The government appeals, pursuant to 28 U.S.C. § 1291 , a decision of the district court granting an estate tax refund of $8,609,393 to the estate of Howard F. Ah-manson (Ahmanson). The government raises issues involving (1) the valuation of assets in the gross estate, (2) the propriety of a charitable deduction, (3) the valuation of property giving rise to a charitable deduction if a charitable deduction is held to be allowable, (4) the availability of a marital deduction or community property exclusion for assets going to Ahmanson’s widow (Mrs. Ahmanson) as the result of a settlement with the estate, and (5) the district court’s rejection of the estate’s request for a jury trial. We agree with the government with respect to certain issues involving the valuation of assets in the gross estate and the valuation of property giving rise to the charitable deduction, as well as certain issues involving the treatment of the settlement. We affirm in part, reverse in part, and remand for further proceedings.

Facts of the Case

Ahmanson died on June 17, 1968. Ah-manson Bank & Trust Company was appointed executor. Ahmanson was the founder of Home Savings & Loan Association. At his death, Ahmanson owned 15 percent of the Home Savings & Loan Association stock. Eighty-one percent of the stock was owned by a holding company, H. F. Ahmanson & Co. (HFA). HFA also owned substantial interests in several additional companies. On the date of Ahman-son’s death, HFA had outstanding 1,000 shares of voting common stock of $10 par value and 106,711 shares of nonvoting common stock of $1 par value. The dividend and liquidation entitlements of these shares were in proportion to their par values. Ah-manson held at his death an income interest in 11,000 shares of nonvoting HFA common stock. He held a controlling interest in the voting HFA common stock (600 shares) in a revocable inter vivos trust, Ahmanson Bank & Trust Company Trust No. 28 (Trust No. 28).

Also in Trust No. 28 were all the shares of Ahmanco Inc. (Ahmanco), a corporate *766 shell with no assets prior to Ahmanson’s death. There were 99 nonvoting shares and 1 voting share of Ahmanco common stock outstanding. The voting share had the same dividend and liquidation entitlements as a nonvoting share. At the moment of Ahmanson’s death pursuant to Declarations of Trust, Ahmanco became unconditionally entitled to the 600 shares of voting HFA common stock.

Under the same Declarations of Trust, the Ahmanson Foundation (Foundation), 1 an exempt charitable organization pursuant to 26 U.S.C. § 501 (c)(3), became entitled to the 99 nonvoting shares of Ahmanco. The 1 voting share of Ahmanco was to remain in Trust No. 28. The right to vote that share was to pass to Ahmanson’s son and his male issue.

Also material to this appeal is Ahmanson Bank & Trust Company Trust No. 1 (Trust No. 1), a revocable inter vivos trust created by Ahmanson that became irrevocable at his death. Trust No. 1 provided, among other things, for a marital trust and for the payment of certain expenses of the estate. The residue of the corpus was to pass to the Foundation.

In its estate tax return the estate attributed no value directly to the 600 HFA shares held in trust. Instead the estate reported the date of death value of the 1 voting share of Ahmanco at $148,676 and of the 99 nonvoting shares at $13,667,549. After audit, the Commissioner determined that the value of the 600 voting shares of HFA was $22,699,611.49. This was divided between the voting share of Ahmanco ($2,880,580.69) and the 99 nonvoting shares ($19,819,030.80). The Commissioner’s increased valuations were due, in part, to the addition of a factor for the control position of the 600 HFA shares and the 1 voting share of Ahmanco.

The government argued during trial that the 600 HFA shares should have been included in the gross estate at a value reflecting the control premium and that the Ah-manco shares had no value because the corporation was a mere shell at the moment of death. The estate argued, and the district judge agreed, that neither the value of the 600 HFA shares nor the value of the 99 Ahmanco shares given to the Foundation were relevant in computing the taxable estate, apparently on the assumption that the charitable deduction for the 99 shares of Ahmanco would equal the contribution of those shares to the gross estate. (Nevertheless, the district judge valued the 99 Ahmanco shares to reflect a 3 percent discount due to their nonvoting status.) Under the method used by the district judge of computing the taxable estate directly, only the 1 share of Ahmanco voting stock remaining in Trust No. 28 had to be valued, based on its pro-rata interest in the HFA shares. The district judge did not allow for a control premium in making this valuation.

Trust No. 1 provided for a marital trust for the benefit of Mrs. Ahmanson. The trust was to include property having a value equal to the value of Mrs. Ahmanson’s share of the community property, and to have a maximum value of $5,000,000. Mrs. Ahmanson also received various smaller bequests passing under the will, which are not in issue here, and $750,000 from the estate in settlement of Mrs. Ahmanson’s challenge to the testamentary dispositions. Mrs. Ah-manson challenged the validity of the will and the validity of Trust No. 1 and Trust No. 28. On the basis of her community property interests and on the basis of her rights pursuant to former section 41 of the California Probate Code, she attempted to take one half of the property otherwise going to charity. It is undisputed that the resulting settlement was the result of adversary negotiations, with both sides represented by counsel. The parties recited that the settlement was in compromise of Mrs. Ahmanson’s community property claims in the assets of the estate. The $750,000 was distributed to Mrs. Ahmanson from funds included in the gross estate in addition to *767 the $5,000,000 paid under the marital trust. The Ahmansons were married from January 14, 1965, until Ahmanson’s death on June 17, 1968.

