FERNANDEZ, Circuit Judge:

AMERCO and a number of its subsidiaries (AMERCO Group) purchased insurance policies from Republic Western Insurance Company (Republic) and deducted the pre-: miums for income tax purposes. Republic was a subsidiary of AMERCO. The Commissioner of Internal Revenue (Commissioner) determined that because of the relationships among the parties the transactions did not constitute insurance. A notice of deficiency was issued by the Commissioner, and AMERCO petitioned the Tax Court for a redetermination. The Tax Court found that the transactions were insurance. 1 It, therefore, held against the Commissioner, who now appeals. We affirm.

BACKGROUND

The AMERCO Group constitutes the U-Haul system. AMERCO itself is a holding company which owns the stock of a number of subsidiaries. Among those subsidiaries are U-Haul International, Inc., the administrative clearing house, AMERCO Lease Co., which owns much of the U-Haul rental equipment, and numerous other rental companies, repair shops, manufacturing companies and service companies. Approximately 250 of these joined in AMERCO’s consolidated returns for the tax years in question — 1979-85.

Republic was incorporated in 1973. It is a third tier wholly owned subsidiary of AMERCO. It is a property and casualty insurance company, licensed in most states and the District of Columbia. Republic issued insurance policies to members of the AMERCO Group and to unrelated parties. Those policies were issued at normal com *164 mercial rates and were divided into a number of categories by the Tax Court. They included: (1) corporate policies issued to members of the AMERCO Group; (2) workers’ compensation policies issued to members of the AMERCO Group; (3) U-Haul rental system policies, which covered members of the AMERCO Group, independent fleet owners, and truck rental customers; (4) SafeMove and SafeStor policies, which covered U-Haul rental customers; and (5) policies which covered risks entirely unconnected with the U-Haul system.

The Tax Court found that based upon gross premiums insurance written for the AMERCO Group itself, related business, was from 26 percent to 48 percent of Republic's total insurance business. Insurance written for others, unrelated business, constituted the remaining 74 percent to 52 percent.

The Commissioner took the position that the transactions between the AMERCO Group and Republic could not be insurance because Republic is a wholly owned subsidiary of AMERCO and is, therefore, a member of the same economic family as the AMERCO Group. As a result, the Commissioner contended, there could be no risk-shifting or risk-distributing and, absent those, insurance could not exist. The Commissioner takes the same position before us.

JURISDICTION AND STANDARD OF REVIEW

The Tax Court had jurisdiction pursuant to 26 U.S.C. § 6213. We have jurisdiction pursuant to 26 U.S.C. § 7482.

Whether the transactions constitute insurance is a question of law subject to de novo review. Clougherty Packing Co. v. Commissioner, 811 F.2d 1297, 1299 (9th Cir.1987) (based upon stipulated facts it is a question of law whether payments to a captive insurer constitute deductible insurance premiums). Whether certain insurance policies issued by Republic are related or unrelated business is a question of fact reviewed for clear error. See Pomarantz v. Commissioner, 867 F.2d 495, 497 (9th Cir.1988). Although a presumption exists that the Tax Court correctly applied the law, no special deference is given to Tax Court decisions. Clougherty, 811 F.2d at 1299.

DISCUSSION

It is common ground that insurance premiums constitute ordinary and necessary business expenses which can be deducted in arriving at taxable income. See 26 U.S.C. § 162; Treas.Reg. § 1.162-l(a) (as amended in 1988). On the other hand, amounts placed by a company into a self insurance reserve fund cannot be deducted; any deductions must await an actual payment out of that reserve. See Clougherty, 811 F.2d at 1300.

The question before us is an intermediate one: Can insurance premiums paid to a wholly owned subsidiary be deducted or are they more like amounts paid into a self insurance reserve? In order to answer that question it is necessary to determine the nature of insurance, and, more particularly, its nature for income tax purposes. Then, we must see if transactions between members of a corporate family and a subsidiary insurance company can meet that definition.

A. Definition of Insurance.

In setting forth definitions of insurance, we do not start afresh. Rather, we look to a line of cases starting with Helvering v. Le Gierse, 312 U.S. 531, 61 S.Ct. 646, 85 L.Ed. 996 (1941). In Le Gierse, an estate tax case, the Supreme Court described insurance as follows:

We think the fair import of subsection (g) [the estate tax section] is that the amounts must be received as the result of a transaction which involved an actual “insurance risk” at the time the transaction was executed. Historically and commonly insurance involves risk-shifting and risk-distributing. That life insurance is desirable from an economic and social standpoint as a device to shift and distribute risk of loss from premature death is unquestionable. That these elements of risk-shifting and risk-distributing are essential to a life insurance contract is agreed by courts and commentators.

