Opinion for the Court filed by Circuit Judge STEPHEN F. WILLIAMS.
Concurring opinion filed by Circuit Judge CLARENCE THOMAS.
STEPHEN F. WILLIAMS, Circuit Judge:In its Opinion No. 240, the Federal Energy Regulatory Commission picked the figure 15.1% as the return on equity for Tennessee Gas Pipeline Company for the period June 1, 1982 through January 31, 1983. Tennessee Gas Pipeline Co., 32 FER.C 11 61,086 (1985). Because of a serious methodological error — FERC had derived one of the key ingredients in its calculation from a logically irrelevant prior period — we reversed and remanded, saying: “[s]uch result-oriented manipulation of an objective ratemaking calculation is patently arbitrary and capricious decisionmaking.” Public Service Comm’n of New York v. FERC, 813 F.2d 448, 465 (D.C.Cir.1987). On remand, the Commission again picked *33515.1%. Tennessee argues not only that the decision on remand is arbitrary and capricious, but that it is so egregious as to require us to set the rate of return ourselves. We agree that the Commission erred fatally and that we must reverse; as ratemaking is for the Commission and not for us, however, we must again remand. ******
In the early 1980s Tennessee filed a number of general rate increases under § 4 of the Natural Gas Act, 15 U.S.C. § 717c (1988), which the Commission consolidated. A major issue, and the one that survives through to this case, was choosing the appropriate rate of return on equity, an inevitable component of cost-of-service rate-making. Recognizing that utility investors must be allowed an opportunity to earn returns sufficient to “attract capital,” Federal Power Commission v. Hope Natural Gas Co., 320 U.S. 591, 605, 64 S.Ct. 281, 289, 88 L.Ed. 333 (1944), and “to compensate [the] investors for the risks assumed,” id., the Commission endeavors to set a utility’s rate of return on equity at its cost of equity capital. “The cost of capital is the minimum rate of return necessary to attract capital to an investment.” A. Lawrence Kolbe et al., The Cost of Capital: Estimating the Rate of Return for Public Utilities 13 (1984).
In Opinion No. 190, Tennessee Gas Pipeline Co., 25 FERC ¶ 61,020 (1983), the Commission addressed the two periods immediately preceding the present one and set the cost of equity capital (and hence the return on equity) at 15.95%. It reached this number by taking the midpoint of a “zone of reasonableness.” Id. at 61,097. The Commission evidently found the lower bound of this zone in its staff’s recommendation of 15%,1 and the upper bound in a discounted cash flow (“DCF”)2 analysis of Tennessee’s parent company, Tenneco, Inc. Id. at 61,093.
Opinion No. 190 did not reach the period at issue here, but an Administrative Law Judge sought to apply its methodology to the period. Tennessee Gas Pipeline Co., 25 FERC 1163,052 (1983). She believed that the 15% lower bound had been the result of risk premium analysis, which typically takes the risk-free rate of return on U.S. government bonds and adds an estimated premium for the greater risk of the particular stock. Accordingly, she used a kind of “reverse engineering” to arrive at a lower bound of 13.2%.3 She set the upper bound at 16.93%, the unrevised DCF figure from Opinion No. 190. Id. at 65,170. Finally, she chose the midpoint of this zone of reasonableness, 15.1%. Id.
In Opinion No. 240, Tennessee Gas Pipeline Co., 32 FERC 1161,086 (1985), the Commission expressly adopted the AU’s analysis and conclusion, noting “[t]he close proximity of the issuance of Opinion No. 190 to [the AU’s] decision.” Id. at 61,225. On Tennessee’s request for rehearing, the *336Commission rejected Tennessee’s argument that it was improper to use DCF figures from prior periods. Tennessee Gas Pipeline Co., 33 FERC 1161,005 (1985). We found the use of obsolete data arbitrary and capricious, and reversed and remanded. Public Service Comm’n of New York v. FERC, 813 F.2d at 462-65.
