id at 6.

This policy has tremendous utility for leasing companies as it helps them stabilize cash flows, increasing income predictability.

To determine if this was an insurance risk, the court looked at the following factors.

• Is there a common law or statutory history of courts treating a risk as insurance? The R.V.I. court noted that New York and Connecticut statutorily defined this coverage. Washington State precedent approving this policy was also quoted.
• Did the state regulators approve the policy? All companies include a policy sample (along with an actuarial analysis) of their proposed policies. Connecticut approved the policy as part of the review process.
• Did the company's annual auditor approve the policies? Every year, accountants and actuaries provide various opinions on the financial state of insurance companies. These professionals must formally attest to various aspects of the company's operations and structure, including a determination that the insurer "bore a significant 'insurance risk.'"

In R.V.I., the court determined that R.V.I.'s residual value policy complied with these three factors, making these risks valid insurance risks.

Paradoxically, the court in Avrahami v. Commissioner, 149 T.C. 144 (2017), which discussed various insurance policies issued by the micro-captive, did not use the above methodology. Instead, the court noted whether the US Treasury believed a policy was valid, followed by a discussion of pricing. It could be argued the discussion in that case was more focused on pricing. Yet, several discussions noted the parties agreed that a specific coverage was valid. There was no discussion as to why the RVI methodology was not used.

Is There Both Risk Shifting and Risk Distribution?

Risk shifting occurs when the financial harm caused by a fortuitous risk moves from the original corporate location to another company. Consider the following examples.

• A company with 100 employees purchases workers compensation insurance from an unrelated insurer. When an employee is injured on the job, the insurer pays the losses.
• A company with 100 trucks purchases commercial auto coverage. When a truck gets into an accident, the insurer pays for the loss.

In both cases, an independent third party pools the risk with similar unrelated exposures to distribute the loss over a large body of insureds. In each case, risk has shifted.

However, self-insurance raises several thorny and nuanced complications. From a tax perspective, a captive looks suspiciously like an accounting reserve achieved through a series of preplanned steps. ¹ Regardless of the path, reserve contributions are nondeductible. ² The lack of nonparent risks adds to the complications ³ as does the insured's control over the funds. ⁴ For these reasons, captives faced an uphill battle in Tax Court. The US Treasury was not alone in its skepticism; Irving Pfeffer noted in Insurance and Economic Theory, "Self-insurance is no more insurance than is education or any other uncertainty-reducing institution." ⁵

The Treasury announced its opposition to all self-insurance structures in Revenue Ruling 77-316, which was based on three fact patterns.

• A parent company purchases insurance from a wholly owned subsidiary.
• A parent company purchases insurance from a third-party insurer, which then reinsures some of the risk with a captive owned by the parent.
• A parent company purchases insurance from a captive, which then reinsures some or all of its risk with a third-party insurer.

Only two sentences in the ruling provide any hint at the underlying reasoning:

In each situation described, the insuring parent corporation and its domestic subsidiaries, and the wholly owned "insurance" subsidiary, though separate corporate entities, represent one economic family with the result that those who bear the ultimate economic burden of loss are the same persons who suffer the loss.

This hints that the captive is a de facto accounting reserve disguised as an insurance company—an observation bolstered by the ruling's concern that the parent company maintained "control of the corporate family members." Revenue Ruling 77-316 also intimates that the captive transaction violates anti-abuse law by noting the reasoning is based on the "economic reality of each situation described." ⁶ Unfortunately, the ruling provides no additional detail supporting its contentions, leaving that to the courts.

The Treasury's victories in the earliest captive cases exclusively focused on risk shifting and were based on 77-316's "economic family" doctrine. Four observations are warranted before explaining the reasoning of those decisions.

• The Internal Revenue Service (IRS) was attacking large—and, in some cases, familiar—companies such as the Humana Hospital Chain, Beech Aircraft, and the Carnation Company.
• The programs insured standard insurance risk, such as property, ⁷ workers compensation, ⁸ and product liability. ⁹
• All were formed for nontax reasons. ¹⁰
• Several employed routine insurance structures such as fronting ¹¹ or excess of loss arrangements. ¹²

It is difficult to conceive that companies of such size and sophistication, insuring common risks for legitimate business reasons and employing standard insurance mechanisms, would somehow be unable to self-insure some or all of their insurance risk. Yet, the courts ruled in five decisions that the challenged structures lacked risk shifting.

The first two victories—Carnation and Stearns-Roger—contained an unintentional assist from the taxpayers in the form of a parental guarantee. Each company guaranteed they would contribute additional capital to their respective captives should those funds be needed to pay claims. ¹³ This created a circular flow of funds that voided risk transfer. Should a captive seek additional parental capital, the parent would provide the same but then quickly receive the contribution back in the form of a claim payment.

