115 F.3d 430
79 A.F.T.R.2d 97-2724, 97-1 USTC P 50,440
INDIANAPOLIS LIFE INSURANCE COMPANY and subsidiary,
UNITED STATES of America, Defendant-Appellee.
United States Court of Appeals,
Argued April 24, 1997.
Decided May 20, 1997.
Rehearing Denied July 18, 1997.
Theodore J. Nowacki, Bose, McKinney & Evans, Indianapolis, IN, Theodore R. Groom, Robin L. Greenhouse, Groom & Nordberg, Washington, DC, Matthew J. Zinn (argued), J. Walker Johnson, Steptoe & Johnson, Washington, DC, for Plaintiff-Appellant.
Judith A. Stewart, Office of United States Attorney, Indianapolis, IN, David I. Pincus (argued), Edward T. Perelmuter, Department of Justice, Tax Division, Appellate Section, Washington, DC, Janet Reno, U.S. Attorney General, Office of United States Attorney General, Washington, DC, for Defendant-Appellee.
John G. Kester, James T. Fuller, III, Dan S. Sokolov, Williams & Connolly, Washington, DC, for Amicus Curiae First Allmerica Financial Life Insurance Company.
William B. Harman, Jr., Davis & Harman, Washington, DC, for Amicus Curiae Stock Company Information Group.
Before EASTERBROOK, KANNE, and EVANS, Circuit Judges.
EASTERBROOK, Circuit Judge.
Life insurance companies invest their premium income. Some of the earnings are held in reserve to pay the death benefits or annuities promised in the policies. Earnings in excess of the amounts needed for these purposes may be distributed to holders of whole life or "participating" policies as dividends. These dividends are not taxed at the insurance company level; the insurer is treated like a mutual fund, which passes income through to the investor. For mutual insurance companies, some of the dividends may not be what they seem. Policyholders of a mutual insurer also are nominal "owners" of the company. Perhaps some of their premiums are equivalent to purchase of the insurer's stock. Then dividends could be a combination of return on investment by the insurer (taxed only at the policyholder level) and return on investment in the insurer (taxed at both levels, because corporations cannot deduct dividends on their own stock).
Section 809 of the Internal Revenue Code, 26 U.S.C. § 809, establishes a complex formula for allocating distributions to mutual policyholders between investment earnings and implicit dividends on imputed stock ownership. Simplified radically, the statute provides that when a rolling average of reported earnings of major stock life insurance companies exceeds the earnings of a mutual insurer, the mutual insurer must treat about 90 percent of the excess (multiplied by the firm's equity base) as taxable income. The difference between the stock rate and the mutual rate is the return on policyholders' imputed equity investment, and these sums cannot be deducted as dividends. What happens if the mutual's income exceeds the stock-company average? May the mutual deduct the difference? Stated otherwise, can the excess of the stock rate over the mutual rate be negative? The Commissioner of Internal Revenue believes that the answer is no. 26 C.F.R. 1.809-9(a). The only court of appeals that has considered the issue agrees. American Mutual Life Insurance Co. v. United States, 43 F.3d 1172 (8th Cir.1994). If the Commissioner is wrong, the United States must refund about $5 billion to mutual insurers. The district court concluded that the Commissioner is right, although not for the Commissioner's preferred reasons. 940 F.Supp. 1370 (S.D.Ind.1996). The district court disapproved the analysis in American Mutual Life but decided against the taxpayer for a reason the Commissioner declines to defend on appeal. The Treasury Department's own understanding of § 809 has fluctuated. For a brief time it sided with the mutual insurers but then switched views; in a report to Congress the Secretary of the Treasury critiqued both the design and operation of § 809 and called for revision, which has not taken place. Final Report to The Congress on Life Insurance Company Taxation 23-36 (1989).
