delivered the opinion of the Court.

The issue in this case is whether a bequest in trust providing for the monthly payment to decedent’s widow of a fixed amount can qualify for the estate tax marital deduction under § 2056 (b)(5) of the Internal Revenue Code of 1954, 26 U. S. C. § 2056 (b)(5). That section allows a marital deduction from a decedent’s adjusted gross estate of up to one-half the value of the estate *215in respect to specified interests which pass to the surviving spouse. Among the interests which qualify is one in which the surviving spouse “is entitled for life to . . . all the income from a specific portion [of the trust property], payable annually or at more frequent intervals, with power in the surviving spouse to appoint . . . such specific portion . ...”1

At the date of decedent’s death, the value of the trust corpus created by his will was $69,246. The will provided that his widow should receive $300 per month until decedent’s youngest child reached 18, and $350 per month thereafter. If the trust income were insufficient, corpus could be invaded to make the specified payments; if income exceeded the monthly amount, it was to be accu*216mulated. The widow was given power to appoint the entire corpus by will.2

On decedent’s estate tax return, his executor reported an adjusted gross estate of $199,760. The executor claimed the maximum marital deduction of one-half the gross estate, $99,875, on the ground that qualified interests passing to the wife exceeded that amount. The value of the property which passed to the widow outright was $41,751. To this the executor added the full value of the trust, $69,246. The Commissioner, however, determined that the trust did not qualify for the marital deduction because the widow’s right to the income of the trust was not expressed as a “fractional or percentile share” of the total trust income, as the Treasury Regulation, § 20.2056 (b)-5 (c), requires. Accordingly, the Commis*217sioner reduced the amount of the allowable deduction to $41,751. The resulting deficiency in estate tax was paid, a claim for refund was disallowed, the executor sued in District Court for refund, and the District Judge gave summary judgment for the executor. On appeal, the Court of Appeals for the Third Circuit, sitting en banc, reversed, with three judges dissenting. Because of an acknowledged conflict between the decision of the Third Circuit in this case and that of the Seventh Circuit in United States v. Citizens National Bank of Evansville, 359 F. 2d 817, petition for certiorari pending, No. 488, October Term, 1966,3 we granted certiorari. 385 U. S. 967. We reverse.

*218The basis for the Commissioner’s disallowance lay in Treasury Regulation § 20.2056 (b)-5 (c). This inters pretative Regulation purports to define “specific portion” as it is used in §2056 (b)(5) of the Code: “A partial interest in property is not treated as a specific portion of the entire interest unless the rights of the surviving spouse in income . . . constitute a fractional or percentile share of a property interest . . . .” The Regulation specifically provides that “if the annual income of the spouse is limited to a specific sum . . . the interest is not a deductible interest.”4 If this Regulation properly implements the Code, the trust in this case plainly fails to qualify for the marital deduction. We hold, however, that in the context of this case the Regulation improperly restricts the scope of the congressionally granted deduction.

In the District Court, the executor initially claimed that the entire trust qualified for the marital deduction simply because, at the time of trial, the corpus had not yet produced an income in excess of $300 per month, and that the widow was therefore entitled “to all the income from the entire interest.” The District Court rejected this contention, observing that the income from *219the corpus could exceed $300 per month, and in that event the excess would have to be accumulated. The executor’s alternative claim, which the District Court accepted, was that the “specific portion” of the trust corpus whose income would amount to $300 per month could be computed, and a deduction allowed for that amount.5

Resolution of the question in this case, whether a qualifying “specific portion” can be computed from the monthly stipend specified in a decedent’s will, is essentially a matter of discovering the intent of Congress. The general history of the marital deduction is well known. See United States v. Stapf, 375 U. S. 118, 128 (1963). The deduction was enacted in 1948, and the underlying purpose was to equalize the incidence of the estate tax in community property and common-law jurisdictions. Under a community property system a surviving spouse takes outright ownership of half of the community property, which therefore is not ificluded in the deceased spouse’s estate. The marital deduction allows transfer of up to one-half of noncommunity property to the surviving spouse free of the estate tax. Congress, however, allowed the deduction even when the interest transferred is less than the outright ownership which community property affords. In “recognition of one of the customary modes of transfer of property in common-law States,” 6 the 1948 statute provided that a bequest in trust, with the surviving spouse “entitled for life to all the income from the corpus of the trust, payable annually or at more frequent intervals, with power . . . *220to appoint the entire corpus” 7 would qualify for the deduction.

