Today, the D.C. Circuit will hear en banc oral argument on the legality of the Clean Power Plan, an EPA rule that proposes to reduce carbon dioxide emissions from electrical power generation by a third over the next 15 years by, essentially, mandating the shutdown of much of the country's coal-fired generation. The EPA claims it has authority for this measure under Section 111(d) of the Clean Air Act, which, at most - I say at most because I'm bracketing the central issue in the case for the moment - provides regulatory authority to set "standards of performance" for existing (i.e., not new) sources of air pollutants that are neither among the six so-called criteria pollutants listed under Section 108 of the Act, nor are hazardous air pollutants (pollutants that cause cancer or other serious health effects) regulated under Section 112 of the Act. (I will refer, throughout, to sections of the Act; to find any of these sections in the U.S. Code, go to Title 42 and add 7300 to the section number.)
There are many serious questions about whether a mandate, broadly speaking, to shut down coal-fired generation and trade it for renewable sources is a "standard of performance" for coal-fired generation or, to quote the Act's definition of "standard of performance," a "best system of emission reduction which . . . the Administrator determines has been adequately demonstrated." Inasmuch as the Clean Power Plan's "standard of performance" consists of requiring the owners of fleets of generators to engage in a complex cap-and-trade-like system, there are serious questions about whether the Clean Power Plan's standards of performance are standards of performance for sources, as 111(d) requires, or the entities that own them. It is my sense, however, formed only on an admittedly cursory read of the briefs, that while the EPA may have the worse of these points absent deference, it can probably show sufficient ambiguity on them to survive Chevron, at least before the D.C. Circuit as presently constituted. The more dangerous question in the case for the EPA is whether, under a snippet of text often referred to as the 111(d) exclusion, coal-fired plants' emission of carbon dioxide is excluded from 111(d) regulation altogether.
I said above that, at most, 111(d) extends to pollutants that are neither 108 criteria pollutants nor 112 hazardous air pollutants (or as they are commonly known, HAPs). That is because, under at least the codified version of 111(d)'s exclusion, the EPA has authority to set standards of performance for sources of "any air pollutant  for which air quality criteria have not been issued [and] which is not included on a list published under section 7408(a) of this title [108(a) of the Act] or emitted from a source category which is regulated under section 7412 of this title [112 of the Act]." (My interpolation of an "and" replaces an "or" that all parties agree cannot be taken literally, as it would render the exclusion meaningless.) The references to pollutants for which criteria have not been issued and that have not been listed under 108 excludes criteria pollutants; the reference to pollutants emitted from source categories regulated under 112 excludes, at the very least, HAPs.
What, however, the exclusion of sources of pollutants "emitted from a source category which is regulated under " also seems to exclude is not just the HAPs themselves regulated in 112, but any sources of non-HAP pollutants that also emit HAPs and are regulated under 112, or indeed any category of HAP-emitting, 112-regulated source. That is to say, if coal-fired plants emit a HAP, and are therefore regulated under 112, EPA cannot regulate carbon dioxide emissions from coal-fired plants under 111(d). On first, second and third blush, this appears to be the only possible reading of the language. "Which is regulated under " clearly modifies the immediately preceding phrase "source category," not the "air pollutant" mentioned 31 words prior. What Congress excluded from 111(d) regulation was plainly, it would seem, not air pollutants regulated under 112, but "any air pollutant . . . . emitted from a source category," which source category, in turn, is regulated under 112 because it emits HAPs.
When the EPA first proposed the Clean Power Plan, it responded to this point in a rather foolish way. Prior to the 1990 amendments to the Clean Air Act, 111(d) only excluded from regulation, insofar as 112 went, any pollutant "included on a list published under" 112(b)(1)(A). Section 112 was renumbered in the 1990 amendments, such that, absent a substantive amendment to the scope of the 111(d) exclusion, the reference to 112(b)(1)(A) had to become a reference to 112(b). Hence, the Senate, which had no plans to broaden the 111(d) exclusion, passed a so-called conforming amendment modifying the cross-reference - specifically, "striking '112(b)(1)(A) and inserting in lieu thereof '112(b).'" Had this been all that happened in 1990, the 111(d) exclusion would still only exclude Section 112 HAPs, not any non-HAP emissions from sources that emit HAPs.