Facts of this case matter little, but we summarize them to show how the question arises. During the 1986 taxable year, Indianapolis Life paid or accrued about $21 million in policyholder dividends. Section 808(c)(1) allows an insurer to deduct these dividends; § 808(c)(2) and § 809(a)(1) require a mutual insurer to reduce the dividends deduction by the "differential earnings amount." This amount depends on the relation between the earnings rate of the stock life insurance industry and the post-dividend earnings rate of the mutual life insurance industry, with adjustments to normalize these returns to the 1984 level (when Congress enacted § 809). 26 U.S.C. § 809(d). Earnings of mutual companies are computed year-by-year, but trail the current year. Earnings of stock companies are averaged over three years. When calculating its taxes for 1986, Indianapolis Life was required to use the mutual rate for 1984, compared to the stock rate for 1984-86. It is not possible for the IRS to complete the calculations to determine these industry-wide earnings rates before insurers must remit their taxes. So the IRS publishes a preliminary figure, which mutual insurers use. The preliminary figure for the 1986 tax year showed that the mutual rate was less than the stock rate by an amount that required Indianapolis Life to reduce its dividends deduction by approximately $10.5 million, which it did.
During the next year, the IRS provided a final calculation called the "recomputed differential earnings amount." This calculation showed that the mutual earnings rate actually had been higher than the three-year average of the stock earnings rate (at least after normalization). Under § 809(f)(2), when the "differential earnings amount" exceeds the "recomputed differential earnings amount" calculated under § 809(f)(3)--which happened for the 1986 tax year--"such excess shall be allowed as a life insurance deduction for the succeeding taxable year." Indianapolis Life ran the calculations using the "recomputed differential earnings amount" for 1986 and took a "life insurance deduction" of $12.2 million for 1986 on its 1987 return. This meant that the dividend deduction of $10.5 million from 1986, plus the life insurance deduction from 1987 (attributable to 1986 results), exceeded by about $1.7 million the policyholder dividends paid or accrued in 1986. This happened because the "excess" of the stock rate over the mutual rate was negative, and operation of the statutory formula produced a deduction larger than dividends paid--just as the formula reduces the dividends deduction when the stock rate is greater than the mutual rate.
The Commissioner disallowed the extra $1.7 million deduction, leading to this suit in which Indianapolis Life seeks a refund. Because the refund claim depends on loss carrybacks, several tax years are in question, and the sum in dispute turns out to differ slightly from $1.7 million. But the legal question is clear. If operation of the § 809 formula produces a negative number (when the mutual rate exceeds the stock rate), may mutual insurers deduct from income earned in a year more than they paid that year in policyholder dividends? The Commissioner insists that the answer is "no" because "negative excess" is contradictory, but the concept of negative numbers, although rejected by the Greeks and Romans, has been accepted in western culture since the 16th Century. Elliott Mendelson, Number Systems and the Foundations of Analysis (1973). Today mathematics even comprises irrational, complex, and imaginary numbers, of which all too many appear on tax returns. The Commissioner also observes that § 809 is captioned "[r]eduction in certain deductions" and § 809(a)(1) reads: "In the case of any mutual life insurance company, the amount of the deduction allowed under section 808 shall be reduced (but not below zero) by the differential earnings amount." As the Commissioner sees things, § 809 reduces deductions but never increases them. But complex statutes often do things that are hard to encapsulate in a caption. Undoubtedly the principal effect of § 809 is dividend reduction. Does § 809(f)(2)--which authorizes "life insurance deductions" rather than dividend reductions--sometimes produce a deduction that exceeds dividends paid? That question cannot be answered by examining the caption of § 809 or the text of § 809(a)(1).
Nor can it be answered by reference to arguments about "windfalls." The Commissioner says that any deduction exceeding dividends paid would be a "windfall," but this is an unhelpful concept in the law of taxation. A greater deduction means a lower tax, but the mutual insurer will continue to be a net taxpayer. How high that tax will be depends on the interaction of many provisions, and a particular tax shield is just part of the edifice rather than a "windfall." For the period 1984-87, § 809 decreases Indianapolis Life's deductions by $15,918,822 if we accept its position, and by $17,602,644 if we accept the Commissioner's. A "negative excess" that smoothes the imputed dividend stream over time is no more a "windfall" than is an operating loss carryback or carryforward. Many parts of the tax code break the link between the calendar or fiscal year and the computation of tax; § 809 itself, with its rolling average of stock-company earnings and the look-back provision for recomputed differentials in § 809(f), is such a section.
