dissenting:
Southern Pacific Transportation Co. (“SPT”) petitions this court for reversal of an Interstate Commerce Commission (“ICC”) order denying SPT’s application to abandon a branch railroad line. I would reverse and remand, on the grounds that the ICC’s decision failed to establish a rational link between facts found and decision made on two issues: Opportunity costs, and prospects for growth in traffic.
I
From Folsom Junction, California, near Sacramento, a branch railroad line winds some thirty-eight miles into the foothills of Sierra Nevada mountains to the small town of Placerville. A century ago, the line carried gold-seekers into the hills. Today, it carries more prosaic cargoes of lumber. Approximately two freight trains per week use the Placerville branch, which is serviced by SPT.
Traffic on the line has declined precipitously, from 2,500 carloads of freight in 1975 to 341 in 1986. Believing the decline irreversible, SPT filed an application with the ICC to abandon the line on December 30, 1986. Pursuant to 49 U.S.C. § 10904, the ICC has the authority to decide whether a railroad may abandon any part of its system. SPT based its application on data regarding the line’s finances, and on inferences based on those data. The decline in carloads was well documented. SPT attributed the decline to general difficulties in the lumber business, and to a shift of lumber transportation to trucks. SPT made a profit on the Placerville branch of $184,019 in the base year of July 1985 through July 1986.
SPT foresaw major losses in the future for several reasons, however. First, several of its most important customers had recently ended operations or shifted entirely to trucking. Second, major maintenance was needed to keep the line operating. Third, SPT calculated large “opportunity costs.” Opportunity cost is the economic loss incurred by committing assets to an inefficient use, when they could be used for other purposes generating a higher rate of return. SPT calculated its opportunity cost of continuing to operate the Placerville branch at $1,442,637 per year. SPT further asserted that the loss to Placerville businesses of abandoning service would be minimal, since Placerville enjoys good highway access, allowing goods to be trucked to (or from) Placerville directly to customers, or to the Sacramento railhead for transshipment. In short, SPT argued that it should not be required to operate an inefficient and unnecessary rail line indefinitely.
Certain shippers using, or allegedly planning to use, the Placerville branch protested the application. They were joined by the county of El Dorado, California. Some of these shippers informally projected very large increases in their use of rail transportation.
On July 27, 1987, by a three to two vote, the ICC denied SPT’s application. Southern Pacific Transp. Co., No. AB-12 (ICC July 27, 1987) [Hereinafter SPT I], The ICC engaged in a balancing analysis that weighed the harm to the railroad of continued service against the harm to the community of abandonment. The ICC noted that the line was still profitable, and that projected increases in shipments could offset the loss of some major customers.
The general thrust of the ICC’s opinion was that the burden of proof was on SPT, 49 U.S.C. § 10904(d)(1), and that SPT had failed to assert sufficient facts regarding its projected losses, the availability of alternate transport, and the lack of adverse community impact to meet its burden. The ICC considered opportunity costs, noting that the costs as calculated by SPT and by the protestants varied by a factor of ten, due primarily to disputes over the value of the roadbed land. The ICC did not venture into this morass to attempt to calculate opportunity cost itself; accepting SPT’s *846figure for the purpose of argument, the ICC held that even this cost did not outweigh the factors adverse to abandonment. Two commissioners sharply dissented on all the issues.
SPT appealed the denial administratively. The ICC considered the appeal but denied it, basically for the same reasons asserted in the original opinion. Southern Pacific Transp. Co., No. AB-12 (ICC Nov. 10, 1987) [Hereinafter SPT II]. SPT timely filed a petition in this court for review of the ICC’s order. We have jurisdiction. 28 U.S.C. §§ 2321, 2342(5), 2344; Fed.R.App. P. 15.
