149 T.C. No. 7



UNITED STATES TAX COURT. Docket Nos. 17594-13, 18274-13. Filed August 21, 2017.

Ps claimed deductions under I.R.C. section 162 on their 2009 and 2010 tax returns for amounts paid by their passthrough entities to captive insurance company C wholly owned by PW and to off-shore company A which reinsured a portion of its risk with C. R denied the deductions and determined that C’s elections under I.R.C. section 831(b) to be treated as a small insurance company and under I.R.C. section 953(d) to be taxed as a domestic corporation were invalid, as the amounts paid did not qualify as insurance premiums for federal income tax purposes. R also determined that amounts transferred out of C were distributions to Ps, not loans, and that Ps were liable for accuracy-related penalties under I.R.C. section 6662(a) .

Held : Amounts paid to C and A are not insurance premiums for federal income tax purposes and are not deductible under I.R.C. section 162 .

Held , further , C’s I.R.C. section 831(b) and section 953(d) elections are invalid for 2009 and 2010.

Held , further , the amount transferred directly from C to PW is an ordinary dividend.

Held , further , the amount transferred indirectly from C to Ps is not taxable to the extent it is a loan repayment, but the excess is either taxable interest or an ordinary dividend.

Held , further , Ps are not liable for accuracy-related penalties under I.R.C. section 6662(a) except in relation to the amounts determined to be ordinary dividends or taxable interest.

Tim A. Tarter and Kacie N.C. Dillon , for petitioners. *

Brandon A. Keim , Doreen M. Susi , Steven I. Josephy , and John W. Stevens , for respondent.

HOLMES, Judge: Benyamin and Orna Avrahami own three shopping centers and three thriving jewelry stores. In 2006 they spent a little more than $150,000 insuring them. In 2009 this insurance bill soared to more than $1.1 million and it flew even higher, to more than $1.3 million, in 2010. The Avrahamis were paying the overwhelming share of these big bills to a new insurance company called Feedback that was wholly owned by Mrs. Avrahami. Yet there were no claims made on any of the Feedback policies until the IRS began an audit of the Avrahamis’ and their various entities’ returns. With money flooding in and none going back out to pay claims, Feedback accumulated a surplus of more than $3.8 million by the end of 2010, $1.7 million of which ended up back in the Avrahamis’ bank account—as loans and loan repayments, say the Avrahamis; as distributions, says the Commissioner. Also included in Feedback’s surplus was $720,000 that the Avrahamis’ jewelry stores sent down to a Caribbean company for terrorism-risk insurance. The full $720,000 then flew right back to Feedback after—the Avrahamis argue—it distributed enough risk for the whole plan to constitute insurance as that term is commonly understood.


A. The Avrahamis and Their Businesses

Benyamin Avrahami was born in Iran but was raised in Israel where his family fled religious persecution. He immigrated to the United States in 1974, went to college, and obtained degrees in both business administration and gemology as well as a real-estate license. He met and married Orna Avrahami, who was born and raised in Israel but moved to the United States in 1980. The couple now live near Phoenix, Arizona, and have three adult children.

In 1980 Mr. Avrahami decided to go into business with his brother, so he created American Findings Corporation (American Findings). 1 As its name implies, American Findings started out as a supplier of findings—the components that go into finished pieces of jewelry including clasps, split-rings, solder, and settings for stones. A few years later, however, American Findings bought an existing but financially troubled jewelry store named London Gold and got out of the wholesale findings business. The Avrahamis are talented businesspeople. They turned London Gold around, and now American Findings (d.b.a. London Gold) operates—and operated during the years at issue in these cases—three successful retail jewelry stores that employ 35 people in the Phoenix metropolitan area.

In addition to their jewelry stores, the Avrahamis own several commercial real-estate companies. There are six involved in these cases:

BYS Company, ACC (BYS), 2 which owns and operates a retail shopping center in Tempe, Arizona;

Chandler One, LLC (Chandler One), 3 which owns a commercial building in Chandler, Arizona, and leases the space to three tenants—one of the jewelry stores owned by American Findings, a vitamin store, and a wireless carrier;

Junction Development, LLC (Junction Development), which is in Scottsdale, Arizona, and leases space to another of the jewelry stores owned by American Findings;

O & E Corporation (O&E), 4 which owns a shopping center in Phoenix, Arizona;

White Mountain Equities, L.L.C. (White Mountain), 5 which owns land in Show Low, Arizona; and

White Knight Investment, A.C.C. (White Knight), 6 which owns a large commercial strip mall in Tempe, Arizona, and leases the space to several tenants including a charter school.

While the Avrahamis are consulted on major decisions like new tenants and costly repairs, they hire out the responsibility for the day-to-day operations of Chandler One, O&E, and White Knight to a management company. In 2006 American Findings, Chandler One, O&E, and White Knight (collectively, Avrahami entities) deducted a combined total of a little more than $150,000 in insurance expenses.

B. The Advisers

By 2007 the Avrahami entities were flourishing and the Avrahamis were in need of some advice. They turned to Craig McEntee, who had been their trusted CPA for about 25 years. Upon McEntee’s recommendation, the Avrahamis retained Neil Hiller for some estate-planning services. Hiller is a Phoenix-based lawyer who practices in estate planning, employee benefits, and tax.

Around the same time, McEntee also suggested that a captive insurance company might be a good fit for the Avrahamis and recommended that they consult Celia Clark. Clark, who graduated from a well-regarded law school in the Midwest, but who has lived and worked in New York for many years, focuses her practice on tax, trusts, and estate planning. She is the founding partner of Clark & Gentry, PLLC, which was formerly known as the Law Offices of Celia R. Clark, PLLC. 7 Clark first got interested in captive insurance companies in 2002, and her practice grew from there. In 2006 she helped draft captive-insurance legislation for the dual-island Caribbean nation of Saint Christopher and Nevis (St. Kitts). Clark had more than 50 captive insurance clients in St. Kitts by 2007 and more than 75 by 2008. Today a large part of Clark’s practice is the formation and maintenance of such insurance companies.

Before moving forward with Clark, the Avrahamis told Hiller they were considering forming a captive insurance company and asked for his advice. Hiller discussed the idea with the Avrahamis and recommended that they hire Clark, whom he had previously worked with on another captive insurance company matter. The Avrahamis therefore gave the green light for Clark to start reviewing information about their various businesses—to be provided by Hiller and McEntee—and to determine what sort of captive insurance company might work for them. Then, in November 2007, the Avrahamis signed a retainer agreement with Clark in which they agreed that Clark and Hiller would act as co-counsel and provide all legal services relating to the start-up of a captive insurance company in exchange for $75,000. This agreement eventually led to the formation of the Avrahamis’ captive insurance company—Feedback Insurance Company, Ltd. (Feedback).

C. Feedback

Feedback was incorporated in St. Kitts in November 2007. Mrs. Avrahami was its sole shareholder as well as its treasurer and bookkeeper, though both Avrahamis had signature authority over Feedback’s bank account. Feedback also hired a St. Kitts company called Heritor Management, Ltd. (Heritor), to assist with general management, monitor compliance with Kittian regulations, apply for licenses, and process claims. Heritor is owned by Robin Trevors. Before the end of 2007, Feedback applied for and received authorization from St. Kitts to “conduct small group captive insurance business” under the St. Kitts 2006 Captive Insurance Companies Act. In 2008 it also made two elections. The first—filed by Clark on Feedback’s behalf—was an election under section 953(d) to be treated as a domestic corporation for federal income tax purposes, which was approved by the IRS. And the second—filed with its 2007 income tax return—was an election to be taxed as a small insurance company under section 831(b) .

1. 2007 and 2008

In its first two years of operation—2007 and 2008, which are not before us in these cases—Feedback sold property and casualty insurance policies to various entities owned by the Avrahamis. In 2007 these included American Findings, BYS, Chandler One, O&E, White Mountain, and White Knight, but in 2008 only Chandler One, O&E, and White Knight. Feedback also entered into a cross-insurance program to reinsure terrorism insurance for other small captive insurers through a risk-distribution pool set up by Clark exclusively for clients of her firm.

2. 2009 and 2010

In 2009 and 2010—the years at issue in these cases—Feedback continued to sell policies to entities owned by the Avrahamis and to reinsure terrorism polices through one of Clark’s risk distribution programs. Specifically, Feedback issued the following direct policies: 8


Coverage type

2009 premium

2010 premium

2009 limit (occurrence/aggregate

2010 limit (occurrence/aggregate)

American Findings
Business income
Employee fidelity
Litigation expense
Loss of key employee
$1.5M/ $1.5M
Tax indemnity
Total American Findings
Chandler One
Administrative actions
Business risk indemnity
Total Chandler One
Administrative actions
Business risk indemnity
Total O&E
White Knight
Administrative actions


Business risk indemnity


Total White Knight

Despite the formation of Feedback, each of the entities owned by the Avrahamis continued to buy insurance from third-party commercial carriers and made no change to its coverage under those policies after contracting with Feedback. The following charts summarize each entity’s commercial coverage for 2009 and 2010:

American Findings

Coverage term


Coverage type


Limit (occurrence/aggregate)

Jewelers Mutual
Business owners & jewelers block
Jewelers Mutual
Business owners & jewelers block
Chandler One

Coverage term


Coverage type


Limit (occurrence/aggregate)

Commercial general liability
Commercial general liability

Coverage term


Coverage type


Limit (occurrence/aggregate)

Commercial general liability
Business owners
White Knight

Coverage term


Coverage type


Limit (occurrence/aggregate)

Commercial general liability
Commercial general liability
Commercial property/building
Commercial property/business income
The Avrahami entities deducted a total of more than $1.1 million for 2009, and more than $1.3 million for 2010, in insurance expenses. The IRS is not challenging the validity of the Avrahami entities’ commercial policies or the deductibility of those premiums. The premiums paid to Feedback are another matter. The IRS is taking the position that what Feedback sold was not insurance, meaning the premiums are not deductible as ordinary and necessary business expenses.

3. Insurance Policy Pricing

As the Avrahami’s expert witness explained, underwriting “is the process of determining the price, terms and conditions, [and] acceptability of a risk by an insurance company.” In a competitive market, an insurer’s goal is to price policies in such a way that the premiums brought in cover losses and the insurer’s business expenses with enough profit left over to keep investors happy. To accomplish this goal insurance companies typically use both actuaries and underwriters. According to Feedback’s actuary, “the actuaries define the rating scheme and the underwriters make * * * the individual selections and adjustments for the given risks.” An actuary typically starts with published rates and large datasets for particular risks and makes adjustments for policy limits, estimates of the frequency and severity of loss, deductibles, the claims history of a particular customer, and perhaps a dozen or so other factors that can be combined into equations that he uses to set a premium for a particular policy. Actuaries are also supposed to ensure their work is appropriate for its intended use, consider whether their work includes large enough risk classes “to allow credible statistical inferences regarding expected outcomes,” and check the reasonableness of their results. See Actuarial Standard of Practice No. 12: Risk Classification (for All Practice Areas), sec. 3.3 (Actuarial Standards Bd. 2005). 9 No one thinks this process lacks all subjectivity, but the work of an actuary must be reproducible and explainable to other actuaries. See Actuarial Standard of Practice No. 41: Actuarial Communications, sec. 3.2 (Actuarial Standards Bd. 2010).

The actuarial services that Feedback obtained were somewhat different.

4. Feedback Policy Pricing

During 2009 and 2010 Clark hired Allen Rosenbach, an actuary, to price the Feedback policies. 10 Rosenbach first reviewed the work of Feedback’s previous actuary and then began developing his own pricing model for the products Feedback might wish to offer. To assist Rosenbach in his calculations, Clark provided various documents—including Feedback’s business plan, which she drafted, and which detailed the types of coverage Feedback planned to issue, as well as the insurance policy applications from the various Avrahami entities. 11 At a very high level, Rosenbach’s pricing process was to determine a base premium for each policy and then to adjust that base by various factors. 12 Because the premium for each policy was determined using slightly different base rates and factors, we will provide a more detailed explanation policy by policy.

a. Administrative Actions

The Administrative Actions policies covered any legal expenses arising from an administrative action or disciplinary proceeding instituted against the policyholder—Chandler One, O&E, or White Knight. The parties agree that this is an insurable risk and Rosenbach testified that this sort of coverage is available on the commercial insurance market, though it is often part of another policy.

Rosenbach analogized the coverage provided by the Administrative Actions policies to commercial miscellaneous-professional-liability insurance. He therefore started his premium calculations with information found in the public filings of large insurance companies—specifically a January 2005 Chubb filing—because Chubb is a very large commercial writer of insurance. The Chubb filing indicates that the base rate should be calculated using the insured’s gross revenue and its classification into one of four hazard groups according to the level of risk it poses. For example, Rosenbach stated that Chandler One would be considered a “property manager,” which according to a June 2005 Chubb filing falls under hazard group 4. 13 According to the January 2005 Chubb filing, the base rate for an insured in hazard group 4 is a flat $10,400 for its first $250,000 of gross revenue and then $6.70 per thousand of gross revenue for the next $250,000. Rosenbach followed this methodology and calculated a base premium for Chandler One—which for 2009 had expected gross revenue of $470,000—of $11,874. 14

Once the base premium for Chandler One was set, Rosenbach adjusted it by five factors. The first was a claims-made factor of 1.3. Claims-made policies are “[a]n agreement to indemnify against all claims made during a specified period, regardless of when the incidents that gave rise to the claims occurred.” Black’s Law Dictionary 821 (8th ed. 2004). However, a claims-made policy can also include a retroactive date that limits how far back the incident could have happened. Claims-made policies are often contrasted with occurrence policies, which are “[a]n agreement to indemnify for any loss from an event that occurs within the policy period, regardless of when the claim is made.” Black’s Law Dictionary 822 (8th ed. 2004). Rosenbach used a claims-made factor of 1.3 because that was the factor designated in the Chubb filing for a claims-made policy with retroactive coverage for five or more years. And in his view, all of the Feedback policies were claims-made with no retroactive date, meaning the incident or event that caused the insured loss could come from any point in time as long as the claim was made during the policy period. The insuring agreement states that Feedback “agrees to pay to the Insured any legal expense incurred by the insured during the Policy Period, arising from or relating to the defense of any Insured Event as defined hereunder, which Insured Event is instituted against the Insured during the Policy Period.” The policy defines “Policy Period” as “[e]vents occurring and reported from and after 12:01 a.m. December 15, 2009 and prior to 12:01 a.m. December 15, 2010.”

The second factor was a deductible factor of 2.3 and was meant to compensate for the fact that the Feedback policy has a deductible 15 different from that of the base Chubb policy. For example, because the Chandler One policy had a zero deductible, Rosenbach went to the “Deductible Factors” table in the Chubb filing and looked up the factor for a hazard-group-4 insured with no deductible. However, the lowest deductible on the Chubb table is $500 and carries a factor of 2.02, so Rosenbach extrapolated to reach the 2.3 factor he used in his model. The third adjustment was the “increased limit factor” of 1.2, which like the deductible factor accounts for differences between the policy issued by Feedback and the Chubb policy. In this case, the Chubb policy was priced assuming both the aggregate and occurrence limits were $1 million, but the Chandler One policy had an occurrence limit of $1 million and an aggregate limit of $2 million. According to the Chubb filing, this difference in aggregate limits adds 20% to the base premium, thus the factor of 1.2.