However, that was not all that happened. The House, in its bill, more substantively amended the 111(d) exclusion to include the language codified today - the exclusion of pollutants "emitted from a source category which is regulated under 112." Even though the Senate receded to the House amendment in conference, and even though the House amendment mooted the Senate's conforming amendment, which struck a reference to 112(b)(1)(A) in 111(d) that under the House amendment no longer existed, the Senate's conforming amendment found its way into the merged bill passed by both houses and signed by the President. The compilers of the U.S. Code sensibly ignored the amendment, and it was forgotten for 25 years until the EPA famously seized on it to claim that there were two enacted and signed versions of 111(d), and that it had discretion under Chevron to choose which one it would follow.
EPA's arguments on this score were absurd, both because the failure to pull the Senate conforming amendment from the final bill was plainly a drafting error, and because even if Congress did intentionally enact alternate versions of the same statute, delegating the question of which the law would be to an agency is the Platonic ideal of a violation of the otherwise moribund non-delegation doctrine. So, among others, Laurence Tribe argued between the EPA's issuance of its proposed and final rules, in fora including a hopeless petition for review in the D.C. Circuit of the EPA's proposed rule. Largely, I imagine, because of his arguments, the EPA's final rule primarily reasoned that the codified version of 111(d) was ambiguous, saving its discretion to choose between versions of 111(d) as an adventurous back-up theory.
How is "any air pollutant . . . emitted from a source category which is regulated under 112" ambiguous? It is unclear what part of this language, if any, the EPA deems ambiguous - it variously suggests that the exclusion is clear on "who" is regulated under 112 (sources) but not "what" is, or that perhaps "any air pollutant" may really mean "any HAP," though it obviously doesn't mean that with respect to the first half of the 111(d) exclusion that addresses criteria pollutants. But the upshot of its argument is that the exclusion only excludes source categories' 112-regulated emission of HAPs from duplicative 111(d) regulation - not, as 111(d) appears to say, any source emissions of any pollutant from a source category that is regulated under 112 for emitting HAPs.
The obvious rejoinder, which Professor Tribe and his co-counsel forcefully make in reply, is that the EPA is either reading a lot of words into or out of the statute. At one point, the EPA suggests that "any air pollutant . . . emitted from a source category which is regulated under 112" should be read "any HAP . . . emitted from a source category which is regulated under 112." Alternatively, maybe EPA is suggesting the language means nothing more than "any air pollutant . . . which is regulated under 112," though it would surely protest that claim. Or perhaps when the EPA says that its reading addresses "what" is "regulated under 112," not just "who," it is reading the language to say "any air pollutant . . . emitted from a source category which is regulated under 112 with respect to that pollutant" - which would limit the exclusion to 112-regulated HAPs. (It's this, I will argue, that is EPA's real and best argument.)
How, one wants to ask, can EPA justify reading all that into or out of the seemingly rather plain text of 111(d)? How is 111(d) ambiguous on whether it implicitly includes all these missing words, or more implausibly still, implicitly excludes a lot of the words that are actually there? It turns out that sentences of the form of the 111(d) exclusion are routinely ambiguous in just the way the EPA posits.
111(d), again, requires regulation of "any air pollutant . . . which is not . . . emitted from a source category which is regulated under 112." I will suppose that this language is an instance of sentences that take the following general form: do/don't (verb) (noun) that is/isn't (verb optional) from a (noun) which (something). Some examples: don't buy chicken (that is) from a store that has a low sanitation score. Don't buy a dog (that is) bred from a dog who had rabies. Only adopt dogs (that are) from shelters that never kill any animals. Regulate any air pollutant which is not emitted from a source category which is regulated under 112.
As these examples hopefully illustrate, sentences like these have three parts. First, there is a command with respect to some thing. Second, that thing is qualified as being "from" or not "from" something else, sometimes with a verb preceding the from ("bred from," "emitted from"). Third, that something else is in turn qualified in some way.
Most fairly read, EPA's argument is that the third part of the relevant language in the 111(d) exclusion - the part that qualifies the source category as "regulated under 112" - not only qualifies the source category in the second part of the sentence, but implicitly refers back to and is limited by "any air pollutant" in the first part of the sentence. In this way, the exclusion only applies to pollutants emitted from source categories regulated under 112 with respect to those pollutants. Can 111(d) be read this way? To me, this comes down to whether sentences of the form "do/don't (verb) A that is from B that is C" sometimes are commonly understood in context, in conversation or writing, to mean "do/don't (verb) A that is from B that is C with respect to A." Some examples will show that they are.