The 1948 legislation required that the bequest in trust entitle the surviving spouse to “all the income” from the trust corpus, and grant a power to appoint the “entire corpus.” These requirements were held by several lower courts to disqualify for the deduction a single trust in which the surviving spouse was granted a right to receive half (for example) of the income and to appoint half of the corpus.8 Since there was no good reason to require a testator to create two separate trusts — one for his wife, the other for his children, for example — Congress in 1954 revised the marital deduction provision of the statute to allow the deduction where a decedent gives his surviving spouse “all the income from the entire interest, or all the income from a specific portion thereof” and a power to “appoint the entire interest, or such specific portion.” The House Report on this change states that “The bill makes it clear that ... a right to income plus a general power of appointment over only an undivided part of the property will qualify that part of the property for the marital deduction.” 9 The Senate Report contains identical language.10 There is no indication in the legislative history of the change from which one could conclude that Congress — in using the words “all the income from a specific portion” in the statute, or the equivalent words “a right to income . . . over ... an *221undivided part” in the committee reports — intended that the deduction afforded would be defeated merely because the “specific portion” or the “undivided part” was not expressed by the testator in terms of a “fractional or percentile share” of the whole corpus.11

Congress’ intent to afford a liberal “estate-splitting” possibility to married couples, where the deductible half of the decedent’s estate would ultimately — if not consumed — be taxable in the estate of the survivor, is unmistakable. Indeed, in § 93 of the Technical Amendments Act of 1958, 72 Stat. 1668, Congress made “The more realistic rules of the 1954 Code” apply retroactively to the original enactment of the marital deduction in 1948, and opened the statute of- limitations to allow refunds or credits for overpayments.12 Plainly such a provision should not be construed so as to impose unwarranted restrictions upon the availability of the deduction. Yet the Government insists that even where there are well-established principles for computing the principal required to produce the monthly stipend provided for in a trust, a “specific portion” cannot be determined in that way. The “specific portion” must, the Government urges, be expressed in the trust as a fractional or percentile share of the total corpus. The spouse of a testator whose will provides for a specific monthly stipend is deprived of any benefit from the marital deduction, according to the Government’s view. But we can find no warrant for that *222narrow view, in common sense or in the statute and its history.

The Government puts most of its reliance upon a phrase which occurred once in the legislative history of the 1948 enactment. The Senate Report stated that the marital deduction would be available “where the surviving spouse, by reason of her [sic] right to the income and a power of appointment, is the virtual owner of the property.” 13 The Government’s argument is that the deduction was intended only in cases where the equivalent of the outright ownership of a community property State was granted, and that this is what the Senate Report meant by the words “virtual owner.” Actually, however, the words were not used in that context at all. The section of the Report from which those words derive deals with the rule that, with minor exceptions, the marital deduction does not apply where any person other than the surviving spouse has any power over the income or corpus of the trust. It is in this sense that the Report described the surviving spouse as a “virtual owner.” Hence, the Government’s argument that only a grant of the income from a fractional or percentile share subjects the surviving spouse to the vagaries and fluctuations of the economic performance of the corpus in the way an outright owner would be, is simply irrelevant. There is no indication whatsoever that Congress intended the deduction to be available only in such a situation, nor is there any apparent connection between the purposes of the deduction and such a limitation on its availability. Compare Gelb v. Commissioner, 298 F. 2d 544, 550-551 (C. A. 2d Cir. 1962). Obviously Congress did not intend the deduction to be available only with respect to interests equivalent to outright ownership, or trusts would not have been permitted to qualify at all.14

*223The Court of Appeals advanced a somewhat different argument in support of the Government’s conclusion. Without relying upon the validity of the Regulation, the Court of Appeals maintained that a “specific portion” can be found only where there is an acceptable method of computing it, and that no such method is available in a case of the present sort. The Court of Appeals noted that the computation must produce the “ratio between the maximum monthly income [producible by the whole corpus] and the monthly stipend [provided for in the trust].” 363 F. 2d 476, 484. The following example was given:

“If the investment factors involved were constant and it could be determined that the maximum income that could be produced from the corpus in a month was, for example, $500 then the relationship between the $300 monthly stipend and the $500 maximum income would define 'specific portion’ for marital deduction purposes, i. e.:

“$300 being % of $500 then % of $69,245.85 would be the 'specific portion’ of the trust corpus from which the surviving spouse would be entitled to the entire income of $300 monthly under maximum production circumstances.

