United States Court of Appeals
FOR THE DISTRICT OF COLUMBIA CIRCUIT
Argued March 19, 2010 Decided June 18, 2010
No. 09-1202
AEP TEXAS NORTH COMPANY,
PETITIONER
v.
SURFACE TRANSPORTATION BOARD AND UNITED STATES OF
AMERICA,
RESPONDENTS
BNSF RAILWAY COMPANY,
INTERVENOR
On Petition for Review of an Order
of the Surface Transportation Board
William L. Slover argued the cause for petitioner. With him
on the briefs were Kelvin J. Dowd, Robert D. Rosenberg, and
Andrew B. Kolesar III.
James A. Read, Attorney, Surface Transportation Board,
argued the cause for respondents. With him on the brief were
Robert B. Nicholson and John P. Fonte, Attorneys, U.S.
Department of Justice, Ellen D. Hanson, General Counsel,
Surface Transportation Board, Craig M. Keats, Deputy General
Counsel, and Thomas J. Stilling, Attorney.
2
Samuel M. Sipe Jr., Anthony J. LaRocca, and Richard E.
Weicher were on the brief for intervenor BNSF Railway
Company in support of respondent. Kathryn J. Gainey entered
an appearance.
Before: SENTELLE, Chief Judge, HENDERSON and BROWN,
Circuit Judges.
Opinion for the Court filed by Chief Judge SENTELLE.
SENTELLE, Chief Judge: AEP Texas (AEPT), an electric
utility company, uses railroads owned by BNSF Railway
Company (BNSF) to transport coal from mines in Wyoming to
an electric generating station in Texas. BNSF exerts market
dominance over the route used for this coal transportation,
which renders the rates charged subject to review by the Surface
Transportation Board (the Board). Under this authority, the
Board has set a reasonable maximum rate for the transportation
service since 1996. See W. Tex. Util. Co. v. Burlington N. R.R.
Co., 1 S.T.B. 638 (1996). In calculating a reasonable maximum
rate, the Board considers, among other variables, the cost of
equity capital, approximating a reasonable rate of return for
railroad investors. AEPT petitioned the Board for a change in
methodology in the calculation of the cost of equity capital.
While the Board entered some changes, it denied a portion of the
petition requesting a recomputation of that variable for the years
1998–2005. AEPT now petitions for review of that decision of
the Board. We deny the petition with respect to the Board’s
decision not to recalculate that variable with respect to the years
1998–2004. However, as to the 2005 calculation, we hold that
the Board failed adequately to explain its decision. We therefore
vacate the Board’s decision with respect to the 2005 calculation
and remand for further consideration.
3
I. Background
AEPT is an electric utility generating and transmitting
electricity to consumers in central and west Texas. To ship coal
from Wyoming mines to its electric generating station near
Vernon, Texas, AEPT uses railways owned by BNSF. Because
there is “an absence of effective competition from other rail
carriers or modes of transportation for the transportation” of coal
on that route, BNSF is said to exert “market dominance” over
AEPT’s shipments. See 49 U.S.C. § 10707(a). Therefore, the
rates BNSF charges on that route are subject to review by the
Surface Transportation Board under § 10707(b), (c).
If the Board determines a railroad has market dominance,
it can establish a reasonable maximum rate the railroad may
charge for the transportation. § 10707(c). The Board calculates
the reasonable maximum rate using a “constrained market
pricing” methodology set forth in the Coal Rate Guidelines, 1
I.C.C.2d 520 (1985). To evaluate the rate for a specific route
under this methodology, the Board posits a hypothetical railroad
serving a subset of a real railroad’s network. The hypothetical
railroad, called a Stand Alone Railroad (SARR), operates the
route used by the relevant shipper and is presumed to operate at
optimal efficiency. The Board calculates the Stand Alone Cost
for the SARR, which represents the cost of running the
hypothetical railroad and includes a reasonable rate of return on
investment. The return on investment is the rate of return “that
shareholders require to compensate them for the use of their
capital.” Methodology to be Employed in Determining the
Railroad Industry’s Cost of Capital, STB Ex Parte No. 664,
2008 WL 162591, at *1 (Jan. 17, 2008) (“Methodology to be
Employed—2008”). This rate of return is referred to as the “cost
of equity capital,” or the “cost of equity.” The SARR’s Stand
Alone Cost determines the maximum rate the railroad may
charge all the shippers using the routes in the SARR. See BNSF
4
Ry. Co. v. STB, 526 F.3d 770, 777 (D.C. Cir. 2008). This
method ensures that no shipper subsidizes another; “though
some bear a higher share of fixed costs than others, they still pay
no more than what they would for a facility designed to serve
only them.” Burlington N. R.R. Co. v. ICC, 985 F.2d 589, 596
(D.C. Cir. 1993).
