dissenting.
Introduction
Plaintiffs-Appellants are management employees of Conectiv who worked for Delmarva Power & Light or Atlantic City Electric Company. They were participants in “average highest pay” defined benefit pension plans, meaning plans that calculated benefits based on an average of highest pay levels achieved by participants over their careers. Conectiv merged the plans and in 1999 converted the merged plan to a cash balance plan. Plaintiffs sued in 2005, alleging, among other things, that the amendment violated ERISA’s requirement that plan participants receive notice of certain plan amendments. 29 U.S.C. § 1054(h). My colleagues in the majority believe no notice was required. I do. Moreover, I believe the notice given was both untimely and inadequate. I thus respectfully dissent.
*454The Employee Retirement Income Security Act of 1974 (“ERISA”) requires that pension plan sponsors give notice to plan participants before amending their plans in ways that should be expected to reduce significantly then' benefits. Since the level of pension benefits depends in part on wages, whether a significant reduction is expected depends on guesses about future wage levels. The majority’s opinion relies on the belief that because we are not clairvoyant, we must assume that things will stay the same. Because of that belief (which, I assert, is incorrect), the majority prefers an assumption of zero wage growth over one of positive wage growth in comparing the old and new plans. Doing so strays from textbook pension valuation practice and defeats the purpose of comparing future benefits under the two plans. Zero wage growth is not a reasonable expectation and Conectiv makes no allegations to challenge Plaintiffs’ showing that, assuming 4.5% wage growth (the figure used by Conectiv itself in a Form 10-K filing with the Securities and Exchange Commission), a “significant reduction” in the rate of benefit accrual under the new plan should reasonably have been expected.
Not only do I believe notice was required to Plaintiffs, I believe that notice was untimely because the amendment was not adopted before its effective date. The bullet summary “adopted” by Conectiv pri- or to the amendment’s effective date lacked essential details, such as pay and transition credit rates. I also believe that the notice was inadequate because it did not tell participants that they might reasonably expect their benefits to be reduced by the amendment.
Discussion
Reading together 29 U.S.C. § 1054(h) and the relevant regulations as they appeared during the conversion to the cash balance plan, notice is: (1) necessary if a significant reduction in benefits could be reasonably expected to result from the amendment;9 (2) timely if it comes after adoption but more than 15 days before the effective date of the amendment;10 and (3) adequate if it contains the effective date and a summary of the amendment written in a manner calculated to be understood by the average plan participant.11
Significant Reduction
Whether a significant reduction reasonably may be expected turns on the wage growth assumption used in estimating future benefit accrual. Plaintiffs use a 4.5% annual growth rate to show a reduction in the rate of future benefit accrual. Defendants use a 0% wage growth rate to show *455an increase in the rate of future benefit accrual. Compare JA 1797 with JA 0630.
Both the majority and the District Court hold that zero wage growth is the proper assumption. Their position contradicts textbook pension valuation practice and judicial opinions, makes meaningful comparison of plans impossible, and ignores Conectiv counsel’s concession at oral argument that a positive wage growth assumption is necessary. We should hold that a 4.5% wage growth rate assumption was reasonable because Conectiv used this rate in projections of growth in its aggregate labor cost.
The majority’s holding rests on separate grounds from that of the District Court. The latter relied on the statute’s backload-ing provision, which states that “social security benefits and all other relevant factors used to compute benefits shall be treated as remaining constant as of the current year for all years after the current year.” 29 U.S.C. § 1054(b)(l)(B)(iv); Charles v. Pepco Holdings, Inc., 513 F.Supp.2d 47, 54 (D.Del.2007).12 The majority relies on Conectiv’s argument13 on appeal that, since the future is not knowable, it is reasonable to expect that things will stay the same. As a result, zero wage growth is a reasonable assumption in predicting rates of benefit accrual before and after an amendment. Both grounds are flawed.
Addressing the second ground first, if we assume that the future is unknowable, then in theory we must conclude that anything is possible. There is no reason why we should privilege the current wage rate over any other rate.14
Although the future is not knowable with certainty, many physical, social, and economic phenomena, including wage growth, behave in ways that can be predicted with various levels of accuracy.15 Actuaries use data concerning “relevant facts and circumstances” available in the present to produce cost-of-living predictors, mortality tables and salary scales that represent best estimates of future values of pension valuation variables, including wages. See Dan M. McGill et al., *456Fundamentals of Private Pensions 609 (8th ed. 2005) (“McGill”); cf. Maj. Op. 452 (stating that “importantly, ... determination [of future benefit rates] must take into account the relevant facts and circumstances at the time the amendment is adopted ”)(internal quotations omitted)(emphasis in original). Since there is no reason a priori why such estimates for wage growth should be zero, Conectiv must provide some empirical justification for why a zero wage growth assumption is reasonable.
In fact, a zero wage growth assumption is generally unreasonable because wages typically grow over time. A leading pensions textbook states that “[c]ost-of-living increase assumptions in current use generally range from 2 percent to 3 percent” and that “[a]ctuaries usually assume that increases in average compensation will exceed the assumed increase in prices by 0 to 2 percent.” McGill at 609-10.16 That means that in general actuaries assume wage growth of between 2% and 5%, not 0%.
Ignoring wage growth affects the benefit accrual rate under highest pay defined benefit plans, such as those being compared with the cash balance plan in this case. The two highest pay plans at issue here calculate benefits as an increasing function of average pay over the five consecutive years of highest pay.17 As a result, valuing benefits without taking into account future pay increases reduces estimates of future benefits.
Textbook pension valuations of highest pay plans take the sensitivity of average pay plans to wage levels into account by assuming wage growth. In the section on “Valuation Assumptions and Factors” for plans, McGill states that
if the retirement benefits of a pension plan are expressed in terms of final average or highest average earnings, it is necessary to project the current earnings of plan participants to the level expected to prevail during the period of service that will establish the formula salary base — usually the years immediately preceding retirement.... Compensation for an individual participant ... is projected through an assumption about the annual rate at which the individual’s compensation will increase over his or her future working lifetime.
McGill at 609 (emphasis added). This treatise cites to an article stating that
[t]he salary scale is of prime importance in the calculation of liabilities for pension plans provided for salaried employees. Since the compensation of these employees [ie., employees in average earnings plans, as in this case] is customarily increased year by year to match increasing responsibilities, it follows that the pension liabilities are similarly increased, and, in fact, that a proper estimate of the liabilities for the *457pensions cannot be made without an attempt to forecast the salary increment.
William F. Marples, Salary Scales, 14 Trans. Soc. Actuaries no. 38, pt. 1 at 1-2 (1962), cited in McGill at 609 n. 12.
The sensitivity of highest pay plans to wage growth assumptions is so significant that, if our Court were to adopt a zero-wage-growth rule, it would permit employers to carve an entire class of plan amendments out of the notice requirement. Any reduction in benefits that could be accomplished by altering the wage parameter used in the benefit formula would be “invisible” to a court in testing for significant reductions in future benefits. For example, an employer could reduce benefits by amending a final pay plan to become a lowest pay plan by changing the benefit equation to use a participant’s lowest pay over his career as the base for computing pension benefits. Assuming lowest pay of $40,000 and final pay of $70,000, such a change would reduce annual pension benefits at retirement by 43%. Cf. Amara v. Cigna Corp., 534 F.Supp.2d 288, 338 (D.Conn.2008) (“switching from a final or highest average formula to a career average pay formula itself is almost guaranteed to cause a substantial reduction in future benefits, and in fact such a result is commonsensical”). However, a court following the majority’s approach would hold that no reduction in benefits can reasonably be expected under the amendment. Cf. Appellant’s Reply Br. 5-6 (showing how an employer who reduced benefits by 5.3% by changing the wage formula used in calculating benefits under a final pay plan would appear under a zero wage growth rule to have not reduced benefits at all).
Courts have acknowledged both the reality of wage growth and the importance of taking it into account in valuing benefits in traditional defined benefit plans. Hirt v. Equitable Retirement Plan, 441 F.Supp.2d 516, 520 (S.D.N.Y.2006), points out that “[t]he principal feature .of ... defined benefit plans was to weight the compensation earned by participants in their late, pre-retirement years, for these, generally, were the years of highest salaries or commissions.” In Amara, the Court relied in part on an expert report provided by the plan sponsor that assumed 4.5% annual salary increases in holding that an amendment converting a three-year final average pay plan into a cash balance plan was expected to reduce benefits significantly. Amara, 534 F.Supp.2d at 338.
Plaintiffs estimate benefits using the future growth rate in aggregate labor costs assumed by Conectiv in its 1998 and 1999 financial statements. Appellant’s Br. 32; JA 1791. This is reasonable because: (1) the growth rate in aggregate labor costs takes into account the growth of wages of non-union management employees; and (2) future rate projections used by the plan sponsor itself carry weight because the sponsor’s access to data and actuarial expertise makes it best situated to develop accurate estimates of future growth.
Conectiv challenges Plaintiffs’ growth assumption on the ground that aggregate labor cost growth reflects growth in union wages.18 But despite being the party in *458the best position to identify that evidence, Conectiv does not allege that non-senior management employees such as Plaintiffs in fact experience wage growth significantly different from growth in aggregate labor costs. See Appellee’s Br. 30 (stating only that Conectiv used 4.5% wage growth to “project pension funding obligations for its entire workforce”). Absent such a showing, aggregate wage growth is a reasonable predictor of Plaintiffs’ wage growth.
Conectiv argues that “[njowhere ... do Plaintiffs explain how, in 1998, Conectiv was supposed to be able to predict Plaintiffs’ salaries over the next seven years with pinpoint accuracy....” Appellees’ Br. 31. To predict a Plaintiffs wage growth rate with “pinpoint accuracy,” as Conectiv would have the Court require, is not prediction, but clairvoyance. Every prediction is a generalization about the fate of a class or group of similarly situated individuals. Since we do not know the future for certain, we cannot precisely identify how each individual fares. The better the prediction, the smaller the class to which our prediction applies and the closer we come to differentiating fully each individual’s fate. But no prediction succeeds completely. Cf. Hirt, 441 F.Supp.2d at 541 (stating that “[uncertainty in determining expected benefits is inevitable, due to fluctuations in salary levels, [etc.] ....”); see generally Nicholas Rescher, Predicting the Future: An Introduction to the Theory of Forecasting (1997).
The majority suggests that a zero wage rate is a reasonable expectation because “in the preceding five-year period from 1994 to 1999, one Plaintiff saw nearly zero wage increase, from $56,237.84 to $56,824.37, and another received average wage increases of only 1.4% per annum, from $59,267.75 to $63,522.62.” Maj. Op. 453. Extrapolating future wage growth rates for four Plaintiffs over the roughly 15 or more years remaining until they reach retirement age by examining five years of wage growth history for two of them will produce very weak predictions. A small sample size such as the one used by the majority can be highly inaccurate because, among other reasons, wage rates sometimes plateau temporarily as workers move from one pay grade to another. See, e.g., William F. Marples, Salary Scales, 14 Trans. Soc. Actuaries no. 38AB, pt. I at 2 (1962) (“In the larger employments, for consistent and equitable control of salary increments, it has been the practice to establish grades within which progression from the minimum to the maximum salary for the grade occurs in stipulated increments related to the period of service within the grade. The effect on the salary history of the individual employee is a steady increase in compensation with age, with occasional ‘flats’ while a move from one grade to the next is awaited.”).
The majority does not explain why it disregards data for two other Plaintiffs. Those Plaintiffs, Maurice Ward and Thomas Troup, experienced large increases in salary over the same period. Ward’s salary jumped from $58,872 to $69,132 and Fink’s from $53,092 to $66,034, increases of approximately 3.3% and 4.5% per year, respectively. JA 0647 (listing yearly wages for each such Plaintiff from 1994 through 1999 on which annual percentage increase calculations are based). The effect of the majority’s omission of this data is to bias its conclusions in favor of low wage growth.
Unlike the majority, the District Court did not engage in reasonable expectations analysis because it thought that a zero wage growth assumption was mandated by text in the backloading statute. That conclusion is wrong. There is no support for a zero wage growth assumption in the statute or implementing regulations. Assumptions required in testing for backload-*459ing are not the same as those required in calculating rates of future benefit accrual. ERISA’s anti-backloading test seeks to determine whether, all else equal, benefit increases of an unacceptable degree are sensitive to increases in age alone. See Rev. Rul.2008-7, 2008 WL 274325. It therefore holds wage rates (along with everything else but age) constant and compares benefits levels at different ages.
The goal in testing for a significant reduction in future benefits is not to isolate the influence of a single variable such as age, but to compare “reasonable expectations” about how much money plan participants will get in the future under two plans. This requires comparing the behavior of two formulas (average highest pay and cash balance) under reasonable assumptions about changes in the variables (e.g., compensation and mortality) on which they are based. This explains why Congress placed language holding variables constant in a subsection of the backloading provision, but not in any part of the notice provision. Compare 29 U.S.C. § 1054(b)(l)(B)(iv) with 29 U.S.C. § 1054(h).
At oral argument, Conectiv’s counsel conceded that a zero wage growth assumption is inappropriate and neither suggested an alternative nor challenged Plaintiffs’ growth rate assumption.19 Because the 4.5% growth rate used by Plaintiffs was reasonable and the only alternative growth assumption proffered by Conectiv was not, the District Court should, I believe, have granted summary judgment as a matter of law on this issue in favor of Plaintiffs rather than Conectiv.20
If the majority is uncertain whether Plaintiffs’ 4.5% assumption is reasonable because there may be a significant divergence between union and non-union wage growth, as Conectiv suggests, our only option, in view of the inappropriateness of a zero growth assumption, is to remand to the District Court for further findings of *460fact on the appropriate wage growth assumption. Record evidence that should be required on remand includes: (1) opinions of economists and pension actuaries concerning the proper wage growth assumption; and (2) any information, including reports and projections, developed by Towers Perrin and Watson Wyatt for Co-nectiv in connection with the cash balance plan conversion.
With respect to the latter, the District Court quashed subpoenas of Towers Perrin and Watson Wyatt after Co-nectiv objected that such information was irrelevant. Conectiv argued that the “reasonable expectations” standard is objective and as such does not turn on Conectiv’s subjective expectations or state of mind. Br. Supp. Pis.’ Mot. for Continuance Pursuant to Fed.R.CivP. 56(f) or Preclusion Order at 5-7, Charles, No. 05-702, 2007 WL 4426074 (D.Del. Sept. 4, 2007). The District Court should have permitted discovery of projections and other materials relating to future benefits prepared by professionals because these are relevant to establishing “reasonable expectations,” even under an objective standard. Cf. George v. Duke Energy Retirement Cash Balance Plan, 560 F.Supp.2d 444, 482 (D.S.C.2008) (reciting descriptions of analyses by consultant discovered by plaintiffs that reflected future benefits reductions and which plaintiffs considered relevant to “significant reduction” analysis); Amara, 534 F.Supp.2d at 300, 306 (citing to materials used by outside consultants that planned conversion to cash balance plan, although not in the context of “reasonable expectations”).
Given its access to relevant data, resources, institutional experience, and incentive to forecast its own future liabilities accurately, a plan sponsor, as noted, is often in the best position to predict future benefits at the time an amendment is adopted. That makes its expectations relevant to an objective inquiry into reasonable expectations. The irrelevance of subjective expectations means only that a sponsor’s materials need not be the final word in any reasonable expectations inquiry.
Using the 4.5% growth assumption, Plaintiffs calculate that the amendment could be reasonably expected to cause reductions in their monthly benefits at retirement age ranging from 14% to 37%. JA 1797; JA 0975.21 Since Plaintiffs’ growth assumption was reasonable and benefits under that assumption decreased under the amendment, I believe the majority should have examined whether the decreases were “significant.”
Congress has provided no guidance in this context on the meaning of “significant.” Courts have not articulated a rule, though they appear to know a significant reduction when they see it. See Davidson v. Canteen, 957 F.2d 1404, 1407 (7th Cir.1992) (reduction in benefits of almost $13,000 and over $17,000 per year for life significant); Amara, 534 F.Supp.2d at 338 (reduction significant where some participants experienced reductions of 29% or 45% under different interest rate assumptions); Koenig v. Intercontinental Life, 880 F.Supp. 372, 376 (E.D.Pa.1995) (reduction in projected benefits of between 22% and 32% significant); Copeland v. Geddes Fed. Sav. & Loan Ass’n Ret. Income Plan, 62 F.Supp.2d 673, 677 & n. 5 (N.D.N.Y. 1999) (reduction from $458.34 per month until age 65 and $5.00 per month thereafter to $266.87 per month until age 65 and nothing thereafter is significant).
At oral argument, Conectiv claimed that “if the fact of a decrease [in relative bene*461fits] at all depends upon your pay increase assumption, then that is not a significant reduction.” There is no support for such a rule in the statutes or regulations. Given the sensitivity of highest pay plans to wage growth assumptions, such a rule would permit large reductions in benefits to count as insignificant. As discussed above, it would effectively exempt an entire class of plan amendments from the notice requirement.
Rather than choose that rule, courts should anchor their analyses by reference to the purpose of the “significant reduction” standard, which is to trigger notice. They should ask whether the reduction is such that a plan participant would want notice of the reduction. A reduction of pennies per month is not significant. A reduction of $100 per month is likely significant for participants with small monthly benefits. In close cases, the opinion of experts concerning the economic burden of the reduction will be relevant. But this is not a close case. Reductions of between 14% and 37% are significant. As in the decisions noted above, I know a significant reduction when I see one. Summary judgment in favor of Plaintiffs, or at least a remand, is warranted on this issue.
Timeliness
Notice must be timely. It is so if it comes after adoption but more than 15 days before the effective date of the amendment. 29 U.S.C. § 1054(h) (Notice is required “after adoption of the plan amendment and not less than 15 days before the effective date of the plan amendment. ...”).
A committee of Conectiv’s board “adopted” the cash balance plan “in substantially the same form presented in the attachment” on April 28, 1998, more than 15 days before the January 1, 1999, effective date. But the attachment does not contain the final version of the cash balance plan. Rather, it contains a bullet summary that does not contain specific pay, interest, and transition credit percentages, among other terms necessary to calculate benefits under the new plan.22 There is no evidence in the record that the final version of the amendment existed at the time the plan was adopted. Cf JA 0788-89 (employee who worked on the plan conversion stating that he assumes that the document reached its final form in December 1999, nearly a year after Conec-tiv made effective the cash balance plan).
The statute’s requirement that adoption precede notice and the amendment’s effective date ensures that participants have a fixed target to which they can respond. See Curtiss-Wright v. Schoonejongen, 514 U.S. 73, 82, 115 S.Ct. 1223, 131 L.Ed.2d 94 (1995) (“Requiring every plan to have a coherent amendment procedure ... increases the likelihood that proposed plan amendments, which are fairly serious events, are recognized as such and given the special consideration they deserve.”). Adoption reduces oversight costs for participants because it provides them with a signal for when heightened vigilance is warranted and authoritative information regarding what management intends to do.
The bullet summary adopted on April 23, 1998, does not constitute adoption of *462the amendment for purposes of satisfying the notice requirement because the bullet summary was a moving target. It did not fix interest credit, pay credit, or transition credit percentages that a participant would need to determine whether the amendment would reduce benefits. Cf Normann v. Amphenol, 956 F.Supp. 158, 165 (N.D.N.Y.1997) (rejecting notice because it did not mention specific numbers adopted in the amendment). It also did not indicate whether union employees would be covered by the plan, a fact that would have been useful to non-union employees such as Plaintiffs in determining whether the amendment was good for them.
It does not make a difference that the Facts newsletter and Decision Kit mailed to employees after April 28, 1998, but before the effective date, contained specific pay, interest, and transition credit percentages. See JA 0257; JA 1330-34. Since the board did not adopt the numbers in these notices in April, 1998, and the notices themselves purport to describe rather than to be the plan amendment, they provide an ambiguous and incomplete signal of management’s intentions rather than the fixed target adoption requires.
Conectiv argues that adoption of amendments with general terms is permissible. However, in none of the cases it cites is specificity of terms at issue. See Appel-lee’s Br. 37-38. At oral argument, Conec-tiv’s counsel noted that in Haigler v. CIGNA Corp., No. . 8:04-CV-851-T-30, 2006 WL 196936 (M.D.Fla. Jan. 24, 2006), the Court approved adoption of an amendment containing a single line of text. Unlike in this case, that single line contained the specific percentage necessary to gauge the amendment’s effect on the rate of benefit accrual. Id. at *4.
Certification of the actual amendment text on December 10, 1999 (over 11 months after the purported effective date) did not retroactively ratify adoption. Permitting such ratification would defeat the purpose of requiring adoption and notice before the amendment becomes effective. In addition, ratification does not apply because it is “prohibited where the amendment retroactively reduces the intervening rights of third parties, such as plan participants.” Depenbrock v. Cigna Corp., 389 F.3d 78, 83 (3d Cir.2004) (holding impermissible the ratification of cash balance plan conversion amendment because participant had right to receive benefits under pre-amendment plan); Schoonejongen v. Curtiss-Wright, 143 F.3d 120, 124-25 (3d Cir.1998).
As the amendment was not adopted on April 23, 1998, or apparently at any time before its effective date of January 1,1999, notice of its adoption could not have been provided more than 15 days before its effective date. As a result, I believe timely notice was not given. See, e.g., Copeland v. Geddes, 62 F.Supp.2d 673, 678 (N.D.N.Y.1999) (notice after effective date untimely).
Adequacy
Notice was also inadequate. It is adequate if it contains the effective date and a summary of the amendment written in a manner calculated to be understood by the average plan participant.23 The purpose of *463the notice requirement is to ensure that plan participants can defend them interests. See, e.g., Canteen, 957 F.2d at 1409. As a result, notice must cause participants to become aware of the interests that the amendment affects. See Schoonejongen, 514 U.S. at 83, 115 S.Ct. 1223 (“[0]ne of ERISA’s central goals is to enable plan beneficiaries to learn them rights and obligations at any time.”); Hirt, 441 F.Supp.2d at 538 (rejecting notice because “[i]n order to understand the reductions in benefits that would result from its application ... participants would have had to have performed sophisticated calculations and comparisons,” and stating that “[a] notice is intended to give fair warning and fails to do so if it is cryptic, or requires research beyond the document itself’). Notice that does not make clear to participants, either directly or through inferences that the average plan participant can be expected to draw, what interests the amendment affects is not adequate.
Notice must make plan participants aware that benefits can be expected to be reduced under the amendment. This is so because benefits are a participant’s principal interest with respect to a plan and notice is itself triggered when an amendment can be expected to reduce benefits significantly.
Only notice that explicitly tells participants that their benefits can be expected to be reduced is calculated to be understood by the average plan participant. Absent such a statement, promotional materials that discuss only advantages of the amendments and meetings convened to respond to cash balance plan criticism are not notice. Taking actions that might induce suspicious participants to seek expert help, through which they might then learn that benefit reductions are to be expected, is not notice. What is needed is a plain English statement both simple and direct that the cash balance plan will affect at least some of the participants by reducing their retirement benefits.
Several courts have held or implied that notice must contain some warning of expected benefit reductions. In Amara, the Court held that there was misrepresentation sufficient to violate the notice requirement where notices contained numerous statements such as “each dollar’s worth of credits is a dollar of retirement benefits payable to you after you are vested. Under the plan, your benefit will grow steadily throughout your career as credits are added to your account.” The Court noted the absence of any warning about future benefit reductions. 534 F.Supp.2d at 340; compare JA 0256-63, 0265-69, 1330-34 (in which Conectiv’s notices used similar language to that in Amara, such as: “Throughout your Conectiv career, your account grows through additional yearly company contributions, plus interest. When you retire, the current cash value of your account is yours.”); see also Hirt, 441 F.Supp.2d at 538. Indeed, here the District Court stated that “some discussion of the possible adverse affects of the controversial pension plan would have been in order.” Charles, 513 F.Supp.2d at 55 n. 10.
Our Court’s ruling in Register v. PNC Financial Services, 477 F.3d 56 (3d Cir.2007), does not preclude the requirement that notice must contain some warning about benefit reductions. Register held only that notice that did not do “more than tell the participants that the new plan ‘may affect the future rate of benefit accruals’[,] and ‘in some instances may reduce the rate of future Pension Plan benefit accruals, ’” was adequate. Register, 477 F.3d at 73 (emphasis added). While it rejected a requirement that plan sponsors “explain to ... participants that the conversion would ‘significantly reduce[ ] the rate of future *464pension plan benefit accruals for each plan participant,’ ” the Court did not state that notice omitting warning that an amendment “may reduce the rate of future ... benefit accruals” would pass muster. Id. at 73.
Conectiv’s notice was inadequate because it did not warn of potential benefit reductions. None of the seven notices24 identified by Conectiv contained a warning of potential benefit reduction such as that present in Register. See Appellee’s Br. 10-16. Conectiv might argue that The Wall Street Journal articles provided to some employees, and a July 1999 meeting convened to address criticism of cash balance plans in the press, were sufficient to notify Plaintiffs of the negative effect on them pensions. See Appellee’s Br. 15-17, 19. Convening meetings to allay concerns is not a statement that benefit reductions could be expected as a result of the amendment. Notice is not a vehicle for plan promotion or justification but rather for sober disclosure.25
Conclusion
Notice was required because zero wage growth is an untenable assumption and, absent evidence to the contrary, Conectiv’s own 4.5% aggregate labor cost growth projection is reasonable. Notice could not have been timely because the bullet summary of the amendment that was “adopted” before the effective date was insufficiently detailed and the actual amendment does not appear to have been adopted until after the effective date. In any event, notice was not adequate because it failed to state in plain English that the cash balance plan could affect at least some of the participants by reducing their *465retirement benefits. I respectfully dissent.
. A defined benefit plan “may not be amended so as to provide for a significant reduction in the rate of future benefit accrual, unless ... the plan administrator provides a written notice....” 29 U.S.C. § 1054(h). "Whether an amendment provides for a significant reduction in the rate of future benefit accrual ... is determined based on reasonable expectations taking into account the relevant facts and circumstances at the time the amendment is adopted." 26 C.F.R. § 1.411(d)-6. Q & A 7.
. Notice is required “after adoption of the plan amendment and not less than 15 days before the effective date of the plan amendment. ..." 29 U.S.C. § 1054(h).
.The plan administrator must provide "a written notice, setting forth the plan amendment and its effective date to ... each participant in the plan....” 29 U.S.C. § 1054(h)(1996). “The notice does not fail to comply with section [1054(h)] merely because the notice contains a summary of the amendment, rather than the text of the amendment, if the summary is written in a manner calculated to be understood by the average plan participant and contains the effective date. The summary need not explain how the individual benefit of each participant or alternative payee will be affected by the amendment.” 26 C.F.R. § 1.411(d) — 6. Q & A 10.
. This is a reconstruction of the District Court's brief comments on this issue. See Charles, 513 F.Supp.2d at 54-55. The Court first noted that plaintiffs assumed salary increases while Conectiv did not. It then stated that "[c]onsistent with its conclusions regarding ERISA § 204(b)(1)(B) [29 U.S.C. § 1054(b)(1)(B) ], the court concludes that defendant’s expert employed the more appropriate methodology to determine whether, at the effective date of the amendment (January 1999), Plaintiffs’ rate of future benefit accrual would be significantly reduced under the Cash Balance Plan.” Id. at 54. The Court stated that since Conectiv's calculations (using the zero wage growth assumption) had not themselves been challenged, there was “no probative evidence of record” to support Plaintiffs’ claim. Id. at 54-55. The response to this reasoning is addressed below.
. Conectiv states that predicting the effect of an amendment on the rate of future benefit accrual “can only be done on the basis of data existing at the time of the amendment’s adoption, as no Plan sponsor possesses the clairvoyance necessary to know the course of future events.” Appellee's Br. 29.
. The majority’s reasoning is a species of “status quo bias” in human cognition. See, e.g., William Samuelson, et al., Status Quo Bias in Decision Making, 1 J. Risk and Uncertainty 7 (1988).
. People often rely on predictions that the status quo will not continue into the future. When a baseball player times a fly ball, he relies on his predictions of changes in the rate of movement (speed) of the ball. Not being clairvoyant does not mean that it is reasonable for him to assume that once the ball leaves the bat it will continue out of the stadium at a constant speed forever. The player's experience with fly balls combined with the experience of generations of players tell him that there are far more reasonable predictions. The same is true for wage rates.
. The publisher, Oxford University Press, advertises that: “For almost five decades, [McGill] has been the most authoritative text and reference book on private pensions in the world.” Listing on http://www.us.oup.com/ us/catalog/general/subject/Finance/ lnvestments/?view=usa&ci=0199269505 (last visited Oct. 9, 2008).
. The Delmarva Power & Light Company Retirement Plan calculated benefits as an increasing function of average pay over the consecutive five-year period of highest compensation. JA 1155 (defining “Average Annual Earnings” for purposes of calculating the Normal Retirement Benefit of Non-Bargaining Unit Employees). The Atlantic City Electric Company Retirement Plan calculated benefits as an increasing function of average pay over the consecutive five-year period of highest compensation in the ten years preceding retirement. JA 1198 & 1204 (defining "Average Annual Compensation” for purposes of calculating the Normal Retirement Benefit).
. I would be remiss if I did not note that the unions involved here did not accept amending their highest pay plan to a cash balance plan. Defs.’ Answering Br. Opp'n Pis.’ Mot. Class Certification 3, Charles, No. 05-702 (D.Del. Feb. 16, 2007). Indeed, only non-union middle-management, estimated by Plaintiffs to number a few hundred, got stuck with the new plan. Opening Br. Supp. Pls.’ Mot. Class Certification 11, Charles, No. 05-702, 2007 WL 618244 (D.Del. Jan. 11, 2007); but see JA 0051 (cash balance plan covered certain management and unionized employees). Senior management is covered under the old plans or under an executive retirement plan. JA 1369-71.
. Judge Ambro: "It would be ludicrous not to assume that there is growth in earnings. That’s the way America works.”
Susan K. Hoffman, Counsel for Conectiv: "Judge Ambro, I will accept your premise that it is reasonable to assume some pay increase.”
. Notes 7 and 8 of the majority opinion assert that I miss its underlying point that, because wages are inherently unpredictable, using the current wage rate in projecting future benefits is reasonable. See Maj. Op. 453 nn. 7-8. This misses my critique of that position. My contention is that if wages are unpredictable, then neither the current wage rate nor any other is a reasonable predictor of future wages. To maintain that the current wage rate is the proper assumption in calculating future benefits, the majority must mean that wages are predictable. Specifically, they must believe that the current wage rate is a reasonable predictor of future wages. But that is an empirical claim in support of which Conectiv has not pointed to any evidence. As I explain in the body of this dissent, actuarial practice and the courts accept that wages exhibit a positive trend. The fact that median income fell from 2000 to 2008 does not undermine a positive trend assumption in 1998 or 1999. The drop is the “first time on record that income failed to set a new record in an economic expansion.” David Leonhardt, Next Victim of Turmoil May Be Your Salary, N.Y. Times, Oct. 14, 2008.
The majority further argues that I place the burden of proof on Conectiv. I do not. Plaintiffs have the burden of proof and have met it by pointing to 4.5% aggregate labor cost growth projected by Conectiv in its SEC filing. Although the majority may not find Conectiv's SEC filing persuasive (a question that is not at issue on summary judgment), they cannot claim, as they do, that "Plaintiffs have provided no basis in the record to establish the reasonableness of any alternative wage growth rate.” Maj. Op. 453 n. 7. The SEC filing is such a basis and Conectiv has produced no evidence to rebut it. Arguing that a divergence might exist between union and middle-management, non-union wage growth rates sufficient to skew aggregate labor cost growth does not count.
. Percentages provided here are lower than those in the record because they are recalculated as percentage reductions, i.e., using benefits under the old plan as base.
. The attachment contains the following text under the heading "Cash Balance Pension Plan”:
• Effective: 1/1/99
• Basic formula is: End of Year Balance = Beginning of Year Balance + Pay Credits + Interest Credits
• Portable — after 5 year vesting
• Improved employee communications
• Survivor receives the "balance”
• Extensive "Grandfathering”
• Transition Credits
• Business Link
• Useful in Divestitures
• Can be differentiated by SBU. JA 0826.
. The plan administrator must provide “a written notice, setting forth the plan amendment and its effective date to ... each participant in the plan....” 29 U.S.C. § 1054(h). "The notice does not fail to comply with section [1054(h) ] merely because the notice contains a summary of the amendment, rather than the text of the amendment, if the summary is written in a manner calculated to be understood by the average plan participant and contains the effective date. The summary need not explain how the individual benefit of each participant or alternative payee will be affected by the amendment.” 26 C.F.R. § 1.411 (d) — 6. Q & A 10.
. The notices were those of: October 13, 1997 (JA 0246); October 20, 1997 (JA 0252); May 1998 (two notices) (JA 1330-34, JA 0256-63); December 21, 1998 (JA 0265-69); June 23, 1999 (JA 0271-72); and July 1999 (JA 0274-80).
. Although the District Court did not reach the issue, Conectiv argues on appeal that Plaintiffs' claims are barred by the statute of limitations. Appellees' Br. 48-55. The question whether Plaintiffs complied with the statute of limitations is closely related to notice analysis because the statute of limitations only accrues when the employee learns or should have learned that the amendment brought about a "clear repudiation” of rights under the plan. Romero v. Allstate, 404 F.3d 212, 223 (3d Cir.2005) (holding that accrual should not be unwaveringly tied to effective date).
It is difficult, I submit, for the statute of limitations to accrue in benefit reduction cases without a direct statement from the plan sponsor that benefits are being reduced. This is so because average participants cannot be expected to become aware of reductions by inspection of old and new plan terms or benefit formulas. See Hirt, 441 F.Supp.2d at 538. Short of a direct statement from the sponsor, the only way a participant can become aware of a reduction is by chancing upon the opportunity to compare her benefits with those she would receive under the old plan. Cf. In re JP Morgan Chase Cash Balance Litigation, 460 F.Supp.2d 479, 484 (S.D.N.Y.2006)(limitations period could accrue at retirement rather than effective date).
Plaintiffs allege that they became aware of the reductions in 2004 through communication with employees who remained under the old plan. Appellant's Br. 42. They filed suit in 2005, allegedly within the three-year limitations period. See Charles, 2006 WL 1652749 (D.Del.2006) (parties agree on three-year limitations period). Conectiv does not allege that Plaintiffs became aware of the reduction in their future benefits at any earlier date. It argues instead that access to articles critical of cash balance plans in general, and statements by Plaintiffs that they worried about the effect of the amendment in 1999, place the suit outside the limitations period. Appellee’s Br. 52-54. Suspicion and access to critical media does not necessarily amount to awareness that the amendment could be expected to reduce benefits. Because this issue was not addressed either by the District Court or the majority, I would remand it to the District Court.