A demand for a jury trial was made by the estate, but was rejected by the district judge as untimely. The government at no time objected to that ruling in the district court.

Gross Estate

A. Inclusion of Assets.

The chief dispute between the government and the Foundation is over the proper valuation of the contents of Trust No. 28 for the purposes of the gross estate. The government contends that the value of the 600 shares of the HFA stock is includable in the gross estate, along with the value of the 100 shares of Ahmanco stock. However, the government admits that the value of the Ahmanco stock in the gross estate is equal to zero, as Ahmanco had no. assets until it acquired the 600 shares of HFA stock upon the death of Ahmanson. The Foundation argues that it is the value of the Ahmanco stock, understood as holding the 600 shares of HFA, that is to be included in the gross estate, because upon Ah-manson’s death the asset that was transferred to the beneficiaries was not the HFA stock itself, but rather the Ahmanco stock. The district judge agreed with the Foundation, concluding that the transfers of the HFA stock to Ahmanco and of Ahmanco to the beneficiaries must be considered as two phases of a simultaneous and integrated transfer that took place at the moment of death. We conclude, for reasons that will more fully appear in part B, that the district judge correctly decided that the transfers should be seen as integrated. The analysis must proceed further, however, with close attention to the statutory language.

For clarification, we adopt the common shorthand expression, and will say that property is “in the gross estate” if the value of the gross estate includes the value of that property. In these terms both the Ahmanco and the HFA shares will be in the gross estate. By identical language sections 2036(a) and 2038(a)(1) of the Internal Revenue Code of 1954 as amended by the Revenue Act of 1962 provide that: “[t]he value of the gross estate shall include the value of all property to the extent of any interest therein of which the decedent has at any time made a transfer ... by trust or otherwise . . .. ” 26 U.S.C. §§ 2036 (a), 2038(a)(1). In section 2036(a) the condition on transfer which makes the property in-cludable in the gross estate involves, roughly summarized, a retained life estate, and in section 2038 the condition on transfer is revocability. Thus the transfers referred to in both sections are inter vivos transfers. In the present case the conditional nature of the transfers of HFA and Ahmanco stock into Trust No. 28 satisfies the requirements of sections 2036(a) and 2038(a)(1) and thus both the 600 HFA shares and 100 Ahmanco shares must be included in the gross estate.

B. Valuation in General.

Our concern, however, is only about the value of what is in the gross estate. It is undisputed that that value is to be determined at the moment of death. Cf. United States v. Land, 303 F.2d 170, 171-75 (5th Cir.), cert. denied, 371 U.S. 862 , 83 S.Ct. 121 , 9 L.Ed.2d 100 (1962). Therefore we are instructed to determine the value at the moment of death of the 600 shares of HFA stock and the 100 shares of Ahmanco stock. In so doing we must take into account any transformations of the property that are logically prior to its distribution to the beneficiaries. Cf. Provident Nat’l Bank v. United States, 581 F.2d 1081 (3d Cir. 1978). In particular, as the district judge recognized, we must take into account the fact that Ahmanco has acquired the 600 shares of HFA as an asset. In short, we must valúate the HFA and Ahmanco stock as of the moment of Ahmanson’s death, bearing in mind that the HFA shares, in their entirety, have become an asset of Ahmanco. In effect, this is to valuate the Ahmanco stock.

*768 In its brief, the Foundation argues forcefully that the Ahmanco shares should be partitioned into two blocks prior to valuation. That is, the 99 shares going to the Foundation should be valued separately from the 1 share that had a private destination. The Foundation argues that because the estate tax is imposed upon the transfer of property, see, e. g., United States Trust Co. of New York v. Helvering, 307 U.S. 57, 60 , 59 S.Ct. 692, 693-94 , 83 L.Ed. 1104 (1939), the valuation of property in the gross estate must take into account any changes in value brought about by the fact of the distribution itself. It is undisputed that the valuation must take into account changes brought about by the death of the testator. Ordinarily death itself does not alter the value of property owned by the decedent. However, in a few instances such as when a small business loses the services of a valuable partner, death does change the value of property. See United States v. Land, supra, 303 F.2d at 172 . The valuation should also take into account transformations brought about by those aspects of the estate plan which go into effect logically prior to the distribution of property in the gross estate to the beneficiaries. Thus, for example, if a public figure ordered his executor to shred and burn his papers, and then to turn the ashes over to a newspaper, the value to be counted would be the value of the ashes, rather than the papers. Similarly, if a will provides that prior to the distribution of the estate a close corporation owned by the testator is to be recapitalized, with one class of stock in the gross estate exchanged for another, the value of the gross estate would be based on the shares resulting from the recapitalization. Provident Nat’l Bank v. United States, supra, 581 F.2d at 1086-87 .