The Board annually recalculates the industry-wide cost of
capital. This cost of capital is used in various regulatory
capacities, including calculating reasonable maximum rates for
railroads with market dominance. The published cost of capital
includes both the cost of equity capital and the cost of debt
capital. Because the interest rates on borrowed money are easily
observable, the cost of debt capital rarely becomes the subject of
debate. The cost of equity capital, however, is not directly
observable, which forces the Board “to rely on complex finance
models to estimate that component of the cost of capital.”
Methodology to be Employed—2008, 2008 WL 162591, at *1.
Since 1996, the Board has used the Stand Alone Cost
method to set the rates BNSF may charge AEPT1 for use of the
Wyoming to Texas railways. W. Tex. Util. Co. v. Burlington N.
R.R. Co., 1 S.T.B. 638 (1996), aff’d sub nom. Burlington N.
R.R. Co. v. STB, 114 F.3d 206, 209 (1997). As dictated by the
Coal Rate Guidelines, the Board determines the Stand Alone
Cost for the route AEPT uses by calculating, among other
things, the cost of capital for the relevant SARR, called the
Texas & Northern Railroad (TNR). However, because SARRs
are hypothetical, the Board rarely has enough data to estimate an
individualized cost of capital for a specific SARR each year.
Therefore, for most of its Stand Alone Cost determinations,
1
At the time, the complaining shipper was actually AEPT’s
predecessor, West Texas Utilities Company. For convenience,
however, we refer to the company as AEPT for all time periods.
5
including the ones at issue in this case, the Board uses the
estimated cost of capital it publishes each year for the whole
railroad industry.
Obviously, the calculation of the industry-wide cost of
equity capital is a significant factor in the determination of
reasonable rates, as the cost of equity is a large component of
the overall cost of a railroad. See Railroad Cost of
Capital—2008, STB Ex Parte No. 558 (Sub No. 12), 2009 WL
3052742, at *10 (Sept. 25, 2009). But the cost of equity also has
recursive implications. The Board averages each year’s cost of
equity into an historical cost of capital. The average historical
cost then affects the Board’s forecast of growth, which in turn
affects the cost of equity in future years.
Because the methodology employed to estimate the cost of
equity can substantially affect the rates a railroad can charge,
and because the cost of equity is difficult to estimate, the
Board’s choice of methodology has provoked controversy for
over a decade. During the Board’s 1997 proceedings for
calculating industry-wide cost of capital, a trade association of
coal shippers called the Western Coal Traffic League (WCTL)
argued that the Board’s methodology for calculating the cost of
equity capital overestimated actual costs. See Railroad Cost of
Capital—1997, 3 S.T.B. 176, 177 n.1 (July 9, 1998). However,
WCTL did not present any alternative calculations of its own.
Therefore, the Board decided to retain its existing methodology.
That methodology was a single-stage discounted cash flow
(DCF) model, which the Board had been using since its first
annual cost of capital calculation in 1981. Under the DCF
method, the cost of equity is calculated as the discount rate that
makes the present value of the expected returns on a stock,
including dividends and price appreciation, equal to the current
market value of the stock. See, e.g., Railroad Cost of
Capital—1987, 4 I.C.C.2d 621, 625, 626 (1988); Railroad Cost
6
of Capital—2004, STB Ex Parte No. 558 (Sub No. 8), 2005 WL
1534327 (June 21, 2005). In short, the DCF model computes
the cost of equity through a relatively straightforward
manipulation of stock predictions.
In 2005, WCTL again complained about the DCF method
in the industry-wide proceedings, and proposed that the Board
use a capital asset pricing model (CAPM) instead. See Railroad
Cost of Capital—2005, 2006 WL 2692729, at *4 (Sept. 15,
2006). Under CAPM, the cost of equity is equal to the risk-free
rate of return (for example, the rate of return one could expect
from a 20-year Treasury bond) plus a premium associated with
a particular investment. That premium is derived through an
ordinary least-squares regression calculation that takes into
account the overall rate of return on the stock market, systematic
and non-diversifiable risk, and the historical rate of return for a
class of stocks (here, railroad stocks). Methodology to be
Employed—2008, 2008 WL 162591, at *6–8.