Rosenbach’s fourth adjustment was a 10% increase—a factor of 1.1—for “endorsements”. As Rosenbach explained, the magnitude of the endorsement factor was based solely on his professional judgment and accounts for “the breadth of the coverage being broad.” And the final adjustment was a coverage factor of 0.65, which Rosenbach stated was actually made up of two parts—“30 percent of the premium comes from the administrative actions coverage and another 30 percent comes from the coverage of fines and penalties.” In other words, this factor adjusts the base premium because the Administrative Actions policy issued by Feedback was both narrower, because it covered fewer events, and broader, because it covered fines and penalties, than the miscellaneous-professional-liability policy described in the Chubb filing.

To reach the total premium for this type of policy, Rosenbach then multiplied the base premium by the five factors. For example, for Chandler One’s 2009 policy Rosenbach calculated a premium of $30,000. 16 The calculations for the Administrative Actions policies purchased by Chandler One in 2010, O&E in 2009 and 2010, and White Knight in 2010 were performed in an identical manner. Each used the exact same formula and factors—claims made of 1.3, deductible of 2.3, increased limit of 1.2, endorsement of 1.1, and coverage of 0.65—but started with the various insured’s gross revenue when calculating the base premium.

b. Business Risk Indemnity

The Business Risk Indemnity policies generally covered business liabilities caused by “construction defects” or events excluded under the policyholder’s commercial policies—for example, losses from asbestos, climate change, or fungi.

Rosenbach’s pricing model was slightly more complicated for these policies because the premiums required three separate calculations. The first was for the premium associated with coverage for events not covered by—the “major gaps” in—a commercial policy. The second was for excess coverage, under which an insurer agrees to indemnify an insured against a loss only if it exceeds the amount covered by another policy. And the third was for the premium associated with “construction defect” coverage.

The gap-coverage portion of the calculation was similar to the Administrative Actions model in that it started with the base premium from the January 2005 Chubb filing for Miscellaneous Professional Liability coverage and then adjusted for the same five factors. Rosenbach again used a claims-made factor of 1.3 and a deductible factor of 2.3, but the increased limit, endorsement, and coverage factors are slightly different. For example, for Chandler One’s 2009 policy, Rosenbach used an increased limit factor of 1.92 which is the amount indicated in the Chubb filing for a policy with aggregate and occurrence limits of $4 million. He also used an endorsement factor of 1.25 to account for his belief that the Business Risk Indemnity policy covered “much more” than the Chubb base policy. Finally, Rosenbach used a coverage factor of 0.38, which he explained at trial as:

THE COURT: There is a coverage factor next, you had mentioned that before. This one is only .38, all right they don’t add up to one. What’s going on there? You have [a] coverage factor of 0.38 and then [an] adjustment of 0.75?

THE WITNESS: Yeah, the seven is built into the coverage factor, the .75 fee is a point, I take the product of a base adjustment and then I adjust it for the risk.

THE COURT: Okay, I don’t understand that.


THE COURT: What number do you start out with, the .75 or the .38?

THE WITNESS: The .75 is —

THE COURT: Okay, and what do you do with the .75?

THE WITNESS: I actually multiply it by another factor that’s built into the coverage to come up with that —

THE COURT: And what other factor are you multiplying that by?

THE WITNESS: Well in this case it’s roughly .5, 50 percent.


THE WITNESS: That, okay that’s what I’m saying, both the endorsement and the coverage puts together all of the risks that the insured is getting, the major ones. And it adjusts on a relative basis on what’s it’s worth, so it’s really a potpourri of a bunch of factors that come up with that .38.

THE COURT: And again was this based on your experience or is this based on a commercially available filing like this Chubb one?

THE WITNESS: It’s more experience.

With the base premium and five factors, Rosenbach reached a 2009 premium for Chandler One for the gap-coverage portion of the Business Risk Indemnity policy of $31,953. 17

But recall that this is only one part of the calculation. Rosenbach then added $2,500 to account for the excess coverage—“the straight commercial coverages going up in higher limits”—provided by the policy. And he added another $26,438 for “construction defects” coverage because Chandler One’s application indicated it was in the real-estate development and management business and Rosenbach believed that “[a]nything that touches the real estate [Chandler One] can get sued for it.” He arrived at the $26,438 by multiplying Chandler One’s 2009 expected gross revenue of $470,000 by “certain rates that [he] use[s] for construction defect coverage because [he has] more data on that.” 18 Putting all of this together, Rosenbach calculated a total premium for Chandler One of $61,000. 19 An identical calculation was done for the Business Risk Indemnity policy purchased by O&E in 2009 except that Rosenbach started with O&E’s gross revenue and added no additional premium for construction-defect coverage.

In 2010 the Business Risk Indemnity policies were calculated in a similar manner, but again with a few adjustments. For example, the endorsement factor was dropped to 1.0 and the coverage factor was increased to 0.5. 20 And for Chandler One’s 2010 policy, Rosenbach also dropped the occurrence and aggregate policy limits to $3 million reducing the increased limit factor to 1.7, lowered the charge for the excess coverage to $1,000, and increased the premium for the construction defects coverage from $26,438 to $62,730. Rosenbach explained that this sharp increase in premiums related to construction defects was due in part to 50% higher expected gross revenues and in part because “the increased limit factor was applied to the base rate and the calculation, where in the prior years it wasn’t. So it was a, it was a tweak in the model to incorporate some more of a line risk.”

c. Business Income

The Business Income policy covered any amount of business income (limited to “overhead expenses”) that American Findings lost as the result of reputational damage or new competition.

To come up with the premiums for American Findings’s 2009 Business Income policy Rosenbach started out with its gross revenue and multiplied it by 7.5%. As he explained, the 7.5% represents his assumption that “you might only expect one policy limit loss every 20 years * * * [which] would turn into a five percent expected loss, and that expected loss grossed up for expenses and risk will give you a seven and a half percent rate.” Then Rosenbach adjusted this amount by an increased limit factor and claims-made factor in the same manner as for the other policies previously discussed. Finally, he multiplied by an “adjustment factor” of 0.9 and an “other” factor of 0.165. Together, he said, these “judgmental factors” accounted for financial stability, size, profitability, entry into the market, additional coverages, and a 10% surcharge for “competition and the reputational damage.” Multiplying all of these parts together, Rosenbach calculated a premium of $271,000. 21

American Findings’ 2010 policy was calculated in the same manner, but the “adjustment factor” was decreased to 0.65 and the “other” factor was increased to 0.2325. The reason for these differences is not clear from the record, as Rosenbach said that he used the same methodology for 2010 as he had for 2009.

d. Employee Fidelity

The Employee Fidelity policy covered losses to American Findings caused by fraudulent or dishonest acts committed by one of its employees acting alone or in collusion with others. The parties agree this is an insurable risk.

The 2009 Employee Fidelity policy was priced a little differently because it represents coverage that exists in the commercial market and Rosenbach testified that he was able to follow the rating methodology of a commercial carrier—specifically a Chubb employee-fidelity crime-theft filing. 22 For each factor, therefore, he used his judgment and what he knew about the American Findings policy to select a factor from the defined range for that type of factor in the Chubb filing. By multiplying all of the factors times the base premium—also derived from the Chubb filing—Rosenbach reached a premium of $71,000. 23 The calculation of the 2010 premium was exactly the same, but it started with a different base premium.

e. Litigation Expense

The Litigation Expense policy covered any expenses American Findings incurred in defending itself in a legal proceeding, in prosecuting a third party over a matter pertaining to the business, or in obtaining “any legal consultation pertaining to the business.”

For the 2009 Litigation Expense policy Rosenbach started out with an exposure base of $276,000, which he said was “basically a function of the underlying expected losses of the given lines of business in the model with an estimate for things outside of the model.” Or in other words “it’s an estimate from all different sides of all different exposures that could impact the litigation. So, what would feed into it would be parts of, part of the premium for administrative action, part of the premium for crime, part of the premium for, for all different other coverages.” Then Rosenbach multiplied by an “expected loss ratio” of 90%, which accounts for his belief that Feedback would have to pay back 90% of the premiums it collected in the form of reimbursements for losses covered by the policy because “most of the captives have a ten percent expense ratio.” The next adjustment was a 30% charge for “allocated loss adjustment expense.”

Rosenbach testified that this adjustment represents “[a]nything associated with settling claims”—including legal fees—as a portion of the total loss. Then he multiplied by another factor of 0.6 because “we’re only covering legal expense, the loss is everything beforehand, is covering all the loss and loss adjustment expense, we have to get just the, just the legal fee piece out of it.”

Rosenbach’s model also reflects an expense ratio of 10%, 24 an increased limit factor of 1.0, and a claims-made factor of 1.3. He testified that to reach the total premium of $65,000 calculated by his model “it should just be the product of the factors. * * * Probably one minus the expense ratio, times the 0.6, times the 1.3.” 25 The 2010 Litigation Expense policy was calculated in the exact same manner, but with a nearly 60% higher exposure base. Rosenbach explained this increase:

A: It’s an evaluation of all the other coverages that are part of the model and not part of the model and it was adjusted to reflect more exposure.

Q: So it was a model adjustment?

A: In the model, you can adjust for the different contributions to the lines. So it’s me applying the model in a little different light than they did in the past.[ 26 ]

f. Loss of Key Employee

American Findings’ Loss of Key Employee policy covered lost business income (limited to “overhead expenses”) resulting from the temporary or permanent departure—including voluntary departure—of either Mr. or Mrs. Avrahami. “Overhead expenses” is defined as American Findings’ “projected ordinary operating expenses exclusive of distributions to shareholders or owners, dividends, interest and principal payable to shareholders, owners, or affiliates, income taxes, amortization and depreciation, for the period commencing on the date of loss and ending at the termination of the Policy Coverage Period.” This type of policy is not generally available in the commercial insurance market.

To calculate the 2009 premium for the Loss of Key Employee policy, Rosenbach started with projected gross income of $11 million and multiplied it by an event rate of 5% and an “extra expense factor” of 1.15. The 5% is another judgment call and is supposed to represent the expected losses under this type of coverage. And the 1.15 “extra expense factor” accounts for “the cost of finding replacements” in Rosenbach’s experience. He then multiplied by an adjustment factor of 1.5 for “a disability add on” and another factor of 0.5 for the assumption that “the duration of a claim won’t last a full year, it’ll only last half a year.” Multiplying all of the factors together Rosenbach reached a preliminary premium of $474,000. 27 Next he apportioned this preliminary premium to the key employees covered by the policy—the Avrahamis—whose salaries were 18.1% of American Findings’ total payroll expense. This led to a final premium of $86,000. 28 The 2010 premium was calculated exactly the same way.

g. Tax Indemnity

That brings us to perhaps the most peculiar policy of all—the Tax Indemnity policy. This policy supposedly covered additional taxes, interest, and penalties that American Findings might become obligated to pay as the result of an “adverse resolution” of a position taken on its tax return—with exclusions for fraud, criminal conduct, or a willful violation of the law. This type of policy is, as one might expect, not generally available in the commercial insurance market.

Here Rosenbach started with the $2 million policy limit and multiplied it by an event rate of 7.5%. Rosenbach testified that he derived the 7.5% from an IRS study on audit results. The origins of this number remain murky. We find it more likely than not to be the ratio of deficiencies—including interest, penalties, and additions to tax—to gross income reported on all returns aggregated for the country as a whole, but adjusted for the size of the entity and the potential for a multiyear audit. Next Rosenbach multiplied by an endorsement factor of 1.0 and an “experience factor” of 0.5. The experience factor was a partially subjective adjustment that accounted for American Findings’ ratio of deductions to total revenue and the consistency of its tax returns from year to year. This led Rosenbach to a total premium of $75,000 for 2009. 29 The 2010 premium was exactly the same.

5. The “Target”

In total, the Avrahami entities paid Feedback premiums for their direct policies of $730,000 in 2009 and $810,000 in 2010. Each year Clark told Rosenbach that the Avrahamis had a “target premium” of $840,000 for their direct policies and $1.2 million for total premiums—the direct policy premiums plus $360,000 in premiums for terrorism insurance from Pan American. See infra p. 32. After completing his calculations, Rosenbach would send them back to Clark for comments. While not related to the years at issue in these cases, Rosenbach testified that an email chain between himself and Clark relating to the 2011 policy premiums was representative of the sort of comments he received. Those emails show that Rosenbach initially proposed total direct premiums for 2011 of $899,000, but was then asked to add a “proration factor.” The email back from Clark says: “It looks like we’re still pretty far even with 10 months prorated for the old policies. I think we should go back to full years for all the policies with $840,000 as the target.” Rosenbach replied with new calculations when he had “dropped the limits as suggested” and calculated total direct premiums of $835,000. Once the premiums were finalized, Clark would go back and draft the actual policies.

D. Pan American

In addition to its direct policies, Feedback also participated in “risk distribution” programs. In 2007 and 2008 the program was a cross-insurance pool, which offered Clark’s clients the opportunity to purchase terrorism insurance from other clients’ captive insurance companies. For example, in 2007 entities owned by the Avrahamis paid $360,000 into the pool for terrorism coverage 30 and Feedback received $360,000 for insuring other members of the pool.

In 2009 and 2010, however, Feedback started participating in a risk-distribution program through Pan American Reinsurance Company, Ltd. (Pan American). Pan American was incorporated in January 2009 in St. Kitts and during the years at issue was an insurer licensed in and regulated by the Island of Nevis. It had four shareholders—Diana Gentry (2.5%), Carl Gentry (2.5%), Laurence Mohn (47.5%), and Sheila Trevors (47.5%). Diana and Carl are Clark’s children, and neither ever communicated with Pan American’s management or the other shareholders about business matters. Mohn was a “courtesy director” and testified that “Nevis required someone with insurance experience that really was my only, the only reason why I was involved.” He had no duties, no involvement with Pan American’s day-to-day operations, and no regular communications with anyone at Pan American. And Sheila Trevors is the wife of Robin Trevors—the owner of Feedback’s management company, Heritor. Heritor’s sister company, Heritage Services, Ltd. (Heritage), is the registered agent and insurance manager of Pan American.

1. The Structure

The aim of Pan American was simple—distribute risk. Clark told her clients:

As you know, your captive insurance company is required to “distribute risk” in order to be treated as an insurance company for tax purposes. The IRS considers this requirement to be satisfied if a significant portion of the insured risk borne by your company is spread among one or more insureds that are unrelated to your company. Case law has established that 30% of the total premiums received by an insurance company represents a significant portion of its risk.

Pan American, therefore, was intended to connect Clark’s clients with other businesses that operate small insurance companies so they could spread their risk to each other by buying and reinsuring terrorism insurance. 31 Pan American would sell policies to participating businesses and then reinsure—or “cede”—all of the risk through the participating insurance companies pursuant to a Terrorism Risk Quota Share Reinsurance Agreement. Each of the participating insurance companies would pay premiums for terrorism coverage to Pan American, which would deposit them in a trust account, and then return an amount almost equal to what it had received to each of the reinsuring companies. For its services, Pan American did not charge a ceding commission; rather it received a portion of the fixed “all-inclusive” $5,000 fee that Clark charged each of her clients for participating in the program.