We must distinguish, however, the effect of “predistribution” transformations and changes in value brought about by the testator’s death, from changes in value resulting from the fact that under the decedent’s estate plan the assets in the gross estate ultimately come to rest in the hands of different beneficiaries. The estate tax is a tax upon a transfer as the Foundation contends. However, it is a tax on the privilege of passing on property, not a tax on the privilege of receiving property. “The tax is on the act of the testator not on the receipt of the property by the legatees.” Ithaca Trust Co. v. United States, 279 U.S. 151, 155 , 49 S.Ct. 291, 291-92 , 73 L.Ed. 647 (1929) (discussing Revenue Act of 1918, 40 Stat. 1057 ).

YMCA v. Davis, 264 U.S. 47, 50 , 44 S.Ct. 291, 292 , 68 L.Ed. 558 (1924) (discussing Revenue Act of 1918, 40 Stat. 1057 ). There is nothing in the statutes or in the case law that suggests that valuation of the gross estate should take into account that the assets will come to rest in several hands rather vthan one.

To take into account for valuation purposes the fact that the testator’s unitary holding has become divided in the hands of two or more beneficiaries, would invite abuse. For instance, a testator with two equally valuable pieces of real property could give equal undivided shares in each to both of two beneficiaries. Because undivided shares of real property frequently sell at a discount, the total value of the gross estate, under the theory offered by the Foundation, would be less than the value of the two parcels in the hands of the testator. The two beneficiaries could later exchange shares, each ending up the outright owner of a parcel. We may imagine that this would accomplish the original purpose of the testator, with a considerable tax savings. Estate planners would implement such a tax-avoidance scheme whenever at least one of the assets in the gross estate has a diminished value if divided among two or more beneficiaries. As there is nothing in either the language of the statutes or the underlying theory of the estate tax that requires the existence of this loophole, we shall not impute it to Congress.

*769 During oral argument the Foundation made a new argument, the possibility of which it appeared to have denied in its brief. Its new contention was that the value of the 99 nonvoting shares of Ahman-co is the same whether or not those shares are held by the same person holding the 1 voting share, and thus valued as one block rather than two. The Foundation asserted at oral argument that the disposition of the Ahmanco shares was irrelevant to their valuation. Although this argument is in apparent conflict with the main thrust of the Foundation’s argument in its brief, we believe that it merits serious consideration.

The Foundation asserted in oral argument that there is abundant evidence in the record supporting the proposition that the value of the Ahmanco shares is the same whether considered in one block or two. That is, it argued that there is sufficient evidence that the 3 percent discount applied by the district judge in valuing the nonvoting shares would be appropriately applied even if those shares were held in the same block as the sole voting share. Before we examine the state of the record, however, we must determine what the district judge found on this point, as that will determine our standard of review.

The district judge, in his findings of fact and conclusions of law, treats separately the 99 nonvoting shares of Ahmanco and the 1 voting share. This strongly suggests that he valuated the shares in two blocks rather than one. However, as he did not explicitly state this, we may consider the possibility that he intended to find that the discount would apply to the 99 shares even if sold in a block with the 1 voting share.

If the district judge had made the factual finding that the 100 shares of Ahmanco sold as a block would be subject to a discount for their nonvoting component, that finding would be subject to review under the clearly erroneous standard. Fed. R.Civ.P. 52(a); Sutton v. Atlantic Richfield Co., 646 F.2d 407, 412 (9th Cir. 1981). If the district judge had so held, we believe that it would be clear error. As the Foundation stated in its brief, all of the experts approached the 1 voting share and the 99 nonvoting shares of Ahmanco as separately marketable assets. The record simply does not contain support for the proposition that nonvoting shares are sold at a discount when sold together in a package with sufficient voting shares to give control. Indeed, it seems difficult to imagine how there could be a discount in such a case, as the buyer of the package would have the ability simply to recapitalize and remove the distinction in the classes of stock and thus the need for the hypothesized discount affecting the portion of the shares which theretofore were without voting rights. We must therefore reject the Foundation’s suggestion that the Ahmanco stock would be subject to a discount if all 100 shares were valued as a block.

We therefore conclude that for purposes of valuing the gross estate of the decedent, the 100 shares of Ahmanco stock (representing the value of the 600 HFA shares that constitute the sole asset of Ahmanco) should not be viewed for valuation purposes as two separate marketable assets as they would be viewed in the hands of their respective recipients under the terms of the will. Rather, the 100 shares of Ahmanco stock should be viewed in the hands of the testator and thus given a value equal to the price which the 600 shares of HFA stock would have in an exchange between a willing buyer and seller.

C. Control Premium.

The government contends that the district judge undervalued the 600 shares of HFA stock by failing to find that they would command a control premium. The 600 shares of HFA stock constituted a controlling interest in the Ahmanson group of enterprises. The valuation of stock is a question of fact. Penn v. Commissioner, 219 F.2d 18, 20-21 (9th Cir. 1955). We therefore review it under the “clearly erroneous” standard. Fed.R.Civ.P. 52(a); Sutton v. Atlantic Richfield Co., supra, 646 F.2d at 412 .