Citing a sparse record, the Board decided not to change its
cost of capital methodology for 2005. See Railroad Cost of
Capital—2005 at *5–6. It did, however, start a separate
rulemaking proceeding to determine, among other things,
whether switching from DCF to a new method was warranted.
Methodology to be Employed in Determining the Railroad
Industry’s Cost of Capital, 2006 WL 2692727 (Sept. 15, 2006)
(“Methodology to be Employed—2006”). After a notice and
comment period in which various interested parties including
WCTL, BNSF, and AEPT participated, the Board adopted
CAPM for calculating cost of equity capital for 2006 and 2007.
Methodology to be Employed—2008, 2008 WL 162591, at *1;
see also AEP Tex. N. Co. v. BNSF Ry. Co., STB No. 41191 (Sub
No. 1), 2007 WL 2680223 (Sept. 7, 2007), at *4 (“AEP Texas
and BNSF were among the interested parties that filed”
comments). For its 2008 calculation, the Board decided to
7
calculate the cost of equity by using both CAPM and an updated
DCF model and averaging the outputs of the two methods. Use
of a Multi-Stage Discounted Cash Flow Model in Determining
the Railroad Industry’s Cost of Capital, 2009 WL 197991, at *1
(Jan. 23, 2009).
While the Board proceeded with its rulemaking, WCTL
filed a petition to this court arguing the Board’s use of the DCF
model for its 2005 calculation was arbitrary and capricious
under the Administrative Procedure Act, 5 U.S.C. § 706(2)(A).
Just before oral argument, the Board released its decision to use
CAPM for 2006, and in oral argument counsel for the Board
conceded that the DCF methodology was flawed. However, the
Board also argued that, because of its recent decision to change
the methodology for future years, the proper way to resolve the
issue of whether to use CAPM for prior years’ calculations
would be through administrative proceedings reopening past
years’ decisions. The administrative proceedings could either
address the industry-wide proceedings in 2005 and prior years,
or deal with individual rate cases. If there was enough evidence
to establish that using DCF in prior years was material error, the
Board would retroactively adjust the calculations. This court
therefore denied the petition before it, stating that the 2005 cost
of capital decision was “now ripe for review pursuant to a
petition to reopen under 49 U.S.C. § 722(c).” W. Coal Traffic
League v. STB, 264 F. App’x 7, 8–9 (D.C. Cir. 2008). That
provision allows any interested person to petition to reopen an
STB decision where there is “material error, new evidence, or
substantially changed circumstances.” 49 U.S.C. § 722(c).
However, neither WCTL nor any other party petitioned to
reopen the 2005 industry-wide cost of capital decision.
While WCTL was challenging the industry-wide cost of
capital calculations, AEPT was engaged for several years in its
own rate case before the Board, challenging the reasonableness
8
of the rates set by the Board for the TNR railroad. AEPT
argued, among other things, that the DCF-derived cost of equity
calculations were flawed, and particularly disputed the 2005
figure once the Board published it. See AEP Tex. N. Co. v.
BNSF Ry. Co., STB Docket No. 41191, 2007 WL 2680223, at
*86 (Sept. 7, 2007). In 2006, the Board held the proceeding in
abeyance while it resolved the industry-wide rulemaking. Major
Issues in Rail Rate Cases, STB Ex Parte No. 657 (Sub No. 1),
2006 WL 513502, at *2 (Feb. 27, 2006). In 2007, the Board
decided, inter alia, that the DCF-derived cost of equity estimates
were reasonable. AEP Tex. N. Co. v. BNSF Ry. Co., STB
Docket No. 41191, 2007 WL 2680223, at *19–20 (Sept. 7,
2007).
