The Commissioner appeals the district court's decision allowing tax deductions for contributions to an individual defined benefit pension plan. The district court found that the actuarial assumptions used in creating the plan were reasonable in the aggregate and were the actuaries' best estimate. We reverse that decision.

In 1984, the law firm of Rhoades, McKee, Boer, Goodrich & Titta created an individual defined benefit plan for one of its partners, Dale Rhoades. This type of plan differs from a traditional pension plan in that it specifies the benefits that will be available after retirement. Rhoades funded the plan with money he would otherwise have received as income. Because the law firm was a general partnership until 1988, Rhoades deducted the contributions for 1986 and 1987 on his personal income tax. Having incorporated in 1988, Rhoades, McKee deducted contributions to the Plan for that year on its corporate tax return.

As the result of an audit in 1991, the Internal Revenue Service disallowed the reported deductions for these years, 1986-88. On April 4, 1991, the IRS issued two Notices of Final Partnership Administrative Adjustments that disallowed contributions to the plan of $151,905 for 1986 and $99,473 for 1987. On May 6, 1991, the IRS issued a Notice of Deficiency to Rhoades, McKee disallowing contributions to the plan in the amount of $94,543 in 1988. Rhoades, McKee deposited $137,619 with the IRS representing the amount of the disallowance of plan contributions for 1986 and 1987, and paid taxes, penalties, and interest for 1988 in the amount of $77,076. After a refund on taxes paid for 1988 was denied, Rhoades, McKee filed suit in district court to refund the taxes. Then, for reasons unexplained in the record, the IRS issued a check representing a full refund of $51,188 for the 1988 tax deposited, and $5,990.83 in interest. Consequently, Rhoades, McKee sought a refund of the remaining $163,507 on deposit.

Because a defined benefit plan provides for a specified benefit at retirement, there are a number of uncertainties connected with its funding. Actuarial assumptions are therefore critical to assure the plan's promised benefits are fulfilled. During the three years involved here, Rhoades, McKee used two different actuaries to make the assumptions necessary to calculate contributions to the plan, one for the 1986 assumptions, and another for 1987 and 1988. The four assumptions at issue are: retirement age, the pre-and post-retirement interest rates, and the mortality tables. The retirement age for all three years was assumed to be sixty. The pre-retirement interest was assumed to be six percent for 1986 and 1987, and 7.75% for 1988. The post-retirement interest was assumed to be five percent for all three years. The mortality tables used were 1983 IAM (female) for 1986, 1983 IAM (male) with a two year set-back for 1987, and 1983 IAM (male) with no set-back for 1988.

Section 404(a)(1) of the Internal Revenue Code allows deductions for contributions to an individual defined benefit plan, so long as the underlying actuarial assumptions are reasonable. The reasonableness standard is defined in 26 U.S.C. Sec. 412(c)(3). This appeal raises the question of the proper interpretation of Section 412(c)(3), which at the time Rhoades made these contributions stated

For purposes of this section, all costs, liabilities, rates of interest, and other factors under the plan shall be determined on the basis of actuarial assumptions and methods ... which, in the aggregate are reasonable (taking into account the experience of the plan and reasonable expectations) and which, in combination, offer the actuary's best estimate of anticipated experience under the plan.

26 U.S.C. Sec. 412(c)(3) (1982). This section establishes a two-part test for actuarial assumptions: (1) they must be reasonable; and (2) they must be the actuary's best estimate.

The Commissioner claimed below that the deductions were not allowable under Section 404(a) because the assumptions used by the actuaries were unreasonable and did not represent their best estimate of anticipated experience under the plan. The district court rejected these assertions and adopted a standard of review by which actuarial assumptions will not be retroactively changed unless they are found to be "substantially unreasonable." The court reasoned that a degree of deference should be given the assumptions used by the actuary in computing plan contributions. The court then examined each of the assumptions to determine if it was reasonable and represented the actuaries' best estimate. Finding that none of the assumptions were substantially unreasonable, the court concluded that they were reasonable in the aggregate. The court then evaluated the assumptions as a whole and found that in combination they represented the actuaries' best estimate of anticipated experience under the plan.

Section 412(c)(3) has been the subject of two other circuits' recent decisions. These opinions have addressed the same questions presented here: the "substantially unreasonable" standard of review, and the meaning of the "best estimate" language. As the facts of each case are similar to the instant one, we begin by looking to those decisions as a guide in making our own.

