dissenting.
I agree with the majority that the petition for review presents a controversy within the judicial power, but that is as far as our agreement reaches. The majority says that the question on the merits is “whether it was ‘necessary’ for Monsanto to have an extension of the original waiver when the company discovered in August 1990 that its control technology did not perform as predicted”. Opinion at 1204. Putting things in this way shows principally that the wrong question begets the wrong answer.
Under 42 U.S.C. § 7412(c)(l)(B)(ii) (1988) the Administrator of the EPA “may grant a waiver permitting [a stationary source] a period of up to two years after the effective date of a standard to comply with the standard, if he finds that such period is necessary for the installation of controls” and that steps in the interim will “assure that the health of persons will be protected from imminent endangerment.” “May” grant, not “must” grant. A finding that more time is “necessary” to comply with the rules is a necessary but not a sufficient condition of a waiver. The applicant must persuade the Administrator that a favorable exercise of discretion is appropriate. Statutes of this kind are common. Like rules of judicial procedure that permit judges to grant litigants more time, they permit the person in charge to exercise discretion. Nothing in § 7412(c)(l)(B)(ii) or in any of the regulations sets out criteria under which anyone is entitled to a waiver. The Administrator “may” grant waivers but would be within her rights to limit them to extraordinary cases. Chevron U.S.A. Inc. v. Natural Resources Defense Council, Inc., 467 U.S. 837, 842-44, 104 S.Ct. 2778, 2781-82, 81 L.Ed.2d 694 (1984); United Auto Workers v. Dole, 919 F.2d 753, 755-58 (D.C.Cir.1990); Wisconsin Electric Power Co. v. Reilly, 893 F.2d 901, 907 (7th Cir.1990); Ethyl Corp. v. EPA 541 F.2d 1, 34 (D.C.Cir.1976) (en banc). Cf. INS v. Jong Ha Wang, 450 U.S. 139, 144-45, 101 S.Ct. 1027, 1031, 67 L.Ed.2d 123 (1981). Monsanto does not cite, and I could not find, any case holding that the EPA must grant more time than the 90 days the Clean Air Act prescribes as the norm.
As it happened, the EPA gave Monsanto more than the 90 days. It allowed 11 months — all the time Monsanto originally requested — to comply with the standard. Monsanto believes that the agency’s failure to grant an additional 13 months, up to the statutory maximum, is arbitrary. A water scrubbing system returns benzene to the storage vessel for reuse. Congress believes that such systems are preferable to end-of-the-pipe devices that capture the hazardous substance and either destroy it and create byproducts (as incineration does) or mix it with other substances for disposal. See 42 U.S.C. § 13101 (Supp. II), added by the Pollution Prevention Act of 1990; 42 Ú.S.C. § 7401(c) (Supp. II), added by the Clean Air Amendments Act of 1990. After it discovered that the water scrubbing system it had installed removed 80% of the benzene rather than the necessary 95%, Monsanto told the EPA that it was considering both incineration and carbon adsorption as supplementary technologies to remove the remaining benzene. After Monsanto settled on carbon adsorption — which leaves a slurry of carbon, benzene, and other substances that must be buried or otherwise disposed of — the EPA turned down its request for more time. Ac*1209cording to the EPA’s letter denying the request for additional time, Monsanto could have installed carbon adsorption units at the beginning, bringing itself into compliance with the benzene standard in less time than the 24 months it wanted. As Monsanto portrays these events, the EPA thus “penalized” it for its initial choice of water scrubbing.
There might be a problem if the'EPA had said something like: We will give you 24 months if you want to install carbon adsorption technology (dirty) but only 11 months if you want to use water scrubbing (clean). What it said, however, is that Monsanto could have 11 months to install water scrubbing and no additional time to add carbon adsorption. Any thumb on the scale favored the cleaner technology. Actually, however, there was no thumb at all. The EPA did not favor either technology; it allowed Monsanto to choose and then insisted that Monsanto keep its word that it would comply by August 1990. If it is apt to apply the word “penalty” to this sequence, then the event penalized is failure. The EPA gave Monsanto enough time to install a control technology of Monsanto’s choosing. When this strategy flopped, the EPA concluded that Monsanto rather than the public should pay the price. Although Monsanto contends that it acted in the best of faith — that computer modeling showed that water scrubbing would work, and that it was dismayed when the predictions did not come true — the EPA may insist, as pollution control statutes generally do, on results. Union Electric Co. v. EPA, 427 U.S. 246, 257-60, 96 S.Ct. 2518, 2525-27, 49 L.Ed.2d 474 (1976). An A for effort may affect the selection of a penalty in an enforcement proceeding (or may influence the exercise of prosecutorial discretion), but it does not compel the EPA to give a polluter the maximum waiver permitted by law.