In 2008, AEPT petitioned for reconsideration pursuant to 49
U.S.C. § 722(c) on various issues. The Board reopened the case
only with respect to the cost of equity. AEP Tex. N. Co. v. BNSF
Ry. Co., STB Docket No. 41191, 2008 WL 2216062, at *7–9
(May 27, 2008). In the reopened case, AEPT argued that the
rates for the years 2000 to 2005 should be recalculated using
CAPM. AEPT also requested that the Board either recalculate,
using CAPM, the historical cost of capital figures used to
forecast growth, or discard altogether the 1998 to 2005 numbers
in the historical average. See Opening Fourth Supplemental
Evidence of Complainant AEP Texas North Company in
Response to the Board’s Decision on Reconsideration 15, 31
tbl.4 (Aug. 8, 2008) (“Opening Fourth Supp. Evid.”); AEP Tex.
N. Co. v. BNSF Ry. Co., STB Docket No. 41191, at 11 (May 15,
2009) (“2009 Decision”).
In May 2009, the Board denied AEPT’s petition for
reconsideration, rejecting the use of CAPM for any of the years
from 2000 to 2005. 2009 Decision at 4, 10. The Board also
declined to change or excise the 1998 to 2005 DCF calculations
in its historical cost of capital average. Id. at 11. In its decision,
9
the Board concluded that “it would be poor public policy to
depart from the previously published figures” for two reasons,
id. at 8. First, the railroad industry and its investors relied on the
published figures. Changing the calculations retroactively
would undermine those expectations as well as erode confidence
in future cost of capital calculations. Id. Second, the Board
concluded that the DCF calculations were reasonable. Even
though CAPM is now the industry norm, the DCF method
produced numbers “easily within the range produced by [] other
finance models” in every year but 2005. Id. at 10 (citing id. at
Chart 1). And as for the 2005 number, “the figure does not vary
significantly more than other models that produce the highest or
lowest estimate in a given year.” Id.
AEPT petitions this court under the Administrative
Procedure Act.
II. Standard of Review
AEPT brings its petition under 5 U.S.C. § 706(2)(A), which
requires courts to set aside an agency decision if it is “arbitrary,
capricious, an abuse of discretion, or otherwise not in
accordance with law.” This standard requires a “rational
connection between the facts found and the choice made.”
Motor Vehicle Mfrs. Ass’n v. State Farm Mut. Auto. Ins. Co.,
463 U.S. 29, 43 (1983) (“State Farm”) (quoting Burlington
Truck Lines, Inc. v. United States, 371 U.S. 156, 168 (1962)).
This court must vacate the Board’s denial of AEPT’s petition if
the Board made “a clear error of judgment” or if it “entirely
failed to consider an important aspect of the problem.” Id.
(quotation marks omitted). However, because the Board acts
“at the zenith of its powers” when it sets rail rates, Burlington N.
R.R. Co. v. STB, 114 F.3d 206, 210 (D.C. Cir. 1997) (quoting
Am. Trucking Ass’ns v. United States, 627 F.2d 1313, 1320
(D.C. Cir. 1980)) (internal quotation mark omitted), we
10
recognize that the Board is “entitled to particular deference.” Id.
As long as the Board “has rationally set forth the grounds on
which it acted, . . . this court may not substitute its judgment for
that of the agency.” BNSF Ry. Co. v. STB, 453 F.3d 473, 480
(D.C. Cir. 2006) (quoting McCarty Farms v. STB, 158 F.3d
1294, 1301 (D.C. Cir. 1998)).
In addition, the history of the cost of equity methodology
debate presents a concern about what standard the Board was
required to apply in its decisionmaking process. Because AEPT
requests that the Board recalculate past years’ rates with the
methodology adopted in the industry-wide proceedings in 2006
and 2007, both the Board and AEPT discuss the issue in terms
of retroactive application of a new rule. See 2009 Decision at 7.
In its decision denying AEPT’s petition, the Board therefore
analyzed AEPT’s petition using a balancing test similar to one
created by this court to evaluate whether a change in
methodology can be given retroactive effect. Citing Williams
Natural Gas Co. v. FERC, 3 F.3d 1544 (D.C. Cir. 1993), the
Board decided to “conduct the kind of balancing test described
by the court in Williams Natural Gas.” 2009 Decision at 8.
However, the Board did not apply the actual analysis outlined by
that case, which was a five factor test originally set forth in
Retail, Wholesale & Department Store Union v. NLRB, 466 F.2d
380, 390 (D.C. Cir. 1972). See Williams Natural Gas, 3 F.3d at
1553–54. Instead, the Board created its own two factor analysis,
weighing “the degree of reliance by the railroad industry on our
prior findings, and whether the prior findings appear to be
within the bounds of reasonable predictions for the industry’s
cost of equity.” 2009 Decision at 8.