Taxpayers argued the Treasury was mischaracterizing their wholly owned insurers as an accounting reserve, ignoring the corporate separateness of these structures. ¹⁴ Their cases were primarily based on the doctrine expressed in Moline Properties, Inc. v. Commissioner, 319 U.S. 436, 438-439 (1943), which holds the following:

The doctrine of corporate entity fills a useful purpose in business life. Whether the purpose be to gain an advantage under the law of the state of incorporation or to avoid or to comply with the demands of creditors or to serve the creator's personal or undisclosed convenience, so long as that purpose is the equivalent of business activity or is followed by the carrying on of business by the corporation, the corporation remains a separate taxable entity.

The courts disagreed, ruling that they were simply reclassifying the transaction for its substance, which is a common occurrence in tax law. ¹⁵

In justifying their decisions that no risk shifting existed, the courts eventually settled on the balance sheet theory, which is based on the following steps.

1. The parent company owns all the captive's stock.
2. When the captive pays a claim, the value of its stock declines by the amount of the claim payment.
3. This decline in stock value is reflected on the parent company's balance sheet.
4. Because the stock declines by the same amount as the claim payment, and that decline is reflected on the parent company's balance sheet, there is no risk shifting.

Though simplistic, this idea is consistent with a very basic definition of book value as taught in accounting. (Generally speaking, "book value" is assets minus liabilities.) The balance sheet theory is limited to situations where the parent company is the sole owner of the captive's stock. When the captive is owned by multiple corporate subsidiaries, risk shifting exists. ¹⁶

The IRS withdrew the general application of the economic family doctrine in Revenue Ruling 2001-31, writing, "No court, in addressing a captive insurance transaction, has fully accepted the economic family theory set forth in Rev. Rul. 77-316." This is an oddly disingenuous statement; it is true that no court formally agreed to use the term "economic family"—and one court strenuously objected to using the term. ¹⁷ But the concept that risk did not shift so long as the parent owned all the stock of the captive was the basis for five decisions, no matter the name used to describe the underlying legal theory. And a concurring opinion in Clougherty noted that Carnation was an economic family case in all but name. ¹⁸

As no court formally ruled against the economic family doctrine, and since there are still five cases that hold it a valid reason to disallow a payment to a captive insurer, it is conceivable the IRS could reassert it in future litigation or in an audit with the intent of forcing the taxpayer to expend funds in its defense.

Defining Risk Distribution

There are two distinct historical threads defining risk distribution. The first began in the early 20th century writings of insurance academics who argued insurers must pool enough risk for the law of large numbers (LLN) to apply to the statistical results. But as early as 1923, Charles O. Hardy, in Risk and Risk Bearing, noted that Lloyd's—the largest alternative insurance market—was not LLN compliant. Later insurance authors also noted the LLN fallacy.

The second thread originated in captive insurance jurisprudence, where the tax courts defined risk distribution in accounting terms, ¹⁹ requiring third-party funds to comprise a minimum percentage of every claim payment. More recent decisions use the concept of statistically independent risk in which the courts count the number of things insured to determine if the pool is large enough for risk distribution to exist.

Allan Willet's The Economic Theory of Risk and Insurance contains one of the earliest conceptual mentions of risk distribution. He noted insurance requires "the combination of the risks of many individuals," which "brings about … a reduction in uncertainty"—an idea that is intuitively satisfying. If a single homeowner self-insures fire risk, they are potentially liable for the house's fair market value. By combining their funds with nine other homeowners, the possibility of multiple policy limit events decreases, and this process continues as the size of the pool increases.

Several early insurance authors argued this was an example of the LLN. But this only applies if the observations being compared have the same statistical distribution. For example, a fair (equally weighted) coin flip will result in 50 percent heads and 50 percent tails if performed enough times. The ratio may be different for a smaller number of attempts—it may be 70 percent heads after 10 tries, 65 percent heads after 20 tries, etc., but at some point, the total results of the number of tries would be 50/50.

But what if the second coin is materially different from the first? For example, the second coin is weighted so that heads occurs 60 percent of the time while tails occurs 40 percent of the time, resulting in a 60/40 frequency distribution. These two coins have different statistical distributions, meaning the LLN does not apply between them.

The inherent complexity of insurance risks magnifies this problem. For example, an actuary determines that a trucking company's 1,000 trucks will have an average of 50 accidents per year with an average severity of $5,000 for a total projected loss of $250,000. But during the policy year, the location where the trucks drive experiences 5 inches of additional rainfall. Because of this, the frequency and severity of accidents are likely to increase. Hence, the results during the policy year will differ from those projected by the actuaries.

This explains why the following insurance authors noted the shortcomings of the LLN relative to insurance.