On remand, the Commission again used the ALJ’s reverse-engineered risk premium figure of 13.2% as its lower bound and used an updated DCF figure of 18.79% as its upper bound. Tennessee Gas Pipeline Co., 46 FERC 1161,089 at 61,383 (1989) (“Order on Remand”). While recognizing that the midpoint of this zone of reasonableness was 15.99%, id., it instead chose its old favorite, 15.1%, id. at 61,384. To justify choosing well below the midpoint, the Commission noted that the price of Tenneco’s stock rose in the six months following the end of the relevant period, so that its dividend yield (dividend divided by price) fell. Id. at 61,383. Use of this out-of-period data would, the Commission noted, result in a DCF or upper limit of only 16.84%. Id. (Explicit use of that figure as the upper bound would have yielded a midpoint of 15.02%,4 a rate quite similar to what the Commission adopted.)
On Tennessee’s request for rehearing, the Commission acknowledged that its Order on Remand failed to articulate its rationale, and explained that its reference to the out-of-period data was “made only to show the existence of a lag in the decline in dividend yields following the decline in interest rates.” Tennessee Gas Pipeline Co., 49 FERC 1161,392 at 62,420 (1989) (“Order Denying Rehearing”). On the basis of this supposed lag, the Commission explained that it made a “pragmatic adjustment.” Id. at 62,419. Thus the supposed lag — the proposition that investors are not able to account fully for the effects of a decline in interest rates on their investment alternatives until some six months or more after those rates are published — is the sole basis for the Commission’s use of a rate of return below the midpoint.
* * He * * *
The Commission’s approach to estimating the cost of equity capital appears to be a two-step process, in which it first frames a zone of reasonableness with the estimation tools of its choice. Then, in the absence of evidence that leads the Commission to prefer one estimate over the other, it sets the rate of return at the average of those boundary figures. If “other factors” warrant a preference one way or the other, the Commission makes a suitable “pragmatic adjustment”.
We have no quarrel with this general methodology. Even if we did, we have no authority to insist that the Commission use “any single formula or combination of formulae.” Hope, 320 U.S. at 602, 64 S.Ct. at 287. But the notion of lawfulness requires insistence that the chosen framework not collapse in practice into a standardless exercise of Commission discretion resting on no more than an assertion of “expertise”. See, e.g., Greater Boston Television Corporation v. FCC, 444 F.2d 841, 850-52 (D.C.Cir.1970). We therefore turn to an evaluation of the “pragmatism” behind the adjustment.
The trend of Tenneco’s dividend yield diverged from the trend of interest rates during the disputed period. One held steady, the other fell. The Commission introduced its “lag” theory as the explanation. That solution, however, disregards the evidence and violates the principles underlying the Commission’s own methodology.
As the Commission noted, there is a direct relationship between interest rates and dividend yields. Everything else being equal, a decline in interest rates means a decline in dividend yields, as stocks and bonds compete for investors’ capital. A “drastic drop in interest rates attracts] capital away from bonds and into stocks, causing a rise in stock prices and a decline in dividend yields____” Order on Remand, 46 FERC at 61,383; see also Roger *337A. Morin, Utilities' Cost of Capital 30 (1984).
Dividend yields can thus be expected to move with interest rates so long as other things remain equal. In the case of a divergence, such as the Commission found for Tenneco in the disputed period, the natural inquiry is whether factors special to Tenneco (or to some broader segment of industry that shared its trend) could account for the divergence. In other decisions the Commission has recognized exactly this point:
[Although changes in long-term interest rates often correspond closely with changes in the cost of common equity for the utility industry as a whole, they do not necessarily correspond with changes in the cost of common equity for a particular public utility. Moreover, they do not reflect company-specific changes in business or financial risk____
Orange & Rockland Utilities, Inc., 45 FERC ¶ 61,252 at 61,754 (1988) (footnote omitted).
Thus, implicit in the Commission’s assertion that the “decline [in interest rates] was not yet reflected in the relevant [dividend yield] data,” Order on Remand, 46 FERC at 61,383, is a finding of fact that nothing (save interest rates) changed during the disputed period. Under the Natural Gas Act the Commission’s factual findings are conclusive only “if supported by substantial evidence.” 15 U.S.C. § 717r(b) (1988). The Commission has not pointed to any evidence that would support this finding.