In its petition to this court, AEPT argues the Board
misapplied the retroactivity analysis. AEPT quotes
Consolidated Edison Co. v. FERC, 315 F.3d 316 (D.C. Cir.
2003), to posit that the Board should have looked not at
11
industry’s reliance on the DCF model and the reasonableness of
said methodology, but only at whether “the parties before the
agency are given notice and an opportunity to offer evidence
bearing on the new standard, and the affected parties have not
detrimentally relied on the established legal regime.” Id. at 323.
(internal citation omitted). But in trying to cabin the Board’s
analysis within that case’s framework, AEPT overestimates the
similarity between this case and Consolidated Edison.
Consolidated Edison addressed whether any legal principle
mandates retroactive application to pending cases of a new
policy statement not given the force of law. 315 F.3d at 319.
In Williams Natural Gas, we considered whether “a rule
announced in an agency adjudication may be given retroactive
effect.” 3 F.3d at 1553 (emphasis added). In this case, the
Board created a new rule through a full notice and comment
legislative rulemaking under 5 U.S.C. § 553. As neither
Consolidated Edison nor Williams Natural Gas is on point, the
Board was certainly not bound to employ the factors outlined in
either of the cases.
It is also worth noting that AEPT does not request a truly
retroactive application of the new rule. The promulgated rule
establishing CAPM for the 2006 and 2007 industry-wide
proceedings did not disturb the industry-wide calculations for
prior years, and in this case the Board did not reopen those
industry-wide calculations. Rather, the Board revisited whether,
in light of “material error, new evidence, or substantially
changed circumstances,” 49 U.S.C. § 722(c), the Board should
recalculate the maximum rate BNSF was allowed to charge
AEPT from 2000 to 2005. AEPT’s petition did not request a
retroactive application of the new rule, but rather argued that the
new rule revealed that using the old method was a mistake.
In sum, this court must simply determine whether the
Board’s evaluation withstands the arbitrary and capricious
12
standard outlined by 5 U.S.C. § 706(2)(A). Because this court’s
precedents on retroactivity do not dictate the Board’s approach,
the use of “the kind of balancing test described by the court in
Williams Natural Gas,” 2009 Decision at 8, neither forecloses
nor guarantees that the Board’s decision satisfies those
requirements.
III. Analysis
AEPT argues the Board’s decision must be overturned
because it is inconsistent with the Board’s prior representations
to this court about how the Board would reassess the DCF-
derived cost of equity estimates. While incompatibility of an
agency’s action with the agency’s prior representations to this
court might be evidence of arbitrary and capricious
decisionmaking, the Board did not act inconsistently in this
instance. Before this court in Western Coal Traffic League v.
STB, 264 F. App’x 7 (D.C. Cir. 2008), the Board did represent
that the use of the DCF methodology in the industry-wide
determination for 2005 did not establish that AEPT necessarily
would have to submit to the DCF-derived number. In its brief
to this court in that case, the Board explained that AEPT could
seek reopening of its then-pending rate case under 49 U.S.C. §
722(c) if AEPT believed the Board’s final rule switching to
CAPM materially affected the outcome of AEPT’s rate case.
And at oral argument, the Board conceded that CAPM was a
better method, but urged the court to allow the Board to address
the issue first through § 722(c) proceedings. When AEPT filed
its petition for reconsideration, the Board did exactly what it
said it would: address the issue. The Board simply arrived at a
result AEPT did not like. But a promise to allow AEPT to avail
itself of the reconsideration process is not the same as a promise
to change the prior years’ rates. The Board has not acted
inconsistently with its prior representations to this court.
13
Therefore, any evidence of arbitrary and capricious
decisionmaking must come from the 2009 Decision itself. As
described above, the Board divided its analysis into two major
sections: reliance by the industry on the previously-published
figures and the reasonableness of the DCF-derived cost of equity
estimates. On the first factor, reliance, the Board concluded that
“there has been significant investment-backed reliance by the
railroad industry on our prior cost-of-capital findings.” 2009
Decision at 8. AEPT argues this was an improper conclusion
because the decision looks at reliance by industry investors in
general, not by BNSF in particular. In addition, AEPT points
out that the Board’s conclusion seemingly contradicts prior
decisions by the Board that say CAPM “dominates the private
sector” and is the “industry norm.” Methodology to be
Employed in Determining the Railroad Industry’s Cost of
Capital, STB Ex Parte No. 664, 2007 WL 2363415, at *4, *5
(Aug. 20, 2007) (“Methodology to be Employed—2007”). How,
argues AEPT, could the industry have relied on the DCF
numbers when the industry was using CAPM? Finally, AEPT
argues that even if investors did rely on the Board’s use of DCF,
there is no evidence such reliance was reasonable. At least by
2005, BNSF and other railroads knew the DCF methodology
was being seriously challenged by WCTL in the industry-wide
proceedings and by AEPT in its individual rate case.
