Opinion for the Court filed by Circuit Judge STEPHEN F. WILLIAMS.
STEPHEN F. WILLIAMS, Circuit Judge:
Long-distance telephone traffic is ordinarily transmitted by a local exchange carrier (“LEC”) from its origin to a long-
distance carrier (or interexchange carrier or “IXC”). The IXC carries the traffic to its region of destination and hands it off to the LEC there. The IXC charges the customer for the call and pays “access charges” to the LECs at either end. In a 1997 rulemaking the Federal Communications Commission amended its methodology for limiting these charges, as applied to the largest LECs. The rule is challenged on one side by á group of LECs, and on the other by one IXC, namely MCI, and an Ad Hoc Telecommunications Users Committee (collectively referred to here as MCI).
In regulating access charges the FCC currently uses a “price cap” method'—mandatory for the largest LECs (the regional Bell operating companies and GTE) and optional for others. Under traditional rate-of-return regulation an agency sets rates calculated to allow the utility to recover its costs, including a reasonable rate of return on investment, with adjustment as needed to reflect cost changes; here, however, it sets rate ceilings and, with some qualifications, allows the utilities to keep whatever profits they can make while charging rates at or under the cap. (A LEC may also file rates above the caps, but for these the review process is cumbersome and the substantive standards stringent.) The price cap system is intended (among other things) to improve the utility’s incentives to cut costs and refrain from overinvestment, incentives that are more blunted under the traditional method. See generally
National Rural Telecom Ass’n v. FCC,
988 F.2d 174, 177-79 (D.C.Cir.1993).
The price caps were initially set at the levels of each carrier’s rates on July 1, 1990. From the outset they have been subject to various annual adjustments, including reduction by a “productivity offset,” or “X-Factor.” See 47 CFR § 61.45. In the order under review, the agency revised the method for determining the X-Factor, eliminated a “sharing” mechanism that forced LECs to return some or all of the profits above specified levels to ratepayers, and required “reinitialization,” i.e., a reduction in the price caps applicable after July 1, 1997 so that they would be calculated as if the new X-Factor had been in effect for the LECs’ 1996 tariff filings.
In the Matter of Price Cap Performance Review for Local Exchange Carriers,
Fourth Report & Order, 12 FCC Red 16,-642 (1997)
(“1997 Order”).
Because the access charges are in the aggregate so enormous, even small changes in the X-Factor have a large monetary value; the LECs claim (without dispute) that each 0.1% change in the factor represents a $23 million change in the industry-wide access charge.
I. The historic productivity component of the X-Factor
The X-Factor is aimed at capturing a portion of expected increases in carrier productivity, so that these improvements, as under competition, will result in lower prices for consumers.
In the Matter of Policy and Rules Concerning Rates for Dominant Carriers,
3 FCC Red 3195, 3394 (1988). Apart from a “consumer productivity dividend” (“CPD”) described below, it is based on an assumption that historic productivity increases will be matched in the future. The agency resolved in the 1997 Order that the X-Factor (apart from the CPD) should be calculated as the sum of the difference in productivity growth and the difference in input price growth between the LECs and the economy as a whole. See 12 FCC Red at 16,680, ¶ 95. It can thus be expressed as follows: X = (A% LEC TFP - A% TFP) -I- (A% U.S. input prices - A% LEC input prices), where TFP = total factor productivity. See 12 FCC Red at ÍO^S.
The formula
may be more readily conceptualized as X = (A% LEC TFP - LEC input prices) - (A% U.S. TFP - A% U.S. input prices).
Several parties submitted estimates of historical X-Factors. In a determination unchallenged here, the FCC accorded the greatest weight to its own estimates, although it also gave “some weight” to AT&T’s estimates (we discuss this decision below). See
1997 Order,
12 FCC Red at 16,695, ¶ 37. The estimates the FCC considered, and the averages of those estimates over specified periods, are the following:
[[Image here]]
1997 Order,
12 FCC Red at 16,696, ¶ 137.
The FCC consulted the moving averages to establish a range of reasonableness from 5.2% to 6.3% and then selected 6.0% as the historical (i.e., non-CPD) component of the X-Faetor. See
id.
at 16,697, ¶ 141. The LECs argue that the FCC did not give a rational explanation of that choice, and we agree. None of the reasons given for choosing 6.0% holds water.