In Vinson & Elkins v. Comm'r of Internal Revenue, 7 F.3d 1235 (5th Cir.1993), the Fifth Circuit affirmed the tax court's finding that the actuarial assumptions underlying contributions to a defined benefit pension plan were reasonable. Vinson & Elkins, a large general practice law firm in Houston, Texas, established individual defined benefit pension plans for many of its partners in 1984. After auditing the plans, the Commissioner of Internal Revenue disallowed deductions for contributions to the plans in 1986 and 1987. The Commissioner argued that the actuarial assumptions were too conservative and were therefore unreasonable. The tax court disagreed, finding the assumptions were reasonable and reinstating the disallowed amounts. Of the legal arguments raised in the Commissioner's appeal, two are the same as those presented here.

The Fifth Circuit found that the best estimate test was procedural in nature, not substantive. That court noted that "[t]he 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." Id., at 1238. To argue that the best estimate provision imposes a second substantive requirement for actuarial assumptions conflicts with the statutory scheme: "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." Id.

The Fifth Circuit also addressed the "substantial unreasonableness" test. In reviewing the actuarial assumptions, the tax court used the same test that the district court applied here, "which barred overturning past actuarial assumptions unless they were substantially unreasonable." Id. at 1240. The Fifth Circuit noted that this test, derived from the legislative history, "may be incorrect, particularly since the statute itself merely provides that assumptions be reasonable, rather than 'not substantially unreasonable.' " Id. However, the court found that the application of this test was harmless error as the tax court expressly found the actuarial assumptions were reasonable.

The Second Circuit has also recently reviewed a similar case concerning the interpretation of Section 412(c)(3). Wachtell, Lipton, Rosen & Katz v. Comm'r of Internal Revenue, 26 F.3d 291 (2d Cir.1994). There, a New York law firm adopted a retirement plan in 1984 for its partners, each of whom was covered by an individual defined benefit plan. The Commissioner challenged that plan. However, the tax court found the actuarial assumptions used to calculate plan contributions were reasonable in the aggregate and represented the actuary's best estimate. Id. The Second Circuit affirmed that decision. In doing so, that court also addressed the best estimate provision and substantial unreasonableness test.

In language almost identical to that used by the Commissioner here, in Wachtell the Commissioner argued that the tax court's interpretation of Section 412(c) was erroneous as a matter of law, which robbed its factual findings of all vitality, and required a remand to reconsider the evidence in light of the correct legal standard. The Wachtell court noted that Congress "was aware of the important role that actuaries would play in plan-funding decisions." Id. at 295. The court agreed with the Vinson decision "that the 'best estimate' requirement is basically procedural in nature and is principally designed to insure that the chosen assumptions actually represent the actuary's own judgment rather than the dictates of plan administrators or sponsors." Id. at 296.

In addressing the substantially unreasonable test, the court found that "the Tax Court erred in extracting from the legislative history of the statute a test that was not included in the statute's text." Id. However, this error did not negate all of the court's factual findings. The tax court found that the assumptions were not substantially unreasonable only after it found that each individual assumption was reasonable and that the assumptions were reasonable in the aggregate. Thus, the substantially unreasonable test did not affect the outcome of the case. Id. at 296-97.

We adopt the reasoning of the Fifth and Second Circuits. The plain language of the statute does not support the application of the substantially unreasonable test. In considering the actuaries' assumptions in this case, the district court reasoned, "[i]f no individual assumption is 'substantially unreasonable,' then the aggregate of the assumptions must be 'reasonable.' " Rhoades, McKee & Boer v. United States, 822 F. Supp. 445, 450 (W.D.Mich.1993). Applying this standard was error. The statute speaks only of whether the assumptions are reasonable in the aggregate.

The United States also argues that the district court improperly interpreted the statute as allowing the actuaries to incorporate a degree of actuarial "conservatism" into their assumptions. This position was considered and rejected by both the Fifth and Second Circuits. "Standard actuarial practice includes a basic tenet of conservatism." Vinson, 7 F.3d at 1239. Indeed, "generally, Sec. 412(c) is not violated when an actuary chooses an assumption that is within the range of reasonable assumptions, even when the assumption is at the conservative end of that range, provided the chosen assumption is the actuary's best estimate of anticipated plan experience." Wachtell, 26 F.3d at 296. Here, the district court recognized that

[i]n all situations in which estimates are required, planning for error is part of the calculation .... the intent of Congress in the field of pension funding was for actuaries to err on the side of over-funding rather than underfunding. Therefore, a degree of conservatism in assumptions is warranted in defined benefit plans.