AEPT’s complaint that the Board looked to the industry
rather than BNSF does not undermine the Board’s analysis,
especially because AEPT did not present any information to the
Board to suggest that BNSF, a large and prominent carrier,
operates differently from the industry as a whole. The Board did
not make BNSF-specific conclusions, but it did extrapolate from
what it knew about investors generally to the case at hand. 2009
Decision at 8. Noting that BNSF itself makes significant capital
investments—over $9 billion between 2004 and 2007—the
Board pointed out that “[r]ailroads and investors . . . make
14
investment decisions based in part on” the published cost of
capital figures. Id. “If we change that figure retroactively here,
we not only undermine settled expectations but we erode
investor confidence in future cost-of-capital findings.” Id.
(emphasis in original). Because investors might not invest
sufficiently in railroads if they have no confidence in the
stability of the Board’s cost of capital figures, the Board decided
to “set aside our cost-of-capital findings only if the prior
published findings are shown to clearly fall outside a reasonable
range.” Id. at 9. It was reasonable for the Board to use its
general understanding about railroad investors in making its
decision in this case concerning two specific parties.
Nor does the Board’s conclusion that railroads and investors
relied on the DCF-derived numbers conflict with the Board’s
published acknowledgment that CAPM has become the industry
standard. Railroads like BNSF could rely on the Board’s use of
DCF to model cost of equity even if they used CAPM internally.
A simple hypothetical illustrates this point: If investors
concluded, using CAPM, that it would cost $1 million to run a
railroad, including the cost of equity, but knew the Board would
allow rates charging up to $1.5 million because of the DCF
calculation, of course they would rely on the Board’s published
figures in deciding whether to invest in the railroad—and decide
that it was a very good deal indeed. The fact that one method is
the industry standard does not mean the industry cannot rely on
the Board’s use of a different method.
AEPT’s final argument about the Board’s analysis of
BNSF’s reliance is more persuasive. The Board’s discussion of
reliance makes no distinction between reliance in one year
versus any other. But the circumstances surrounding the 2005
cost of capital determination are different from other years. In
its 2005 cost of capital determination retaining DCF-derived
figures, the Board explained it was instituting a separate
15
proceeding to review the cost of equity methodology. Railroad
Cost of Capital—2005 at *5. Meanwhile, WCTL petitioned this
court for review of the 2005 calculation, and during oral
argument the Board itself represented it might be appropriate to
reconsider at least the 2005 figures. WCTL v. STB, No. 07-1064,
Oral Argument Tr. at 10, 15. AEPT brought evidence of all of
these circumstances before the Board in the 49 U.S.C. § 722(c)
proceeding. See Opening Fourth Supp. Evid. at 7–8. But in the
2009 Decision, the Board completely failed to address the
unique circumstances of the 2005 cost of capital determination.
The Board did not consider whether railroads and investors
actually or reasonably could have relied on the permanence of
the 2005 cost of capital determination when it was undermined
by shippers in litigation and even by the Board itself. By relying
only on generalized conclusions about how industry players rely
on cost of capital determinations, the Board “entirely failed to
consider an important aspect of the problem,” making its
assessment of reliance for the year 2005 arbitrary and
capricious. State Farm, 463 U.S. at 43.