*770 There was expert testimony before the district court that the 600 shares of HFA would not command a control premium. One aspect of this testimony focused on the fact that the 600 HFA shares represented too small an equity entitlement to warrant a premium. This testimony failed to take proper account of certain nonvoting HFA stock and stock in subsidiaries within the gross estate in valuing the 600 shares of HFA stock. The possibility of a control premium should be assessed in light of all the economic entitlement within the gross estate that is affected by the control premium, on the assumption that a hypothetical buyer of the control shares would also bid for any other property in the gross estate which when combined with the purchase of the control shares would greatly enhance their value. We conclude, however, that despite the failure of the Foundation’s expert witness to consider Ahmanson’s ownership of these other economic entitlements in his calculations, it was not clear error for the district judge to find that there would be no control premium paid for the HFA shares. The government failed to show that a prospective buyer of both the 600 control shares and the other related economic interests in HFA and its subsidiaries would be able to use the voting control of HFA in such a way to assure an increased economic advantage worth paying a premium for.

The government presented evidence that the 600 HFA shares would command a substantial premium in the marketplace because of the fact that the owner of such a control block could exploit corporate assets for personal advantage. It was not clear error for the district judge to reject the notion that such an opportunity for exploitation would not affect the fair market value of the stock. The Foundation presented expert opinion testimony that in a highly regulated business like the savings and loan industry, there are numerous legal restraints which protect against exploitation of corporate assets through such self-dealing.

There was testimony from which the district judge could find that the only substantial avenue for such economic benefit would be through the control group’s replacement of management by a superior management team. However, there was also testimony that the Ahmanson companies were already very well managed.

Thus, it was not clear error for the district judge to find that the ownership of control shares in HFA, under all the circumstances, would provide no economic benefit beyond that of ownership of an equal number of nonvoting shares.

Ill

Charitable Deduction

A. Certainty of the Charitable Deduction.

The government argues that the gift to the Foundation was, at the moment of Ah-manson’s death, too uncertain to permit a charitable deduction. The uncertainty identified by the government is the possibility that Mrs. Ahmanson might succeed in voiding the charitable gift pursuant to section 41 of the California Probate Code.

We need not consider the effect on the charitable deduction had the gift to charity been voidable. The government concedes that it was not voidable under controlling California law. The government argues that the legal theory supporting voidability, though wrong, is plausible enough to make a legal challenge a genuine possibility. The legal challenge, in turn, might have resulted in an incorrect judgment of voidability by a state court or a settlement deflecting part or all of the charitable gift.

We reject the argument that a testator’s estate plan should be penalized, not because it was faulty, but because our brethren or sisters on the state bench might mistakenly find it so. The probability of such error is negligible, and therefore does not foreclose the charitable deduction. See Underwood v. United States, 407 F.2d 608, 610 (6th Cir. 1969).

The probability of a settlement that might affect the gift to charity, evaluated at the time of death, is a considerably less *771 remote theory. An estate may have good grounds for making a compromise with disappointed beneficiaries even when the testator’s dispositions are legally sound. Indeed in the present case such a settlement was made between Mrs. Ahmanson and the estate.

We must distinguish, however, between the effect of an actual settlement upon the charitable deduction and the effect of a probability of settlement at the time of death. An actual settlement that reduces the gift to charity reduces the charitable deduction accordingly. Toulmin v. United States, 462 F.2d 978, 982 (6th Cir. 1972). The government does not contend that the settlement between the estate and Mrs. Ahmanson reduced the gift to charity.

Because an actual settlement may be taken into account in the filing of the estate tax return or on amended return, it is not necessary to police this matter preemptively by denying or limiting the charitable deduction on the basis of an artificial procedure that estimates the probability of such a settlement at the moment of death. Moreover, whenever there is a large gift to charity, there may well be a greater than negligible probability of a compromise, however carefully the estate plan was drawn up. The present case serves as an example. If we were to invalidate the charitable deduction in such cases to the extent of the largest reduction of the charitable gift by a settlement for which there is a greater than negligible probability, we would largely undermine Congress’s purpose in allowing charitable deductions — to encourage testators to make charitable bequests. See Underwood v. United States, supra, 407 F.2d at 610 .

We conclude that there should be no reduction of the charitable deduction because of the alleged uncertainty at the time of Ahmanson’s death of the charitable gift occasioned by the possibility of an unsound legal challenge to Ahmanson’s dispositions or by the possibility of a settlement affecting the charitable gift.

B. Assets of Ahmanco for Charitable Deduction.

The government contends that the charitable deduction for the 99 shares of Ahman-co given to the Foundation should, in any event, be zero, because Ahmanco, as a shell corporation without assets, had no value at the moment of Ahmanson’s death. By this analysis the full value of the assets going to the Foundation is included in the gross estate, but none of it is deemed to go to charity for purposes of the charitable deduction. If this approach were accepted, future testators might conclude that even with the ablest counsel, technicalities of the estate tax could deprive them of credit for bona fide charitable gifts. Such a conclusion would be contrary to Congress’s intent in establishing the charitable deduction.

The government argues that, although it does violence to the substance of the transaction, this analysis is required by the form in which the transaction was structured. We disagree. The problem results only from the form that the government seeks to impose upon the transaction. The government divides the moment of death into two parts. In the first, the death has occurred and the trust has become irrevocable, but only in the second has Ahmanco acquired the HFA stock. As discussed in section II, we reject this metaphysical division of the moment of death, and agree with the district judge that the transfers of the HFA stock to Ahmanco and of Ahmanco to the beneficiaries must be considered as two phases of a simultaneous and integrated transfer at the moment of death. Viewed in this way, the value of the Ahmanco shares for estate tax purposes takes into account the HFA stock acquired at the moment of death.