If the mutuals are right about what § 809(f)(2) does, and if the mutual earnings rate always exceeded the stock earnings rate, then § 809 would authorize deductions exceeding actual dividends year in and year out. That is hardly what Congress expected to happen--the implicit dividend on policyholders' imputed equity investments cannot be negative over the long run--but the operation of many a provision in the Internal Revenue Code surprises its drafters. Some loopholes in the Code are concessions to interest groups but others are just loopholes. Let us bypass fantasies, however; the mutual earnings rate does not exceed the stock rate and is not likely to do so. Historically, the mutual earnings rate has been about 3 percent lower than that of stock life insurance firms. According to the mutuals, allowance of negative excesses is essential to limit the dividend reduction to that 3 percent, on average. An example illustrates:
Mutual Stock Annual Imputed Mutuals' Treasury Year Rate Rate Difference Rate Position Position 12 15 3 18 21 3 6 9 3 1 12 15 3 15 3 3 2 18 21 3 15 "3 0 3 6 9 3 15 9 9 Average 12 15 3 15 3 4
In this example, both mutual and stock company earnings rates fluctuate, with the mutual rate always 3 percent below the stock rate. But because the stock rate is averaged over three years, while the mutual rate is not averaged, in Year 2 the mutual rate of 18 percent exceeds the imputed stock earnings rate of 15 percent. This is the year that, on the mutuals' reasoning, has a "negative excess" (the excess of the stock rate over the mutual rate is negative) and leads to life insurance deductions greater than dividends paid. Recognizing the negative excess produces an average dividend reduction, in the Mutuals' Position column, of 3 percent of the insurer's income per year. Failing to recognize the negative excess leads to an average reduction of 4 percent, even though the stock rate exceeds the mutual rate by only 3 percent. The higher reduction, in the Treasury Position column, is driven by the results in Year 3, when the mutual earnings rate is low but the stock rate remains at 15 percent, as a result of the rolling three-year average. To offset years such as Year 3, Indianapolis Life insists, it is essential to recognize a "negative excess" in Year 2. Rounding negative years up to zero, as the Commissioner does, inevitably leads to an average dividend reduction greater than the long-term excess of the stock rate over the mutual rate.
Logical the mutuals' position may be, but the Internal Revenue Code does not always follow logic. See Howell v. United States, 775 F.2d 887 (7th Cir.1985). The "problem" that the use of a negative excess addresses is built into § 809. It comes from the fact that the stock earnings rate is averaged over three years, while the mutual earnings rate is not averaged. The "negative excess" methodology effectively averages the mutual rate and the stock rate over the same period. Perhaps that would be a good thing to do, but it is not what the Code does. We may assume that the legislature never thought about the effects the difference between the one-year and three-year periods would have, but this does not authorize us to rewrite the Code along the lines Congress might have chosen, had it considered the problem. West Virginia University Hospitals, Inc. v. Casey, 499 U.S. 83, 100-01, 111 S.Ct. 1138, 1147-48, 113 L.Ed.2d 68 (1991); Tafflin v. Levitt, 493 U.S. 455, 461-62, 110 S.Ct. 792, 796-97, 107 L.Ed.2d 887 (1990). What Congress might have done, it did not do; only what Congress actually enacted is the law.