Indeed, the evidence offered by Tennessee strongly suggests changes in company-specific risk. First, the spread between Tenneco’s bond yields and the risk-free yield increased substantially, see Order Denying Rehearing, 49 FERC at 62,421 n. 26, indicating that investors thought Tenneco was becoming riskier. While that may have been due to components of Tenneco’s business other than Tennessee, Tennessee also offered evidence of stiffening competition from oil as a result of declining oil prices (which declined in price from $31.74 to $29.49)5 and from other gas suppliers, with a concomitant likelihood that Tennessee would run up greater liability under the “take-or-pay” contracts by which it secured its gas. Joint Appendix 33-35B. It is thus hard to imagine how the Commission could support a finding that interest rates were the only relevant dynamic force in an otherwise static world. In any event, Tennessee did offer evidence of countervailing risk factors, and it had “a right to have its proof either accepted or appropriately refuted.” Union Elec. Co. v. FERC, 890 F.2d 1193, 1204 (D.C.Cir.1989).
Even if there were no direct evidence of changing risks for Tennessee, however, the Commission’s lag theory would still be in intolerable conflict with the principles the Commission has endorsed in adopting the DCF method itself. DCF analysis works from the proposition that the price of a stock is the current value of all expected future cashflows, discounted at the rate of return.6 The key equation, r = Di/Po + g, employs the current price of the utility, because that price is understood to represent the best possible assessment of the available information about the utility. See, e.g., Morin, Utilities’ Cost of Capital 119-20. If the market is in fact unable to promptly reflect information so widely publicized as risk-free interest rates, DCF theory collapses.
More generally, the Commission’s lag theory implies a frontal assault on “the cornerstone of modern investment theory,” the Efficient Market Hypothesis. Id. at 140 n. 4. In its “semi-strong” form, the hypothesis says that stock prices will react promptly to new public releases of information and thus “fully reflect all public information.” Frank K. Reilly, Investment *338Analysis and Portfolio Management 215 (1989). If, as the Commission urges, the market cannot promptly digest information about interest rates, it must be quite inefficient.
In fact, if the stock market is such a laggard as the members of the Commission say, they would do well to abandon their regulatory work and turn to exploitation of their theory. At the conclusion of any sharp change in interest rates, they could buy stocks or sell them short, as appropriate, and then await the market’s leisurely response — the Commission claims it may take more than six months. See Order Denying Rehearing, 49 FERC at 62,420. That investors can have been so obtuse as to miss this opportunity — and thus compete away the resulting profits — strains credulity-
13] The Commission proffers two authorities for its “lag theory.” First, the Commission points to itself as an expert in the field. The Commission’s expertise lies not in financial theory itself, but in the application of the teachings of financial and economic theory to the setting of rates for regulated utilities. In any event, expertise cannot be used as a cloak for fiat judgments. Second, in its Order Denying Rehearing, the Commission cited J. Cohen, E. Zinbarg & A. Zeikel, Investment Analysis and Portfolio Management 224-30 (5th ed. 1987). Those pages in fact include the assertion that “investors have come to recognize that the stock market tends to move inversely to interest rates, with a lag that varies from three to nine months.” Id. at 225. They also include a graph suggesting such a lag over one specific two-year period. Id.
As a basis for overthrowing the principles underlying DCF and efficient market theory, this short excerpt is weak. First, it offers no explanation for its obvious conflict with those widely held theories. Second, the following page contains a graph covering a 20-year period in which stock and bond rates appear generally to move in lock step. Id. at 226. Third, the excerpt offers nothing on the subject of utility stocks, which, because of their (historically) low levels of risk, have been especially close substitutes for bonds. In fact, the very book cited by the Commission for its lag theory points out that utility stocks are “very sensitive to changes in interest rates, falling in price if interest rates rise sharply and increasing in price if interest rates decline.” Id. at 27. Finally, it ignores direct evidence that the market in fact assimilates interest rate information with lightning speed. Professors Pierce and Roley, for example, found that most of the response of stock prices to unexpected announcements about money supply, inflation, real economic activity, and the discount rate did not persist beyond the day of announcement. Douglas K. Pierce and V. Vance Roley, “Stock Prices and Economic News,” 59 Journal of Business 49 (January 1985).