We’ll use real numbers to show how this worked. Feedback decided to participate in Pan American’s program in 2009 “at $360,000, calculated at 30% of [its] target premiums for 2009, which [was] $1.2 million.” Under the Terrorism Risk Quota Share Reinsurance Agreement, it therefore agreed to a reinsurance premium of $360,000 in exchange for accepting 1.797% of Pan American’s “Ultimate Total Loss for terrorism coverage to the insureds.” In December 2009 American Findings paid Pan American $360,000—the same amount—for “Terrorism Risk Insurance” with a policy limit of $5,525,000 and coverage running from December 15, 2009, to December 15, 2010. Feedback, in turn, received three payments from Pan American—slightly more than $180,000 (50% of its reinsurance premium plus interest) in March 2010; slightly more than $171,000 (47.5% of its reinsurance premium plus interest) in June 2010; and slightly more than $9,000 (2.5% of its reinsurance premium plus interest) in December 2010. The same process repeated itself the next year—American Findings paid Pan American $360,000—this time for up to $5,125,000 of coverage—and Pan American paid Feedback $360,000 plus interest (50% in March 2011, 47.5% in June, and 2.5% in December).

In 2009 Pan American wrote policies for 103 insureds and then reinsured the policies through 85 of Clark’s captive insurance companies; in 2010 it wrote policies for 139 insureds and reinsured through 101 of those companies. Pan American received more than $20 million in premiums in 2009 and almost $23 million in 2010. These amounts were then cycled back whence they came—50% after 90 days and another 47.5% after 180 days. The last 2.5%—about $500,000 in 2009 and $570,000 in 2010—was held back as a “loss reserve” until the policies expired on December 15. Clark told her clients that the loss reserve was intended to build a comfort level for the participants, but that “Nevis law requires loss reserves to be maintained on net premiums only. In the situation outlined here, Pan-American would not be retaining any risk or premiums, and so, technically, would not be required to maintain any loss reserves.” Other than premiums receivable, the only assets reported on Pan American’s tax returns for 2009 and 2010 were cash or cash equivalents of around $200,000 and $390,000, respectively.

2. The Policy

The Pan American risk-distribution program was built around its Terrorism Risk Insurance Pool (TRIP). By its terms Pan American agreed to reimburse policyholders for “losses of ‘property’ and ‘expenses’ resulting directly from an ‘act of terrorism’ occurring during the Indemnity Period.” It was designed to look a bit like the terrorism coverage required to be offered with certain commercial insurance products under the Terrorism Risk Insurance Act of 2002 (TRIA), Pub. L. No. 107-297, sec. 103 , 116 Stat. at 2327. During the years at issue here, both TRIP and TRIA were triggered by an act of terrorism certified by the Secretary of the Treasury with the concurrence of the Secretary of State and the Attorney General that resulted in more than $100 million in losses. 32 As the parties note, however, there are also several key differences. TRIP includes coverage for damage caused by the dispersion of biological or chemical agents, which is excluded under most—if not all—TRIA-backed policies. TRIP excludes acts of terrorism “occurring in a city with more than 1.5 million residents,” though TRIP policies notably leave the term “city” undefined. And TRIP is a stand-alone terrorism insurance policy, meaning it is not tied to any provisions of another policy. 33

3. The Pricing

Clark hired Rosenbach to perform a market survey of commercial terrorism risk insurance to determine a price for the Pan American policies. In the survey Rosenbach looked at several sources including government and industry reports as well as the work of Feedback’s prior actuary. This helped him understand “what was going on in the typical terrorism market” and “what competitors were doing that offered similar coverage.” Combining this understanding with historical information, catastrophe information, the differences between TRIP and other terrorism policies, and his personal judgment, Rosenbach recommended a “rate on line”—the premium divided by the occurrence limit—of 5% to 8% for 2009 and 5% to 9% for 2010. 34 Rosenbach testified that 80% to 90% of these rates are related to the chemical and biological coverage and that the range is applicable to all of the captives participating in the pool regardless of the type of business being insured or its geographic location.

Whether Rosenbach took other provisions of the Terrorism Risk Insurance Insuring Agreement into account in his suggested range of rates is unclear from the record. For example, the Commissioner’s expert witness pointed out that the TRIP policy contains wording that makes it an excess policy, 35 meaning that it would be triggered only if the loss was not covered, or not fully covered, by another policy—e.g., the TRIA coverage included in American Findings’ policy with Jewelers Mutual. The insuring agreement also allows Pan American—if it has impaired solvency—to issue promissory notes to the insured payable over a maximum of three years. 36 And as Rosenbach noted, TRIP policies contain no deductibles, don’t qualify for TRIA—i.e., government-backed—coverage, and to his knowledge would not have been triggered by any event in history. Plus, each reinsurer “is not liable for the subscription of any other participating [reinsurer] who cannot or will not satisfy all or any part of their obligations.”

In addition to its TRIP policy from Pan American, American Findings continued to buy add-on terrorism coverage—backed by the federal government in compliance with TRIA—from its commercial-insurance provider, Jewelers Mutual. The record shows that American Findings paid around $1,500 in 2009 and $1,600 in 2010 for this additional coverage. The Jewelers Mutual policy had a $2 million aggregate limit and specifically excluded coverage for chemical and biological hazards.

E. The Flow of Funds

To summarize a bit, in 2009 entities owned by the Avrahamis paid Feedback $730,000 in premiums for direct coverage, American Findings paid Pan American an additional $360,000 for terrorism insurance, and Pan American paid Feedback $360,000 in reinsurance premiums. Likewise, in 2010 entities owned by the Avrahamis paid Feedback $810,000 in premiums for direct coverage, American Findings paid Pan American $360,000 for terrorism insurance, and Pan American paid Feedback $360,000 in reinsurance premiums. This means the Avrahami entities collectively deducted—as business expenses—insurance premiums of $1,090,000 for 2009 and $1,170,000 for 2010. No claims were filed against Feedback under any of its direct policies in either 2009 or 2010. 37 And no events took place triggering a claim under TRIP in either year. With significant cash coming in the door and none going out to pay claims, Feedback quickly accumulated a surplus. And it used this surplus to transfer funds to Mrs. Avrahami and an entity named Belly Button Center, LLC (Belly Button).

Belly Button was formed in 2007 and is treated as a partnership for tax purposes. It is owned equally by the Avrahami’s three children, but all three testified that they had no knowledge of Belly Button, that they owned Belly Button, or what Belly Button did. We found out: Belly Button owns approximately 27 acres of land in Snowflake, Arizona, which it purchased for around $1,960,000 with about $1.2 million in cash from Mr. Avrahami and the rest with a note payable to the sellers. The $1.2 million from Mr. Avrahami was reported on Belly Button’s tax return as a liability “due to affiliates.” The aim was not to make a gift to Belly Button. Mr. Avrahami—acting as the manager of Belly Button—executed an unsecured promissory note for $1.2 million 38 payable to himself as an individual in or before April 2017. The note carried a 4% simple interest rate.

In March 2008—before the years at issue in these cases—Feedback transferred $830,000 to Belly Button and reported the amount on its tax return under “Mortgage and real estate loans.” 39 The same day Mr. Avrahami—acting again as manager of Belly Button—executed a $830,000 promissory note payable to Feedback. The note required Belly Button to pay the principal and all accrued interest—4.3% per year, compounded annually—in or before February 2018 and was secured by “a Realty Mortgage on certain collateral.” The Realty Mortgage described the land owned by Belly Button in Snowflake, Arizona. That same day Mr. Avrahami also withdrew around $813,000 from Belly Button’s bank account allegedly to pay off the note payable and accrued interest due to the original sellers of the Snowflake land.

Belly Button continued to benefit from its connection to Feedback. In March 2010 Feedback transferred an additional $1.5 million to Belly Button and again reported the amount on its tax return under “Mortgage and real estate loans.” The next day Mr. Avrahami—again on behalf of Belly Button—executed a $1.5 million promissory note payable to Feedback. This note carried an interest rate of 4% per year—simple interest accruing “from time to time”—and was due in March 2020. The note also states that “[t]he Indebtedness is unsecured.” Two days after this transfer of funds from Feedback to Belly Button—and the day after Mr. Avrahami executed the $1.5 million promissory note—the Avrahamis (meaning Mr. and Mrs. Avrahami, not their children, who were Belly Button’s putative owners) transferred $1.5 million from Belly Button’s bank account into their personal one.

Then it happened again. In December 2010, $200,000 went directly from Feedback’s account to Mrs. Avrahami. Yet the transfer was papered just like the transfers that had gone through Belly Button. There was a promissory note due on demand, but no earlier than December 2012, that carried an interest rate of 3% per year, was signed by Mr. Avrahami on behalf of Belly Button, and was reported on Feedback’s 2010 tax return as a mortgage and real estate loan. 40

Insurance regulators often raise their bureaucratic eyebrows at related-party dealings like this. But Feedback did not seek approval from its Kittian regulators for any of these transfers to Belly Button or to Mrs. Avrahami before making them. 41 Clark disclosed the three transfers to Heritor in March 2014, after the Commissioner began his omphaloskeptical review. Heritor communicated the information to the St. Kitts’s Registrar of Captive Insurance Companies in September 2014.

F. The Returns and the Audit

1. Feedback’s Returns

Feedback timely filed its 2009 and 2010 tax returns. On both returns Feedback indicated that it had previously made an election under section 953(d) to be treated as a domestic corporation for federal income tax purposes. And both returns included current-year elections for Feedback to be treated and taxed as a small insurance company under section 831(b) . Feedback’s tax returns reported total assets of almost $2.4 million at the end of 2009 and nearly $3.9 million at the end of 2010, but because of the section 831(b) election it paid income tax only on its investment income—i.e., interest, but not premiums.

The IRS sent Feedback a notice of deficiency in May 2013 that questioned whether Feedback was a valid insurance company and determined that “the amounts characterized as insurance premiums” were income to Feedback under section 61 since Feedback had not established that they were excludable under another provision of the Code. Feedback timely petitioned this Court. 42 The parties stipulated before trial that the “‘Taxable Premiums Earned’ by [Feedback] in the amounts of $1,090,000 and $1,170,000 for taxable years 2009 and 2010, respectively, are not U.S. source fixed or determinable, annual or periodical income under section 881 , or income that is effectively connected with a U.S. trade or business under section 882 .”

2. The Avrahamis’ Returns

The Avrahamis likewise filed 2009 and 2010 tax returns. Incorporated in their returns was the income or loss—reflecting any insurance-expense deduction—passed through to them from numerous partnerships and S corporations, including the Avrahami entities. We summarize those insurance-expense deductions:


American Findings

Chandler One


White Knight

Nowhere on their 2009 or 2010 return did the Avrahamis report the amounts transferred to them from Belly Button—$1.5 million—or from Feedback—$200,000, because they assert that the cash that flowed from Belly Button went just to repay loans. They admit that the $200,000 that went straight from Feedback to Mrs. Avrahami cannot nestle in that nontaxable cubbyhole, but they instead assert that it should have been reported and taxed for 2010 as a qualified dividend.

3. The Audit and the Claims

The IRS began auditing the Avrahamis’ 2009 return in March 2012 and later expanded the audit to include their 2010 return as well as the returns from Feedback and the Avrahami entities. In January 2013 the IRS mailed the Avrahamis documents explaining the examination changes for American Findings, Chandler One, O&E, and White Knight. These documents noted that from Feedback’s inception in 2007 to the end of 2010, Feedback had received premiums totaling almost $3.9 million but had paid no claims. They also noted that one of the nonexclusive factors for determining whether a captive insurance company is a sham is “[w]hether any claims were filed with the captive; if claims were filed – whether the validity of the claims was established before payments were made on them.” This looks like it triggered something: By March 2013 the claims started rolling in from entities owned by the Avrahamis:


Date of claim


Nature of loss


American Findings
Business income/2011-2012
Ring dispute
American Findings
Litigation expense/2011-2012
Ring dispute litigation
White Knight
Business risk/2011-2012
Roof repairs
Junction Development
Business risk/2011-2012
Building repairs
American Findings
Business risk/2013-2014
Water damage
American Findings
Litigation expense/2013-2014
Tax Court litigation
Feedback’s policy was to deal with claims on an “ad hoc basis.” For each claim, Clark determined whether it appeared to be covered, drafted a claim notification, requested a notification extension (if needed), prepared a sworn statement in proof of loss, and sent everything to Heritor along with supporting documents. Heritor then sent a letter back to Clark approving the claims. The Commissioner does question whether several of the claims should have been approved. The Business Risk policies under which the claims were made all contain provisions requiring that Feedback receive the claim notification within the policy period. Yet Heritor granted notification extensions and approved claims filed in April 2013 for policies that ended December 15, 2012.

The Avrahamis weren’t alone in having returns audited because of their interactions with a related microcaptive insurance company. The IRS has applied increased scrutiny to these transactions, adding them to the “dirty dozen” list of tax scams in 2015 and declaring them “transactions of interest” in 2016. See Notice 2016-66 , 2016-47 I.R.B. 745; I.R.S. News Release IR-2015-19 (Feb. 3, 2015). The Avrahamis’, however, is the first section 831(b) case to make it to trial. The Commissioner determined deficiencies of nearly $380,000 for 2009 and $990,000 for 2010, plus almost $275,000 in penalties. These deficiencies are the result of three adjustments:

an increase in the income passed through to the Avrahamis from American Findings, Chandler One, O&E, and White Knight of more than $1 million for both 2009 and 2010; 43

recharacterization of the $1.5 million transfer from Feedback to Belly Button and the $200,000 transfer from Feedback to Mrs. Avrahami as “other income” on the Avrahamis’ 2010 return; and

a computational adjustment that decreased the amount of medical expense deductions allowed each year.

The Avrahamis were Arizona residents when they filed a timely petition with the Court. 44


I. Taxation of Insurance

Amounts paid for insurance are deductible under section 162(a) as ordinary and necessary expenses paid or incurred in connection with a trade or business. Sec. 1.162-1(a) , Income Tax Regs. But amounts set aside in a loss reserve as a form of self-insurance are not. See Harper Grp. v. Commissioner , 96 T.C. 45, 46 (1991), aff’d , 979 F.2d 1341 (9th Cir. 1992); see also Steere Tank Lines, Inc. v. United States , 577 F.2d 279, 280 (5th Cir. 1978); Spring Canyon Coal Co. v. Commissioner , 43 F.2d 78, 80 (10th Cir. 1930). However, neither the Code nor the regulations define “insurance”. Securitas Holdings, Inc. v. Commissioner , T.C. Memo. 2014-225, at *18 . Instead we are guided by caselaw when determining whether insurance exists for federal income tax purposes. The Supreme Court has stated that insurance is a transaction that involves “an actual ‘insurance risk’” and that “[h]istorically and commonly insurance involves risk-shifting and risk-distributing.” Helvering v. Le Gierse , 312 U.S. 531, 539 (1941).

A. Section 831

While the Code permits the deduction of insurance premiums paid , it also taxes insurance premiums received . Insurance companies—other than life-insurance companies, see section 832 —are generally taxed on their income in the same manner as other corporations. See secs. 831(a) , 11. Section 831(b) , however, provides an alternative taxing structure for certain small insurance companies.