II
The ICC long has had the power to decide whether a railroad may abandon part of its network. See Colorado v. United States, 271 U.S. 153, 46 S.Ct. 452, 70 L.Ed. 878 (1926). The ICC must apply a balancing test in which the needs of the community served are weighed against the burden on interstate commerce that results from requiring a railroad to use its resources inefficiently and unprofitably. Id.; Cartersville Elevator, Inc. v. ICC, 724 F.2d 668, 670 (8th Cir.), aff'd on rehearing, 735 F.2d 1059 (1984) (en banc). The ICC presently has the authority to approve an abandonment “only if [it] finds that the present or future public convenience and necessity require or permit the abandonment or discontinuance.” 49 U.S.C.A. § 10903(a) (Supp.1988). The statutory scheme and accompanying regulations require a decision based upon full consideration of the relevant interests of the railroad, its customers, and the communities affected. See 49 U.S.C.A. §§ 10903-04; 49 C.F.R. § 1152 (1988).
We review ICC decisions under the narrow “arbitrary, capricious, an abuse of discretion, or otherwise not in accordance with law” standard of the Administrative Procedure Act.1 5 U.S.C. § 706(2)(A). The scope for review is indeed limited. Although our inquiry is to be searching and careful, we do not substitute our judgment for that of the agency. Bowman Transp., Inc. v. Arkansas-Best Freight System, Inc., 419 U.S. 281, 285, 95 S.Ct. 438, 441, 42 L.Ed.2d 447 (1974). The searching and careful inquiry determines whether the agency has articulated a “rational connection between the facts found and the choice made.” Id. (quoting Burlington Truck Lines v. United States, 371 U.S. 156, 168, 83 S.Ct. 239, 245, 9 L.Ed.2d 207 (1962)); see Glazer Steel Corp. v. ICC, 748 F.2d 1006 (5th Cir.1984) (affirming ICC approval of abandonment under conditions of declining usage and high maintenance cost). We may not weigh the evidence ourselves, but must inquire into whether the agency’s conclusions are rationally supported. Bud Antle, Inc. v. United States, 593 F.2d 865, 869 (9th Cir.1979).
Ill
SPT makes several specific objections to the ICC’s decision.2 SPT first questions how the ICC could rationally have decided the case without calculating a specific figure for the opportunity cost of the branch line. Other circuits reviewing abandonment petitions have held that opportunity costs should be considered by the ICC. International Minerals & Chem. Corp. v. ICC, 656 F.2d 251, 259 (7th Cir.1981); Missouri Pac. R.R. v. United States, 625 F.2d 178, 182 (8th Cir.1980). Here, the ICC declined to determine opportunity costs. The ICC noted that the protestants’ calculation of opportunity cost was $144,000, while SPT’s was $1,442,637 per year. SPT I, supra, at 8. The ICC admitted that either figure represented a “significant” opportunity cost. It held, however, that an exact calculation was not necessary: “Even ac*847cepting the SPT’s figures, those costs would not constitute a reason to grant the abandonment.” Id. (emphasis added). In other words, the ICC held that an assumed $1.4 million yearly opportunity cost to SPT was outweighed by the various intangible factors mitigating against abandonment.
On appeal, the ICC repeated that it had “considered” opportunity costs, but found them to be outweighed by the adverse community impact factors. SPT II, supra, at 8. The ICC also severely compromised its acceptance of SPT’s calculations of opportunity cost by strongly emphasizing the contrary calculations of El Dorado County. Id.
This treatment of opportunity cost was arbitrary and capricious. First, it is unclear, despite the ICC’s apparent acceptance of SPT’s figures, which ones it really relied upon, based on its continued emphasis of the protestants’ lower figures and its lumping together of the various positions as “substantial.” True as this may be, there is a major difference between an opportunity cost less than base year profit, and one approximately eight times greater. The factual basis for the decision was so unclear as to require reversal and remand for a better explanation of which opportunity cost was used.
Second, assuming that the ICC did accept SPT’s figure of $1.4 million, the decision was not grounded by a rational explanation of its factual basis. The ICC is correct in stating that a consideration of opportunity cost need not involve a precise numerical weighing of the cost against all other factors. Data accepted by the ICC that indicate an opportunity cost of $1.4 million per year, however, require more than a conclusory response that they are “outweighed” by intangible (and, in fact, quite speculative) factors. Such a “consideration” is, in fact, no consideration at all.3 The only thing making this a close case is the deferential standard of review, which counsels due respect for the ICC’s authority at every turn. But the ICC has failed to articulate the rational connection between facts and decision that we require to perform our reviewing function. See Bowman, 419 U.S. at 285, 95 S.Ct. at 441. The denial of SPT’s petition should be reversed, and the case remanded for further proceedings, such proceedings to include either 1) a calculation of opportunity cost and weighing of that cost against the other relevant factors, or 2) a thorough explanation of why the factors of continued profitability, adverse community and shipper impact, and alternate transportation services outweigh an opportunity cost to SPT of $1,442,637 per year.