The 1984 amendments were the result of a knock-down, drag-out fight between mutual and stock life insurers, each arguing that amendments were needed to "level the playing field" and make the Code "neutral" between mutual and stock forms of organization. The Treasury, an interested participant, wanted to use the occasion to increase revenues from this industry. The compromise gave none of the players everything it asked for. Asymmetric averaging creates the prospect of dividend reductions larger than the actual earnings difference; but the normalization feature inserted into the Code at the mutuals' behest knocks out some of the reduction. This is not to say, as the district court did, 940 F.Supp. at 1386-88, that the normalization feature justifies what would otherwise be an inequitably high tax on mutuals. It is to say, rather, that the ultimate purpose of a tax code is to raise revenue, and the many rough cuts that result from the political battles about how much will be paid by whom should not be revised, in litigation, to make them look more like one side's idea of an "ideal" tax. Any effort along the lines of § 809 may be doomed to failure, as the Treasury Department believes; if so, then an attempt to equalize things in one respect will accentuate some other disparity. "[P]olicyholder dividends may blend together price reductions, interest payments (reflecting the companies' use of any redundant premiums between receipt and repayment), and equity-like returns. Unfortunately, there has never been a practical or accurate means of determining what portion of policyholder dividends falls in each category." Report at 23-24. Perhaps policyholders should be treated more like partners than stockholders, with all income passed through, while mutuals' debt investors are treated more like stockholders. Perhaps the tax on mutuals' reserves already covers the economic value of imputed dividends. See Michael J. Graetz, Life Insurance Company Taxation: An Overview of the Mutual-Stock Differential, in Life Insurance Taxation: The Mutual vs. Stock Differential 1-9 (Graetz ed. 1986). Perhaps the only "fair" tax on the insurance business is zero; insurance is a form of financial intermediation, and some other financial intermediaries pass earnings through to investors without double taxation at the corporate level. But we are hardly in a position to "level the playing field" on which life insurers, mutual funds, and ERISA pension or annuity plans compete. Our part is simply to enforce the Code as it stands.
None of this, unfortunately, has moved us closer to understanding what the Code as it stands means. The Commissioner's observation that "excess" usually is a positive number, and that it appears frequently in § 809 in a context suggesting a positive meaning, does not establish that the word "excess" in § 809(f) is necessarily positive. As we have stressed, § 809 is long and complex. It would not be surprising for the same word to bear two meanings in different contexts. See Robinson v. Shell Oil Co., --- U.S. ----, 117 S.Ct. 843, 136 L.Ed.2d 808 (1997) (concluding that the word "employee" has different meanings within a single statute). See also Atlantic Cleaners & Dyers, Inc. v. United States, 286 U.S. 427, 433, 52 S.Ct. 607, 608-09, 76 L.Ed. 1204 (1932). The IRS itself has calculated negative "recomputed differential earnings amounts", and we hesitate to say that it was out of bounds in doing so. Still, that it is both linguistically and mathematically possible for the statutory adjustment factor to be negative need not imply that the mutual insurers may deduct more than their dividends. Section 805(a) has something to say about this:
For purposes of this part, there shall be allowed the following deductions:
* * * * * *
(3) The deduction for policyholder dividends (determined under section 808(c)).
* * * * * *
Except as provided in paragraph (3), no amount shall be allowed as a deduction under this part in respect of policyholder dividends.
Section 809(f)(2), on which Indianapolis Life relies, is in the same part of the Code as § 805. So if a deduction under § 809(f)(2) is "in respect of policyholder dividends", it is limited by the amount of dividends paid or accrued.
Indianapolis Life submits that the "life insurance deduction" authorized by § 809(f)(2) is not "in respect of policyholder dividends". In support it offers not only the difference in language ("life insurance deduction" versus "deduction for policyholder dividends") but also the fact that, under § 809(f)(1), when the "recomputed differential earnings amount" exceeds the "differential earnings amount", the "excess shall be included in life insurance gross income for the succeeding taxable year." These details of § 809(f) show, according to Indianapolis Life, that although the effect of the "differential earnings amount" under § 809(c) and § 808(c)(2) may be to reduce the dividends deduction--and therefore is "in respect of policyholder dividends" under § 805(a)--the "recomputed differential earnings amount" operates directly on income. If the recomputed amount exceeds the original amount, the difference goes straight to the bottom line as income; if the recomputed amount is less than the original amount, the difference is subtracted from income as a "life insurance deduction." Either way, dividends paid are irrelevant, making § 805(a) irrelevant too.