Congress added this section to the Code as part of the Tax Reform Act of 1986 (TRA), Pub. L. No. 99-514, sec. 1024 , 100 Stat. at 2405, but a tax break for qualifying insurance companies is not a new idea. The Revenue Act of 1924, ch. 234, sec. 231(10) , 43 Stat. at 283, exempted certain mutual-insurance companies 45 from tax if 85% or more of their income was collected “for the sole purpose of meeting losses and expense.” In 1942 the exemption provisions were revised, making only small mutual insurers—those with interest, dividends, rents, and premiums not exceeding $75,000 for the taxable year—exempt from tax and establishing an alternate taxing structure for insurers with income between $75,000 and $125,000. See Revenue Act of 1942, ch. 619, sec. 165(a) and (b) , 56 Stat. at 872; Tax Reform Proposals—XXII: Hearing Before the S. Comm. on Finance, 99th Cong. 129 (1985) (Statement on Behalf of the National Association of Mutual Insurance Companies (NAMIC)). The new limit aimed to favor small and local mutual-insurance companies, practically all of which would fall under the $75,000 threshold. See S. Rept. No. 77-1631 (1942), 1942-2 C.B. 504, 616. In 1962 the threshold for complete tax exemption was raised to $150,000 and the range for alternative taxation was changed to $150,000 to $500,000. See Revenue Act of 1962, Pub. L. No. 87-834, sec. 8(a) , 76 Stat. at 997; see also secs. 821(c) , 501(c)(15) (1962). Nonlife mutual-insurance companies were sorted into three categories: (1) those with gross receipts not exceeding $150,000, which were tax exempt; (2) “small mutuals” with gross receipts between $150,000 and $500,000, which were taxed only on their investment income; and (3) “ordinary mutuals” with gross receipts over $500,000, which were taxed on both investment and underwriting—i.e., premium—income. See Taxation of Property and Casualty Insurance Companies: Hearing Before the S. Comm. on Finance, 98th Cong. 10 (1983) (Overview).

The next overhaul of these Code provisions occurred in 1986 when Congress repealed and redesignated the Code sections governing the taxation of mutual-insurance companies— sections 821 to 826 , respectively—and instead imposed a single taxing structure for all nonlife insurance companies—mutual and stock—under section 831 . TRA sec. 1024 . This included the addition of section 831(b) to the Code, which provided alternative taxation to all small nonlife insurance companies, but with a simplified structure and higher limits. See S. Rept. No. 99-313, at 512 (1985), 1986-3 C.B. (Vol. 3) 1, 512 (explaining the reason for the change in law was to simplify the taxation of small insurance companies and remove the distinction between small stock and mutual insurers).

This brings us up to the years at issue in these cases. For 2009 and 2010, if a nonlife insurance company had gross receipts that did not exceed $600,000—or $150,000 for mutual-insurance companies—and met certain premium percentage requirements, then it was exempt from tax. Sec. 501(c)(15) (2010). Otherwise, if it had net written premiums (or, if greater, direct written premiums) that did not exceed $1.2 million for the year, then it could elect to be taxed under section 831(b) and be subject to tax only on its taxable investment income. Sec. 831(b)(1) and (2) . 46

B. Captive Insurance Companies

Insurance taxation gets slightly more complicated when the insurer and the insureds are related because the line between insurance and self-insurance begins to blur. A pure captive insurance company is one that insures only the risks of companies related to it by ownership. See Hosp. Corp. of Am. v. Commissioner , T.C. Memo. 1997-482, 1997 WL 663283, at *2. The concept is often attributed to Fred Reiss who, in the 1950s, in response to skyrocketing commercial-insurance prices, figured out that Youngstown Sheet and Tube could set up its own insurance company to insure its coke and iron mines. See Christopher L. Kramer, “Ohio Agent Credited With Captives”, National Underwriter (Mar. 10, 2003), https://www.captive.com/docs/default-source/default-document-library/ohio-agent-credited-with-captive-movement.pdf. The concept spread, and the IRS started challenging whether the payments between a company and its captive were deductible insurance expenses. See, e.g ., Rev. Rul. 77-316 , 1977-2 C.B. 53. When the issue has come to us, we have applied and construed the Supreme Court’s definition of insurance in Le Gierse and its four nonexclusive criteria. To be considered insurance the arrangement must:

involve risk-shifting;

involve risk-distribution;

involve insurance risk; and

meet commonly accepted notions of insurance.

See Rent-A-Center, Inc. v. Commissioner , 142 T.C. 1, 13 (2014); see also R.V.I. Guar. Co. v. Commissioner , 145 T.C. 209, 225 (2015); Harper Grp ., 96 T.C. at 58; AMERCO & Subs. v. Commissioner , 96 T.C. 18, 38 (1991), aff’d , 979 F.2d 162 (9th Cir. 1992); Securitas , at *18.

In AMERCO , 96 T.C. at 42, we found that transactions between AMERCO—and its subsidiaries, which were in the business of renting U-Hauls—and its indirectly wholly owned captive insurance company were for insurance for federal tax purposes. In that case, the captive was regulated “under standard state insurance laws” and issued a wide variety of insurance policies. Id . at 36-37. Some of the policies were issued to AMERCO and its subsidiaries and covered risks such as commercial perils, pollution, and worker’s compensation. Id . at 25-26. But over 50% of the captive’s gross written premiums came from unrelated parties including some of AMERCO’s customers. Id . at 36-37. We held that on the basis of the potential hazards faced by the insureds, insurance risk—not merely investment risk—was present. Id . at 39. We also found that there was risk-shifting—as opposed to merely the building of a reserve for losses—because the captive wrote insurance contracts, collected premiums, paid losses, was regulated by several states, and was “a separate, viable entity, financially capable of meetings its obligations.” Id . at 40-41. We also found risk distribution because the captive’s “insurance business was diverse, multifaceted, and * * * involved a substantial amount of outside risks.” Id . at 41. Finally, we found on the basis of all the facts and circumstances, including the captive’s licensing as an insurance company in 45 states, that the transactions at issue in AMERCO involved insurance in the commonly accepted sense. Id . at 37, 42.

Applying these criteria, we also found in favor of the taxpayer in Harper Grp ., 96 T.C. at 60. Harper was a holding company whose subsidiaries were in the business of international shipping and which indirectly owned a captive insurance company licensed in and regulated by Hong Kong. Id . at 47, 49. The captive was created to provide marine legal-liability insurance to Harper’s subsidiaries, and after its incorporation Harper terminated its substantially similar commercial-carrier coverage. Id . at 49-50. The captive also derived approximately 30% of its premium revenue from a large number of unrelated parties—specifically the customers of Harper’s subsidiaries—who purchased shipper’s-interest cargo insurance. Id . at 50, 59-60. We found that the insureds faced substantial potential liability in the shipping business—liability that had previously been covered by a commercial policy—so this arrangement involved insurance risk. Id . at 58. We found that risk-shifting occurred because the captive was not a sham, conducted a bona fide insurance business, charged arm’s-length premiums, and was financially capable of satisfying the claims made against it. Id . at 59. We also found the policies that the captive offered were for insurance in its commonly accepted sense. Id . at 60. But for risk distribution we focused on the number of unrelated insureds covered by the captive and what proportion of the captive’s premium income they represented. Id . at 59-60. We held that a “relatively large number of unrelated insureds compris[ing] approximately 30 percent of [the captive’s] business” amounted to “a sufficient pool of insureds to provide risk distribution.” Id .

More recently, we held that payments to a Bermudian captive from its bother-sister companies—i.e., all owned by the same corporate parent—were deductible insurance expenses. Rent-A-Center , 142 T.C. at 25. In Rent-A-Center the captive provided workers’ compensation, automobile, and general-liability coverage only to its parent’s other subsidiaries. Id . at 5-6. We found that the captive was not a sham because it was formed “to reduce [Rent-A-Center’s] insurance costs, obtain otherwise unavailable insurance coverage, formalize and more efficiently manage its insurance program, and provide accountability and transparency relating to insurance costs.” Id . at 11. The Commissioner conceded in that case that Rent-A-Center faced insurable risks. Id . at 12-13. We found that there was risk-shifting—despite the captive’s parental guaranty and lack of unrelated insureds—because a covered loss did not economically affect the insureds. Id . at 21-22. We also explained that “[r]isk distribution occurs when an insurer pools a large enough collection of unrelated risks.” Id . at 24. In Rent-A-Center the captive had “a sufficient number of statistically independent risks” because it provided workers’ compensation, automobile, and general liability policies that covered more than 14,000 employees; 7,100 vehicles; and 2,600 stores. Id .

C. Microcaptives

Combining these two concepts—captive insurance companies and the section 831(b) tax advantages for small insurance companies—yields the “microcaptive”. In theory, a microcaptive could be run in the manner of the captives in Rent-A-Center or Harper Group , albeit on a much smaller scale. But it might also be run so that related parties pay the captive deductible insurance premiums of just under $1.2 million a year. In turn the captive might pay out few if any claims, might make a section 831(b) election so it pays tax only on its investment income, and might quickly build up a large surplus. Then, if the captive were to be licensed and regulated in a jurisdiction with extremely low reserve requirements and loose rules on related-party transactions, it might lend its surplus back to its affiliates. This might generate nearly $1.2 million in tax deductions while arguably only moving money from one pocket to another. Or perhaps the captive could be owned by a Roth IRA, which might mean it could make large dividend payments to its stockholder, creating a form of deductible, yet tax-free, retirement savings. Or perhaps the captive could be owned by its business owner’s children or an irrevocable family trust, which might enable the avoidance of future gift and estate taxes.

There may be other variations on this theme; the Commissioner, however, finds none of them pleasing.

II. The Parties’ Arguments

In these cases, the Commissioner’s position is that the amounts paid to Feedback and Pan American are not deductible business expenses because the arrangements lack all four criteria necessary to be considered “insurance” for federal tax purposes. He argues that several of Feedback’s policies included uninsurable risks; that Feedback failed to distribute risk because it had an insufficient pool of insureds; that risk was not shifted because neither Feedback nor Pan American was financially capable of meeting its obligations; and that the arrangements did not embody common notions of insurance because Feedback and Pan American did not operate like insurance companies and their premiums were not determined at arm’s length. In the Commissioner’s view this means that Feedback was not an insurance company and that the funds transferred out of Feedback that eventually ended up in the Avrahamis’ bank account—whether directly or indirectly via Belly Button—must be ordinary income to them.

The Avrahamis and Feedback, on the other hand, argue everything they did complied with the Code and relevant caselaw. Their position is that Feedback is a valid insurance company that qualified and properly elected to be taxed under section 831(b) ; that all the policies covered insurable risks; that all premiums were actuarially determined; and that Feedback distributed risk by ensuring at least 30% of its premium income came from unrelated parties participating in the Pan American program. They say this means the insurance-expense deductions claimed by the various Avrahami entities were proper and that all the distributions to Belly Button were valid, documented loans. (The amount distributed directly to Mrs. Avrahami, on the other hand, they concede should have been reported and taxed—at qualified dividend rates, they argue—and would have been but for an error by their CPA.)

III. Were the Policies Contracts for “Insurance”?

These cases turn on whether the transactions at issue involved “insurance” for federal tax purposes. Precedent tells us to answer this question by considering all the facts and circumstances and deciding whether the arrangement involves risk shifting, risk distribution, and insurance risk, and meets commonly accepted notions of insurance. See Rent-A-Center , 142 T.C. at 13-14; Securitas , at *18. We will start with risk distribution.

A. Risk Distribution

Risk distribution is one of the common characteristics of insurance identified by the Supreme Court in Le Gierse , 312 U.S. at 539, and occurs when the insurer pools a large enough collection of unrelated risks. Rent-A-Center , 142 T.C. at 24; Securitas , at *25. The idea is based on the law of large numbers—a statistical concept that theorizes that the average of a large number of independent losses will be close to the expected loss. See Securitas , at *25-*26; Malone & Hyde, Inc. v. Commissioner , T.C. Memo. 1993-585, 1993 WL 516207, at *14 n.23, rev’d on other grounds , 62 F.3d 835 (6th Cir. 1995). As the Ninth Circuit explained in Clougherty Packing Co. v. Commissioner , 811 F.2d 1297, 1300 (9th Cir. 1987), aff’g 84 T.C. 948 (1985), “[b]y assuming numerous relatively small, independent risks that occur randomly over time, the insurer smoothes out losses to match more closely its receipt of premiums.”

1. The Parties’ Arguments

The Avrahamis and Feedback claim their arrangement distributed risk in two independently sufficient ways. First, they rely on Securitas and the Sixth Circuit’s opinion in Humana Inc. v. Commissioner , 881 F.2d 247 (6th Cir. 1989), aff’g in part, rev’g in part 88 T.C. 197 (1987), to argue that Feedback created a sufficient pool of risks merely by issuing seven types of direct policies—each covering a variety of risks—to the Avrahami entities. Second, they argue that Feedback adequately distributed risk through the Pan American program by reinsuring over 100 geographically diverse third parties. They read Harper Group to establish a rule that risk is distributed whenever premiums from unrelated insureds constitute more than 29% of a captive’s premiums and point out that Feedback received over 33% of its premiums in 2009 and almost 31% of its premiums in 2010 from Pan American participants.

The Commissioner doesn’t think it is quite that simple. He argues that Feedback directly insured only three related entities in 2009 and four in 2010, too few compared with the numbers in our precedents where we found there to be insurance and wholly insufficient to benefit from the law of large numbers. He also claims that Feedback and the Avrahamis misread Harper Group by focusing solely on the percentage of premiums that came from third parties when the number of unrelated insureds—over 7,500—was equally important. Finally, the Commissioner argues that the Pan American program doesn’t distribute risk because the terrorism policies weren’t covering relatively small, independent risks.

2. Affiliated Entities

We will look first at the argument that Feedback achieved sufficient risk distribution by issuing policies to the Avrahami entities. It is certainly possible for a captive to distribute risk by insuring only its brother-sister businesses. The Sixth Circuit said as much in Humana , 881 F.2d at 257, and we subsequently agreed. See Rent-A-Center , 142 T.C. at 24 (“A captive may achieve adequate risk distribution by insuring only subsidiaries within its affiliated group.”); Securitas , at *26-*27. But because the captive must still have a large enough pool of unrelated risks, the question is, what is “large enough”?