Monsanto contends, and the majority believes, that the EPA misunderstood the original possibilities. Suppose Monsanto had two choices: (a) install water scrubbing equipment and, if this is not completely successful, add a small carbon adsorption unit to remove benzene from gasses that get past the scrubber; (b) install a larger carbon adsorption unit as the sole pollution control method. If these two strategies are equally costly and take an equal time to accomplish, then the former should be preferred because it produces less toxic waste and because the initial scrubbing stage can be put in place faster, reducing aggregate emissions.- Monsanto submits that the EPA acted arbitrarily in denying it enough time to complete the second stage of strategy (a), when it would have afforded the 24 months needed for strategy (b). Yet nothing in the EPA’s decisions suggests that it would have allowed 24 months for strategy (b). To the contrary, the EPA believed that strategy (b) would itself have taken only 11 months, making a full 24-month waiver unnecessary. Monsanto thinks that the EPA’s belief about the time that would have been necessary to install carbon adsorption from the start is unsupported, making its decision an abuse of discretion. Monsanto did not furnish the EPA with the facts necessary to support its position, however. Instead of providing, say, an engineer’s report .showing why strategy (b) would have taken 24 months, Monsanto submitted this rhetoric:
A second incorrect assumption that the U.S. EPA relies on is that “the record indicates that Monsanto could have been in compliance within six months after December, 1989.” It bases this conclusion on the fact that “once Monsanto decided to use a carbon absorption [sic] unit as its contrbl technology, it could project that it would have the unit ‘up and running’ within four months, and that it would be in final compliance within six months.” This- ignores the fact that the modifications to the water scrubbers remain in place and are the primary method of control. The carbon adsorption unit is additional control so that compliance with the Benzene NESH-AP is demonstrable. The four month and six month deadlines refer to the installation of additional control equipment, not the installation of a carbon adsorption unit as the primary method of control. Because a carbon adsorption unit as the primary method of control would have to be geometrically larger, the time for design and installation would have been geometrically longer.
Letter of August 19, 1992, at 9, with emphasis and brackets in original. What does it mean to say: “Because a carbon adsorption *1210unit as the primary method of control would have to be geometrically larger, the time for design and installation would have been geometrically longer.”? How much larger? How much longer? The letter does not say. Why should the EPA believe that there is a geometric relation between the size ■ of a mechanism and the time needed to install it? The letter does not say. The majority writes: “It is unfortunate that Monsanto was not more specific as to how much longer it would take to install the larger system.” Opinion at 1207. “Unfortunate” is an understatement. How can it be an abuse of discretion to turn down a flatulent request? Monsanto bore the burden of persuasion; it submitted nothing but hot air (presumably benzene-free) and so must lose.
Undergirding the majority’s opinion is an independent evaluation of the merits of different pollution-control strategies. Two judges believe that Monsanto “made a scientifically and environmentally sound decision to proceed with the water scrubber system” (opinion at 1206) and that the EPA’s view is “short-sighted and bad environmental policy” (opinion at 1206). Yet the record in this case does not demonstrate that Monsanto’s system is “sound” or that the EPA’s view is “bad environmental policy”. It contains essentially no evidence on these subjects (although Monsanto’s brief is full of self-congratulation, which my colleagues have swallowed). We are not engineers and are in no position to evaluate the evidence it does contain, and at all events we are not the persons to whom Congress delegated the estimation of costs and benefits.
The EPA may have much to answer for in its design of benzene control rules. According to the Office of Management and Budget, the several benzene NESHAPs create costs as high as $168.2 million per premature death averted. Regulatory Program of the United States Government, April 1, 1991— March SI, 1992 at 12. Costs in this range likely imperil moré people than they protect. Higher income is associated with better nutrition and medical care; regulations creating costs exceeding $7.6 million per life (directly) saved may well yield greater indirect loss of life. See Stephen Breyer, Breaking the Vicious Circle 23 (1993) (citing empirical studies). Nonetheless, Congress vested in agencies rather than the judiciary the task of maximizing the benefits of safety regulation. American Dental Ass’n v. Martin, 984 F.2d 823, 825-27 (7th Cir.1993). Monsanto does not challenge the NESHAP for benzene storage vessels. We must assume, therefore, that expeditious compliance is desirable, and we must accept the EPA’s judgment that speedy compliance has benefits exceeding the costs of using somewhat “dirtier” control strategies. The greater the gains from the rule, the more a rational person would sacrifice to achieve compliance sooner; yet Monsanto does not assess the safety effects of the benzene storage vessel rule. By refusing to extend the waiver, the EPA put Monsanto to a choice. Delay created the- possibility of fines. It could spend more to expedite compliance, or it could pay the fines. Monsanto elected not to speed its efforts; now we excuse it from paying fines. Given the structure of the Clean Air Act, an incentive to comply sooner cannot be an abuse of discretion.
The EPA was entitled to be stingy when evaluating Monsanto’s second application. The EPA need not continue granting extensions to a firm that bets on the wrong technology, as Monsanto did. My colleagues explain why they would have given Monsanto more time; they do not demonstrate that the Administrator’s contrary decision was an abuse of discretion.