312 U.S. at 539, 61 S.Ct. at 649.

*165 We have often referred to this definition, with particular emphasis on the risk-shifting and risk-distributing aspect.. See Clougherty, 811 F.2d at 1301; Carnation Co. v. Commissioner, 640 F.2d 1010, 1012 (9th Cir.), cert. denied, 454 U.S. 965, 102 S.Ct. 506, 70 L.Ed.2d 381 (1981). In this case, the Tax Court identified three principles at the heart of the Le Gierse definition: “(1) [t]hat an insurance transaction must involve ‘insurance risk;’ (2) that insurance involves risk-shifting and risk-distributing; and (3) that, in the absence of a statutory definition, ‘insurance’ is to be defined in its commonly accepted sense.” 96 T.C. at 38. It supplemented these with the reflection that “matters of Federal income taxation must be resolved with principles of Federal income taxation borne in mind.” Id. We agree with this formulation, which supplements what we have said before. While the Tax Court avoided calling this a definition, we do not think that the three articulated principles places it at odds with our prior cases. Rather, they underscore the fact that many considerations can come into play when one attempts to decide whether a deduction of a purported insurance premium will be allowed.

Here, as in other cases, the focus is on risk-shifting and risk-distributing, but that does not mean that additional factors could not be important at times. Here, as the Tax Court determined, there was an insurance risk involved—the AMERCO Group undoubtedly faced potential hazards from its operations which constituted insurable risks. By the same token, there could be no real doubt that Republic was engaged in the insurance business' in the commonly accepted sense. The Commissioner does not contest those determinations. Thus, we must turn to an analysis of risk-shifting and risk-distributing.

B. Risk-Shifting and Risk-Distributing.

Courts have not spilled a great deal of ink in defining risk-shifting as opposed to risk-distributing. That is probably because in most instances the facts which demonstrated that one did not exist also demonstrated that the other did not. See Carnation, 640 F.2d at 1013. Nevertheless, it is fair to say that “ ‘[r]isk-shift-ing’ means one party shifts his risk of loss to another, and ‘risk-distributing’ means that the party assuming the risk distributes his potential liability, in part, among others. An arrangement without the elements of risk-shifting and risk-distributing lacks the fundamentals inherent in a true contract of insurance.” Beech Aircraft Corp. v. United States, 797 F.2d 920, 922 (10th Cir.1986); see also Robert E. Keeton and Alan I. Widiss, Insurance Law § 1.3(b)(2) (1988).

(1) Risk-Shifting.

In the germinal case of Le Gierse, the Court described a classic situation where no risk-shifting could take place. There the. 80-year-old insured purchased both an insurance policy on her life and an annuity. The premiums were fixed in a way that assured she would retain all of the risk of her own untimely death. The insurance company assumed none of the risk attendant upon that event, and the only result of the combined transactions would have been to remove assets from her taxable estate. Not surprisingly, the Court found that no insurance existed, even though the company itself was, no doubt, an insurance company which had written an insurance policy. In effect, there was no risk-shifting, and, for that matter, no risk-distributing.

We confront the same sort of problem when a parent purchases insurance from its wholly owned subsidiary, and that subsidiary is a captive insurance company. By “captive insurance company” we mean one which is organized for the purpose of insuring the liabilities of the parent and its affiliates. See Clougherty, 811 F.2d at 1298 n. 1.

Since at least 1977 the Commissioner has taken the position that where the only insurance written by a captive insurance company is for the parent and its affiliates, there is no risk-shifting and hence no insurance. See Rev.Rul. 77-316, 1977-2 C.B. 53. The reason is, the Commissioner says, that they represent one economic family unit so that the risk of loss is still borne by that *166 unit no matter what happens. That approach is not entirely without merit. Nevertheless, it presents a difficulty with the general tax law principle of treating subsidiary corporations as if they were truly separate entities. See Moline Properties, Inc. v. Commissioner, 319 U.S. 436, 439, 63 S.Ct. 1132, 1134, 87 L.Ed. 1499 (1943). We have discussed that difficulty previously. Clougherty, 811 F.2d at 1302-05. While in Clougherty we pointed out that looking to the realities of a transaction between corporations and their subsidiaries does not necessarily run afoul of Moline, we also found it unnecessary to accept the economic family concept for the purpose of deciding the case then before us. Id. at 1305. The reason we did not need to accept the economic family theory is also the reason that we and other courts have hitherto refused to allow the deduction of premiums paid to captive insurance companies. In those cases, not unlike Le Gierse, in economic reality the transaction was not insurance.

In Clougherty itself, for example, the parent set up and capitalized the captive insurance company. The parent then purchased insurance from an unrelated company which immediately reinsured most of the loss with the captive. We agreed that deduction of the portion of the premium ceded to the captive should be disallowed. 811 F.2d at 1307.