“Though in reality it might take the entire corpus to produce the monthly stipend, or even the necessity to invade corpus might be present, nevertheless ... it could be said, after computing the theoretical maximum income, that the surviving spouse’s income interest of $300 monthly represented the investment of % of the corpus. ‘Specific portion’ would then be accurately defined for marital deduction purposes.” (Italics in original.) 363 F. 2d, at 484, n. 17.

*224The Court of Appeals concluded, however, that the computation could not be made because “[t]he market conditions for purposes of investment are unknown” and, therefore, there are no constant investment factors to use in computing the maximum possible monthly income of the whole corpus. 363 F. 2d, at 484.

It is with this latter conclusion that we disagree. To be sure, perfect prediction of realistic future rates of return15 is not possible. However, the use of projected rates of return in the administration of the federal tax laws is hardly an innovation. Cf. Gelb v. Commissioner, 298 F. 2d 544, 551, n. 7 (C. A. 2d Cir. 1962). It should not be a difficult matter to settle on a rate of return available to a trustee under reasonable investment conditions, which could be used to compute the “specific portion” of the corpus whose income is equal to the monthly stipend provided for in the trust. As the Court of Appeals for the Second Circuit observed in Gelb, supra, “the use of actuarial tables for dealing with estate tax problems has been so widespread and of such long standing that we cannot assume Congress would have balked at it here; the United States is in business with enough different taxpayers so that the law of averages has ample opportunity to work.” 298 F. 2d, at 551-552.

The Government concedes, as it must, that application of a projected rate of return to determine the “specific portion” of the trust corpus whose income is equal to the monthly stipend allotted will not result in any of the combined marital estate escaping ultimate taxation in either the decedent’s or the surviving spouse’s estate. The Government argues, however, that if analogous *225actuarial methods were used to compute as a fixed dollar amount the “specific portion” as to which a qualifying power of appointment is given, where the power in fact granted extends to the whole corpus but the corpus is subject to measurable invasions for the benefit, for example, of a child, the result, in some cases, would be to enable substantial avoidance of estate tax. Whether, properly viewed, the Government’s claim holds true, and, if so, what effect that should have upon the qualification of such a trust, is a difficult matter. Needless to say, nothing we hold in this opinion has reference to that quite different problem, which is not before us. Cf. Gelb v. Commissioner, supra.

The District Court used an annuity-valuation approach to compute the “specific portion.” This was incorrect. The question, as the Court of Appeals recognized, is to determine the amount of the corpus required to produce the fixed monthly stipend, not to compute the present value of the right to monthly payments over an actuarially computed life expectancy. Accordingly, we reverse and remand for further proceedings in conformity with this opinion.

Reversed and remanded.

whom Mr. Justice Black and Mr. Justice Harlan join, dissenting.

“Resolution of the question in this case, whether a qualifying 'specific portion’ can be computed from the monthly stipend specified in a decedent’s will, is,” says the Court, “essentially a matter of discovering the intent of Congress.” Ante, p. 219. Substituting “exclusively” for “essentially,” I entirely agree with the Court’s statement of the case. “The deduction was enacted in 1948, and the underlying purpose was to equalize the incidence of the estate tax in community property and common-law jurisdictions.” Ante, p. 219. Again I agree. But I must *226differ with the Court in its determination that the intent of Congress leads to the result the Court today reaches. For allowing the trust before us to qualify for the marital deduction will inevitably lead to the ironic and unjustified result of giving common-law jurisdictions more favorable tax treatment than community property States.

The Court holds that the widow in this case had an interest in “all the income from a specific portion” of the trust because the stream of payments to her could be capitalized by the use of assumed interest rates. This capitalized sum is then said to constitute the “specific portion” which qualifies for the marital deduction. A corollary of the Court’s theory is that a trust which gave the widow the right to the income from a fixed amount (in dollars) of corpus and the right to appoint the entire corpus would support a marital deduction.1 But if such a bequest qualifies, then one which limits her power of appointment to only that amount of corpus with respect to which she has income rights will also qualify for the marital deduction. For under the statute, the survivor must have only the right to “all the income from a specific portion . . . with power in the surviving spouse *227to appoint . . . such specific portion.”2 (Emphasis added.) The way in which such an estate allows a tax avoidance scheme not available to a community-property couple can be easily illustrated.