C. Treatment of Ahmanco Shares in the Gross Estate and for the Charitable Deduction.

The Foundation argues that it makes no difference if we conclude, as we did in section II, that the gross estate should include the value of the 600 HFA shares in the hands of Ahmanson, because the 99 *772 nonvoting shares must have the same value for the charitable deduction as they have in the gross estate. The Foundation argues that inconsistent valuations, for these two purposes, would be incompatible with the orderly administration and application of the estate tax law. There is, certainly, an initial plausibility to the suggestion that fairness dictates that the same method of valuation be used in computing the gross estate and the charitable deduction. This initial plausibility, however, does not survive a close second look.

The statute does not ordain equal valuation as between an item in the gross estate and the same item under the charitable deduction. Instead, it states that the value of the charitable deduction “shall not exceed the value of the transferred property required to be included in the gross estate.” 26 U.S.C. § 2055 (d). Moreover, the statutory scheme specifically requires a lower valuation for the charitable deduction than for the same item within the gross estate under certain circumstances. If the alternate valuation date is used and the property becomes more valuable by virtue of a contingency occurring between the date of death and the alternate valuation date, the higher value is included in the gross estate, but the lower value is used in computing the charitable deduction. 26 U.S.C. § 2032 (b).

In light of the purpose of the charitable deduction to encourage gifts to charity, it seems doubtful that Congress intended to give as great a charitable deduction when the testamentary plan diminishes the value of the charitable property as it would when the testamentary plan conveys the full value of the property to the charity intact. That is, the intent of encouraging charitable gifts suggests the further policy of encouraging greater rather than lesser charitable gifts. By severing the voting power of the stock from its economic entitlement, and giving only the economic entitlement to charity, Ahmanson reduced the value of the stock to the charity. In the present case, the district judge found that the reduction in value was relatively small. Under other circumstances, however, the reduction in value might be substantial. The proper administration of the charitable deduction cannot ignore such differences in the value actually received by the charity.

Thus there are compelling considerations in conflict with the initially plausible suggestion that valuation for purposes of the gross estate must always be the same as valuation for purposes of the charitable deduction. When the valuation would be different depending on whether an asset is held in conjunction with other assets, the gross estate must be computed considering the assets in the estate as a block. Otherwise, as discussed above, the testator would be able to produce an artificially low valuation by manipulatively disbursing complimentary assets into the hands of different beneficiaries — only to have those beneficiaries recombine the assets in their more valuable arrangements at some later time. The valuation of these same sorts of assets for the purpose of the charitable deduction, however, is subject to the principle that the testator may only be allowed a deduction for estate tax purposes for what is actually received by the charity — a principle required by the purpose of the charitable deduction.

Therefore the district judge erred in concluding that the valuation of the 99 nonvoting shares of Ahmanco stock would be the same for the purpose of the charitable deduction and for the purpose of the gross estate. The district judge should recompute the taxable estate, beginning with a value in the gross estate equal to the 100 shares of Ahmanco undiminished by the 3 percent reduction for the nonvoting status of the 99 shares. The charitable deduction should then be computed on the basis of that 3 percent decrease in value that resulted from the severance of the voting rights from these 99 shares.

The Marital Deduction or Community Property Exclusion

Although the government in its original audit allowed a marital deduction for the *773 $750,000 payment by the estate to Mrs. Ahmanson in settlement of her claims against the estate, the government contends that the estate was not entitled to the deduction. The government raises the issue of nondeductibility by way of a defense to the estate’s suit for refund, in the nature of a setoff. The government argues that the marital deduction or community property exclusion is limited to the value of qualifying property that Mrs. Ahmanson had a right to receive under state law at the time of Ahmanson’s death. The Foundation contends that the deduction or exclusion is available for property transferred from the estate pursuant to a bona fide settlement, as well as any amounts to which Mrs. Ah-manson was actually entitled under state law at the time of Ahmanson’s death. The first question that we must resolve, therefore, is whether the marital deduction or community property exclusion requires an underlying enforceable claim by a decedent’s spouse or whether a good faith settlement is sufficient to qualify for the deduction.

The marital deduction and community property exclusion are technically different. The marital deduction is a subtraction from the gross estate made in calculating the taxable estate. 26 U.S.C. § 2056 . The surviving spouse’s community property is excluded from the gross estate and thereby from the taxable estate. Northeastern Pennsylvania Nat’l Bank & Trust Co. v. United States, 387 U.S. 213, 219 , 87 S.Ct. 1573, 1576-77 , 18 L.Ed.2d 726 (1967) (dictum). Because the effects of the two are the same, and because they are intended to play parallel and complementary roles, insuring equal treatment as between community property and non-community property states, id., we have assumed, along with the parties, that the legal issues will be the same in this case whether the $750,000 settlement implicates the marital deduction or the community property exclusion.