A. Devaluation of 1986-95 and 1991-95 averages
First, in choosing a point within the range of reasonableness, the FCC determined that it was “reasonable to place less weight” on two lowest averages, the ones for 1986-95 and 1991-95. It said that the first, 1986-95, “is heavily influenced by the improbably low 1986 estimate of-0.5 percent.”
Id.
at 16,697, ¶ 139. But the Commission gave no reason for condemning the 1986 estimate as “improbable,” and mere divergence from the other numbers does not justify such a conclusion. See Thomas H. Wonnacott & Ronald J. Wonnacott,
Introductory Statistics for Business and Economics
497 (2d ed.1977). The FCC invokes our cases upholding the elimination of outlying data points, but in them the agency explained why the outliers were unreliable or their use inappropriate. See
Bell Atlantic Tel. Cos. v. FCC,
79 F.3d 1195, 1202 (D.C.Cir.1996) (study indicated outlier erroneous);
Association of Oil Pipe Lines v. FERC,
83 F.3d 1424, 1434 (D.C.Cir.1996) (skewed data distribution required outlier elimination to avoid windfall profits to many oil pipelines).
As to the 1991-95 average, the Commission said it was the one “most affected by the low 1992 estimate,” which it in turn diagnosed as “an artifact of a one-year jump in the measured productivity of the national economy as economic activity increased, rather than a change in the growth rate of LEC productivity or input prices.”
1997 Order,
12 FCC Red at 16,-697, ¶ 139. This is mystifying. If the productivity component of the X-Factor is to reflect the difference between LEC and
overall productivity growth, a proposition that is built into the Commission’s formula, see
1997 Order,
12 FCC Red at 16,785, there seems no reason to slight a datum because its anomalous character stems from the unusual magnitude of the second term rather than of the first.
B. Alleged upward trend
In justification of its choice of 6.0% the FCC also cites an upward trend in the X-Factor during the last years it surveyed. See
1997 Order,
12 FCC Red at 16,697, ¶ 139 (“[F]rom 1993 onward there has been an upward trend in the X-Factor”);
id.
at ¶ 141 (“[T]here appears to be a strong upward trend in productivity growth from 1992 to 1995”).
The FCC’s reliance on the upward trend necessarily reflects the (unexplained) assumption that the trend will continue, at. least in the immediate future. Explanation might be reasonably omitted if there were no obvious reason to doubt continuation of an observed trend. But two such reasons exist.
First, the trend appears to be part of a cyclical pattern. Although the X-Factor did increase steadily in the 1992-95 period, it also decreased from 1990 to 1992, after rising from 1986 to 1990. See Table 1,
supra.
Perhaps there was reason to believe that there would be no cyclical downturn during the expected life of this X-Factor determination, which was to be reviewed about two years after being made. See
1997 Order,
12 FCC Red at 16,707, ¶ 166. But the FCC offered no such reason.
Second, the X-Factor is calculated as the sum of two components, neither of which followed a trend during the period in question. In fact, their year-to-year fluctuations swamped the trend increments:
1997 Order,
12 FCC Red at 16,785. Where’s the trend? As the underlying variables appear to be thrashing about wildly, the FCC’s conclusion that the trend in the difference between the two had some predictive value requires explanation.
