SNEED, Circuit Judge:

The Commissioner of Internal Revenue assessed deficiencies against a partnership, Vinson Elkins (V E), disallowing a portion of the deductions, which were passed through to its partners, for contributions to defined benefit pension plans. The Tax Court found that the actuarial assumptions underlying those contributions were reasonable, and decided in favor of the partnership. The Commissioner appeals. We affirm.

I. [2] FACTS AND PROCEEDINGS BELOW

V E is a large general practice law firm. Commencing in 1984. V E established individual defined benefit pension plans (IDBs) for many of its partners.

Under these plans, each partner made annual contributions to his own IDB such that accumulated contributions, plus the expected income to be earned on those contributions, would suffice to pay the defined periodic benefit amounts when that partner retired. Section 404 of the Internal Revenue Code allows employers to deduct such contributions. 26 U.S.C. § 404 (1988). V E, as a partnership, listed contributions on its partnership, return which were then passed through to V E partners, who claimed them on their individual returns.

Computing the annual contribution necessarily involved estimating such variables as the expected income on plan contributions and the expected retirement age and longevity of the covered partners. Section 412(c)(3) provides that these estimates are to be determined by actuaries, but requires that actuarial estimates be "reasonable," either individually or in the aggregate, and that they overall represent the actuary's "best estimate of anticipated [plan] experience." Id. § 412(c)(3).

For the 1986 and 1987 fiscal years, V E's actuary calculated plan contributions by estimating (1) a five percent annual return on plan assets, (2) a retirement age of 62, and (3) annual postretirement administrative expenses of five percent of benefit costs. In order to estimate the number of years each retiring partner would live to receive benefits, the actuary used mortality tables established in 1971 by insurance companies to estimate the life expectancy of annuity purchasers.

Beginning in 1986, V E added a preretirement death benefit option to its IDB plans. Under this option, a partner's beneficiaries would receive a defined benefit if that partner died before retiring. Forty-nine partners elected this option. In order to calculate death benefit contributions, V E's actuary used mortality tables prepared in 1958 by life insurers.

Based on these assumptions, the actuary computed aggregate plan contributions of approximately $10.8 million and $5.1 million for 1986 and 1987, respectively. Upon audit, the Commissioner disallowed approximately $8 million and $3.5 million, respectively, of the 1986 and 1987 deductions, arguing that V E's actuarial assumptions were too conservative and were therefore unreasonable. V E's challenged these findings before the Tax Court, which found that V E's assumptions were reasonable and reinstated the disallowed amounts. Vinson Elkins v. Commissioner, 99 T.C. 9, 1992 WL 162641 (1992). The Commissioner appeals.

II. [9] JURISDICTION AND STANDARD OF REVIEW

This court has jurisdiction under 26 U.S.C. § 7482(a).

Deciding the appropriate standard of review both begins and ends this case. If, as V E contends, the Commissioner is merely challenging the Tax Court's factual finding of reasonableness, we can only overturn for clear error. See Jerome Mirza Assocs. v. United States, 882 F.2d 229, 230 (7th Cir. 1989), cert. denied, 495 U.S. 929, 110 S.Ct. 2166, 109 L.Ed.2d 496 (1990).

Under a clear error standard, the Tax Court judgment should be affirmed. The Tax Court carefully weighed expert testimony from both sides, along with evidence of industry practice, in finding that each of V E's assumptions were reasonable. Its well-considered findings are not clearly erroneous.

On the other hand, the Commissioner argues that the Tax Court applied both an improper legal standard and a proper one improperly. Specifically, the Commissioner asserts that the Tax Court: (1) reviewed V E's assumptions for reasonableness without determining whether they were also the "best" estimates, (2) assessed the IDBs in the aggregate without accounting for variations among the plans, (3) incorrectly applied a "substantial unreasonableness" test, and (4) erred in allowing V E to use different mortality assumptions for pension and death benefit calculations.

