concurring in part and dissenting in part.
The court holds that the Plan Administrator did not abuse its discretion in its calculations of lump sum benefit payments between August 1, 1994, and August 1, 1996. The court further concludes that, as of August 1, 1996, the actuarial table the Administrator was using was inapplicable and the Administrator’s continued reliance on it was in conflict with Plan language. I agree that the Administrator abused its discretion as of August 1, 1996, and therefore I concur in part B. of the court’s opinion. Because I conclude that the 1993 regulatory amendments applied to the Plan as of August 1, 1994, and the Administrator’s failure to apply the changes was an abuse of discretion, I dissent from part A. of the court’s opinion.
Plain language and simple logic reveal that the 1993 regulatory amendments are applicable to this case and that plaintiffs are entitled to the one year set-back in calculating their benefits. Part 2619 of the Pension Benefit Guaranty Corporation regulations governed the valuation of plan benefits in single-employer plans. Sub-part C, entitled “Trusteed Plans,” was largely reshaped by the 1993 amendments. It began with the following section:
2619.41 Purpose and scope.
This subpart sets forth the rules for calculating the value of a benefit to be paid a participant or beneficiary under a terminating pension plan that has been or will be placed into trusteeship by the PBGC. Sections 2619.43 — 2619.48 set forth rules for calculating the value of a benefit to be paid as a lump sum. Section 2619.49 sets forth rules for calculating benefits to be paid as an annuity.
29 C.F.R. § 2619.41 (1994). Contrary to the text of Section 2619.41, however, the rules for calculating lump sum benefits were actually found in Section 2619.49. Sections 2619.43 through 2619.48 discussed mathematical symbols and terms (§ 2619.43), immediate annuities (§ 2619.44), deferred annuities (§ 2619.45), early retirement benefits (§ 2619.46), death benefits (§ 2619.47), and withdrawal of employee contributions (§ 2619.48). Section 2619.49 was entitled “Valuation of benefits to be paid as annuities,” but it contained the only reference to mortality tables for lump sum payments:
(a) General rule....
For valuing benefits to be paid as lump sums, and for determining whether the lump sum value of a benefit exceeds $3,500, the plan administrator shall value benefits in the same manner as benefits to be paid as annuities except that the mortality assumptions set forth in Tables I through VI of appendix A to this part ... shall apply....
29 C.F.R. § 2619.49. Table I of Appendix A, the table the Administrator had always used to calculate lump sum payments, thus continued to be applicable because it was the only one published by the Pension Benefit Guaranty Corporation for healthy male lives and the Plan, of course, specified assumptions for healthy male lives. Table I retained its title as “Mortality Rates for Healthy Male Participants,” and in fact it was exactly the same table as was in use before the 1993 amendments. Compare 29 C.F.R. pt. 2619, app. A, Table I (1992) with 29 C.F.R. pt. 2619, app. A, Table I (1994). The only difference between the pre-1993 table and the one *911adopted in 1993 is that lump sum payments gained a one year set-back in the new version through amended language in the preamble to the Mortality Rate Tables:
For determining whether the value of a benefit is $3,500 or less under § 2613.8(b)(1) and for calculating the value of a lump sum benefit, the PBGC will employ the mortality assumptions set out in Tables I through VI of this Appendix by substituting x +1 for x throughout the appropriate male mortality table. The Tables provide, for each age (denoted by x), the probability that a person assumed to be living at that age will not survive to attain age x +1.
29 C.F.R. pt. 2619, app. A, Table I (1994). As a result, Table I now assumed individuals would live one year longer than before, thereby increasing lump sum payments.
The 1993 regulatory amendments not only added the one year set-back in calculating benefits, but they also directed that Table I (though titled “Healthy Male Participants”) be the source of unisex calculations for the payment of lump sum benefits. 58 Fed.Reg. 50,812 (Sept. 28, 1993). The Pension Benefit Guaranty Corporation stated the following in the Background section of the announcement of the final rule changes:
[U]nder the proposed regulation, the PBGC would continue to value lump sums under trusteed plans using its historical assumptions, except that its mortality assumption would be from a unisex table....