The experts in these cases would have us focus on the number of insureds. David Babbel—the Commissioner’s expert witness with a background in insurance and financial economics whom we find to be credible—testified at trial that “depending on the nature of the risk” a captive would likely need to insure a minimum of 35 sibling entities to get good distribution of risk. 47 In his opinion the captive must insure this minimum number to start to benefit from the law of large numbers and to get sufficient pooling of risk. The Avrahamis’ expert witnesses, Daniel Spragg and Sheila Small, expressed a similar opinion—but with a lower number—in their expert report. Citing Revenue Ruling 2002-90 , 2002-2 C.B. 985, they explained: “It is commonly cited that there must be at least 12 affiliated corporate policyholders that pool their risk in a pure captive arrangement and in an association captive arrangement, there must be at least seven policyholders.” 48 Since Feedback was insuring only three affiliated entities in 2009 and four in 2010, we don’t need to decide which of these experts is correct. Instead we will simply agree that when analyzing the number of related insured companies, Feedback failed to adequately distribute risk. 49

We also want to emphasize that it isn’t just the number of brother-sister entities that one should look at in deciding whether an arrangement is distributing risk. It’s even more important to figure out the number of independent risk exposures . For example, in R.V.I . we found risk distribution not just because an insurance company insured 714 different unrelated parties, but also because it issued 951 policies covering more than 750,000 vehicles, 2,000 real estate properties, and 1.3 million equipment assets in 7 different geographic regions. R.V.I ., 145 T.C. at 228-29. And in Rent-A-Center we found “a sufficient number of statistically independent risks” when the captive provided workers’ compensation, automobile, and general liability policies that covered more than 14,000 employees, 7,100 vehicles, and 2,600 stores in all 50 states. 142 T.C. at 24. We also held that risk distribution was achieved when a captive provided worker’s compensation coverage for more than 300,000 employees, automobile coverage for more than 2,250 vehicles, and other coverages for more than 25 separate entities. See Securitas , at *26. Even in Humana , 881 F.2d at 257, where the Sixth Circuit held that risk distribution could occur when a captive “insures several separate corporations within an affiliated group”, the captive insured more than 20 corporations operating more than 60 hospitals with more than 8,500 beds. See Humana Inc. v. Commissioner , 88 T.C. 197, 199-202 (1987).

We find that Feedback’s risk exposures fall short of these situations. It issued seven types of direct policies—five to American Findings and two each to Chandler One, O&E, and White Knight. American Findings’ policies covered 3 jewelry stores, 2 key employees, and around 35 employees. The remaining policies covered three commercial real estate properties, all in metropolitan Phoenix. While we recognize that Feedback is a microcaptive and must operate on a smaller scale than the insurance companies in R.V.I . or Rent-A-Center , we can’t find that it covered a sufficient number of risk exposures to achieve risk distribution merely through its affiliated entities.

3. Pan American

Recall, however, that the Avrahamis also argue that Feedback distributed risk by participating in the Pan American program and reinsuring third-party risk. They stress that we found risk distribution in two prior cases by looking at the percentage of the captive’s gross premium income received from unrelated insureds. Specifically, in AMERCO , 96 T.C. at 36-37, “outside insurance constituted over 50 percent of [the captive’s] gross written premiums” and in Harper Group , 96 T.C. at 52, at least 29% of the captive’s gross premium revenue came from unrelated parties. 50 The Avrahamis encourage us to use the same reasoning here. We will decline to do so because we can’t read these cases—and others like them—that narrowly.

In cases where we have found there to be risk distribution because a captive insured a sufficient number of unrelated parties, we also found the policies issued to those unrelated parties covered insurable risks, successfully shifted the risks to the captive, and satisfied all the commonly accepted notions of insurance. See, e.g., Rent-A-Center , 142 T.C. at 21-25; Harper Grp ., 96 T.C. at 58-60; AMERCO , 96 T.C. at 39-42; Securitas , at *7-*10. Therefore, before we can say whether Feedback distributed risk through the Pan American program, we must decide if Pan American itself was a bona fide insurance company. See Rent-A-Center , 142 T.C. at 10. In determining whether an entity is a bona fide insurance company we have looked at such factors as:

whether it was created for legitimate nontax reasons;

whether there was a circular flow of funds;

whether the entity faced actual and insurable risk;

whether the policies were arm’s-length contracts;

whether the entity charged actuarially determined premiums;

whether comparable coverage was more expensive or even available;

whether it was subject to regulatory control and met minimum statutory requirements;

whether it was adequately capitalized; and

whether it paid claims from a separately maintained account.

Id . at 10-13. We will take a few of these in turn.

i. Circular Flow of Funds

Recall that Pan American was structured in such a way that a small business would pay Pan American a premium for coverage up to a certain limit. Then Pan American would turn around and reinsure all the risk it had assumed, making sure that the captive related to the small business received reinsurance premiums equal to those paid by that small business. Under the 2009 program, for example, American Findings paid Pan American $360,000 for up to $5,525,000 in terrorism coverage and Pan American paid Feedback $360,000 for reinsuring 1.797% of Pan American’s total losses. The same was true under the 2010 program—American Findings paid Pan American $360,000 for up to $5,125,000 in terrorism coverage and Pan American paid Feedback $360,000 for reinsuring 1.56% of Pan American’s total losses. While not quite a complete loop, this arrangement looks suspiciously like a circular flow of funds. The end result of two years in the Pan American program was the transfer of $720,000 from an entity owned 100% by the Avrahamis to one owned 100% by Mrs. Avrahami.

ii Unreasonable Premiums

Recall also that Pan American was structured around a single type of insurance policy—terrorism coverage known as TRIP. While the parties agree that terrorism is an insurable risk, their views on the reasonableness of the policy terms and premiums charged are drastically different. The Avrahamis take the position that the premium rate—6.5% of the loss limit in 2009 and 7% in 2010—was reasonable because they fell within an actuarially determined range. Specifically, they argue that Rosenbach—the actuary hired by Clark—performed a market study and “based upon his experience and research” recommended a rate on line between 5% and 8% of the policy limit in 2009 and between 5% and 9% in 2010. The Commissioner, however, has serious qualms about Rosenbach’s method. He points out with considerable persuasiveness that Rosenbach’s study included no rates from insurers that actually provided similar coverage. Rosenbach’s study did not address, much less quantify, how the differences between Pan American’s TRIP coverage and commercially available TRIA coverage might affect premiums. And in the end he recommended a one-size-fits-all rate for all of Clark’s scores of clients. There are some rather obvious questions here: For example, Pan American charged the same rate to all of its participants regardless of their geographic location, yet Rosenbach’s report cited an Insurance Services Office source indicating the commercial market price for terrorism risk insurance subject to TRIA can be up to 10 times more expensive for policyholders in large cities like New York or Washington. Pan American participants also paid the same rate even if they—like American Findings—had terrorism coverage under another policy, even though the TRIP policy was written as an excess-coverage policy, in which the risk assumed by Pan American was only the portion of a loss not covered, or not fully covered, by another policy.

In analyzing the reasonableness of Pan American’s premiums, Babbel’s report compared Pan American’s terrorism risk rate to that of commercial carriers. While we recognize that Pan American was operating on a smaller scale than other insurers providing stand-alone terrorism coverage and that the TRIP policies contained some unique terms, we find Babbel’s insight instructive and a good reasonableness check. For starters, American Findings paid Jewelers Mutual more than $1,500 in 2009 and $1,600 in 2010 for add-on terrorism coverage with a policy limit of $2 million and exclusions for chemical and biological hazards. This amounts to a rate on line—the premium divided by the limit—of less than 0.081%. On the other hand, American Findings paid Pan American $360,000 in both 2009 and 2010 for stand-alone terrorism coverage with policy limits of around $5.5 million in 2009 and $5.2 million in 2010. This amounts to a rate on line of between 6.5% and 7%—or approximately 80 times more than the rate under the Jewelers Mutual policy. Some of this difference can be explained by TRIP’s additional coverage of chemical and biological attacks—Rosenbach testified that 80% to 90% of the rate range he proposed was related to the chemical and biological coverage. But this doesn’t account for TRIP’s policy terms that cut in the other direction, such as the provision excluding coverage for any attack occurring in any city with more than 1.5 million residents, the provision giving TRIP excess status to any other policy, or the provision allowing for the payment of claims with a promissory note payable over a maximum of three years.

Babbel also compared Pan American’s rates to the stand-alone terrorism premium rates published in 2013 by Marsh Risk Management Research. 51 As he explained, Marsh “reported that median stand-alone terrorism insurance premium rates for the real estate industry was $40 per million of TIV [total insurable value] in 2010, while for the retail/wholesale sector, the median rate was $24 per million of TIV in 2010.” Total insurable value (TIV) is the “full value of insured’s covered property, business income values, and other covered property interests.” Using the $8 million in total assets reported on American Findings’ 2010 tax return as a proxy for its TIV, this means American Findings paid Pan American $45,000 per million of TIV. 52 Even if American Findings’ TIV was higher—perhaps to account for TRIP’s inclusion of coverage for both the loss of property and business income that resulted directly from a covered act of terrorism—the premiums charged by Pan American are grossly excessive.

iii. Arm’s-Length Contracts

Babbel also questioned the contracts with Pan American and concluded that “[u]nder this combination of conditions, no reasonable, profit-seeking business would enter into a contract with the terms of the Pan-American coverage with an insurer absent certain beneficial tax considerations” (these conditions being that Pan American was charging exorbitantly high premiums for coverage that had a very low probability of being triggered and a questionable ability to pay claims). For a claim to be covered under TRIP, the policyholder had to suffer a loss resulting directly from an act of terrorism certified by the Secretary of the Treasury with the concurrence of the Secretary of State and the Attorney General that resulted in over $100 million in losses and that occurred within the United States but not in a city with more than 1.5 million residents. The claim also had to not be covered in full under another policy. Rosenbach testified that he did not know of any event in history that would have met these requirements. And of course no events took place that triggered a claim during the years at issue in these cases.

It is also questionable whether a qualifying loss would have been paid. Pan American received more than $20 million in premiums in 2009 and in turn agreed to insure up to an aggregate of more than $308 million in losses. It ceded 100% of this insurance risk to the participating captives and likewise passed on 100% of its premiums—50% within 90 days, another 47.5% within 180 days, and the last 2.5% when the policy expired on December 15. This means that by July 2010—more than 180 days into the 2009 policy period—Pan American had only about $500,000 of premium revenue on hand to pay claims. If a covered claim came in, Pan American would have to go to each of the participating reinsurers to get the necessary cash; and if a reinsurer wouldn’t or couldn’t pay, Pan American would have to foot the bill itself—an event that would be more probable if, like Mr. Avrahami, other captives owners would “freak out” if they lost money in the Pan American program. We understand that Pan American was adequately capitalized under Nevisian law, which required only $75,000 in capital because Pan American ceded 100% of its risk. See Nevis International Insurance Ordinance 2009, c. 7.07(N), sec. 9(1)(b)(iii) . But we need not pretend that a company as thinly capitalized as Pan American, with directors and a management team substantially weaker in numbers and ability than those of a normal reinsurer, would be hard pressed to enforce the cession agreements against the scores of captive insurers it might have to go after.

iv. Bona Fide Insurance Company

The Avrahamis nevertheless argue that Pan American was a fronting company whose business purpose was to issue terrorism risk insurance and then fully cede—or reinsure—that risk to other insurers. Both parties agree that fronting companies are a real thing in the insurance industry. See, e.g., Hanover Ins. Co. v. Urban Outfitters, Inc ., 806 F.3d 761, 764 n.3 (3d Cir. 2015) (explaining that a fronting company issues fronting policies, which are “a risk management technique in which an insurer underwrites a policy to cover a specific risk but then cedes the risk to a reinsurer”); Wal-Mart Stores, Inc. v. Crist , 855 F.2d 1326, 1330 n.3 (8th Cir. 1988) (describing an entity as a “fronting company” when it permitted the use of its insurance policy forms for a percentage fee). The Commissioner, on the other hand, says Pan American wasn’t operated like a fronting company and was formed merely as a mechanism for Clark’s clients to meet their risk-distribution requirements. The Avrahamis’ expert witnesses testified that fronting companies generally receive a fee of around 5% of premiums as a “ceding commission” and noted that it is always a percentage rather than a fixed fee. A fronting company, while not taking on any insurance risk, still must be compensated for the services it provides, including the assumption of credit risk—the risk that the reinsurers can’t or won’t pay their portion of an insured loss, forcing the fronting company to come up with the money to settle a claim. Pan American, however, did not charge a ceding commission. Instead, 100% of the premiums it received were passed through to the reinsurers and its only compensation was an “insurance pool administration fee” paid by Clark of around $120,000 in 2009 and $410,000 in 2010. 53 These amounts came out of the $5,000 “all-inclusive” annual fee paid to Clark by each of the businesses participating in the Pan American program.

We won’t condemn Pan American solely for its atypical fee structure, but that structure came combined with excessive premiums, an ultralow probability of a claim ever being paid, and payments of a circular nature. We have to make a finding of fact on this. Is such an entity engaged in insurance or is it just part of a tax-reduction scheme papered to look like an entity engaged in insurance? The answer is that more likely than not, Pan American is not a bona fide insurance company. We know it was regulated under the laws of the Island of Nevis and met that loosely regulated regime’s low capitalization requirements, but that is not enough.

Because we find that Pan American was not a bona fide insurance company, we cannot find that the policies it was issuing were insurance, which in turn means Feedback’s reinsurance of those same policies did not distribute risk. Therefore, neither through its affiliated entities nor through Pan American did Feedback accomplish sufficient risk distribution for its arrangements to be considered “insurance” for federal income tax purposes.

B. Insurance in the Commonly Accepted Sense

The absence of risk distribution by itself is enough to sink Feedback. See AMERCO , 96 T.C. at 40 (holding that risk-shifting and risk-distributing “are necessary to the existence of insurance” (emphasis added) (citing Gulf Oil Corp. v. Commissioner , 89 T.C. 1010, 1023 (1987)); see also Humana , 881 F.2d at 257 (“Risk transfer and risk distribution are two separate and distinct prongs of the test and both must be met to create an insurance contract.”). But the cases tell us that in deciding whether an arrangement is insurance we can also look at whether it looks like insurance in the commonly accepted sense. This is an alternative ground. To analyze it, we look at numerous factors, including whether the company was organized, operated, and regulated as an insurance company; whether the insurer was adequately capitalized; whether the policies were valid and binding; whether the premiums were reasonable and the result of an arm’s-length transaction; and whether claims were paid. See R.V.I ., 145 T.C. at 231; Rent-A-Center , 142 T.C. at 24-25; Harper Grp ., 96 T.C. at 60; Securitas , at *27. We have also looked at whether the policies covered typical insurance risks and whether there was a legitimate business reason for acquiring insurance from the captive. See Hosp. Corp ., 1997 WL 663283, at *26.

1. The Parties’ Arguments

The Avrahamis and Feedback argue that their arrangements clearly satisfied these requirements. They assert that Feedback was organized, operated, and regulated as an insurance company under the laws of St. Kitts. Likewise, they claim Feedback was adequately capitalized because it met or exceeded the minimum capitalization requirements established by St. Kitts. Their position is that all the policies were valid and binding because they contained customary insurance terms and conditions, the premiums were reasonable and determined by an actuary, and all approved losses were paid.

The Commissioner disagrees. He argues that Feedback was organized solely for tax purposes, didn’t operate as an insurance company, lacked arm’s-length premium determinations, failed to get timely approval for transfers to related parties in violation of Kittian regulations, didn’t satisfy Arizona insurance regulations, and issued policies with contradictory provisions.