IV
SPT argues that the ICC improperly ignored its evidence regarding the loss of major customers, and the resulting prospect of major financial losses. The Michigan-California Lumber Co. shipped 96 cars on the Placerville branch in 1985. In September 1986, it abandoned its spur line to the branch and began shipping exclusively by truck. The Certainteed Corp. received 266 carloads of plastic pellets in 1985, nearly half of the branch line traffic. The plant has since closed. SPT argues that these losses ripped the heart out of its business.
In response, certain of the protestants projected replacement use of the line. The Certainteed plant had been purchased by the PW Pipe Co. (“PW”) and had resumed operations. PW forecast possible use of 275 carloads of freight per year. In addition, the Cornett Co. had purchased the Placerville Lumber Co. Cornett projected possible shipment of a remarkable total of 1,000 carloads annually.4 Its predecessor, *848Placerville Lumber Co., shipped exactly three carloads in 1985.
The ICC accepted these projections more or less at face value. In so doing, the ICC failed to make even a minimal determination of credibility. These projections did not deserve the weight they were given— equal to, or more significant than, SPT’s projection of sharply lower traffic volume as a result of the loss of major customers. SPT’s estimate was based on hard data of plant closures and declining traffic. The companies projecting massive future shipments are under no financial obligation actually to live up to their promises. It is in their interest to require the line to be kept open by regulatory means for their possible future use, because they incur no costs by doing so. The ICC failed to explain why it gave these extraordinarily speculative predictions the same credence as projections backed up by a clear economic interest and by historical data.
The ICC held that SPT had failed to meet its burden of showing that these projections were inaccurate. It is difficult to see how SPT ever could do so to the ICC’s satisfaction. SPT pointed out the obvious fact that these projections were pure pie-in-the-sky speculation, especially those of Cor-nett Co. It further showed that both Cor-nett and PW would have to build rail spurs to the branch line to use it at all, and that since opening their doors both have appeared quite content to truck all their shipments.
The ICC responds that since all projections are uncertain, it prefers to maintain the status quo until the level of shipments is better known. The abandonment may be approved at some future date, but once approved, the service will never be restored. The ICC has no authority to require restoration of abandoned service. See Hayfield N. R.R. v. Chicago & North Western Transp. Co., 467 U.S. 622, 633-34, 104 S.Ct. 2610, 2617, 81 L.Ed.2d 527 (1984). The ICC further points out that shippers cannot be expected to invest in spur lines while the future of the branch line is in doubt.5 The ICC’s caution is therefore understandable. On the other hand, of course, requiring monetary expenditures to maintain a branch line that likely will soon be abandoned anyway is inefficient. The argument can he made endlessly that a better decision can be made at some later date. At some point, justice delayed is justice denied.
We defer to predictions based upon an agency’s expertise. See FCC v. National Citizens Comm’n for Broadcasting, 436 U.S. 775, 813-14, 98 S.Ct. 2096, 2121, 56 L.Ed.2d 697 (1978). It is apparent from the record, however, that these predictions were not based on anyone’s expertise. They were sheer speculation, unsupported by any evidence that they would ever come to pass. The ICC did not make any reasonable findings as to their credibility, or establish, in other than a merely conclusory manner, the rational link between reasonable factfinding and ultimate decision that Bowman requires. Bowman, 419 U.S. at 285, 95 S.Ct. at 441.
V
SPT next argues that the ICC gave insufficient weight to its alleged maintenance and rehabilitation costs. These costs should be considered by the ICC. See International Minerals, 656 F.2d at 256-58. The ICC did consider them, and found that the line would remain profitable if short-term maintenance were carried out while *849“unavoidably more speculative” long-term rehabilitation was deferred. SPT’s objections to this finding are insufficient to support reversal in light of our deference to agency factfinding.