All of these arguments initially present legal issues of statutory interpretation, which we must review de novo. Lennox v. Commissioner, 998 F.2d 244, 247 (5th Cir. 1993). We hold that the Commissioner's legal arguments lack merit. As a consequence, there remains only a factual dispute as to which we must affirm the Tax Court.

The legal arguments of the Commissioner fall under four headings. The first concerns the meaning of the "Best Estimate Test"; the second, the "Reasonableness of Uniform Assumptions" in fixing contributions; third, the questionable "Substantially Unreasonable Test"; and fourth, the "Reasonableness of Death Benefit Assumptions" used in this case. Each shall be addressed separately.

III. [16] DISCUSSION

A. The Best Estimate Test

Section 412(c)(3) requires that actuarial estimates be reasonable and "offer the actuary's best estimate of anticipated experience." 26 U.S.C. § 412(c)(3). In reaching its decision, the Tax Court assessed the reasonableness of each individual actuarial assumption and concluded that each item was reasonable. The court rejected the Commissioner's argument that the statutory "best estimate" language imposed an additional test. It held that Congress recognized that always there would be a range of reasonable estimates and decided to give deference to actuarial judgment within that range. Vinson Elkins, 99 T.C. at 56-57. There is no single "best estimate." Accordingly, the court concluded: "The ultimate inquiry remains whether the assumptions chosen are reasonable in the aggregate." Id. at 57.

The Commissioner contends that this interpretation reads the best estimate provision out of the statute. That only leads, however, to the question of what a best estimate test would entail.

The Commissioner asserts that the best estimate test imposes a second substantive hurdle for actuarial valuations to clear. The argument in favor of such an approach is as follows: Congress did indeed recognize that actuaries could come up with a wide range of reasonable estimates, and wanted to ensure that actuaries would neutrally pick the most likely result within the range to prevent either more or less conservative assumptions.

However, this substantive approach conflicts with the "best estimate" provision and with the statutory scheme as a whole. The statute refers to the actuary's best estimate, not that of a court or of outside experts. Further, by entrusting actuaries with the task of determining plan contributions, and by granting the latitude inherent in the statutory reasonableness test, Congress intended to give actuaries some leeway and freedom from second-guessing. See H.R. Rep. No. 807, 93d Cong., 2d Sess. 27 (1974) U.S. Code Cong. Admin.News 4639, 4670, reprinted in 2 Subcomm. on Labor of the Senate Comm. on Labor Public Welfare, 94th Cong., 2d Sess., Legislative History of the Employee Retirement Income Security Act of 1974, at 3115, 3147 (Comm.Print 1976) (rejecting imposition of uniform actuarial methods and assumptions): "[A]ny attempt to specify actuarial assumptions and funding methods for pension plans would in effect place these plans in a straitjacket . . . and would be likely to result in [unreasonable] cost estimates." Adding a second, more rigorous level of substantive review via the best estimate test would frustrate that goal. Moreover, a second substantive test would render the reasonableness test superfluous.

Finally, section 412(c)(3)'s limitations are not only intended as a ceiling on maximum deductions, but also as a floor on minimum contributions. Compare 26 U.S.C. § 412(c)(3) with 29 U.S.C. § 1082(c)(3) ("[m]inimum funding standards"). The statutory language thus serves the competing goals of ensuring adequate funds for retirees while preventing taxpayer abuse. Rhoades, McKee Boer v. United States, 822 F. Supp. 445, 448 (W.D.Mich. 1993). Within the range of reasonableness, Congress assigned the task of balancing these goals to actuaries. We will not narrow the statutory gap between the Scylla of underfunding and the Charybdis of tax penalties.