C. Partial reliance on AT&T estimates
Finally, the LECs argue that in its treatment of AT&T’s X-Factor estimates the FCC “implicitly endorsed methodologies that it had earlier discredited.” LEC Br. at 27. The FCC incorporated the aspects of AT&T’s method that it deemed reasonable into its own method, see
1997 Order,
12 FCC Red at 16,658, ¶ 33, and then gave independent weight to AT&T’s X-Factor estimates in deciding to extend the range of reasonableness upward, see
1997 Order,
12 FCC Red at 16,697, ¶ 140, and to select a value near the top of the range.
Id.
at ¶ 141. We agree that both these uses of AT&T’s estimates appear irrational; any differences between the FCC’s and AT&T’s estimates presumably resulted from elements of AT&T’s analysis that the FCC specifically rejected. The FCC’s argument that AT&T’s estimates were “helpful” because AT&T’s methodology was “similar,” FCC Br. at 37, fails to overcome that logic. If there is an explanation — for example, conceivably the Commission gave some weight to AT&T’s conclusions out of concern for the risk that it had erred in rejecting specific elements of AT&T’s analysis — the FCC has failed to mention it.
The Commission having failed to state a coherent theory supporting its choice of 6.0%, we remand for further explanation.
II. Consumer productivity dividend
The second component of the X-Factor is a “consumer productivity dividend” (“CPD”) of 0.5%. At the time of the 1990 order instituting price-cap regulation, the FCC “expected ... that incentive regulation would result in greater productivity gains than rate of return regulation,”
Bell Atlantic,
79 F.3d at 1198, and instituted the CPD, as it said, to “assure that the first benefits of price caps flow to customers in the form of reduced rates,”
In the Matter of Policy and Rules Concerning Rates for Dominant Carriers,
5 FCC Red 6786, 6799, ¶ 100 (1990)
(“Price Cap Order”).
It retained the 0.5% CPD without specific explanation in a 1995 interim rule,
Bell Atlantic,
79 F.3d at 1204, and retained it again in the current rule. See
1997 Order,
12 FCC Red at 16,690, ¶ 123.
The LECs challenge the 0.5% CPD as based on an “obsolete” justification. The Commission’s earlier data on historic productivity improvement derived from the rate-of-return era, so an adjustment to reflect the expected incentive effects of price caps was in order; but the posN1990 data presumably reflect those effects.
FCC counsel responds that the agency believes that an innovation in the current rule — the Commission’s elimination of the “sharing” of profits exceeding certain benchmarks — will give the LECs still further productivity incentives, and that the FCC relied on that in retaining the CPD. Even if the agency relied on this justification (which the LECs dispute), it never explained retention of the old percentage, a retention that required some comparison of the current change with the initial one in terms of their likely impacts on productivity. Thus we must remand for an explanation of the Commission’s choice of the amount — 0.5%.
The LECs claim that the FCC did not rely on the expected effects of sharing elimination and that it gave no other reason justifying the retention of
any
CPD. We do not reach these arguments because the FCC will be able to give a clearer statement of its reasons in the remand on the amount and since the LECs do not dispute the argument FCC’s counsel is presently making — that it is defensible to include a CPD corresponding to whatever productivity increase may be expected from the elimination of sharing.
III. Elimination of sharing
Before the rule at issue in this case, the FCC’s price cap regime included a “sharing” mechanism, which mandated LEC rate reductions sufficient to return profits above specified levels to their customers, the IXCs. The most recent sharing regime, enacted in the 1995 interim order, made the sharing obligation dependent on the X-Factor, imposing no obligation on firms choosing a 5.3% X-Factor, and the following on ones choosing 4.7% and 4.0%:
In the Matter of Price Cap Performance Review for Local Exchange Carriers,
10 FCC Red 8961, 9058, ¶ 222
(“Performance Review Order”)
(1995). Attacking the Commission’s elimination of the “sharing” mechanism, MCI first claims that the statutory mandate of “just and reasonable” rates, 47 U.S.C. § 201(b), requires the FCC to impose a mechanism to prevent “unreasonable” returns. In the absence of any indication that Congress directly addressed the issue, we defer to the FCC’s interpretation of the Communications Act unless it is unreasonable. See
Chevron U.S.A. Inc. v. NRDC,
467 U.S. 837, 104 S.Ct. 2778, 81 L.Ed.2d 694 (1984). MCI cites no authority rejecting an FCC interpretation of the statute contrary to the one MCI advances, and in
Time Warner Entertainment Co. v. FCC,
56 F.3d 151 (D.C.Cir.1995), we endorsed a pure price cap regime with no sharing provision in
the face of a statutory mandate to ensure “reasonable” basic cable rates. See
id.
at 162,164-74.