A. Is Enforceability Required?

The Foundation relies upon a line of cases which originated from Lyeth v. Hoey, 305 U.S. 188 , 59 S.Ct. 155 , 83 L.Ed. 119 (1938), where the Supreme Court held that money received in settlement of a will contest should count as “acquired by gift, bequest, devise, or inheritance ...” pursuant to section 22(b)(3) of the Revenue Act of 1932. Id. at 196-97 , 59 S.Ct. at 159-60. The Foundation would draw from this case the principle that the proceeds of a bona fide arm’s-length settlement should be characterized, for estate tax purposes, as would a transfer based upon the alleged right (in this case a claim based on a community property interest) underlying the claim that stimulated the settlement.

The Foundation also relies upon Revenue Ruling 66-139, 1966-1 CB 225, which holds that a settlement made to a surviving spouse in compromise of a claim made in good faith for a dower interest, following arm’s-length negotiations, qualifies for the marital deduction. The ruling involved a challenge to an antenuptial agreement under which the widow waived her dower rights. The ruling found it sufficient to determine that the challenge was colorable. It did not inquire into whether it was, in fact, valid.

The government relies upon Commissioner v. Estate of Bosch, 387 U.S. 456 , 87 S.Ct. 1776 , 18 L.Ed.2d 886 (1967). In Bosch, the Court reviewed two cases each of which involved the question of what effect must be given to a state trial court decree where the issue decided by the state court is determinative of federal estate tax consequences. In both cases the state law question adjudicated by the trial court affected the eligibility of the federal estate tax marital deduction. In one of the two cases reviewed in Bosch, a state probate court interpreted the decedent’s will, and applied state statutes, so as to remove the burden of paying estate tax from that portion of the estate going to the spouse. The effect of this was to diminish the gifts to other beneficiaries, and to enlarge the gift to the spouse. The estate therefore claimed a larger marital deduction. The Supreme Court reversed, holding that when federal estate tax liability turned upon a determination of state law *774 (in this case, whether the decedent’s will directed the non-application of a state death tax proration statute), federal courts would not be bound unless the determination was rendered by the highest court of the state.

The government argues that if a state court adjudication as a result of a good faith adversary proceeding is not binding for estate tax purposes, then a fortiori a private good faith settlement cannot be either. We conclude that the government is correct. It is not sufficient that the assets are transferred from the estate to the spouse as the result of the spouse’s prevailing in an adversarial confrontation with the estate. Justice Harlan, in dissent, would have limited the Bosch principle by attributing “conclusiveness to the judgment of a state court, of whatever level in the state procedural system, unless the litigation from which the judgment resulted does not bear the indicia of a genuinely adversary proceeding.” Id. at 481 , 87 S.Ct. at 1791 (Harlan, J., dissenting). It is Justice Harlan’s alternative that corresponds to the position taken by the Foundation. That alternative was before the Court, and rejected.

The Court in Bosch based its holding, in part, on an examination of the legislative history surrounding the enactment of the marital deduction. The Court found that Congress intended the marital deduction to be strictly construed and applied and in citing the Senate Finance Committee stated that “ ‘proper regard’ not finality, ‘should be given to interpretations .of the will’ by state courts and then only when entered by a court ‘in a bona fide adversary proceeding.’” Id. at 464, 87 S.Ct. at 1782. In addition, the Court observed that Congress placed restrictive limitations on the allowance of the marital deduction as set out in sections 2056(b), (c), and (d) in order to eliminate loopholes that might jeopardize the federal revenues. Id. Thus the majority concluded that the test of “passing” for estate tax purposes should be whether the interest reaches the spouse pursuant to state law, correctly interpreted — not whether it reached the spouse as a result of a good faith adversary confrontation. We cannot conclude that Congress’s concern that the availability of the marital deduction be cautiously guarded, as recognized by the Court in Bosch, would apply any less to cases of settlement than to cases of lower state court adjudication. See Sobeloff, Tax Effect of State Court Decisions—The Im pact of Bosch, 21 Tax Lawyer 507, 523 (1968).

We cannot square Revenue Ruling 66-139 with Bosch. Bosch would require that the interest be enforceable; the revenue ruling appears only to require that the state law claim be sufficiently plausible to support a good faith arm’s-length settlement. For this reason, the revenue ruling, which predates Bosch, is of no effect. Landau v. Riddell, 255 F.2d 252, 254-55 (9th Cir. 1958) (Commissioner not estopped from changing his interpretation of the law when his change of position corrects a mistake of law); 10 Mertens, Law of Federal Income Taxation § 60.16 (1976); see Automobile Club v. Commissioner, 353 U.S. 180, 183 , 77 S.Ct. 707, 709 , 1 L.Ed.2d 746 (1957).

Lyeth v. Hoey dealt with the characterization for income tax purposes, of assets received by the taxpayer. In light of the holding of Bosch, we cannot extend the entire Lyeth v. Hoey doctrine, with its emphasis on good faith settlement, to the characterization of “passing” for the estate tax marital deduction. It is useful to consider the language of Lyeth v. Hoey, in order to see just how much of the doctrine of that case is applicable to the estate tax marital deduction.

Lyeth v. Hoey, supra, 305 U.S. at 196 , 59 S.Ct. at 159 .