VI
The ICC found that SPT had not met its burden of demonstrating that the hardship to it of continued service outweighed the detriment to the community of abandonment. SPT contests this holding as not reasonably grounded in fact. It is true that PW and Cornett began operations while this case was under review and do not use rail at all. On the other hand, the ICC could reasonably have inferred that a loss of all rail links to Placerville would have an adverse effect on the line’s remaining customers. This is a case where both sides present reasonable arguments, but differ completely in their view of the facts. In such a case we must defer to the ICC’s judgment unless there was clear error, which there was not. See Texas, 642 F.2d at 89.
The same analysis applies to the dispute over alternative means of transportation. See Georgia Public Service Comm’n, 704 F.2d at 545 (alternate means of transportation must be economically feasible). SPT contended that Placerville is well served by highways connecting it to the Sacramento railhead. The ICC decided that SPT had not adequately demonstrated the economic feasibility of trucking lumber to Sacramento or elsewhere. Given the current use of trucks by most Placerville lumber companies, this decision is certainly open to question, but it is not so clearly wrong that we must reverse it.
CONCLUSION
The ICC failed to give a rational explanation for its decision that various unquantified costs to Placerville shippers outweighed $1.4 million in annual opportunity costs in the balance test. The ICC further failed to establish, by means of minimum findings as to credibility, a rational reason for its acceptance of highly speculative projections of future traffic growth.
I would reverse and remand on the issues of opportunity costs and projections of growth. The ICC should have the opportunity to make additional findings of fact as to the specific amount of opportunity cost and its relationship to the other factors mitigating against abandonment in the balance test. If the ICC prefers to continue its apparent acceptance of SPT's calculation of opportunity cost, it should better explain why a yearly cost of $1.4 million is outweighed by the stated factors. I would further remand to permit the ICC to make findings as to the credibility of the projections of future traffic provided by certain shippers, and to explain the weight accorded to such projections. I concur in the court’s affirmance of the ICC’s order on the other issues argued by SPT.
. Some courts have applied the "unsupported by substantial evidence" lest of 5 U.S.C. § 706(2)(E). E.g., Texas v. United States, 642 F.2d 87, 89 (5th Cir.1981). It is questionable whether section 706(2)(E) applies to an abandonment case, since no hearing on the record was required or held. See Georgia Public Service Comm’n v. United States, 704 F.2d 538, 542 n. 8 (11th Cir.1983); Illinois v. United States, 666 F.2d 1066, 1071-72 (7th Cir.1981). I use only the "arbitrary and capricious” test.
. I use "decision" to mean both SPT I and SPT II, unless indicated otherwise.
. The dissenting ICC Commissioners pointed out that this opportunity cost differed by orders of magnitude from that found to justify abandonment in other ICC cases.
. In materials filed with the ICC in response to SPT’s appeal, Cornett changed its estimate to 250 carloads per year, without any explanation of the 75 percent reduction. PW also estimated 250 carloads annually. See SPT II, supra, at 3-4. The ICC held that the projections were not “speculation” because the shippers had "reaffirmed” their "serious ... efforts to increase traffic." Id. at 4. In other words, the shippers had simply repeated their hypothetical plans to use rail transportation at some unspecified time in the future. The ICC also noted that “since *848SPT has not established that current operations are unprofitable, the shippers have the opportunity to increase traffic to anticipated levels.” Id. This puzzling non sequitur defies logical analysis.
. This argument holds SPT hostage indefinitely. As long as SPT wants to abandon service, the shippers need not invest in the improvements that would allow them to use the line. Therefore, they may merely assert that they want to use the line at some future date and, as long as it wants to abandon, SPT can never rebut their speculative projections. SPT apparently can only prove its case for abandonment by withdrawing its application and waiting several years to see if any traffic materializes — subject, of course, to the counter-argument that if abandonment has once been attempted, the line’s future always remains in doubt because the attempt is likely to be repeated. See J. Heller, Catch-22 (1955).