Assume a trust estate of $200,000, with the widow receiving the right to the income from $100,000 of its corpus and a power of appointment over that $100,000, and the children of the testator receiving income from the balance of the corpus during the widow’s life, their remainders to vest when she dies. Now suppose that when the widow dies the trust corpus has doubled in value to $400,000. The wife’s power of appointment over $100,000 applies only to make $100,000 taxable to her estate.3 The remaining $300,000 passes tax free to the children. Contrast the situation in a community property State. The wife’s 50% interest in the community property places $200,000 of the expanded assets in her estate and taxable as such; only $200,000, therefore, passes directly to the children. Thus, the Court’s interpretation of “specific portion” affords common-law estates a significant tax advantage that community property dispositions cannot obtain.

By changing “specific portion” from the fractional share, which is both described in the Treasury Regulation and used as the basis for community property ownership, into a lump sum bearing no constant relation to the corpus, the Court allows capital appreciation to *228be transferred from the wife’s to the children’s interest in the estate without any tax consequence. Thus, today’s decision is directly opposed to what we have previously recognized as the purpose of the marital deduction:

“The purpose ... is only to permit a married couple’s property to be taxed in two stages and not to allow a tax-exempt transfer of wealth into succeeding generations. Thus the marital deduction is generally restricted to the transfer of property interests that will be includible in the surviving spouse’s gross estate.” United States v. Stapf, 375 U. S. 118, 128.

The reference in the legislative history of the 1948 Act to the wife’s “virtual owner [ship]” of the interest qualifying for the deduction is explained by the purpose discerned in Stapf, supra.4 For only if she is the “virtual owner,” will the wife’s interest appreciate with the rest of the trust. Similarly, the congressional committee reports, in limiting their examples of “specific portions” to fractional shares, manifest an understanding that no tax avoidance was to be allowed via the marital deduction.5 In no other manner could Congress have “equalize[d] the incidence of the estate tax in community property and common-law jurisdictions,” as the Court so aptly puts it.

In ruling as it does today the Court not only frustrates the basic purposes of the marital deduction, it also ignores or brushes aside guideposts for deciding tax cases that have been carefully established in prior decisions of this Court. Thus, a 10-year-old interpretation of the statute contained in the Treasury Regulations is held invalid, although we have consistently given great weight *229to those regulations in the interpretation of tax statutes. See, e. g., United States v. Stapf, 375 U. S. 118, 127, n. 11.

Of even greater importance is the sharp change of attitude toward the marital deduction which today’s decision heralds. The Treasury’s interpretation of “specific portion” is held invalid because “Congress’ intent [was] to afford a liberal ‘estate-splitting’ possibility.” This finding of “liberalism” in the marital deduction leads the Court to reason that “[p]lainly such a provision should not be construed so as to impose unwarranted restrictions upon the availability of the deduction.” Ante, p. 221. But we have previously construed the marital deduction to mean what it says and have not discerned a liberal intent that allows us to write new words into the statute, as the Court does here in changing “specific portion” to “ascertainable amount.” For example, in Jackson v. United States, 376 U. S. 503, 510, eight members of the Court, speaking through Mb. Justice White, declared that “the marital deduction . . . was knowingly hedged with limitations” by Congress, and “ [t] o the extent it was thought desirable to modify the rigors of [such limitations], exceptions . . . were written into the Code.” Thus, the lesson announced in Jackson, but ignored today, was that “[c]ourts should hesitate to provide [other exceptions] by straying so far from the statutory language.” Cf. Meyer v. United States, 364 U. S. 410. One looks in vain through the Jackson, Meyer, and Stapf opinions, supra, for the roots of the liberalism which the Court today finds bursting forth from the marital deduction.

With this change in approach, uncertainty is now introduced into one of the areas of the law where long-range reliance upon the meaning of a statute is essential. Estate planners and tax lawyers are technicians schooled to view the marital deduction as a tightly drawn, precise provision. They are now shown a totally new statute *230that is to be construed in the manner of a workman’s compensation act. See Jackson v. Lykes Bros. S. S. Co., 386 U. S. 731. Such a construction will hardly promote “[t]he achievement of the purposes of the marital deduction [which] is dependent to a great degree upon the careful drafting of wills.” Jackson v. United States, 376 U. S., at 511.

Believing today’s decision to be at odds with the statutory purpose and the consistent interpretation of the marital deduction, I respectfully dissent.