The Board’s second consideration, the reasonableness of the
DCF calculations, also included flaws, especially with respect
to the 2005 figure. The Board began its analysis by addressing
generally the arguments AEPT made in the reopened case about
the superiority of CAPM over DCF. AEPT’s evidence broadly
stated that CAPM is more accurate than DCF, that in the
Methodology to be Employed—2008 decision the Board itself
had concluded that DCF overstated the cost of equity, and that
applying CAPM would be consistent with the Board’s decision
to use CAPM in future years. Opening Fourth Supp. Evid. at
15-21. The Board responded by acknowledging that “CAPM is
a more modern and better accepted method for estimating the
cost of equity than the single-stage DCF model.” 2009 Decision
at 9. However, the Board explained, old calculations are not
invalidated just because methods improve—the Board’s switch
16
in 2006 to CAPM for the cost of equity calculation did not
invalidate the use of DCF in prior years, just as the Board’s
decision to use a blended CAPM/DCF approach in 2008 did not
invalidate the use of CAPM alone in 2006 and 2007. Id. AEPT
argues that the Board did not consider changed circumstances or
new evidence as it should have under 49 U.S.C. § 722(c), but
these statements by the Board in its decision are evidence of the
Board’s consideration. The fact that the Board did not agree that
the changed circumstances warranted changing prior years’
calculations does not by itself mean the Board acted arbitrarily
or capriciously or failed to consider seriously AEPT’s evidence.
However, the rest of the Board’s analysis of reasonableness
stands on less solid ground. In its discussion, the Board relied
heavily on a chart it created comparing the cost of equity
estimates for the years 1994 to 2007 derived by five different
methodologies: the Board’s calculations (a DCF model through
2005, then CAPM in 2006 and 2007); CAPM; and three
commercially produced methodologies published by
Ibbotson/Morningstar, including a single-stage DCF model, a
multi-stage DCF model, and a model called “3-Factor Fama-
French.” 2009 Decision at 9 n.25, 10 & Chart 1. The chart
reveals the fluctuating estimates produced by each methodology,
and demonstrates the large variation possible in outcomes. In
some years all five methodologies produce similar costs of
equity, while in other years they vary greatly, but no method
consistently produces either the highest or lowest estimates.
Each methodology sometimes produced estimates on the higher
end of that year’s range, and sometimes produced estimates on
the lower end of that year’s range.
Referencing the chart, the Board concluded “that various
reasonable and commercially available financial models produce
a range of estimates for the cost of equity. . . . Simply because
one estimate is the highest or lowest in a given year does not
17
mean that it is invalid or even the least accurate.” Id. at 10.
Certainly this is true, but it only establishes that one cannot
determine if the Board’s figures were invalid just by comparing
them to the four other models’ figures. It is problematic, then,
that the Board apparently decided that such a comparison was all
that was necessary to conclude its estimates were reasonable:
The chart also reveals that the Board’s prior determinations
provide a reasonable estimate of the cost of equity for the
hypothetical SARR posited in this case. For every year
except 2005, the Board’s estimate falls easily within the
range produced by the other finance models. . . . Yet even
then [in 2005], the figure does not vary significantly more
than other models that produce the highest or lowest
estimate in a given year. Thus, we do not regard the
increase as sufficiently large to justify setting aside the
industry’s expectation that we would use that finding as the
target rate of return for that year.
Id.
The Board’s analysis contains serious flaws. First, the
record before the Board included no information about any of
the four methods the Board used as comparisons to its own
published cost of equity figures. Instead, it appears that the
Board chose to use the Ibbotson/Morningstar models of its own
accord, saying that though “the Morningstar cost-of-equity
estimates were not submitted by either party in this proceeding,
we take official notice of these publicly available cost-of-equity
estimates for the railroad industry.” Id. at 9 n.25. But by
choosing to use estimates not in the record, the Board foreclosed
any opportunity for AEPT to refute the validity or usefulness of
those models, or to offer counterexamples of other
methodologies. In addition, the Board never explained why
those methods were chosen or how they work, or even whether
18
they are a representative sample of the types of methodologies
used by the railroad industry or its investors. And the Board
never explained why in some years the Morningstar single-stage
DCF model and the Board’s single-stage DCF model produce
such different cost of equity figures.
Most problematic, however, is the Board’s use of the chart
to justify its 2005 calculation. The block quote above represents
the entirety of the Board’s analysis about that year. The Board
concluded that the 2005 calculation “does not vary significantly
more than other models,” but gives no information about how it
arrived at this conclusion. As far as can be ascertained from the
decision, the Board’s analysis consisted of nothing more than an
estimation measured by a cursory glance at the graph. Even in
its briefing to this court, the Board employed a perfunctory
analysis, largely resting its argument about reasonableness on
the assertion that “a review of the chart shows visually [that] the
Board’s 2005 DCF figure is not out of line” with the estimates
produced by the other methodologies. Resp. Br. at 40.