In our earlier decision in Carnation we were faced with the same kind of situation. There, too, the parent, Carnation Corporation,- created a captive insurance company. There, too, the parent insured with an unrelated entity which, in turn, reinsured most of the risk with the captive. There, too, we disallowed the premium payment. 640 F.2d at 1013.

In both Clougherty and Carnation we determined that there had been no shifting of the risk from the parent to the captive. As a matter of economic reality every dollar paid out by the captive was a dollar out of the parent’s pocket from whence it came in the first place. In fact, the arrangement was little different from a reserve fund held by the parent itself; only corporate formalities distinguished it. Those were not enough. We could as easily have said that there was no risk-distributing either because the only risks involved were those of the parent. In Carnation we did say just that. 640 F.2d at 1013.

Cases from other circuits have reached the same conclusion when faced with this simple captive problem and for much the same reasons. For example, in Stearns-Roger Corp. v. United States, 774 F.2d 414 (10th Cir.1985), the parent created a captive which was wholly owned by it and one of its subsidiaries. The parent capitalized the captive and the captive only wrote insurance for the parent. The court found no risk-shifting. In reality, the parent was self-insured. Id. at 416. Similarly, in Beech Aircraft the captive insurance company wrote insurance almost exclusively for the parent — .5 percent was written for others. 797 F.2d at 921-22. The court found that for practical purposes Beech paid its own claims. Id. at 922; see also Mobil Oil Corp. v. United States, 8 Cl.Ct. 555, 561, 567-68 (1985) (parent’s premium deduction disallowed although some portion of captive’s premium came from others); cf. Humana Inc. v. Commissioner, 881 F.2d 247, 257 (6th Cir.1989) (parent’s insurance premium could not be deducted but those of brother-sister corporations could be).

Thus, although many cases, including ours, have disallowed premium deductions in captive insurance company situations, none have found it necessary to rely upon the Commissioner’s economic family theory. When courts have been required to consider it, they have noted its difficulties and have not followed it. See Sears Roebuck & Co. v. Commissioner, 972 F.2d 858, 861 (7th Cir.1992); Ocean Drilling & Exploration Co. v. United States, 24 Cl.Ct. 714, 729 (1991); AMERCO, 96 T.C. at 41 and cases cited. We see no reason to follow it now.

That, however, leaves the question of what distinguishes this case from those which have come before. Of course, the major distinction is that Republic does a substantial unrelated insurance business. The Tax Court found that made all the difference in the world. The Commissioner *167 contends that it really makes none at all. We agree with the Tax Court.

In Rev.Rul. 88-72, 1988-2 C.B. 31, 32, the Commissioner essentially took the position that the mere fact that outside insurance business was done by a captive could have no effect on risk-shifting from the parent or its subsidiaries to the captive. That was based, in part, on the Commissioner’s economic family theory. It was also based on the Commissioner’s view of the nature of insurance.

The Tax Court has determined that when a parent purchases a policy from a captive which has a larger pool of risks from others, the parent has, indeed, shifted its risk to the captive insurer. Of course, the Commissioner is not wholly wrong, for it is undoubtedly true that a decrease in the value of the captive can result in a decrease' in the value of what the parent holds — the captive’s stock. But that is too narrow a focus because there is much more in the insurer’s pool than the parent’s premium- dollars. To put it another way, as the Tax Court noted in a related case, “risk transfer and risk distribution' are two sides of the same coin which as an integrated whole constitute ‘insurance.’ ” Harper Group v. Commissioner, 96 T.C. 45, 59 (1991). That is to say, when the pool consists of a substantial amount of dollars, and risk, from those outside the parent and its affiliates, there is a true shift of the risk, even though the parent could suffer, somewhat if the captive made a payment on account of an insured’s loss. In a real sense, the parent’s risk has been placed with the captive and thence spread among all of those in the pool.

The Commissioner claims that this improperly collapses the ideas of risk-shifting and risk-distributing. What the Commissioner does not recognize, however, is the fact that the pool itself would not exist were it not for those who have purchased policies from the insurer. The existence of that pool enables every insured to have its risk spread among all of the participants. The parent’s situation is no different.

Indeed, it would be more accurate to say that the Commissioner appears to conflate actual insurance risk and speculative risk. The insurance risk is the possibility that a particular event for which an insured will be held liable will occur. Of course, from the standpoint of the insured there can be no profit from that risk. The only possible outcomes are loss or no loss. It is that risk which must be transferred to the insurer if true insurance is to be involved.

Speculative risk, on the other hand, is merely investment risk, and it can produce profit or loss. An insurance company, for example, may earn or lose money based upon the outcome of its investment, underwriting, adjusting, and management activities. When it does, the value of its stock will increase or decrease, and if the parent owns that stock it, too, will suffer gain or loss.