Next, MCI argues that elimination of sharing was arbitrary and capricious. But the agency advanced two sound rationales for its decision. First, it found that “sharing severely blunts the efficiency incentives of price cap regulation by reducing the rewards of LEC efforts and decisions.”
1997 Order,
12 FCC Red at 16,700, ¶ 148. When all profits are taken away, a firm has no incentive to make them; when some proportion is taken away, firms will avoid at least some otherwise desirable choices with a prospect of enhancing profit but a risk of loss. Second, the FCC found that eliminating sharing would remove the incentive to shift costs to services that are subject to sharing and away from services that are not, thus cross-subsidizing the latter.
1997 Order,
12 FCC Red at 16,700, ¶ 148;
id.
at 16,701, ¶ 151. MCI does not contest these effects, nor does it question the Commission’s argument that monitoring to catch them would be administratively burdensome and would increase its reliance on obsolete embedded accounting costs.
Id.
at 16,701-02, ¶¶ 151-52.
Finally, MCI contends that it was arbitrary and capricious for the FCC to scuttle sharing but at the same time retain its “low-end adjustment,” which gives the LECs some pricing leeway to prevent their returns from falling below a given level. There is clearly a literal asymmetry in protecting LECs in lean conditions while not constraining them in unexpectedly fat ones. But the FCC gave a good reason for creating this asymmetry—the Constitution’s takings clause, which forbids the imposition of confiscatory rates without just compensation.
See 1997 Order,
12 FCC Red at 16,704, ¶ 157;
Duquesne Light Co. v. Barasch,
488 U.S. 299, 307-08, 109 S.Ct. 609, 102 L.Ed.2d 646 (1989). The Commission thus avoided raising a non-trivial constitutional question, one that has no analogy at the upper end of the range of allowable rates. See
Time Warner,
56 F.3d at 170.
IV. Interstate v. total-company productivity
MCI argues that in calculating the X-Factor the FCC arbitrarily used the LECs’ productivity in all their telecommunications business rather than productivity only in their interstate operations. Again, we disagree. The FCC reasonably concluded that “the record before us does not allow us to quantify the extent, if any, to which interstate productivity growth may differ significantly from total company productivity growth,”
1997 Order,
12 FCC Red at 16,686, ¶ 110, and this determination was enough to justify using the total company data.
In the first place, it is not clear that “interstate productivity,” as opposed to total company productivity, is measurable, or even economically well-defined. This is so because direct productivity measurement requires measurement of inputs, and there is no obviously meaningful way to segregate LEC interstate and intrastate inputs because, as is undisputed, “interstate and intrastate services are usually provided over common facilities.”
1997 Order,
12 FCC Red at 16,685, ¶ 107. The Commission had previously recognized this analytical difficulty, questioning “whether it would be possible to develop separate production functions for interstate and intrastate services,”
id.,
and it never unambiguously declared the issue resolved.
The Commission nonetheless declared itself ready to consider some adjustment if it were shown that inclusion of intrastate data systematically biased the X-Factor estimate downward.