Of course the Commission is free to adopt a minority position in the financial and economic communities. Tennessee does not argue that the 14th Amendment enacted the Efficient Market Hypothesis, or DCF for that matter. Nor, indeed, does it claim the Natural Gas Act or the Administrative Procedure Act did so. If the Commission proposes to reject either the Efficient Market Hypothesis or DCF methodology, we therefore assume that it is free to do so. But it must say so, and, if the rejection is inconsistent with prior decisions, explain the change. Greater Boston, 444 F.2d at 852.
In fact, the Commission appears quite wedded to DCF analysis and to efficient market theory as its theoretical mainstay. In Montaup Electric Co., 38 FERC ¶ 61,252 (1987), for example, the Commission adopted the DCF methodology over risk premium analysis for a period of rapidly declining interest rates and reasoned that “a market-oriented analysis such as a DCF analysis accounts for all risk factors perceived by investors.” Id. at 61,866. At about the same time as this court’s first remand, it published its third annual “Generic Determination of Rate of Return on Common Equity for Public Utilities,” in *339which it defended the use of the DCF methodology against attacks based on criticism of the Efficient Market Hypothesis. Order No. 461, III FERC Regulations Preambles U 30,722 (1987). It declared enthusiastically, “The concept of an efficient market is astonishingly simple and remarkably well supported by the facts.”7
Indeed, if there is any method on which the Commission has recently frowned, it is risk premium analysis. Thus in Montaup Electric Co., it said, “Because a risk-premium analysis can accentuate erratic market conditions and tends to over-emphasize recent market changes by producing too high a return when capital costs are steeply rising and too low a return when they are steeply falling, this type of analysis must be used with caution.” 38 FERC at 61,869 n. 101 (emphasis added). It echoed the same theme in Pennsylvania Power Co., 26 FERC. 1t 61,354 (1984): “[W]hile the risk premium analysis is one tool which may aid us, it has become apparent that this approach can produce distortions during certain time periods, most particularly when interest rates are steeply rising or steeply declining.” Id. at 61,779 (emphasis added). See also 51 Fed.Reg. 359-60 (1986) (Order No. 442) (Commission “is reluctant to place any great weight on risk premium analyses in general other than those based on a simple ranking of securities.”).
Thus, although the Commission has extolled DCF analysis on efficient market premises, here it slighted the DCF upper bound, invoking a theory that undercuts those premises; and although it has deprecated risk premium analysis, it in effect increased the relative weight of that analysis (or of a lower bound that purported to be based thereon) in precisely the circumstances that it has said require skepticism.
The Commission points us to Boston Edison Co. v. FERC, 885 F.2d 962 (1st Cir. 1989), as proof of its consistency. The case is readily distinguishable. There an AU used data ending in April 1985 to calculate a zone of reasonableness for the company’s rate of return. Id. at 966-67. The Commission, acting in 1988, noted that interest rates had dropped 2.59 points between the six months ending April 1985 and the six months ending March 1988. Dividend yields had also declined. Id. Accordingly, the Commission sought to update the AU’s 1985 analysis for purposes of choosing the rate of return for the future period, starting in March 1988. It did so by using the drop in interest rates to justify moving to the lower end of the previously calculated zone of reasonableness. There being no claim that the change in interest rates was offset by any other factor, the First Circuit had no difficulty upholding the Commission’s action. Id. at 967; cf. Union Electric Co. v. FERC, 890 F.2d 1193, 1201-04 (D.C.Cir.1989) (rejecting a comparable updating where the utility offered evidence that company-specific factors made it inappropriate). Here, by contrast, the Commission seized on data from after the period for which it was setting rates, to undermine the implications of data that were timely and, by its own precepts, controlling (at least in supplying the top of the zone of reasonableness).