In light of this analysis, we find that the best estimate test is procedural, as opposed to substantive, in nature. The statute refers to the actuary's best estimate, which implies a procedural approach. One goal of such an inquiry would be to determine whether assumptions truly came from the plan actuary or whether they were instead chosen by plan management for tax planning or cash flow purposes. See Huber v. Casablanca Indus., Inc., 916 F.2d 85, 93 (3d Cir. 1990) (finding that withdrawal liability calculation was not actuary's best estimate, because management pressured actuary to revise initial figures), cert. dismissed, ___ U.S. ___, 113 S.Ct. 1070, 122 L.Ed.2d 497 (1993); see also H.R. Rep. No. 807, supra, at 95, reprinted in 2 Legislative History, supra, at 3215 (finding it "inappropriate for an employer to substitute his judgment . . . for that of a qualified actuary," and "contemplat[ing] that if such a circumstance were to arise an actuary would have to refuse giving his favorable opinion with regard to the plan").

The relevant legislative history of ERISA, although brief, supports our analysis. The best estimate provision was added by a House-Senate conference committee. The related House Conference Report declares that, in requiring a single "best" estimate, the conferees intended that plans use one set of assumptions for all purposes, such as funding, reports to the Department of Labor, and financial reporting. H.R.Conf.Rep. No. 1280, 93d Cong., 2d Sess. 285 (1974) U.S. Code Cong. Admin.News 4639, 5038, reprinted in 3 Legislative History, supra, at 4277, 4552; cf. Concrete Pipe Prods., Inc. v. Construction Laborers Pension Trust, ___ U.S. ___, ___-___, 113 S.Ct. 2264, 2284-85, 124 L.Ed.2d 539 (1993) (observing that the requirement of consistency among actuarial applications, combined with professional and technical standards for actuaries, serves to minimize risk of bias).

The Commissioner does not challenge V E's procedure, nor argue that its estimates did not come from an actuary, nor contend that V E used different methods for tax purposes as opposed to financial reporting or funding purposes. Rather, the Commissioner argues that the resulting contributions were substantively over-conservative. This merely restates the arguments raised and lost under the reasonableness standard by the Commissioner.

The Commissioner also argues that the Tax Court's ignorance of the best estimate provision colored its view of the reasonableness test, allowing it to accept overly conservative estimates. This misreads the Tax Court opinion. It suggests that the Tax Court found a most likely estimate, then discounted that estimate to support V E's figures. In reality, however, the court did not determine a single most likely figure. Instead, it determined a reasonable range and found that V E's choice of conservative estimates within that range was reasonable based on expert testimony, the lack of actual plan experience, industry practice, and ERISA's purpose to ensure adequate pension funding. See Vinson Elkins, 99 T.C. at 36-41 (evaluating interest rate assumption).

The Tax Court's determination that a conservative approach as to interest rates and life expectancy can be a reasonable one was entirely proper. See Combs v. Classic Coal Corp., 931 F.2d 96, 101 (D.C. Cir. 1991) (finding that conservative interest rate assumption of 5.5 percent in withdrawal liability calculation was justified, in part by lack of participant data). Contrary to the Commissioner's position, reasonableness does not require precision. See Concrete Pipe, ___ U.S. at ___, 113 S.Ct. at 2286 (interpreting statutory reasonableness test for ERISA withdrawal liability calculations): "In practical terms it is a burden to show something about standard actuarial practice, not about the accuracy of a predictive calculation." Standard actuarial practice includes a basic tenet of conservatism; at trial, even the Commissioner's experts said as much. See Vinson Elkins, 99 T.C. at 27, 36. Indeed, some margin for error inheres in every estimation process. Rhoades, McKee Boer v. United States, 822 F. Supp. 445, 454 (W.D.Mich. 1993).

B. Reasonableness of Uniform Assumptions

The Commissioner next argues that the Tax Court erred by evaluating V E's interest rate assumption in the aggregate for all 132 IDBs without accounting for differences in investment mix between the individual plans. This contention lacks merit. Section 412(c)(3) does not require that actuarial assumptions be individualized, meticulous or precise; only that they be reasonable. See 26 U.S.C. § 412(c)(3). This should come as no surprise. At some point, an actuary must turn to statistical averages to predict events, such as a given person's lifespan, for which individual data is not available. See Rhoades, McKee Boer, 822 F. Supp. at 454. At trial, V E explained that individualized determinations would be infeasible and unreasonable given the lack of prior plan experience. The Tax Court implicitly accepted this analysis in accepting V E's uniform interest rate of five percent. To the extent the Commissioner is challenging a factual finding, we hold that it was not clearly erroneous.