This proposed distinction between the original and the recomputed differential earnings amounts gives bureaucratic delay a $5 billion legal effect. For if the IRS could calculate the earnings of the mutual and stock industries faster, then the "differential earnings amount" and the "recomputed differential earnings amount" would be the same, and § 805(a) would prevent the use of any "negative excess" to augment the deduction. Indianapolis Life believes that this is not right, and that even the first calculation under § 809(c) could exceed dividends paid or accrued. It bases this contention on the fact that § 809(a)(2) provides: "In the case of any mutual life insurance company, if the differential earnings amount exceeds the amount allowable as a deduction under section 808 for the taxable year (determined without regard to this section), such excess shall be taken into account under subsections (a) and (b) of section 807." Translated to English, this means that if the amount of dividends deduction to be disallowed exceeds dividends actually paid or accrued, the difference increases the insurer's taxable income by reducing reserves under § 807. It follows, the mutual carriers believe, that the differential-earnings apparatus ignores dividends from beginning to end. The amount included in, or excluded from, net income never has anything to do with the dividends the insurer pays, but only with calculating a tax on implicit dividends on imputed stock ownership. But that's the rub. Section 805(a) refers to deductions "in respect of policyholder dividends"--not necessarily dividends paid, but just "policyholder dividends." Section 809 came about because Congress believed that some dividend payments to mutuals' policyholders are disguised distributions of profits policyholders make in their role as the mutuals' equity investors. Additions and subtractions under § 809, and the limitation in § 808(c)(2), are therefore necessarily "in respect of policyholder dividends", and deductions are capped at the level of dividends.
Whatever ambiguity remains is dispelled by 26 C.F.R. § 1.809-9(a): "Neither the differential earnings rate under section 809(c) nor the recomputed differential earnings rate that is used in computing the recomputed differential earnings amount under section 809(f)(3) may be less than zero." When the Internal Revenue Code is ambiguous, a court must respect the Treasury Department's interpretive regulations, if they "implement the congressional mandate in some reasonable manner." National Muffler Dealers Ass'n v. United States, 440 U.S. 472, 477, 99 S.Ct. 1304, 1307, 59 L.Ed.2d 519 (1979), quoting from United States v. Cartwright, 411 U.S. 546, 550, 93 S.Ct. 1713, 1716, 36 L.Ed.2d 528 (1973), quoting from United States v. Correll, 389 U.S. 299, 307, 88 S.Ct. 445, 449-50, 19 L.Ed.2d 537 (1967). See also Cottage Savings Ass'n v. CIR, 499 U.S. 554, 560-61, 111 S.Ct. 1503, 1507-08, 113 L.Ed.2d 589 (1991). "Ambiguous" is the reading of §§ 805, 808, and 809 most favorable to Indianapolis Life; given the number of times § 809 uses "excess" in a way that must be greater than zero, the mutual insurers cannot (and do not) contend that that word in § 809(c) and § 809(f)(3) bears no reading other than one allowing a "negative excess". Although Indianapolis Life tells us that § 1.809-9(a) should be thrown out because it is just a crutch one litigant uses to support its position, and that it was issued in anticipation of litigation, these arguments prove too much: they would require courts to disregard all federal tax regulations, because the Treasury is vitally interested in every tax case, and most tax questions sooner or later come to litigation. Regulation 1.809-9(a) is no worse than most, and better than some. Cf. Smiley v. Citibank (South Dakota), N.A., --- U.S. ----, ----, 116 S.Ct. 1730, 1733, 135 L.Ed.2d 25 (1996) ("Nor does it matter that the regulation was prompted by litigation, including this very suit."). Given the conclusion of the eighth circuit in American Mutual Life Insurance that the statutes of their own force preclude a "negative excess", it is hard to call the regulation unreasonable. Our rationale is in the end different from both the eighth circuit's and the district court's, but the conclusion is the same, and the judgment is