Thus the Commission’s theory for departing from use of the midpoint of the zone of reasonableness is inconsistent with its basic and strongly held ratemaking theories. We are not saying that there are no legitimate bases for departing from the midpoint; we are merely saying that this is not one. In fact, the theory on which the parties seem to regard the midpoint as the proper norm — that Tennessee is “of average risk” — is a bit puzzling to us. Both boundaries of the zone of reasonableness, the 13.2% “risk premium” figure and the 18.79% DCF calculation, are estimates of Tennessee’s cost of equity capital. Neither *340is a measure of industry rate of return or of Tennessee’s relative risk within the industry. Thus it is unclear why a Commission assessment of Tennessee’s risk in relation to other natural gas companies is helpful in choosing a place within the zone of reasonableness. The existence of the zone of reasonableness range reflects the lack of precision of the estimation tools, not variability within the industry.
A better reason for treating the midpoint as a starting place is that an average is an obvious place to begin when there is no information that would incline the decision-maker to prefer one estimate over another. See Boston Edison Co., 42 FERC 1161,374 at 62,094 (1988) (“a simple average of the two [boundaries of the zone of reasonableness] accords equal weight to each position and represents a rational basis for selecting a particular value within the zone.”). Without some reasoned basis for such a preference, any departure from the midpoint is arbitrary. There may well be valid reasons for a departure, but the Commission has not shown us any.
* * * * # *
The case is remanded for proceedings consistent with this opinion.
So Ordered.
. The methodology behind this figure is unclear, but in adopting it the Commission noted "the relatively high yields on risk-free U.S. Treasury bonds during this period (which averaged 13.8%), the recommendations of the parties (Staffs 15% and Tennessee’s 17%), and the returns we recently allowed for pipelines,” and concluded “that the zone of reasonableness ... is 15.0% to 16.9%____” Id. at 61,097.
. The Discounted Cash Flow model flows from the classical valuation theory "that the value of a financial asset is determined by its ... ability to generate future cash flows.” Roger A. Morin, Utilities' Cost of Capital 74 (1984). Thus, “[t]he fundamental value of [any] asset is the discounted sum of all future income flows that will be received by the owner of the asset." Id.
.She took the lower bound of 15% from Opinion No. 190 and subtracted the then prevailing government bond rate of 13.8%, to arrive at an implicit "risk premium” of 1.2%. She then added this risk premium to the 12.0% government bond rate prevailing in the disputed period. Id. at 65,169. The resulting 1.2% premium is far lower than normal ones. See, e.g., Morin, Utilities’ Cost of Capital 181 (risk premium in electric utility industry ranged from 2.36% to 5.26% between 1974 and 1983 with an average equity risk premium of 3.78%); Frank K. Reilly, Investment Analysis and Portfolio Management 46 (1989), citing Roger G. Ibbotson and Rex A. Sinquefield, Stocks, Bonds, Bills, and Inflation: 1987 Yearbook (average equity risk premium of stocks over long-term government bonds for the period 1926-1987 of 5.6%).
. (13.2% + 16.84%)/2 = 15.02%.
. Energy Information Administration, Monthly Energy Review, January 1984, at 84 (Average Composite Refiner Acquisition Cost of Crude Oil for July 1982 and February 1983).
. This can be stated as Po = Di/(r — g), i.e., the price of a stock equals the value of next year’s dividends divided by the cost of capital net of the steady future growth rate of dividends. See Kolbe, The Cost of Capital 54; Morin, Utilities' Cost of Capital 82. This can then be restated to focus on what the regulator is seeking to discover, the cost of capital: r = Di/Po + g.
. Id. at 30,500, quoting R. Brealey and S. Myers, Principles of Corporate Finance 266 (1984). See also Order No. 461-A, 38 FERC ¶ 61,160, pp. 61,441-42. The Commission castigated Tennessee for not pointing to "any record evidence or explanation that would substantiate why investors would require dividend yields for the locked-in period that are almost 2 percentage points ... higher than those found by the Commission to be appropriate ... in Opinion No. 190.” Order 240-B, 33 FERC f 61,005 at 61,009 (1985). Given the Commission’s endorsement of the efficient market hypothesis, Tennessee’s proffer of DCF figures was just such evidence.