2. Organization, Operation, and Regulation

We will start with organization, operation, and regulation as an insurance company. The parties agree that Feedback was incorporated and regulated as a captive insurance company in St. Kitts. The Commissioner, however, argues that Feedback was also subject to, but in violation of, Arizona insurance regulations. Whether the Commissioner is correct will not change the outcome of these cases, so we’ll leave this question for another day—or perhaps leave it to the Arizona Department of Insurance. The more interesting question is whether Feedback was operated as an insurance company and in making this determination we “must look beyond the formalities and consider the realities of the purported insurance transactions.” Id . at *24 (citing Malone & Hyde, Inc. v. Commissioner , 62 F.3d 835, 842-43 (6th Cir. 1995), rev’g T.C. Memo. 1989-604).

We have to find that Feedback’s operations left something to be desired. It dealt with claims “on an ad hoc basis.” It invested only in illiquid, long-term loans to related parties and failed to get regulatory approval before transferring funds to them. And we will not overlook the fact that the Avrahami entities made no claims whatsoever against Feedback from its inception in 2007 until March 2013—two months after the IRS sent the Avrahamis documents about the audits of the returns of American Findings, Chandler One, O&E, and White Knight that suggested Feedback was a sham. And even the claims Feedback did receive it dealt with in questionable ways. Most of the claims were approved despite being filed late—the policies required that Feedback be notified within 30 days of the loss “as a condition precedent to payment of any benefit hereunder.” And Heritor approved White Knight’s claim for a new roof despite a lack of evidence about the cause of the damage. If the damage was caused by a fortuitous event—i.e., hail damage—then Heritor should have considered whether White Knight’s commercial policies with Allied Insurance would have covered the loss since White Knight’s Business Risk Indemnity policy from Feedback offered only excess coverage. And if the damage was caused by normal wear and tear—i.e., not a fortuitous event—then Feedback probably shouldn’t be covering the loss because it is a normal business expense and not an insurance risk. See Commissioner v. Treganowen , 183 F.2d 288, 290 (2d Cir. 1950), rev’g 13 T.C. 159 (1949).

Feedback also made investment choices only an unthinking insurance company would make. For example, in 2010 Feedback reported two types of assets on its tax return—more than $1.35 million in cash and $2.53 million in “mortgage and real estate loans.” The $2.53 million comprised an $830,000 secured promissory note from Belly Button with no payments due until February 2018, a $1.5 million unsecured promissory note from Belly Button due in March 2020, and a $200,000 unsecured note from Belly Button—even though the funds went directly to Mrs. Avrahami—due no earlier than December 2012. This meant that by the end of 2010 more than 65% of Feedback’s assets were tied up in long-term, illiquid, and partially unsecured loans to related parties. And, despite Kittian regulations requiring advance approval for any loan to a parent or affiliated person, see Captive Insurance Companies Act of 2006, ch. 21.20 sec. 15.6 (2006), Heritor didn’t disclose these loans to St. Kitts’ Registrar until September 2014. If something catastrophic happened to Belly Button, its ability to repay the loans would be impaired and make Feedback’s ability to pay on any claims doubtful. Even if Feedback was organized and regulated as an insurance company, we find it was not operated like one.

3. Capitalization

Next we look at capitalization. The parties agree that Feedback met the minimum capitalization requirements of St. Kitts. But the question is whether meeting Kittian minimums is the same as being adequately capitalized. Our caselaw implies that the answer is yes. See R.V.I ., 145 T.C. at 231 (holding that insurers who met the minimum capital requirements of their regulators were adequately capitalized); Rent-A-Center , 142 T.C. at 13, 23-24 (finding that a captive was adequately capitalized when it met Bermuda’s minimum statutory requirements after consideration of a parental guaranty); Harper Grp ., 96 T.C. at 50, 60 (holding a captive that was in compliance with Hong Kong’s minimum capital law was adequately capitalized); Securitas , at *23 (finding a reinsurer was adequately capitalized when it met Vermont’s minimum premium-to-surplus ratio); Malone , 1993 WL 516207, at *15. We won’t upset that consensus here.

4. Valid and Binding Policies

Caselaw is slightly less vague on what makes a policy “valid and binding.” Our Opinion in R.V.I . suggests policies are valid and binding if the insureds filed claims for all covered losses and the captive paid them. See R.V.I ., 145 T.C. at 231. And in Securitas we found valid and binding policies when “[e]ach insurance policy identified the insured, contained an effective period for the policy, specified what was covered by the policy, stated the premium amount, and was signed by an authorized representative of the company.” Securitas , at *28. Here the policies issued by Feedback and Pan American identified the insured, stated the premium amount, and were signed by Heritor or Heritage as the authorized insurance manager. But some of the policies were less than a model of clarity. The Avrahamis assert that all the policies issued by Feedback were claims-made policies, yet the policies say otherwise. For example, Chandler One’s 2009 Administrative Action policy states that Feedback “agrees to pay to the Insured any legal expense incurred by the insured during the Policy Period, arising from or relating to the defense of any Insured Event as defined hereunder, which Insured Event is instituted against the Insured during the Policy Period.” The policy then defines “Policy Period” as “[e]vents occurring and reported from and after 12:01 a.m. December 15, 2009 and prior to 12:01 a.m. December 15, 2010.” By its plain terms the policy limits coverage to legal expenses incurred and reported between December 2009 and December 2010, terms indicative of both a claims-made policy—the claim must be reported during the policy period—and of an occurrence policy—the claim must occur during the policy period.

5. Reasonableness of Premiums

The next question is whether Feedback’s and Pan American’s premiums were reasonable and the result of an arm’s-length transaction. Clark did hire Rosenbach to help price the policies, and he testified extensively about his process; but because he priced captives only for Clark and his explanations were often incomprehensible, we don’t find them persuasive. We find instead that the premiums were utterly unreasonable. In 2006—before Feedback was formed—the Avrahami entities spent about $150,000 on insurance. With Feedback in the picture, their insurance bills soared to more than $1.1 million in 2009 and more than $1.3 million in 2010. While the Avrahami entities were paying Feedback and Pan American just shy of $1.2 million a year, they were also maintaining their commercial coverage for less than $90,000 a year.

Starting with Feedback’s direct policies, recall that Rosenbach’s process was to determine a base premium for each policy and then to adjust that base by various factors. While he derived parts of his calculations from a standard Chubb filing, he consistently chose inputs that generated higher premiums. For example, Rosenbach used the hazard-group classifications from a June 2005 Chubb filing, which placed “property managers” like Chandler One, O&E, and White Knight into a riskier—and therefore more expensive—group, even though the January 2005 Chubb filing from which he pulled other adjustment factors put them in a less risky group. This decision alone has a ripple effect. All else being equal, if Rosenbach had used hazard group 2 instead of 4 when he determined the premium for Chandler One’s 2009 Administrative Action policy, he would have calculated a base rate of $4,991 54 instead of $11,874, which would reduce the total premium by more than half. He likewise chose a claims-made factor of 1.3—the highest in the Chubb filing and representing retroactive coverage for five or more years—on the assumption that all of the Feedback policies were claims-made with no retroactive date. But as we have already found several of the policies had terms limiting coverage to events occurring during the policy period. And the Chubb filing states that if there is no retroactive coverage the claims-made factor should be 1.0—a 30% reduction from what Rosenbach used.

Apart from the factors suggested in the Chubb filing, Rosenbach also made adjustments based on his professional judgment—most without a coherent explanation. For example, he increased Chandler One’s 2009 Business Risk Indemnity premium by $2,500—$1,000 in 2010—for providing excess general-liability coverage but offered no rationale for the amount chosen. In the same calculation, Rosenbach also increased the premium by more than $26,000—and more than $62,000 in 2010—for “construction defects” coverage. But Chandler One wasn’t constructing anything; it just owned and leased commercial real estate. Rosenbach also struggled to explain the premium calculation for the Litigation Expense policy, which included a subjective “base exposure” and several subjective factors that Rosenbach testified were multiplied together to reach the premium amount. As the Avrahamis had to explain in their reply brief, however, to reach the premium that Rosenbach calculated the base exposure has to be multiplied by all of the factors except the expense ratio and then divided by one minus the expense ratio. In 2010 Rosenbach came up with a nearly 60% higher premium for the Litigation Expense policy because he “appl[ied] the model in a little different light.” Similarly, Rosenbach was unable to explain the 7.5% event rate used in the Tax Indemnity calculation or even what it represented. Without a comprehensible explanation we can’t find these premium amounts justified.

We also note that Rosenbach performed his premium calculations using applications, financial statements, tax returns, and a “business plan” that Clark provided and that summarized the key terms of the policies. The business plan indicated that the Loss of Key Employee policy insured American Findings against “a reduction of its business income resulting from the voluntary departure of any of its key employees, not including shareholders.” Yet the $86,000 premium calculated by Rosenbach for 2009 is based on the assumption that Mr. and Mrs. Avrahami—American Findings’ shareholders—would be the only covered key employees. We find from all this that Rosenbach’s calculations aimed not at actuarially sound decision-making but at justifying total premiums as close as possible to $1.2 million—the target—without going over. To do so he would add in a proration factor or drop the policy limits until he reached his goal.

6. Payment of Claims

Finally, we will look at whether Feedback paid its claims. The simple answer is yes. Once the Avrahamis knew the IRS was looking, they began to file claims against Feedback—six in total. Five have been paid and one was still pending at the time of trial.

Does this add up to “insurance in the commonly accepted sense?” We find that the answer is “no”. Although Feedback was organized and regulated as an insurance company, paid the claims filed against it, and met the minimal capitalization requirements of St. Kitts, these insurance-like traits cannot overcome its other failings. It was not operated like an insurance company, it issued policies with unclear and contradictory terms, and it charged wholly unreasonable premiums.

C. Not Insurance

Because we find that Feedback failed to distribute risk and was not selling insurance in the commonly accepted sense, we need not decide whether its transactions involved insurance risk or risk shifting. See, e.g., Clougherty , 811 F.2d at 1300 n.5 (“While the Supreme Court in Le Gierse expressly stated that insurance must exhibit both the shifting and the distributing of risk, the resolution of that case depended only on the absence of risk shifting. Here, too, we need not consider whether Clougherty’s captive insurer arrangement exhibited risk distribution because we conclude that Clougherty did not shift its risk.”). Even if they did, our decision would remain the same. Despite the attempts of Feedback to make its transactions look like traditional insurance and take advantage of the apparent loophole at the intersection of section 831 and captive insurance caselaw, the premiums paid to Feedback and deducted by the Avrahami entities are not “insurance” for federal tax purposes. Because we find that Pan American was not a bona fide insurance company, the premiums paid to it and deducted by American Findings are also not for “insurance” for federal tax purposes. See Rent-A-Center , 142 T.C. at 10. In light of our holding we need not address the Commissioner’s other arguments—i.e., that the amounts deducted as insurance expenses should be disallowed under the economic-substance, substance-over-form, and step-transaction doctrines.

IV. Effect on Feedback

Our holding has two major consequences. The first is on Feedback’s elections. Recall that Feedback elected to be treated as a domestic corporation for federal income tax purposes under section 953(d) and elected to be taxed as a small insurance company under section 831(b) . To qualify under section 831(b) a company must first be an “insurance company,” which the Code defines as “any company more than half of the business of which during the taxable year is the issuing of insurance or annuity contracts or the reinsuring of risks underwritten by insurance companies.” Secs. 816(a) , 831(c) . Since Feedback’s policies were not contracts for insurance, it does not fall within this definition because less than half of its business—in fact none of its business—was the issuance of insurance contracts. Its section 831(b) election is therefore invalid for the years at issue in these cases—2009 and 2010. 55 Likewise, to make a valid section 953(d) election, a controlled foreign corporation—as defined in section 957(a) —must qualify under part I (life insurance companies) or II (other insurance companies) of subchapter L. Sec. 953(d)(1)(B) . To qualify for either part a company must meet the section 816(a) definition of “insurance company,” which Feedback did not. See secs. 816(a) , 831(c) . Feedback’s section 953(d) election is therefore likewise invalid for 2009 and 2010.

Although we find that Feedback’s elections are invalid and it is therefore a foreign corporation for federal income tax purposes, we still have to figure out what this means for the taxation of the amounts Feedback received. It appears, however, the parties have resolved this issue themselves by stipulating that the “‘Taxable Premiums Earned’ by [Feedback] in the amounts of $1,090,000 and $1,170,000 for taxable years 2009 and 2010, respectively, are not U.S. source fixed or determinable, annual or periodical income under section 881 , or income that is effectively connected with a U.S. trade or business under section 882 .” This means Feedback isn’t subject to tax under those two sections, which govern the taxation of foreign corporations. See secs. 881(a)(1) , 882(a)(1) .

V. Effect on the Avrahamis

The second major consequence is that if the payments aren’t for insurance, then they are not ordinary and necessary business expenses and may not be deducted under section 162(a) by American Findings, Chandler One, O&E, and White Knight. See sec. 1.162-1(a) , Income Tax Regs. We therefore sustain the Commissioner’s determination to adjust the Avrahamis’ income from Schedule E, Supplemental Income and Loss, by disallowing these deductions.

A. Ordinary Dividend

We also must decide the tax effect of the transfers from Feedback to Belly Button and Mrs. Avrahami. Recall that during the years at issue in these cases Feedback transferred $200,000 directly to Mrs. Avrahami. While the Avrahamis and Feedback originally claimed this $200,000 was a loan and was backed by a demand promissory note, they now concede that it was a dividend and was mistakenly left out of their taxable income for 2010. They do argue that it should be taxed as a qualified rather than ordinary dividend. Section 1(h)(11) provides preferential tax rates for “qualified dividend income” if the dividend is received from a domestic corporation or a qualified foreign corporation. Since we found that Feedback’s section 953(d) election was invalid, it is not a domestic corporation. It is also not a qualified foreign corporation because it was not incorporated in a possession of the United States or eligible for benefits under an income tax treaty. See sec. 1(h)(11)(C) . We therefore hold that the $200,000 transferred to Mrs. Avrahami should be taxed at ordinary income rates.

B. Bona Fide Loans

In 2010 Feedback also transferred $1.5 million to Belly Button in exchange for an unsecured promissory note with an interest rate of 4% per year—simple interest accruing “from time to time.” Two days later the Avrahamis transferred$1.5 million from Belly Button’s bank account into their personal one. The Avrahamis argue that these are two separate nontaxable transactions. They assert that the first was a bona fide loan between Feedback and Belly Button and the second was the repayment of the $1.2 million loan—used by Belly Button to purchase land in May 2007—between Belly Button and Mr. Avrahami. The Commissioner counters that together the two transactions were really just a taxable distribution from Feedback to the Avrahamis.