Bosch is perfectly consistent with the Hoey proposition that the judgment of a lower state court should have the same effect as a compromise agreement. There is no reason to doubt that this principle retains validity in the estate tax marital deduction setting. See Craven, Tax Effect of State Court Decisions —The Bosch Case, 2 Real Property, Probate & Trust J. 457, 462 (1967). Bosch, however, is inconsistent with the proposition that “if in any appropriate proceeding” instituted by a spouse, the spouse “had recovered judgment for a part of the estate,” that recovery would qualify as “passing” within the meaning of the estate tax statute. The Bosch opinion relies on both broad and narrow grounds of decision and' could be read as establishing a new rule governing the effect of state court decisions in all federal tax adjudications. See Craven, supra, 2 Real Property, Probate & Trust J. at 457. We need not decide the effects of contested settlements, or state court decisions, in other particular tax areas such as the situation presented to the Court in Lyeth v. Hoey (characterization of money received in settlement of a will contest as income). Since the Court in Bosch clearly grounded part of its decision on the need to protect the estate tax marital deduction, under even the most narrow reading of Bosch, either a good faith settlement or a judgment of a lower state court must be based on an enforceable right, under state law properly interpreted, in order to qualify as “passing” pursuant to the estate tax marital deduction.

B. Was the Right Enforceable?

In short, the government’s right to a setoff depends on whether Mrs. Ahmanson had an enforceable right, prior to the settlement, to at least $5,750,000 under California law. The district judge concluded that Mrs. Ahmanson had such a right. Because of the district judge’s wholesale adoption of the estate’s proposed findings, we must engage in “a more careful analysis of adopted findings than we would have had they been authored by the district judge.” Hagans v. Andrus, 651 F.2d 622, 626 (9th Cir. 1981); see United States v. El Paso Natural Gas Co., 376 U.S. 651, 656-57 , 84 S.Ct. 1044, 1047-48 , 12 L.Ed.2d 12 (1964). The standard of review, however, remains the clearly erroneous standard. Hagans v. Andrus, supra, 651 F.2d at 626 .

The district judge’s finding of an enforceable right of Mrs. Ahmanson to more than $5,750,000 was apparently based upon the following subsidiary findings. First, the district judge found that the fair market value of Mrs. Ahmanson’s community property interest was in excess of $750,000. Second, the district judge found that it was the intent of Ahmanson that the marital trust not be reduced by the value of Mrs. Ahmanson’s community property interest in the assets of Ahmanson’s gross estate.

If this second finding is that Ahmanson intended that Mrs. Ahmanson should receive all of her community property interest in addition to the marital trust, then it is clearly erroneous. Insofar as it appears from the record, all of Mrs. Ahmanson’s community property interest of any significance was swept into the trusts by means of which Ahmanson carried out his testamentary plan. There was no provision for those assets to reach Mrs. Ahmanson except through the marital trust. Moreover, the marital trust explicitly included the value of Mrs. Ahmanson’s community property interest as the first component of the marital trust calculation. The district judge then is certainly correct that there was no intention on Ahmanson’s part that the value of the marital trust be reduced by the value of Mrs. Ahmanson’s community property. Quite the contrary, the marital trust was, in fact, the conduit by which Ahmanson intended that Mrs. Ahmanson’s community property share should pass to her. Therefore, it would be clear error to find that Ahmanson’s intent was for his *776 wife to receive community property by some mechanism other than the marital trust.

In addition, it is quite clear that Ahman-son intended that a maximum of $5,000,000 should pass via the marital trust. 2 It is inescapable that, but for some minor amounts, any community property passing to Mrs. Ahmanson, in excess of or in addition to the $5,000,000, must do so outside the structure of the Ahmanson will and trusts.

Does California law provide that Mrs. Ahmanson has an entitlement to community property passing outside the will and trusts as well as to the property passing via those routes? If so, is she entitled to at least $750,000 outside the will and trusts in addition to the $5,000,000 under the marital trust?

Our examination of the record convinces us that these questions were not properly framed before the district court and that the record in its current form does not support the conclusion of the district judge that Mrs. Ahmanson had an enforceable right to at least $5,750,000. We therefore remand for a determination of the relevant California law and, if necessary, findings of fact on these questions.

It is possible that this case involves a community property election. If Ahmanson attempted to “dispose of his wife’s share of the community property as well as his own, naming her as one of the takers,” he may well have set up a community property election. See Tyre v. Aetna Life Ins. Co., 54 Cal.2d 399, 405 , 6 Cal.Rptr. 13 , 353 P.2d 725 (1960). Arguably the effect would be to require Mrs. Ahmanson to “elect between her community property rights and her husband’s gift.” Id. If this is the fashion in which California law would treat this case, then Mrs. Ahmanson would not be entitled to both the $5,000,000 under the terms of the trust and $750,000 outside the testamentary dispositions.

We have phrased these possibilities in hypothetical language, because we do not wish to foreclose the parties from legal argument on these points, or to intrude upon the district judge’s prerogatives as the primary interpreter, in our federal courts, of the law of the state in which he sits. We do, however, believe that the possibility of a community property election must be given serious attention on remand.