The Commissioner argues, in effect, that because the parent retains speculative risk it cannot have transferred the insurance risk. That argument has weight in the case of a captive insurer which writes insurance entirely, or almost entirely, for the parent. In that event, the insurance risk is not transferred because the parent is in the same economic position whether it holds a policy or not. In addition, because only the parent’s money is in the pool, the parent has no speculative risk other than that involving its own funds, wherever they are.

On the other hand, when there is unrelated insurance business, there is a pool to which the parent’s own risk can be transferred, and the mere fact that a payment from that larger pool may affect the value of the insurer is not dispositive. Again, there would be no pool at all if insurance were not being sold to others.

Nor can we accept the Commissioner’s position that there is no insurance unless all effects of a possible risk of loss have been removed from the insured. That overbroad position flies in the face of the reality of the insurance market. It takes no real account of mutual insurance companies, where policyholders suffer losses each time an amount is paid out of the pool, whether that amount is for their own insurance risk or someone else’s. It also fails to take account of the well-known phenome *168 non of retrospectively rated policies, where the insured will often ultimately bear a large part of the insurance risk. See Sears Roebuck, 972 F.2d at 862. 2 Perhaps it would not be amiss to note that, at bottom, one can never eliminate all of one’s risk— there is always a possibility that the insurance company will become insolvent.

In fine, we are satisfied that a parent and its subsidiaries can shift risk to a captive insurer, where that insurer has significant unrelated business. In this case, the Tax Court determined that if only AMER-CO were looked at, its share of Republic’s business was under one percent and if the whole AMERCO Group were considered, the share was from 26 percent to 48 percent of that business. That, said the court, left sufficient unrelated business to create a true pool and to allow for true risk-shifting. In that determination, the Tax Court did not err.

(2) Risk-Distributing.

As we have already pointed out, risk-distribution looks at the transaction from the standpoint of the insurer. Here, where a substantial part of the insurer’s business comes from sources unrelated to the parent and its subsidiaries, it was proper for the Tax Court to decide that there was sufficient risk distribution. The distribution aspect is rather apparent. As the Tax Court found, Republic’s “insurance business was diverse, multifaceted, and ... involved a substantial amount of outside risks. More of the money in its pool came from outside unrelated insureds than came from AMERCO & Subsidiaries.” 96 T.C. at 41. That determination was not clearly erroneous. 3

CONCLUSION

As precedent requires, we have discussed this case from the standpoint of risk-shifting and risk-distributing. However, we also agree with the Seventh Circuit that discussions of this area might seem less abstruse if we asked ourselves a somewhat different question: “Suppose we ask not ‘What is insurance?' but ‘Is there adequate reason to recharacterize this transaction?,’ given the norm that tax law respects both the form of the transaction and the form of the corporate structure.... For whether a transaction possesses substance independent of tax consequences is an issue of fact_” Sears Roebuck, 972 F.2d at 864. The Seventh Circuit held. that, because nothing of substance would differ from the purchase of insurance from an unrelated insurance company, a trier of fact could conclude that the captive — Allstate—had furnished the parent — Sears—with insurance. Here, for the same reasons, we are satisfied that a trier of fact could, as the Tax Court did, conclude that the captive— Republic — furnished the parent — AMER-CO — and the AMERCO Group with insurance. Here, as there, the captive performed all of the functions of an insurer. The decision of the Tax Court was not clearly erroneous when viewed from this vantage point

At any rate, we conclude that it is possible to have a true insurance transaction between a, corporation and its wholly owned subsidiary insurance company if that captive company does substantial unrelated business. Likewise, it is possible for other-members of the corporate group to have true insurance transactions with the captive company in that instance. 4 When that is so, the insurance premiums paid by the parent and its subsidiaries are deductible.

AFFIRMED.

Notes

. AMERCO & Subsidiaries v. Commissioner, 96 T.C. 18 (1991).

. At argument the Commissioner stated that retrospectively rated policies may not be insurance. We need not decide that issue here. The argument, however, does seem to underscore the Commissioner’s very limited view of what does constitute insurance. That view bespeaks an exceedingly narrow concept of when a corporation is entitled to deduct expenses which are typically considered ordinary and necessary in the conduct of its business.

. The Commissioner attacks a number of the Tax Court's factual determinations regarding the allocation of Republic’s premium income between related and unrelated insureds. We • have reviewed the Commissioner's contentions and find no clear error in the Tax Court’s determinations.

.We need not, and do not, decide whether brother-sister corporations can deduct premiums, even in the absence of business unrelated to the corporate family. See Humana, 881 F.2d at 252-57.