This is a common chew toy in tax litigation. How do we distinguish bona fide loans from distributions? A loan is “an agreement, either express or implied, whereby one person advances money to the other and the other agrees to repay it upon such terms as to time and rate of interest, or without interest, as the parties may agree.” Welch v. Commissioner , 204 F.3d 1228, 1230 (9th Cir. 2000) (quoting Commissioner v. Valley Morris Plan , 305 F.2d 610, 618 (9th Cir. 1962)), aff’g T.C. Memo. 1998-121. We therefore ask whether both parties had an actual intent to establish a debtor-creditor relationship at the time the funds were advanced. Talmage v. Commissioner , T.C. Memo. 2008-34, 2008 WL 440831, at *24 (citing Estate of Chism v. Commissioner , 322 F.2d 956, 960 (9th Cir. 1963), aff’g Chism Ice Cream Co. v. Commissioner , T.C. Memo. 1962-6), aff’d , 391 F. App’x 660 (9th Cir. 2010); see also Kaider v. Commissioner , T.C. Memo. 2011-174, 2011 WL 2976203, at *5. We also ask whether the debtor intended to repay the loan and whether the creditor intended to enforce its repayment. See Kaider , 2011 WL 2976203, at *5; see also Beaver v. Commissioner , 55 T.C. 85, 91 (1970); Fisher v. Commissioner , 54 T.C. 905, 909-10 (1970). Precedent gives us lots of factors to look at to answer these questions:

the ability of the borrower to repay;

the existence or nonexistence of a debt instrument;

security, interest, a fixed repayment date, and a repayment schedule;

how the parties’ records and conduct reflect the transaction;

whether the borrower had made repayments; and

whether the lender had demanded repayment.

Kaider , 2011 WL 2976203, at *5-*6 (citing Welch , 204 F.3d at 1230-31; Frierdich v. Commissioner , 925 F.2d 180, 182-85 (7th Cir. 1991), aff’g T.C. Memo. 1989-393; Mann Constr. Co. v. Commissioner , T.C. Memo. 1999-183); see also Stanley v. Commissioner , T.C. Memo. 2016-196, at *8 .

1. Ability To Repay and Debt Instruments

Whether a borrower has the ability to repay is determined by whether there was “a reasonable expectation of repayment in light of the economic realities of the situation.” Fisher , 54 T.C. at 910. The borrower’s inability to repay at the time it received the funds suggests the parties did not intend for the transaction to be a bona fide loan. Kaider , 2011 WL 2976203, at *6. In 2007 after Belly Button borrowed the $1.2 million from Mr. Avrahami, it had assets of around $1,980,000 and outstanding loans of the same amount. By 2010 its liabilities exceed its assets by more than $50,000. 56 But assuming the land appreciated—which Mr. Avrahami believed at the time that it would since it was across the street from a lot owned by Walmart—there was a reasonable expectation that Belly Button could satisfy its debts and accrued interest by selling the land. This makes them look like loans.

The second factor does too: Both of the transactions at issue here—the transfer from Feedback to Belly Button and from Mr. Avrahami to Belly Button-were papered properly by promissory notes signed by Mr. Avrahami.

2. Security, Interest, and Repayment Schedules

The third factor is slightly more complicated. While both promissory notes indicated the interest rate to be charged and the date for repayment, neither was secured. The promissory note from Belly Button to Feedback required repayment of the $1.5 million principal plus accrued interest—simple interest at 4% per year—on March 17, 2020. The note from Belly Button to Mr. Avrahami required repayment of the $1.2 million principal plus accrued interest—simple interest at 4% per year—on April 30, 2017. The Avrahamis insist that the $1.5 million note from Belly Button to Feedback was secured by a mortgage on Belly Button’s land, but the plain terms of the promissory note indicate otherwise. It specifically states: “No Security. The Indebtedness is unsecured.” The $1.2 million note contains an identical statement. Although the $1.2 million note called for 4% interest, when it was allegedly paid in 2010 neither the Avrahamis nor Belly Button reported any interest paid. While the absence of security weighs against finding a loan, the overall factor is more neutral because defined interest and repayment schedules were included in both notes. But the casual attitude toward the payment and reporting of interest gives us great pause.

3. Parties’ Conduct

The next factor is how the parties’ records and conduct reflect the transactions. In 2007 when the $1.2 million note was executed, Belly Button recorded it on its tax return as $1,201,000 “due to affiliates.” The $1,201,000 was the full amount Belly Button received from Mr. Avrahami, yet the note evidences a $1,000 lower amount for no clear reason. Then in 2010 Belly Button transferred $1.5 million from its bank account into the Avrahamis’ bank account, claiming this was a repayment of the $1.2 million loan. 57 Yet Belly Button’s 2010 tax return shows only a $670,000 reduction in the amount owed to Mr. Avrahami and no indication that any interest had been paid. 58 The recording of the $1.5 million note has similar problems. While the full $1.5 million was reported on Feedback’s 2010 tax return under “mortgage and real estate loans,” Belly Button reported only a $670,000 increase to its note-payable account in 2010. 59 Our great pause grows into skepticism.

4. Repayments and Demands

We will look at the fifth and sixth factors together. Both repayments and demands for repayment are evidence that the parties intended a bona fide loan. See, e.g., Welch , 204 F.3d at 1230-31. Here, the $1.2 million note was removed from Belly Button’s books—which suggests it was repaid, but without the payment of accrued interest or any demand for it. The $1.5 million note, on the other hand, calls for payment of the principal and all accrued interest 10 years after the date of the note. Although no repayments or demands for repayment had been made by the time of trial in these cases, none were due. These factors therefore neither help nor hurt the Avrahamis. See, e.g., Kaider , 2011 WL 2976203, at *9.

We also consider the testimony of the borrower and lender—here, Mr. Avrahami. He did indeed insist that the transactions were intended to be loans. The problem here is that Mr. Avrahami himself was on all sides of both transactions. He was—as an individual—the lender on the $1.2 million note and—as a manager of Feedback—the lender on the $1.5 million note. He was also the signer on behalf of the borrower—Belly Button—for both notes.

Our skeptical view turns into a jaundiced eye here. As we said in Holden v. Commissioner , T.C. Memo. 2015-131, at *27 : “The ultimate question, which is a factual one, is whether the parties intended a loan when the funds were advanced.” See also Miele v. Commissioner , 56 T.C. 556, 567 (1971), aff’d , 474 F.2d 1338 (3d Cir. 1973). This question becomes even more important when the Commissioner has proposed an alternative characterization. See Holden , at *27 n.11; see, e.g., Calloway v. Commissioner , 135 T.C. 26, 27 (2010) (stock sale proceeds), aff’d , 691 F.3d 1315 (11th Cir. 2012); Teymourian v. Commissioner , T.C. Memo. 2005-232 (constructive dividends). Here, the Commissioner argues that Belly Button was a mere conduit for the transfer of a $1.5 million distribution from Feedback to the Avrahamis. He encourages us to apply the substance-over-form and step-transaction doctrines to construe the transfer as a constructive dividend to Mrs. Avrahami. We will nevertheless decline to do so.

It’s a fairly close question, but the economic reality is that when Belly Button accepted the $1.2 million from Mr. Avrahami, it had the ability to repay, reasonably expected its land would appreciate, and evidenced its obligation to repay with a debt instrument that specified the applicable interest rate and repayment schedule. The same facts lead us to conclude that there really was a $1.5 million loan from Feedback to Belly Button as well. But we also find that the difference between these amounts ended up in the Avrahamis’ account. We’ll hold everyone to the terms of these facially valid notes and the bank records we have: $1.2 million of the $1.5 million transfer from Belly Button to the Avrahamis is a nontaxable loan repayment and should not be included in their gross income. The extra $1,000 from Mr. Avrahami to Belly Button will be treated as a nontaxable capital contribution. See sec. 1.118-1 , Income Tax Regs. The $300,000 difference will be split between interest to the Avrahamis calculated according to the $1.2 million note’s terms (we’ll leave the math for the parties to do); and a taxable distribution to Mrs. Avrahami—taxable as an ordinary dividend for the reasons we’ve already recited. 60

VI. Penalties

Out last issue is whether any of this triggers an accuracy-related penalty under section 6662(a) . This section imposes a 20% penalty on the portion of the underpayment of tax attributable to any substantial understatement of income tax or negligence or disregard of rules or regulations. Sec. 6662(a) and (b)(1) and (2) . Negligence includes any failure to make a reasonable attempt to comply with the provisions of the Code; and disregard includes any careless, reckless, or intentional disregard. Sec. 6662(c) . The Commissioner has the initial burden of production to show that the understatement of income tax was substantial. Sec. 7491(c) . This is a math question: An understatement of income tax is substantial if it exceeds the greater of $5,000 or “10 percent of the tax required to be shown on the return.” Sec. 6662(d)(1)(A) . Here, the Avrahamis’ corrected tax liability for 2009 is more than $475,000, but they originally reported a tax liability of just shy of $100,000. Likewise, for 2010 their corrected tax liability reflects adjustments to income of over $1.6 million, but they originally reported a tax liability of less than $30,000. These understatements are well over $5,000 and greater than 10% of the tax required to be shown on the return, so the Commissioner has met his burden.

A. The Insurance Deductions

The Avrahamis can avoid accuracy-related penalties by showing either that they had “substantial authority” or that they acted with reasonable cause and in good faith. See secs. 6662(d)(2)(B)(i) , 6664(c)(1) ; sec. 1.6664-4(a) , Income Tax Regs. They try both, but we begin with reasonable cause and good faith. In determining whether this exception applies we examine all the relevant facts and circumstances, including whether they relied in good faith on professional advice and whether their reliance was reasonable. See sec. 1.6664-4(b) and (c) , Income Tax Regs.; see also United States v. Boyle , 469 U.S. 241, 250-251; Neonatology Assocs., P.A. v. Commissioner , 115 T.C. 43, 98 (2000), aff’d , 299 F.3d 221 (3d Cir. 2002). Reliance on an adviser is reasonable if:

the adviser was a competent professional who had sufficient expertise to justify reliance;

the taxpayer provided necessary and accurate information to the adviser; and

the taxpayer actually relied in good faith on the adviser’s judgment.

Neonatology Assocs ., 115 T.C. at 99. The Avrahamis argue that they relied in good faith on the professional advice of McEntee, Hiller, and Clark. We have no doubt that all three were competent professionals with sufficient expertise—McEntee is a CPA with more than 25 years of experience, Hiller is a reputable Phoenix-based estate planning attorney with more than 30 years of experience, and Clark is a New York lawyer with more than 35 years of experience who focuses her practice on small insurance companies. We also find that the Avrahamis provided their advisers with all the relevant data needed to assess the correct level of tax. The parties dispute, however, whether the Avrahamis actually relied in good faith on the advice of McEntee, Hiller, and Clark.

1. Promoters

We have already held that a taxpayer cannot reasonably rely in good faith on an adviser who is a “promoter” of the disputed transaction. See 106 Ltd. v. Commissioner , 136 T.C. 67, 79-80 (2011), aff’d , 684 F.3d 84 (D.C. Cir. 2012); Neonatology Assocs ., 115 T.C. at 98. A promoter is “an adviser who participated in structuring the transaction or is otherwise related to, has an interest in, or profits from the transaction.” 106 Ltd ., 136 T.C. at 79 (citing Tigers Eye Trading, LLC v. Commissioner , T.C. Memo. 2009-121) (adopting the definition of promoter from Tigers Eye for cases involving an adviser offering the same tax shelter to numerous parties). We find that the Avrahamis cannot reasonably rely on Clark because she is a promoter of the microcaptive transaction. She structured the captive-insurance-company transaction, drafted the purported insurance policies, and designed and set up the failed risk-distribution programs, which she sold to more than 100 of her clients. She also profited a great deal from the transaction, receiving a portion of both an initial $75,000 engagement fee and the annual flat fees. The Avrahamis could not reasonably rely on her.

Hiller is a closer call. He recommended Clark to the Avrahamis, acted as co-counsel during Feedback’s start-up phase, received a portion of the $75,000 start-up fee, and assisted the Avrahamis in documenting the 2008 loan from Feedback to Belly Button. But that appears to be the extent of his involvement. Hiller testified that he did not provide advice on the capitalization of Feedback; had no involvement in selecting an actuary to determine premiums and underwrite the polices; and played no role—not even as a reviewer—in the formation or documentation of the transactions with Pan American. He also had a prior relationship with the Avrahamis, did not give unsolicited advice, and—aside from the start-up fee—billed the Avrahamis on an hourly basis. These are all signs that a tax adviser is not a promoter, see Countryside Ltd. P’ship v. Commissioner , 132 T.C. 347, 352-53 (2009), and we find that he was not a promoter. The same goes for McEntee—he recommended Clark and prepared the tax returns for the Avrahamis and all of their entities, but we find that he was not a promoter because he did not structure the captive-insurance-company transaction or otherwise have an interest in it.

2. Reasonable Reliance

Our analysis does not end here. Our finding that the advice from Hiller and McEntee is not disqualified as advice from a promoter doesn’t necessarily mean that it was reasonable to rely on their advice or that the Avrahamis did in fact reasonably rely on their advice in good faith. The majority of the accuracy-related penalties imposed on the Avrahamis is related to the improper deduction of purported insurance premiums by the Avrahami entities, which calls into question all of the advice they received about captive-insurance-company transactions. At trial the Avrahamis admitted that all of the advice “during the formation of Feedback” came from Clark and Hiller, and McEntee himself testified that the extent of his involvement with Feedback was to provide bookkeeping services and sign tax returns. McEntee’s advice is therefore not a good defense because we find that the Avrahamis did not rely on it when they decided to set up and use Feedback.

Hiller is another matter. Although he did not specifically advise the Avrahamis about brother-sister or parental captive arrangements, Hiller credibly testified that he advised them on how captives worked, the structure of Feedback, and what kind of investments their captive could make. Mr. Avrahami also credibly testified that he decided to move forward with Clark because Hiller “gave his blessing.” Therefore, we find that the Avrahamis did rely—and reasonably relied—on Hiller’s professional advice.

3. Good Faith

But was that reliance on Hiller in good faith? Here we are sympathetic to the Avrahamis’ position. This is a case of first impression—we have discovered no other case addressing microcaptives and the interplay among sections 162 , 831(b) , and 953(d) . And we have previously declined to impose accuracy-related penalties when there is no clear authority to guide taxpayers. See Petersen v. Commissioner , 148 T.C. ___, ___ (slip op. at 29) (June 13, 2017); Williams v. Commissioner , 123 T.C. 144, 153 (2004); Hitchins v. Commissioner , 103 T.C. 711, 719-720 (1994); see also Foster v. Commissioner , 756 F.2d 1430, 1439 (9th Cir. 1985), aff’g in part, vacating in part 80 T.C. 34 (1983). We accordingly find the Avrahamis acted reasonably and in good faith when relying on Hiller’s professional advice, and we will not impose section 6662(a) penalties to the extent the underpayments are attributable to the nondeductibility of the premiums paid to Feedback.

B. Loans

The rest of the accuracy-related penalties imposed on the Avrahamis are related to the Commissioner’s reclassification of loans as taxable distributions from Feedback. We have held that both the $1.2 million loan from Mr. Avrahami to Belly Button and the $1.5 million loan from Feedback to Belly Button were bona fide . Therefore, the section 6662(a) penalties should not be imposed to the extent they are related to repayment of the $1.2 million loan. But there is no reasonable explanation for why any underpayment flowing from the failure to report $300,000 in interest and dividends for the overrepayment of that loan with the $1.5 million from Feedback should not be penalized. The Avrahamis don’t even make such an argument, so the Commissioner wins on this one.