Rhoades, 822 F. Supp. at 454. Further, the best estimate test is procedural only, and does not place a second substantive hurdle in the path of actuarial assumptions. Vinson, 7 F.3d at 1238; Wachtell, 26 F.3d at 296. Rather, it is "principally designed to insure that the chosen assumptions actually represent the actuary's own judgment rather than the dictates of plan administrators or sponsors." Wachtell, 26 F.3d at 296; see also Vinson, 7 F.3d at 1238. Section 412(c) allows actuaries to incorporate a degree of conservatism in their estimates. We agree with the Fifth and Second Circuits on this issue and therefore affirm the district court in this regard. Having resolved these two overarching issues, we now consider the district court's decision regarding the assumptions on retirement age and mortality tables.

We find the assumptions on retirement age and the use of the 1983 IAM (female) mortality table in 1986 were objectively unreasonable. In discussing the reasonableness of the actuaries' assumptions, the district court stated that "the reasonableness of assumptions is objective." Rhoades, 822 F. Supp. at 451. However, the court then applied a subjective standard in determining that the retirement age assumption was reasonable. The actuaries assumed that Rhoades would retire when he reached sixty. Even if this assumption was initially reasonable, a 1987 amendment to the plan made it unreasonable. The amendment increased the retirement benefit, but, because of the Tax Reform Act of 1986, Rhoades could not receive this increased benefit if he retired at sixty. Thus, objectively, Rhoades could not receive the increased benefits unless he worked until he was sixty-three. The district court relied on the actuaries' mistaken belief that the benefits would still be payable at age sixty. It found that "given the premise they were operating from, their conclusions were not unreasonable." Id., at 452. The court's reasoning subjectifies an objective reasonableness standard. Here, the retirement age assumption is objectively unreasonable after the 1987 amendment to the plan.

In reviewing the mortality table used for 1986, the district court found that the difference between using a male and a female table was approximately four years. The court then rejected the Commissioner's argument that this discrepancy was outside the range of reasonableness and concluded that the mortality assumptions used were reasonable. When considering whether this mortality table also represented the actuary's best estimate for that year, the court noted that

the use of a female table for a male beneficiary may make sense when the payout form is joint and survivor 100% annuity, because after Mr. Rhoades' death, Mrs. Rhoades would continue to receive full benefits if she were living.... However, there was no testimony that this is what [the actuary] had in mind when he switched to the female table.

Rhoades, 822 F. Supp. at 456. The actuary for 1986 testified that typically a female table would be used for a male in excellent health, but he had never discussed Rhoades' health. The actuary for 1987 was "uncomfortable" using the female table, so he switched to the parallel male table with a two year set-back.

The district court reasoned that "[t]he only explanation for [the 1986 actuary's] use of a female mortality table seems to be an effort to increase contribution levels." Id. The court concluded that this table switching was motivated by improper considerations and could not be countenanced as a "best estimate." Nevertheless, the court determined that "the assumptions in combination represented the actuaries' best estimate of anticipated experience under the plan." Id. at 457. From this, it is apparent that the district court aggregated the assumptions across all three of the challenged years. The decision to thus aggregate the assumptions was error. From the statute, it appears Congress intended that only the assumptions for a given year should be aggregated. See 26 U.S.C. Sec. 404(a)(1) (referring to "limitations as to the amounts deductible in any year" and stating that the actuarial assumptions used "shall be those used for such year under section 412"). The contributions made by Rhoades and Rhoades, McKee were calculated and deducted on a yearly basis in 1986, 1987, and 1988. Thus, the assumptions underlying those contributions should also be measured on a yearly basis. The district court's failure to do so was error.

For the foregoing reasons, we AFFIRM the district court's acceptance of a degree of actuarial conservatism with respect to the "best estimate" standard; REVERSE the district court with respect to the "substantially unreasonable" standard; REVERSE as to the retirement age assumption, and REMAND for a determination, considering each year separately, of whether the assumptions, in the aggregate, were reasonable with respect to 1987 and 1988; and AFFIRM the district court's conclusion that the use of the female mortality table in 1986 did not represent the actuary's best estimate. We do, however, REMAND this issue for a determination whether the 1986 assumptions were reasonable in the aggregate in light of the problems with the mortality table and any potential problems with the retirement age.