58 Fed.Reg. 50,812 (Sept. 28, 1993). The announcement recounted comments that had been made following the publication of the proposed rule changes and repeated this conclusion:
Two commenters addressed the desirability of adopting unisex mortality assumptions for lump sum valuations in PBGC-trusteed plans; both favored the proposed change. The final rule implements the change published in the proposed rule.
58 Fed.Reg. 50,814 (Sept. 28, 1993). This explains the directive in the preamble to the Mortality Rate Tables that calculations for benefits for “a person” would be based on “the appropriate male mortality table.” 3
To summarize, the 1993 amendments left Table I in place but added a one year set-back to the calculation and made the table applicable to both men and women. Although the Plan continued to use Table I, it did not follow the regulatory change that added one year to the mortality assumptions in calculating benefits. The Plan has never claimed regulatory ambiguity as a reason for its refusal to add the one year set-back. However, the court now holds the Administrator’s conduct was reasonable because the “administrative pronouncements as a whole” are ambiguous. The record reveals that the Plan asserted a much more simplistic explanation during the administrative appeal process that documents the Administrator’s explanations of its conduct.
The Blandin Plan directs the Administrator to calculate lump sum benefits using *912the Pension Benefit Guaranty Corporation assumptions that are applicable for healthy male lives for terminations of trusteed single employer plans with insufficient funds. Citing that language, the Participants made a claim for additional benefits on the ground that the “applicable” assumptions changed with the adoption of the one year set-back effective August 1, 1994. In its letter denying the Participants’ claim for additional benefits, the Administrator acknowledged that the Pension Benefit Guaranty Corporation had changed its assumptions so that the “Healthy Male Table” no longer existed, but stated that it had exercised its discretion to continue using the old Table I because the Plan contained a reference to “healthy male participants.” In this 2003 letter, the Administrator did not distinguish between the regulatory amendments of 1993 and 1996. The Administrator also wrote that “an additional indication of the intent of the Plan language” was found in a labor contract between the Company and some of its unionized employees because it contained a lump sum example that used the original Table I.
The Participants appealed that denial, and the Administrator once again denied the claim. This time, the Administrator specifically acknowledged that the original Table I entitled “Mortality Rates for Healthy Male Participants” no longer existed and that it “became necessary to interpret the [Plan language] to supply a mortality table.” Without explaining how or why it had arrived at its interpretation, the Administrator wrote that its “continued use of the same historical mortality table” had resulted in the plan having been “administered uniformly as to all participants, in all years.” The Plan asserts that its decision provided consistent benefits over time between participants, but the Plan calls for the Administrator to use “applicable” assumptions rather than the same assumptions.
The letter offered no new or repeated basis for its denial, but it contained a curious statement that “the PBGC no longer covers terminations of trusteed single employer plans with insufficient assets; rather, the PBGC covers pension plan terminations on either a voluntary termination basis or a distress termination basis.” That statement is simply incorrect. The agency still covers the termination of trusteed single employer plans with insufficient assets. See 29 C.F.R. § 4044.1(b) (valuation of benefits necessary to determine amount of plan asset insufficiency); § 4044.2(b) (“Trusteed plan means a single-employer plan which has been placed into trusteeship by PBGC.”). In addition, the statute had spoken of voluntary and distress terminations since 1986, so no new terminology was introduced that would have justified the Administrator’s position. See Single-Employer Pension Plan Amendments Act of 1986, Pub.L. No. 99-272, § 11007, 1986 U.S.C.C.A.N. (100 Stat.) 244. Although the record does not supply this detail, it is possible that this erroneous understanding of the regulations contributed to the Administrator’s denial of the claim.