C. Ordinary Dividend

We also have the penalty related to the $200,000 distribution directly from Feedback to Mrs. Avrahami, which we held should have been reported as an ordinary dividend. The Avrahamis argue that this was an inadvertent omission caused by their accountant’s error. The Avrahamis, however, are sophisticated and successful business owners, and we find they acted at least in careless disregard of the Code in their treatment of this item. See sec. 6662(b)(2) and (c) . They controlled Feedback’s bank account, and they caused the $200,000 to be transferred directly from Feedback to Mrs. Avrahami. They knew or should have known that the transfer should not have been reported—as they originally claimed—as an additional loan from Feedback to Belly Button. See sec. 1.6664-4(b) , Income Tax Regs. We are therefore not persuaded that they acted with reasonable cause and in good faith in failing to report their receipt of the $200,000. We will sustain the Commissioner’s imposition of a section 6662(a) penalty to the extent it relates to the $200,000 distribution.

Decisions will be entered under Rule 155 .

Matthew J. Howard as attorney for the Self-Insurance Institute of America, Inc., filed a brief as amicus curiae .

If a business meets the requirements of section 1361 , it may elect to be treated as an “S corporation” and pay no corporate tax. Secs. 1362(a) , 1363(a). An S corporation’s income and losses, like a partnership’s, flow through to its shareholders, who then pay income tax. See sec. 1363(b) . American Findings was originally established as a C corporation, but elected to become an S corporation-made an “S election”—effective for 2008. It is now wholly owned by the Avrahamis. (All section references are to the Internal Revenue Code in effect for the years at issue, and all Rule references are to the Tax Court Rules of Practice and Procedure, unless otherwise indicated.)

BYS was incorporated in 1992 and made an S election effective that same year. Mr. Avrahami owns 85% of BYS, and the remaining 15% is held equally (5% each) by trusts for his three children.

Chandler One was formed as a limited liability company in 2003. It is treated as a partnership for tax purposes and the Avrahamis are both general partners, each with a 50% interest.

O&E was incorporated in 1991 and made an S election effective that same year. Mr. Avrahami is the sole shareholder of O&E.

White Mountain was formed as a limited liability company in 1997 and is treated as a partnership for tax purposes. The Avrahamis each own 39.5% of White Mountain, and each of their three children own 7%.

White Knight was incorporated in 1993 and made an S election effective that same year. Mr. Avrahami is the sole shareholder of White Knight.

We will use “Clark” to refer both to Ms. Clark and her firm.

The policy periods for all of Feedback’s direct policies during the years at issue started December 15 of the stated year and ended one year later. (Except for one—the 2010 Tax Indemnity Policy—that actually says December 15, 2010 , to December 15, 2010 , but we assume that’s a typo.)

“The Actuarial Standards Board (ASB) is vested by the professional actuarial societies with the responsibility for promulgating Actuarial Standards of Practice (ASOPs) for actuaries providing professional services in the United States. Actuaries are required to follow the ASOPs by their actuarial societies.” Acuity, A Mut. Ins. Co., & Subs. v. Commissioner , T.C. Memo. 2013-209, at *13 .

Rosenbach was something of a captive underwriter. He testified that in 2009 and 2010 he prepared premium estimates for more than 50 but fewer than 80 captives. Most, if not all, were Clark’s clients.

Rosenbach was also given “a summary of the terms and conditions of the standard policies” to aid in his pricing calculations, but he never reviewed the actual policies.

The simplified version of Rosenbach’s model that was admitted into evidence in these cases did not identify any of the commercial filings that he relied on in creating his premium determinations. Rosenbach testified that supplemental documentation would have to be provided for another actuary to review the model.

The January 2005 Chubb filing—from which Rosenbach’s pricing model got its base rate calculation and adjustment factors—actually classified “property managers” into hazard group 2, not 4.

The $11,874 is calculated as $10,400 + (($470,000 – $250,000) / 1,000 x 6.70).

In insurance, a deductible is “the portion of the loss to be borne by the insured before the insurer becomes liable for payment.” Black’s Law Dictionary 444 (8th ed. 2004).

The $30,000 is the rounded product of the base premium and the five factors. In other words, $11,874 x 1.3 x 2.3 x 1.2 x 1.1 x 0.65 = $30,462, which rounded to the nearest thousand is $30,000.

The $31,953 is calculated as $11,874 x 1.3 x 2.3 x 1.92 x 1.25 x 0.375. We note that for Rosenbach’s calculation to work out, the coverage factor must be 0.375 (0.75 x 0.5) and references to 0.38 are likely the rounded version of this number.

Rosenbach’s construction defect rate appears to be around 5.625% ($470,000 x 5.625% = $26,438) of Chandler One’s gross revenue, yet he testified that construction defect rates typically run “one to four percent of asset value.”

The $61,000 is the rounded sum of the three separate calculations. In other words, $31,953 + $2,500 + $26,438 = $60,891, which rounded to the nearest thousand is $61,000.

The reason for these changes is not clear from the record. When asked whether his 2010 factors were determined in the same fashion as in prior years, Rosenbach responded: “Yes, absolutely.”

The $271,000 is the rounded product of the gross revenue and the five adjustments. In other words, $11,000,000 x 0.075 x 1.7 x 1.3 x 0.9 x 0.165 = $270,753, which rounded to the nearest thousand is $271,000.

This Chubb filing was not offered into evidence by either party and is not a part of the record.

The exact calculation of the $71,000 is not clear from the record. Rosenbach’s model shows a base premium of $2,675 and then eight factors—2.0, 1.366, 1.3, 1.35, 2.0, 2.609, 1.265, and 0.9. The product of these nine numbers is $76,194. However, in their reply brief the Avrahamis explained—albeit with a few typos—that the correct math is (2.0 x 1.3 x 2.0 x 2.609 x 1.265 x 0.9) x (1.366 + 1.35 – 1.0) = (15.446 x 1.716) = 26.506. And 26.506 x $2,675 = $70,904, which rounded to the nearest thousand is $71,000.

The expense ratio is the complement—a math term for the difference between an amount and 100%—of the 90% expected loss ratio.

The exact calculation of the $65,000 is not clear from the record. The product of all of the factors—but using one minus the expense ratio per Rosenbach’s testimony—is $52,313 ($276,000 x 90% x 30% x (100% – 10%) x 60% x 1.0 x 1.3). However, in their reply brief the Avrahamis explained that what Rosenbach meant was that he took the product of all of the factors except the expense ratio and then divided by one minus the expense ratio. Essentially ($276,000 x 90% x 30% x 60% x 1.0 x 1.3) / (100% – 10%) = $64,584, which rounded to the nearest thousand is $65,000.

We note that Rosenbach said “than they did in the past” even though 2009 was the first year that American Findings purchased a Litigation Expense policy and he was the actuary that made the model.

The $474,000 is the rounded product of the gross income and the four adjustments. In other words, $11,000,000 x 0.05 x 1.15 x 1.5 x 0.5 = $474,375, which rounded to the nearest thousand is $474,000.

The $86,000 is the rounded product of the preliminary premium and the percentage of payroll expense attributed to the Avrahamis. In other words, $474,000 x 18.1% = $85,794, which rounded to the nearest thousand is $86,000.

The $75,000 is the product of the limit and the three adjustments. In other words, $2,000,000 x 0.075 x 1.0 x 0.5 = $75,000.

Specifically, BYS paid $30,000; Chandler One paid $40,000; O&E paid $30,000; White Mountain paid $20,000; White Knight paid $40,000; and American Findings paid $200,000.

“Reinsurance is an agreement between an initial insurer (the ceding company) and a second insurer (the reinsurer), under which the ceding company passes to the reinsurer some or all of the risks that the ceding company assumes through the direct underwriting of insurance policies. Generally, the ceding company and the reinsurer share profits from the reinsured policies, and the reinsurer agrees to reimburse the ceding company for some of the claims that the ceding company pays on those policies.” Trans City Life Ins. Co. v. Commissioner , 106 T.C. 274, 278 (1996).

TRIA was amended by the Terrorism Risk Insurance Extension Act of 2005, Pub. L. No. 109-144, sec. 6 , 119 Stat. at 2662, which added the $100 million loss requirement starting January 1, 2007. TRIA was also extended by the Terrorism Risk Insurance Program Reauthorization Act of 2007 (TRIPRA), Pub. L. No. 110-160, sec. 3(c)(3) , 121 Stat. at 1839, which extended the $100 million loss requirement through 2014. We will refer to the 2002 act including its amendments and extensions as TRIA.

The Avrahamis were apparently confused about the terms of this policy. Mr. Avrahami was under the impression that Feedback was at risk only for the amount of premiums put into Pan American—$360,000 each year—and testified that it would “be weird” to lose money and if Feedback did he would “freak out”.

The rate on line eventually set for all TRIP policies was 6.5% in 2009 and 7% in 2010.

Section E.15 of the Terrorism Risk Insurance Insuring Agreement states: “Except as otherwise stated in this paragraph, this policy does not apply to loss recoverable or recovered under other insurance or indemnity. If the limit of the other insurance or indemnity is insufficient to cover the entire amount of the loss, this policy will apply to that part of the loss, other than that falling within any Deductible Amount, not recoverable or recovered under the other insurance or indemnity.”

Rosenbach and the Commissioner’s expert witness both credibly testified that they had never seen such a provision in an insurance policy before.

Although outside the years at issue in these cases, we note that the first claim against Feedback was filed in March 2013.

Why the note payable was for only $1.2 million when the amount lent was $1,201,000 is unclear from the record.

On its 2009 (and only on its 2009) tax return Belly Button shows this $830,000 as “loans to shareholders.” We surmise this was just a scrivener’s error (or perhaps a Freudian slip). This does not affect the outcome of these cases.

Although outside the years at issue in these cases, we note that in July 2012 the Avrahamis transferred nearly $207,000 from their bank account back into Feedback’s. Feedback also transferred $1,739,000 to BYS in 2012. This transfer was evidenced by a $1,775,000 promissory note with a 4% interest rate and a due date of April 2022. The note was signed by Mr. Avrahami on behalf of BYS and stated that it was “secured by the Deed of Trust.” However, such a Deed of Trust—if it exists—is not part of the record in these cases.

And they should have. St. Kitts’s Captive Insurance Companies Act of 2006, c. 21.20, sec. 15.6 , provides that “[a] captive insurance company may not make a loan to or an investment in its parent company or affiliated persons without prior written approval of the Registrar, and any such loan or investment shall be evidenced by documentation approved by the Registrar.”

At the time it filed its petition, Feedback had no “principal place of business or principal office or agency in any judicial circuit.” Therefore, absent an agreement between the parties, Feedback’s case is appealable to the Tenth Circuit because it filed its tax returns with the IRS Ogden, Utah, office. Sec. 7482(b)(1)(B) .

This increase in income was caused by the disallowance of the Avrahami entities’ insurance expense deductions for the amounts they paid to Feedback and Pan American.

Absent an agreement by the parties, the Avrahamis’ case is appealable to the Ninth Circuit. See sec. 7482(b)(1)(A) .

Mutual-insurance companies—as opposed to stock-insurance companies-generally have the following characteristics: “(1) Common equitable ownership of assets by members; (2) the right of policyholders to be members to the exclusion of others and to choose management; (3) a sole business purpose of supplying insurance at cost; and (4) the right of members to the return of premiums which are in excess of the amount needed to cover losses and expenses.” Okla. State Union of the Farmers Educ. & Coop. Union of Am. v. Commissioner , 68 T.C. 651, 664 (1977); see also Rev. Rul. 74-196 , 1974-1 C.B. 140.

The 2015 amendments to section 831(b) increased the premium ceiling to $2.2 million—adjusted for inflation—and added new diversification requirements that an insurance company must meet in order to receive the favorable tax treatment of subsection (b). See Consolidated Appropriations Act, 2016, Pub. L. No. 114-113, sec. 333 , 129 Stat. at 3106.

The Avrahamis objected to Babbel’s testimony on risk distribution, asserting that he failed to form an opinion on risk distribution in his expert report and citing Rule 143(g). We disagree—Babbel’s expert report did touch on risk distribution in its discussion of the law of large numbers and the need for a sufficiently large number of risk exposures.

We note that Rev. Rul. 2002-90 , 2002-2 C.B. 985, discusses a situation in which the IRS would consider amounts paid to a captive by its 12 sibling entities to be deductible insurance premiums. Where Spragg and Small derive their “seven policyholder” analysis is unclear from the record.

Even the Avrahamis’ expert witnesses stated: “Since there were not enough brother-sister companies to support that risk distribution structure Feedback needed to find an alternative solution.”

In finding we did not err in Harper Grp. v. Commissioner , 96 T.C. 45 (1991), aff’d , 979 F.2d 1341 (9th Cir. 1992), the Ninth Circuit summarized existing precedent: “Prior cases which have found true insurance have also included higher percentages of unrelated business than those found here. See Sears Roebuck & Co. v. Commissioner , 972 F.2d 858, 860 (7th Cir. 1992) (99.75 percent from others); Ocean Drilling & Exploration Co. v. United States , 24 Cl. Ct. 714, 730 (1991) (44 percent to 66 percent from others). Cases which have found no true insurance have found much lower percentages of unrelated business. See, e.g., Beech Aircraft Corp. v. United States , 797 F.2d 920, 921-22 (10th Cir. 1986) (.5 percent from others); Gulf Oil Corp. v. Commissioner , 89 T.C. 1010, 1028 (1987) (2 percent from others), rev’d in part on other grounds , 914 F.2d 396 (3d Cir. 1990); Clougherty Packing Co. v. Commissioner , 811 F.2d 1297, 1299 (9th Cir. 1987) (none from others).” Harper Grp ., 979 F.2d at 1342.

Marsh is one of the world’s largest insurance brokers and has extensive captive insurance experience.

American Findings’ 2010 premium payment to Pan American of $360,000 divided by 8 (number of millions of TIV) is $45,000.

Pan American ceded more than $20 million of premiums in 2009 and almost $23 million in 2010. Therefore, the fixed fees Pan American received were equivalent to approximately 0.6% of premiums in 2009 and 1.8% in 2010.

The $4,991 is calculated as $4,375 + (($470,000 – $250,000) / 1,000 x 2.80).

We do not have jurisdiction over Feedback’s elections for years other than 2009 and 2010 since they were not part of the notice of deficiency and are not before us. See Rule 13.

We note that Belly Button’s 2010 tax return reports only $2,031,000 of liabilities—$1.5 million in notes payable and $531,000 “due to related party.” Yet by the end of 2010 it had outstanding promissory notes payable to Feedback of $2,330,000—$830,000 from 2007 and another $1.5 million from 2010.

We note that this “repayment” was made less than three years after the execution of the $1.2 million promissory note, so even with simple interest accrued at 4% per year on the unpaid principal the repayment shouldn’t have exceeded $1,344,000.

The rest of the $1.5 million transfer appears to be reflected on Belly Button’s 2011 tax return where it reported a further $531,000 reduction in the amount due to related parties and a new $299,000 asset under “loans to partners (or persons related to partners)” but still no indication of any interest payment. ($670,000 + $531,000 + $299,000 = $1.5 million.)

Belly Button seems to have recorded the rest in 2011, when it reported an additional increase to its notes payable of $830,000. We are not sure why it did this—Belly Button is a cash-basis taxpayer, and the record contains the bank statement that plainly states that Belly Button received $1.5 million in March 2010.

No one questioned whether Feedback had sufficient earnings and profits for this distribution to be a dividend. See sec. 316 .