• The LLN didn't apply to coverage purchased form Lloyd's. (Charles O. Hardy, Risk and Risk Bearing, 1923.)
• "There is no line of insurance written which is able to meet the complete set of tests implied by any of the Law of Large Numbers." (Pfeffer, Insurance and Economic Theory, p. 43.)
• "Neither the insurer nor the self-insurer is capable of collecting exactly homogenous risks in sufficient numbers to make theoretical predictions of absolute certainty." (Robert Goshay, Corporate Self-Insurance and Risk Retention Plans, S.S. Huebner Foundation for Insurance Education, 1965, p. 17.)
• "An insurer almost always will experience a deviation from its expected loss ratio if the risks—collectively—differ substantially in capacity for producing loss from those in the group which was the source of the loss ratio." (Donald L. MacDonald, Corporate Risk Control, 1966, p. 60.)
• "Even if we properly evaluate the mean frequency and the mean severity, the actual observed results will generally vary from the underlying mean." (Foundations of Casualty Actuarial Science, third edition, 1996, p. 61.)
• "The law of large numbers has two significant limitations. First, no matter how many past losses have occurred and have been used to compute expected losses for a current insurance policy period, there is some chance that the actual losses for that period will not follow past patterns. Second, and more important, underlying conditions affecting loss experience might change, so the future no longer comes close to repeating the past. (Michael Elliot, George Head, and James D. Blinn, Risk Financing, American Institute for Chartered Property Casualty Underwriters, Inc., 2005, p. 231.)
• "The Law of Large Number only applies to 'extreme' (very large) situations." (Scott E. Harrington and Greg Niehaus, Risk Management and Insurance, second edition, McGraw-Hill, 2003.)

Two texts combined explain the insurance industry's more nuanced understanding of pooling and its implications for the total risk insured. Scott E. Harrington and Greg Niehaus observed that, while pooling does lower risk, it doesn't provide so much stability that variability becomes negligible. Neil Doherty added the key observation that, as the pool increases, the risk per policy decreases. ²⁰ The two observations form the basis of insurance's modern understanding of pooling: While grouping polices lowers risk especially at the policy level, there is still sufficient loss variability that the insurer needs to mitigate risk.

While insurance practitioners used statistics to analyze and understand pooling, the Tax Court viewed risk distribution as either an accounting function or natural occurrence based on the sheer size of the insurer.

The Tax Court and Treasury originally believed risk distribution mandated the presence of third-party moneys, making an implied analogy that since a standard insurer's claims payments are comprised of funds from multiple unique insureds, so must a captive's be as well. This reasoning is the natural outgrowth of the government's original contention that a captive is simply a reserve account, where a single company is the only source of funds. The government's early victories were based on this theory, but once third-party funds were not inconsequential, the court ruled for the taxpayer.

The court recently adopted a new concept: statistically independent risk, which is derived from the following Clougherty citation:

By assuming numerous relatively small, independent risks that occur randomly over time, the insurer smoothes out losses to match more closely its receipt of premiums.

Source: Clougherty Packing Co. v. Commissioner, 811 F.2d 1297, 1300 (9th Cir. 1987).

Here, the court simply counts the number of insured items, such as each truck for commercial auto or each person for workers compensation. This concept is closer to the underwriting policy employed for some insurance policies.

Both Tax Court methods have a fatal flaw: The court offered no mathematical justification for its rulings based on either theory. Instead, each judge gave a simple thumbs-up or thumbs-down. There was no mention of the pool's variance, standard deviation, or coefficient of variation. When compared to the discussion of the pool contained in insurance texts cited above, these methodologies are clearly wanting.

The biggest problem faced by the insurance industry is there are two distinct methods of understanding risk distribution: one based on actuarial concepts and calculations and one centered strictly on a judge's opinion. More concerning is there is no overlap between these different schools of thought. Finally, and perhaps most concerning, is the lack of mathematical support for either government position. Risk distribution is a mathematical concept, and it must be expressed as such.

Is It Insurance in Its Commonly Accepted Sense?

This is the third Harper test factor. Here, the court analyzes routine operations such as claims handling, investment activity, policy drafting, and premium determination to see if the captive was operated like a large, publicly traded insurer. As most of the pre-micro-captive cases involved very large companies, the decisions contained a perfunctory analysis of this factor. This paragraph from Harper is illustrative:

[The captive] was both organized and operated as an insurance company. It was regulated by the Insurance Registry of Hong Kong. The adequacy of Rampart's capitalization is not in dispute. The premiums charged by Rampart to its affiliates, as well as to its shippers, were the result of arm's length transactions. The policies issued by Rampart were valid and binding. In sum … the arrangements between the Harper domestic subsidiaries and Rampart constituted insurance, in the commonly accepted sense. ²¹

Source: Harper Grp. v. Commissioner, supra, 96.

Due to the insureds' sophistication in the earlier cases, such matters were handled in due course.

Conclusion

The law is far from clear for most of the Harper factors. While the R.V.I. court used a sound methodology to determine if insurance risk existed, the Avrahami court simply noted whether the parties agreed if a policy was valid, offering no explanation for ignoring the RVI methodology. Several cases loosely based on the economic family doctrine still stand for the proposition that no risk shifted to the captive. Ideally, this issue should be settled, but the existence of precedent keeps the door open to the challenge. Risk distribution is perhaps the biggest problem as there are two distinct theories that do not overlap. While it's good that we have negative examples for insurance in its commonly accepted sense, the opposite of these facts does not necessarily mean a positive outcome.