1997 Order,
12 FCC Red at 16,686, ¶ 109. AT&T offered claims of faster interstate productivity growth. It based these on an assumption of equal growth rates for interstate and intrastate inputs, but it offered no explanation why that assumption was economically justified, much less one so compelling that it would be error for the FCC to reject it. See
AT&T Comments, CC Docket No. 94-1, App. A at 23-30, 72-78;
1997 Order,
12 FCC Red at 16,686-87, ¶ 110.
MCI argues that in the original 1990 LEC price cap order the Commission inferred faster productivity growth in interstate services from the undisputed fact of faster output increase in that sector. See
Price Cap Order,
5 FCC Red at 6798, ¶ 92 (“[T]he more rapid growth in interstate usage results in higher apparent interstate productivity growth.”). This assumption should have continued, it says. But the 1990 method of measuring productivity had not depended on the measurement of inputs at all; the Commission had simply inferred productivity growth from prior trends in
rate
reductions.
1997 Order,
12 FCC Red at 16,648, ¶ 8. Given the shift to direct focus on input changes (a move that no one questions) and the uncertainty over interstate input trends, we do not see why the agency should have been bound to retain the ^assumption of faster interstate productivity growth. On this record, therefore, we do not find it unreasonable for the agency to have relied on total company productivity despite its theoretical shortcomings.
V. Reinitialization
“Reinitialization” is the name for the Commission’s setting a current price cap at what it would have been if past X-Factors had been different. For instance, if the price cap starts at 100 and the X-Factor is 1% for the first three years, the cap would stand at approximately 97 at the end of those years. 100 -(3xl) = 97. (The figure is only approximate because of compounding.) If the regulator then changes the X-Factor to
2%
and imposes full reinitialization, it would revise the cap to about 94 for the year immediately following. 100 -(3 x 2) = 94. In the 1997 Order, the FCC ordered reinitialization for one year, 1996. See 12 FCC Red at 16,-714, ¶ 179. Under our simple example, then, the cap would fall to approximately 96. 100 - (2 x 1) [two years’ reduction of 1%] - (1 x 2) [one year’s reduction of 2%] = 96.
Both the LECs and MCI challenge this decision, seeking to have it modified to favor their respective interests.
A. Reinitialization based on CPD
The LECs challenge the FCC’s requirement that they include the CPD in the X-Factor used for reinitialization. In Part II, we explained the need to remand the case for further explanation of size of the CPD. We agree with the LECs that if the FCC retains the CPD because of the productivity benefits expected from the elimination of sharing, no element of reinitialization based on the CPD will be appropriate in the absence of evidence linking productivity gains to the anticipation of sharing’s elimination; the companies could not have responded to that incentive before its creation.
B. Disparate impact of uniform reinitialization
The LECs argue that reinitialization fell more harshly on carriers that chose low X-Factors with high sharing obligations for 1996 than on ones that chose high X-Factors. As a result of reinitialization, the low X-Factor carriers lost some of the future benefits of that choice, but were not in a position to recover any of sharing costs that they may have borne because of it. Reinitialization .imposed no such asymmetry on companies that had elected a high X-Factor. The LECs’ specific complaint is that this was “an important aspect of the problem” before the Commission, which it was obliged to discuss. See
Motor Vehicle Mfrs. Ass’n v. State Farm Mut. Auto. Ins. Co.,
463 U.S. 29, 43, 103 S.Ct. 2856, 77 L.Ed.2d 443 (1983).
The Commission argues that it failed to discuss the disparity because the LECs never brought the subject up. It cites § 405 of the Communications Act, 47 U.S.C. § 405, which bars review of an
issue on “which the Commission ... has been afforded no opportunity to pass,” see also
United States v. FCC,
707 F.2d 610, 619 (D.C.Cir.1983), unless the petitioners sought rehearing before the Commission— which the LECs did not. The LECs in turn say they couldn’t have afforded the Commission a chance to pass on it; the Commission had never given notice of any intent to order reinitialization.
Section 405’s “no opportunity to pass” clause does not in terms exclude instances where the lack of opportunity is due to some fault of the Commission — such as its springing a novelty at the last minute. But we need not sort that out here, because we find no fault in the Commission’s procedure. Reinitialization may not have been a subject on which the Commission explicitly elicited comment in its notices for this rulemaking, but the prospect surely brooded over the proceeding. In its 1995 mid-course correction of the price caps it had ordered reinitialization — -in a form, in fact, that fell
only
on those LECs that had chosen a low X-Factor in exchange for greater risk of sharing, and not at all on those that had chosen a high one.