Before this court, AEPT attempted to discredit the Board’s
2005 figure by deriving the arithmetic mean, standard deviation,
and 90% confidence interval of the five figures included in the
Board’s graph, and showing that the Board’s figure falls outside
the confidence interval. The Board counters with the fact that
the 2005 CAPM calculation in its graph also falls outside the
confidence interval—although we note that the CAPM figure the
Board criticizes must stem from a different CAPM model than
the one actually employed by the Board in 2006 and 2007, as the
Board’s published figures for 2006–2007 differ from the data
captioned as “CAPM” for those years. The Board also claims
that other statistical tests better suited to small sample sizes
indicate that the Board’s 2005 figure was not outside a
reasonable range.
19
We will not attempt to decide the merits of the
methodologies. To paraphrase this court’s admonishment to
parties in a slightly different context, we do not sit as a panel of
statisticians, but as a panel of generalist judges. See City of LA
v. U.S. Dep’t of Transp., 165 F.3d 972, 977 (D.C. Cir. 1999)
(“[W]e do not sit as a panel of referees on a professional
economics journal . . . .”). We do note, however, that these
arguments encapsulate exactly the kind of analysis in which the
Board should have engaged before concluding that its 2005
calculation did “not vary significantly more” than other methods
in other years. We recognize the difficulty of determining
whether a model produces estimates so inaccurate as to be
invalid, especially for a value as elusive as the cost of equity.
But that does not mean the Board was free to choose methods
for comparison without opportunity for comment by the parties
and without any apparent rigor in its analysis. The Board’s
particularly cursory analysis of the 2005 cost of equity estimates
constitutes arbitrary and capricious decisionmaking.
This leaves one final issue: whether the Board acted
arbitrarily or capriciously when it decided not to recalculate its
historical cost of equity average used to forecast growth. The
Board declined to restate the past years’ cost of equity estimates
using CAPM or to use only the published 2006 and 2007
estimates in its historical average. Instead, the Board elected to
maintain its normal practice of using the average of the
historical cost of capital figures starting with the construction
start date of the SARR. 2009 Decision at 11. In so deciding, the
Board cited two reasons. First, because the Board concluded
there was “no basis to restate the 2005 estimate” to readjust
BNSF’s actual rates for those years, the Board saw no reason to
restate the estimates for the purpose of forecasting. Id. Second,
averaging all historical figures reduces the impact that any one
year’s aberrant estimate would have on the overall forecast.
Using the DCF-derived estimates in the forecasts rather than
20
using only the 2006 and 2007 numbers decreases the chance that
the forecast will be skewed if the 2006 and 2007 calculations
produced flawed estimates.
These are reasonable justifications by themselves, and we
hold that the Board’s decision is valid insofar as it declines to
throw out all years’ estimates except those for 2006 and 2007.
We also hold the Board’s decision to be adequate with respect
to the years prior to 2005. But the Board also explicitly
depended on its prior conclusions about the industry’s reliance
on and the reasonableness of the 2005 estimate. Because those
conclusions were arbitrary and capricious, the Board must
reassess its use of the 2005 DCF-derived figure for historical
averaging. If on remand the Board is swayed by AEPT’s
evidence with regard to the 2005 estimate for use in calculating
that year’s maximum rate, we see no reason why the Board
would not also be swayed to adjust the 2005 estimate it uses in
the cost of equity forecasts for the TNR.
IV. Conclusion
In its introduction to its analysis of the continued validity of
the DCF-derived costs of equity, the Board explained that “the
exact cost of equity in a given year remains an essentially
unknowable number and any method we adopt will produce only
an estimate.” 2009 Decision at 7. We recognize that truth, but
even an imperfect estimate must be justified to satisfy 5 U.S.C.
§ 706(2)(A). We hold that, with respect to all years but 2005,
the Board met that standard because the decision exhibits a
“rational connection between the facts found and the choice
made.” State Farm, 463 U.S. at 43 (quotation omitted).
However, the Board failed to address the concerns raised
regarding the 2005 cost of equity. Because this constitutes a
failure “to consider an important aspect of the problem,” id., we
vacate the decision and remand to the Board to reassess its
21
decisionmaking for the 2005 cost of equity estimate.
So ordered.