The Commissioner also argues also argues that V E, and the Tax Court, should have adjusted the interest rate assumption to incorporate actual returns earned by the IDBs in their first three years of existence, which averaged well above the assumed five percent. We disagree. In funding a long-term plan, V E was entitled to look at long-term trends rather than the then-prevailing higher rates. See Masters, Mates Pilots Pension Plan v. USX Corp., 900 F.2d 727, 733-34 (4th Cir. 1990) (evaluating interest rate used in withdrawal liability calculation); Board of Trustees, Michigan United Food Commercial Workers Unions v. Eberhard Foods, Inc., 831 F.2d 1258, 1262-63 (6th Cir. 1987) (same). Time and declining interest rates have proven the wisdom of V E's patience.

C. The Substantial Unreasonableness Test

The Commissioner insists that the Tax Court incorrectly used a "substantial unreasonableness" test, which barred overturning past actuarial assumptions unless they were substantially unreasonable. This test, which the Tax Court derived from legislative history, may be incorrect, particularly since the statute itself merely provides that assumptions be reasonable, rather than "not substantially unreasonable." See 26 U.S.C. § 412(c)(3). Even so, any error was harmless. The court expressly found that V E's assumptions were reasonable. See Vinson Elkins, 99 T.C. at 59. Only then did it further conclude that the assumptions were not substantially unreasonable. See id.

D. Reasonableness of Death Benefit Assumptions

Treasury Regulation 1.412(c)(3)-1(f) allows qualified pension plans to fund "ancillary benefits," such as preretirement death benefits. See Treas. Reg. § 1.412(c)(3)-1(f). Taxpayers may choose one of two options for computing ancillary benefit deductions: (1) "using the same method used to compute retirement benefit costs," Treas. Reg. § 1.412(c)(3)-1(f)(1), or (2) using the premium paid under an insurance contract, see id. § 1.412(c)(3)-1(f)(3). V E chose the first option.

In estimating premium costs, V E's actuary used a 1958 life insurance mortality table. The Commissioner notes that V E used a 1971 annuity table, which estimated longer lives, for computing pension costs, while using shorter estimated lives to calculate life insurance premiums. The charge is that V E was trying to have it both ways: using shorter estimated lives to calculate life insurance premiums while using longer lives to calculate pension benefit periods, thereby inflating both.

We cannot say that the use of two tables is unreasonable as a matter of law. The Treasury Regulation itself provides only that the method of computing ancillary benefits be consistent with the rest of the plan, not that the individual assumptions be identical. See Treas. Reg. § 1.412(c)(1)-1(f)(1); see also id. § 1.412(c)(1)-1 ("[T]he specific mortality rate determined to be applicable to a particular plan year is not part of the funding method."). Further, it allows plans to opt out of their usual methods entirely in favor of using insurance contract premiums. See id. § 1.412(c)(1)-1(f)(1). The Commissioner's rigid approach here conflicts with the flexibility allowed by the regulations and by the statutory reasonableness test.

Without a per se test, the Commissioner is again left with the Tax Court's factual finding of reasonableness. The court found that using mortality tables similar to those used by life insurers was a reasonable method of estimating insurance premiums. See Vinson Elkins, 99 T.C. at 53. The court also noted that V E's estimated cost was lower than a comparable insurance company quote. Id. at 51. While acknowledging that V E used different assumptions for pension purposes, the court found the differences acceptable because V E was not estimating its partners' lifespans, but their insurance premiums. Id. at 50, 53. Its factual findings are not clearly erroneous.

AFFIRMED.