Performance Review Order,
10 FCC Red at 9069-73, ¶¶ 245-54. If the perceived asymmetry was as serious as the LECs now make out, we should have expected them to alert the Commission in this proceeding in advance: “If you do a reinitialization, at least avoid the dreadful asymmetry of the 1995 order.” No such alert was sounded.
C. Reinitialization for only one year
MCI claims that the FCC should have reinitialized the X-Factor all the way back to 1991 (the first year of the price-cap regime). It says the agency has a policy of correcting errors in X-Factor determinations and that it decided in the current rule that prior determinations were in error. In the alternative, MCI argues that the FCC should reinitialize back to 1995, the year in which the previous X-Factor was adopted.
In the 1995 interim price cap review, the FCC determined that a single year’s productivity estimate generated by its former method was understated, based in large part on the estimate’s discrepancy with the results of a TFP study. See
Performance Review Order,
10 FCC Red at 9053, ¶ 208. It then calculated a new X-Factor designed to eliminate the effects of the understatement and required LECs to set their price caps as though the new X-Factor had been in effect since the advent of price cap regulation. See
id.
at 9069, ¶ 245. In 1997 the Commission determined that its former method had systematically understated productivity relative to the TFP method, but required reinitialization for one year only. See
1997 Order,
12 FCC Red at 16,713-14, ¶¶ 178-79.
The situations are somewhat similar, but the FCC adequately distinguished them. It rested its 1997 decision to limit reinitialization on the need to “limit harm to LEC productivity incentives that could result from the perception that our regulatory policies unnecessarily lack constancy.”
1997 Order,
12 FCC Red at 16,714, ¶ 179. It seems clear that a second extensive reinitialization would considerably aggravate such a perception. Universal, complete reinitialization would impair the supposed incentive advantages of price caps— which derive from firms’ supposing that their efficiencies will
not
come back to haunt them.
VI. The rule’s effects on small and mid-size LECs
The Independent Telephone and Telecommunications Alliance, an intervenor, argues that the FCC acted arbitrarily and capriciously in establishing a uniform X-Factor for all LECs, regardless of size and economic characteristics, and in failing to consider the disparate impact of its reinitialization requirement on small and mid-size LECs. Because the petitioners here have not raised these issues, ITTA is procedurally barred from arguing them.
See
Illinois Bell Tel. Co. v. FCC,
911 F.2d 776, 785-86 (D.C.Cir.1990).
It is true, as ITTA points out, that this court in
Synovus Fin. Corp. v. Board of Governors,
952 F.2d 426, 434 (D.C.Cir.1991), characterized the rule against consideration of issues raised by intervenors and not by petitioners as “a prudential restraint rather than a jurisdictional bar.” But in deciding to consider the interve-nor’s issue there, the court relied on the fact that the relevant issue was “an essential predicate” to an issue raised by a petitioner.
Id.
That circumstance is certainly not present here. The
Synovus
court offered a second reason to hear the claim — that the intervenor was not “the losing party in the administrative proceeding,” and thus did not have “every incentive to petition for review.”
Id.
Here, ITTA itself claims that it “through its members, participated fully in the proceedings below,” ITTA Reply Br. at 3, and that its “members raised the issue of the necessity of multiple X-Factors,” the very issue it seeks to raise in this court.
Thus, neither of the special circumstances cited in
Synovus
is present. Furthermore, ITTA presents no reason why it could not have petitioned in its own right. We decline to consider its arguments.
Conclusion
The FCC’s decisions to select 6.0% as the first component of the X-Factor and to retain the 0.5% CPD are reversed and remanded to the agency for further explanation; the FCC may of course request a stay of this order pending its reconsideration. The petitions for review are otherwise denied.
So ordered.