The administrative record leads me to conclude that the Administrator’s proffered reason(s) for rejecting the Participants’ claim had no legal basis. The Administrator did not assert during the claim appeal process that its decision was rooted in regulatory interpretation that the 1993 amendments affected only the de minimis function. The Administrator now relies on that as its only argument in favor of its decision being consistent with the Plan language. I have addressed the argument because it is a question of law, but I note the complete discrepancy between the Administrator’s proffered reasons for denial *913in the claims process and in this lawsuit. I am particularly unimpressed by the Administrator’s earlier argument that it was justified in continuing to rely on Table I because the table’s title contained the phrase “healthy male lives” and the Plan used the words “healthy” and “male,” where the regulations made it clear that Table 3 (the applicable table as of 1996) applied to both males and females regardless of its title. The Administrator’s interpretation was contrary to the clear language of the Plan, see Lickteig v. Bus. Men’s Assurance Co. of Am., 61 F.3d 579, 584-85 (8th Cir.1995), and was not a result of ambiguous regulatory language.
The court cites but does not apply the analysis of Finley v. Special Agents Mutual Benefit Association, Inc., 957 F.2d 617, 621 (8th Cir.1992), in its discussion of the reasonableness of the Administrator’s conduct. Our review is limited to the record that the Administrator considered when it denied the Participants’ claims for additional benefits, and we are not to substitute our weighing of the evidence for that of the Administrator. Farley v. Arkansas Blue Cross & Blue Shield, 147 F.3d 774, 777 (8th Cir.1998). We are also guided by a corollary directive that we are not to consider post hoc rationales for the Administrator’s determination. Short v. Cent. States, S.E. & S.W. Areas Pension Fund, 729 F.2d 567, 575 (8th Cir.1984).
I have already answered the ultimate question, whether the Administrator’s interpretation is contrary to the clear language of the Plan. See Lickteig, 61 F.3d at 584-85. There are other parts to the Finley test. Id. at 584. The first asks whether the interpretation was consistent with the goals of the Plan, and I conclude that neither side receives a benefit from this question. The Participants suggest that the Plan’s tax qualification is in jeopardy without evidence to support the assertion.
The second part of the Finley test addresses whether the interpretation renders any language in the Plan meaningless or internally inconsistent. In this case, this question is redundant with the one I have already decided: the interpretation rendered meaningless the word “applicable” because the Administrator chose to continue using an outdated and later nonexistent table.
The third part of the Finley test asks if there is a conflict with the substantive and procedural requirements of ERISA. The Participants argue that the Plan violated ERISA’s requirement that a plan document be clear enough to determine rights under it, while the Plan argues that no ERISA provision is at issue. ERISA contains a number of precatory words that arguably could apply, but the primary issue here is not one of compliance with ERISA. The Administrator’s decision does not conflict with the requirements of the ERISA statute in a manner or to a degree that affects the outcome of this case.
The last part of the test is whether the words at issue have been interpreted consistently. As noted above, the Plan argues that it has applied the Plan consistently by ignoring the changes to the tables. That is not the sort of consistency that the law requires, however. The issue is what table was applicable to participants at the time of their retirement, and that answer changed over time. Accordingly, the Administrator has not interpreted the relevant words consistently.
To summarize, I conclude that the Administrator’s interpretation of which mortality table was “applicable” to the calculation of lump sum benefits as of August 1, 1994, was contrary to the clear language of the Plan and rendered that same language meaningless. Moreover, the Administrator has failed to consistently interpret the *914Plan’s provision so that participants who have chosen lump sum benefits between 1994 and 1996 have been shortchanged. Taken together, these shortcomings render the actions unreasonable.
Because I conclude that the Blandin Paper Company Employees’ Retirement Plan, through its Administrator, abused its discretion by failing to calculate lump sum benefits correctly as of August 1, 1994, I would reverse the district court judgment.
I concur in Part B. of the court’s opinion and dissent from Part A.
. The Plan adopts the assumptions for healthy male participants, thereby distinguishing from assumptions for disabled male participants. The preamble's use of the phrase "appropriate male mortality table” recognizes that there were two kinds of tables.
The Pension Benefit Guaranty Corporation continued to publish analogous tables for female participants because those tables were still used in calculating benefits paid as annuities. The female tables were not to be used for lump sum payments because the preamble to the Mortality Rate Tables directed that lump sum payments only would be determined through use of a unisex table.