C. Who Has Burden of Proof?

The Foundation would have us decide the marital deduction community property exclusion issue on the basis of burdens of proof. Because a sufficiently strong form of the Foundation’s thesis would arguably permit us to dispose of this issue without remand, we must consider this contention.

The leading principle is that “[i]n a refund suit, the taxpayer bears the burden of proving the amount he is entitled to recover.” United States v. Janis, 428 U.S. 433, 440 , 96 S.Ct. 3021, 3025 , 49 L.Ed.2d 1046 (1976); see Fed-Mart Corp. v. United States, 572 F.2d 235, 238 (9th Cir. 1978). If the taxpayer bears the burden of proving the amount he is entitled to recover, then, as a general matter, the taxpayer should also have the burden with respect to all issues bearing upon the recovery amount— including the issues raised when the government asserts as a defense, in the nature of a setoff, the impropriety of some *777 deduction previously allowed with respect to the same return. The Foundation does not dispute that this is so for purposes of the burden of persuasion. It contends, however, that in the special case of an equitable setoff, the burden of going forward should be placed on the government.

The Foundation relies upon a theory of the Court of Claims that in cases in which the government makes a claim for setoff based on the same tax return that the taxpayer bases his suit upon, the government must “ ‘come forward’ with sufficient evidence to show that there is a reasonable basis for its claim .... ” Union Pacific Railroad v. United States, 524 F.2d 1343, 1383 (Ct.Cl.1975), cert. denied, 429 U.S. 827 , 97 S.Ct. 83 , 50 L.Ed.2d 89 (1976). The underlying theory of the Court of Claims’s position is that the government should not be permitted to use setoff defenses as a means of harassing the taxpayer and deterring legitimate claims for refunds. The government must show

Missouri Pacific Railroad v. United States, 338 F.2d 668, 672 (Ct.Cl.1964).

We accept the reasoning of the Court of Claims theory. The government should bear some burden of establishing that its equitable setoff defenses to a tax refund suit are made in good faith and are not completely lacking in substance. If the government fails to make such a showing, then it may be subject to summary judgment.

Thus we agree with the Court of Claims that the government does bear a certain threshold burden. If it is to be called a burden of “going” forward, however, it must be recognized to be a special burden of going forward. Its weight must be understood in terms of its rationale— preventing government misconduct. This rationale does not require that the government introduce sufficient evidence that a trier of fact could find in favor of the government. Instead the government should only be required to show that its claim has sufficient substance that it is made in good faith, rather than for improper purposes of deterrence and harassment. Once the government has established this threshold burden, the burden shifts back to the taxpayer. The taxpayer then has the same burdens of going forward and persuasion as are carried by the taxpayer with respect to other issues in a tax refund suit.

In the present case, the government correctly argued to the district judge that a good faith settlement was insufficient to establish the marital deduction or community property exclusion for the $750,-000 payment from the estate to Mrs. Ahmanson. That correct legal position, together with the will and trust instruments, stating that the value of Mrs. Ahmanson’s community property was to be included in the gift under the will, was sufficient to show the substantial character of the government’s setoff claim. The Bosch case, in light of the record and California law, establishes the very real possibility that the estate would be entitled to the deduction or exclusion for the $750,000 settlement, only if Mrs. Ahmanson’s community property estate equaled or exceeded $5,750,000. Obviously there is a significant possibility that the Ahmansons did not accumulate $11,500,-000 of total community property over the *778 3V2 years of their marriage. We therefore conclude that the government met its burden of “going forward” by showing that there was a substantial issue underlying its setoff defense and thus indicating that such a defense was being used as a shield, rather than a sword.

If there was a failure to carry a burden here, it was a failure on the part of the Foundation to carry its burden of persuasion on the ultimate issue. The record is simply insufficient to establish that Mrs. Ahmanson could have been awarded a total of $5,750,000 from property under control of the estate by a court properly interpreting the law of California. We are reluctant, however, to decide this issue against the Foundation on the basis of the burden of proof because the record suggests that there was confusion on these issues. The parties should be permitted to advance legal arguments and, if necessary, put on proof to permit a proper determination of the question.

Jury Trial

The government failed to argue the jury trial issue in the district court, or indeed even to object to the order of the district judge that the trial would be to the court, in response to the estate’s jury demand. We therefore refuse to review that issue here. See Evans v. Valley West Shopping Center, 567 F.2d 358 , 361 (9th Cir. 1978); Frommhagen v. Klein, 456 F.2d 1391, 1395 (9th Cir. 1972).

Although we do not attribute this motive to the government in the present case, we are appalled at the prospect of a party’s gaining reversal of a trial judgment by means of an objection to the fundamental conduct of that trial, which objection it kept hidden awaiting the outcome of the trial. The requirement that an objection must normally be made in the trial court is intended to prevent one party’s laying this sort of trap for the trial judge and the other party, as well as foreclosing the opportunity of the district judge to correct error called to his attention.

AFFIRMED IN PART; REVERSED IN PART AND REMANDED.

. The Foundation is joined as an appellee with the executor, the Ahmanson Trust Company, as the sole beneficiary of the residue of the estate and payor of the estate taxes and interest in dispute.

. Section Two, Part Four, Article II of the Declaration of Trust provides in part: