dissenting in part:
Normally, we give great deference to an agency’s interpretation of its own regulations. Lyng v. Payne, 476 U.S. 926, 939, 106 S.Ct. 2333, 2341, 90 L.Ed.2d 921 (1986). But deference is not due when the agency’s interpretation flatly contradicts the only sensible reading of an unambiguous regulation. Thomas Jefferson University v. Shalala, — U.S. —, —, 114 S.Ct. 2381, 2386, 129 L.Ed.2d 405 (1994) (agency’s interpretation of its own regulation is not controlling if “plainly erroneous” or “inconsistent with the regulation”). Unlike the majority, I cannot find the term “total compensation” as used in Article I, § 5(a) of the New York Dock conditions at all ambiguous, and I would not defer to what seems to me a most extraordinary and implausible construction of the Commission’s which cannot be justified in result any more than in rhetoric.1
Section 5(a) of the New York Dock conditions sets out a clear and straightforward fixed formula for calculating an employee’s displacement allowance, based on her earn*854ings in a twelve month “test period” immediately prior to displacement:
Each displaced employee’s displacement allowance shall be determined by dividing separately by 12 the total compensation received by the employee and the total time for which he was paid during the last 12 months in which he performed services immediately preceding the date of his displacement as a result of the transaction (thereby producing average monthly compensation and average monthly time paid for in the test period)....
New York Dock, Art. I, § 5(a), 360 I.C.C. 60, 86 (1979) (emphasis added), aff'd sub nom. New York Dock Ry. v. United States, 609 F.2d 83 (2d Cir.1979).
The formula works like this: Begin with the date of the employee’s displacement, and use the preceding twelve months as the “test period.” To determine “average monthly compensation,” take the employee’s total compensation received during the 12-month test period, and divide by 12. Similarly, to determine “average monthly time paid for,” take the total time for which the employee was paid during the 12-month test period, and divide by 12. It is hard to find anything ambiguous there; ordinary terms like “total compensation,” “total time,” and “12 months,” coupled with elementary arithmetic, produce a predictable amount in all eases.
Like any fixed formula for calculating benefits based on time worked, this one may in some few eases produce arguably overly generous or not generous enough results. Although its purpose is to establish a benchmark of “normal” earnings, so as to determine how much an employee has lost from the displacement, the formula makes no adjustment for unusual factors which might make the specified 12-month test period unrepresentative for a given employee. So, for example, an employee who had previously worked twenty years without missing a day, but who is forced to take an unpaid leave during the test period due to illness or family circumstances, will end up with lower “total compensation,” lower “average monthly compensation,” and ultimately lower displacement benefits than one might expect based on her historical “normal” earnings. No adjustment is made for that unfortunate happenstance, even though it might be said that the employer is getting an unjust “windfall” and the employee is suffering an unjust loss.
Similarly, an employee who works an unusually large (or unusually small) amount of overtime during the 12-month test period will end up with “total compensation,” “average monthly compensation,” and displacement benefits at variance with his historical “normal” earnings. In such cases the “windfall” can shift in either direction depending upon whether earnings are abnormally high or abnormally low in the test period. Though this kind of formula may sometimes produce aberrant results in individual cases, it is nevertheless widely used for its compensating virtues of objectivity, uniformity, certainty, and easy administrability.
Despite the lucidity of the formula and its acknowledged potential in some cases for producing atypical results, in either the employer’s or employee’s direction, it is argued that we should engage in its extensive reconstruction when a “literal” application would work against the employer’s interests. In such cases, the Commission and the majority say, “total compensation” should be construed not to mean total compensation as that term is ordinarily understood (i.e., all the compensation an employee receives in the specified period), but instead something akin to “all the compensation the employee receives in the test period, except compensation for any hours in excess of the employee’s ‘normal’ hours.” “Normal” hours are not to be determined solely by reference to the 12-month “test period” specified in the § 5(a) formula; instead, a second, 8-month measuring period prior to the “transaction” is used as the norm. Any hours in the 12-month “test period” in excess of the employee’s “normal” hours as determined from this 8-month measuring period, nowhere mentioned in § 5(a), are presumed to be “transaction-related,” and are deducted from the employee’s “average monthly hours worked.” That second figure, multiplied by the employee’s base rate of pay during the 12-month test period, becomes the new “average monthly compensation,” and the basis upon which the employee’s displacement benefits are calcu*855lated. See American Train Dispatchers Ass’n v. CSX Transportation, Inc., 9 I.C.C.2d 1127, 1129 & n. 6 (1993) (describing methodology).
Whatever other merits or demerits this methodology may have, it certainly does not derive from anything in the language of § 5(a). Indeed, I am hard pressed to see how this new construction can conceivably qualify as a plausible reading of that provision. This is not a case where the term “total compensation” in § 5(a) can bear two plausible meanings, and the Commission has adopted the less obvious one. Rather, the objective formula for calculating displacement benefits so plainly prescribed by § 5(a) is 90% discarded, and an almost totally new methodology substituted in its stead. Under such extreme circumstances, I do not think the Commission’s approach can be fairly said to “interpret,” much less to “draw its essence from” § 5(a) or the New York Dock conditions generally, Chicago & North W. Transp. Co.—Abandonment, 3 I.C.C.2d 729, 735 (1987) (“Lace Curtain”), aff'd sub nom. International Broth. of Elec. Workers v. ICC, 862 F.2d 330 (D.C.Cir.1988).
The majority insists, however, that § 5(a) is ambiguous in light of the “design of New York Dock as a whole.” Maj. op. at 849. In particular, the majority relies on the “narrow definition” of “displaced employee” in § 1(b) as “an employee ... who, as a result of the transaction is placed in a worse position with respect to his compensation and rules governing his working conditions,” New York Dock, Art. I, § 1(b), 360 I.C.C. at 84. The majority says this “reflects an intention underlying the conditions to ensure that employees are no worse off (at least during the protected period) after a consolidation than they would be had the consolidation never taken place.” Maj. op. at 850 (emphasis in original).
This “narrow definition” of a displaced employee, however, is designed only to identify the class of employees eligible for displacement benefits — those injured, ie., “placed in a worse position” or made “worse off’ as a result of the employer’s transaction over a six-year period. There is no question that the dispatchers whose jobs were lost or downgraded by CXT’s consolidation were “placed in a worse position,” and therefore qualify for benefits, however we interpret § 5(a). The § 5(a) formula itself is designed to measure the benefits to which these “displaced” employees are entitled — not to determine whether they are “displaced” and qualify for benefits at all. It is like mixing apples and oranges to use the definition of who is a “displaced employee” and can receive benefits to narrow the plain language of how to calculate the benefits due to such a person. Although § 1(b) reflects a general purpose behind the New York Dock conditions — to assist “displaced employees” by providing them with displacement benefits — it says nothing about how much assistance these employees should receive2; that amount is plainly set out in the formula in § 5(a). I can, in short, find no inconsistency or even tension between § 1(b) and § 5(a), certainly nothing sufficient to render the plain language of § 5(a) ambiguous.
When a provision of law defines an injury and provides that members of the injured class are entitled to compensation, it may be intuitive to expect that the compensation should be commensurate with the injury. But when as here, for other reasons such as predictability, administrability, certainty, etc., a separate provision specifies in plain and unambiguous terms an across-the-board formula for calculating the benefits members of the injured class are to receive, the terms of the benefits provision, not our intuitions, should control.
Actually, the majority’s interpretive argument proves too much. Certainly it is true, as the majority argues, that the New York Dock conditions were not intended to produce “windfalls” for employees — or, we may *856add, for employers either. But if the majority’s reasoning is sufficient to justify overriding a “literal” application of § 5(a) in this case, I do not see any principled ground on which to deny an override of § 5(a) anytime a “windfall” comes the way of either the employer or the employee. And that, of course, would defeat the purpose of the § 5(a) formula entirely, leaving the Commission with discretion in every individual case to decide whether or not the prescribed 12-month test period is truly representative of the employee’s “normal” earnings, or whether § 5(a) should be overridden by some other “fairer” methodology unique to the facts of that case. Such a result is surely contrary to the purposes of New York Dock, which patently seeks to prescribe a uniform, objective formula for determining displacement benefits. For instance, under the majority’s reasoning, these provisions were no more intended to give a “windfall” to an employee who has abnormally large overtime earnings during the test period for reasons unrelated to the transaction causing her displacement — for example, an unexpected spike in business during the test year, or an unusual and temporary shortage of co-workers — than to one who has transaction-related overtime. I can foretell no grounds on which to distinguish between transaction-related and unrelated overtime, since any overtime earnings above a historical average would give the employee a six-year “windfall” arguably inconsistent with the purposes of New York Dock. Indeed, one might think it more unjust — and farther from the purposes of the New York Dock conditions — to reward employees with “windfalls” for extraordinary overtime unrelated to the transaction, since such circumstances are presumably more difficult for the employer to foresee and control than transaction-related overtime in a 12-month test period immediately prior to the employee’s displacement.3
It seems equally certain that the New York Dock conditions were not intended to disadvantage an employee who, through no fault of her own, has abnormally low earnings during the 12-month test period, thus giving her employer a “windfall” by reducing her displacement benefits. Indeed, an employee who loses expected earnings in the “test year” may be banking on making up the loss in subsequent years, and therefore may be doubly “worse off’ when the inflexibility inherent in the § 5(a) formula cuts short her displacement benefits. This result, too, may be the unintended and arguably unjust consequence of a “literal” application of the New York Dock formula. It is just as arguably inconsistent with the “context and purpose” of the regulatory scheme which aims to protect an employee’s “normal” earnings as the majority thinks the present case. But no one suggests a reconstruction of the § 5(a) formula in these cases.
In sum, I find § 5(a) totally unambiguous, but acknowledge it is not without warts. Indeed, some might go so far as to conclude the drafters of the New York Dock conditions *857may have handed the current Commission a bad formula for calculating displacement benefits. Perhaps a different test period than the one plainly specified in § 5(a) would more fairly measure employees’ “normal” earnings.4 Perhaps it would be better to use a formula that explicitly excludes transaction-related overtime from the calculation of displacement benefits. Or perhaps a fixed 12-month test period is too short and too arbitrary in any event, and more flexible individualized methods of calculating “normal” earnings can be devised. The Commission presumably may take appropriate measures to amend or replace the offending provisions. But those measures should not include reading the current provision to say what it plainly does not say.5
I would conclude that Article I, § 5(a) of the New York Dock conditions is not ambiguous, hold that the Commission’s order affirming the arbitrator’s award was arbitrary and capricious, and remand to the Commission for further proceedings. Therefore, I respectfully dissent from that portion of the majority opinion affirming the award.
. I agree with the majority that the New York Dock conditions are not a statute and therefore ordinary standards of statutory interpretation do not apply. See Majority opinion (“Maj. op.”) at 847-48. Here, however, the Commission affirmed the arbitrator's award on the basis of an interpretation of its own regulation that, in my view, is "plainly erroneous” and squarely "inconsistent with the regulation” itself, and therefore not entitled to our deference. See Thomas Jefferson University, — U.S. at —, 114 S.Ct. at 2386; United States v. Larionoff, 431 U.S. 864, 872-73, 97 S.Ct. 2150, 2155-56, 53 L.Ed.2d 48 (1977); Udall v. Tallman, 380 U.S. 1, 16-17, 85 S.Ct. 792, 801, 13 L.Ed.2d 616 (1965); Bowles v. Seminole Rock & Sand Co., 325 U.S. 410, 414, 65 S.Ct. 1215, 1217, 89 L.Ed. 1700 (1945). See also Gardebring v. Jenkins, 485 U.S. 415, 430, 108 S.Ct. 1306, 1314, 99 L.Ed.2d 515 (1988) (agency’s interpretation of its own regulation is entitled to deference "unless ... [an] alternative reading is compelled by the regulation's plain language”).
. Indeed, if § 1(b) did cany any implications relevant to the measure of benefits, I would read it as setting a floor rather than a ceiling on benefits; after all, its concern is that employees not be placed in a worse position during the protected period.
. The Commission relies on the rationale that it would be unjust to allow employees to reap a "windfall” from transaction-related overtime during the 12-month test period, but in actual operation the methodology used here makes no distinction between transaction-related and unrelated overtime. Any overtime in the 12-month test period in excess of the employee’s "normal” earnings in an 8-month measuring period prior to the transaction is deleted from that employee’s "total compensation” for purposes of calculating displacement benefits. See 9 I.C.C.2d at 1129 n. 6, 1135. Thus if during the 12-month test period employee A is forced to work overtime due to employee B's illness, employee A’s “excess” overtime will not count toward her displacement benefits; but employee B's displacement benefits will be based on her abnormally low earnings. The employer is not required to establish any causal nexus between the transaction and the overtime. See id. at 1135. The Commission apparently would shift the burden to the employee to show that overtime in the test period is not transaction-related, but offers no explanation why the evidentiary burden should fall on the employee. See id. In addition, this approach invites an unraveling of the clear and objective methodology of the § 5(a) formula into a detailed, hour-by-hour, fact-based dispute over whether the overtime is, or is not, caused by or "related to” the transaction. I would conclude that the Commission's rationale, based solely on excluding transaction-related overtime, does not adequately explain the methodology actually used in this case. That alone is reason enough to reverse the Commission's action as arbitrary and capricious.
. Underlying the Commission's approach is the idea that the 12-month test period specified in § 5(a) is defective, giving a false indication of an employee’s "normal” earnings in many cases. The test period consists of the 12 months prior to the date the employer’s transaction first adversely affects the individual employee, a period which may embrace temporary fluctuations in earnings due to transitional arrangements necessitated by the transaction itself. To remedy this perceived flaw, the Commission allowed the employer to use a second test period entirely preceding the transaction, overriding the methodology unambiguously prescribed by § 5(a). A more straightforward solution would be to amend § 5(a) to provide for a different test period.
. Nor am I persuaded by the history of previous arbitrators' awards interpreting § 5(a) to exclude transaction-related overtime. If § 5(a) were truly ambiguous, these prior interpretations would carry great weight. But it seems to me that previous arbitrators, like the Commission in this case, disapproved on policy grounds the results unambiguously mandated by § 5(a). I decline to join them in entertaining the fiction that § 5(a) is ambiguous.