First, consider "don't buy chicken from a supermarket that doesn't have a sale." In conversation, this sentence would be universally understood to mean "don't buy chicken from a supermarket that doesn't have a sale on chicken." One might claim that the sentence literally only commands the listener to not buy chicken from a supermarket that isn't running any sale. But no one would understand the sentence that way as (a) read that way it would rule almost no supermarket out and be a pointless thing to say (all supermarkets have sales on something), and (b) whether or not a supermarket has a sale on non-chicken products is irrelevant to whether one should buy chicken there. What is relevant to buying chicken is sales on chicken. Context, then, dictates that "doesn't have a sale" refers back to and is limited by the "chicken" at the beginning of the sentence. Note that this cannot be explained away by something unusual about the grammar or usage of the word "sale"; it is perfectly possible to talk about sales simpliciter in sentences that look a lot like the one in the example. "Don't shop at Neiman Marcus when they don't have a sale" just means "don't shop at Neiman Marcus when they don't have a sale."
Second, consider a more ambiguous case - "don't buy chicken from a fast-food restaurant that has had a recall." Some listeners would interpret this stricture to only apply to restaurants that have had chicken recalls and read the sentence to impliedly say "a recall of chicken"; if a restaurant's had recalls of its lettuce, why should that affect one's chicken purchases? Other listeners would interpret it to apply to any recall; any recall at a restaurant raises doubts in some minds about the safety of buying that restaurant's chicken. Many listeners, I believe, would ask "what do you mean? Any recall or just a recall of the chicken?" In any event, if the speaker clarified and said he only meant chicken recalls, very few people, if any, would feel that the speaker had unambiguously said something different from what he meant.
Third, consider a case where the qualifier in the third part of the sentence almost certainly doesn't refer back to the noun in the first part of the sentence: "don't buy chicken for Chester (who has a severe peanut allergy) from a restaurant that uses peanut oil." A listener who was very un-cautious about peanut allergies, or didn't understand them, might think the speaker meant to only exclude restaurants that use peanut oil in their chicken. But most listeners would understand the risk posed by peanut oil in the kitchen where the chicken was prepared, and take "that uses peanut oil" to not refer back to, and be limited by, "chicken."
What these examples show, I hope, is that "don't (verb) A from a B that C" can impliedly mean (and not just be intended to mean but fail to mean) "don't (verb) A from a B that C with respect to A," and that the substantive content of a sentence with this form will usually tell or help tell us whether "with respect to A" is implied or not. In linguistic terms, the implied "with respect to A" in many sentences like these is a species of conversational implicature, although some theories of linguistics might explain the phenomenon as a convention on which sentences like these just mean (and do not merely imply) an unstated "with respect to A."
Back to the 111(d) exclusion, is the language "any air pollutant . . . emitted from a source category which is regulated under 112" at least ambiguous on whether it means "any air pollutant . . . emitted from a source category which is regulated under 112 with respect to that pollutant"? One standard textualist move would be to argue that this is a very roundabout way of essentially saying "any air pollutant which is regulated under 112," and renders the House amendment in 1990 a strangely pointless exercise in adding unnecessary words to the statute.
Recall, however, "don't buy chicken from a supermarket that doesn't have a sale." That command, which seems ordinary enough, could also be described as a roundabout way of saying "don't buy chicken that isn't on sale"; why mention the supermarket? The speaker that particularly wants to distinguish the right supermarkets from the wrong ones, as people often do in this context, will add the extra words, though they add little extra meaning. Likewise, the drafter that particularly wants to distinguish the right source categories from the wrong source categories will add the extra reference to source category. Given that 111(d) is about regulating sources, not pollutants, it makes all the sense in the world that a drafter would add surplus language that focused the 111(d) exclusion on sources regulated elsewhere. Further, the House amendment would not, on this reading, be entirely pointless, for two reasons. First, it, like the Senate amendment, was rendered necessary by the renumbering of 112; second, by altering the exclusion from pollutants "included on a list published under" 112 to source emissions of pollutants actually "regulated under" 112, the House would have narrowed the scope of the exclusion to source emissions of those pollutants EPA was really regulating, as opposed to any pollutant as to which 112 called for regulation.
Absent the flight to surplusage, the only way I see to deciding whether "air pollutant . . . emitted from a source category which is regulated under 112" can permissibly be read under Chevron to contain an implied "with respect to that pollutant" is to ask whether the statute could make sense that way. For given how common these kinds of implicatures are, the implicature need only be sensible for it to be permissible.
The implicature is obviously sensible; until 1990 111(d) said just about what it would mean now with the implied "with respect to that pollutant." Then, it excluded source emissions of HAPs listed under 112; now, with the implicature, it would exclude source emissions of pollutants emitted from sources that were regulated under 112 for emitting those pollutants, because those pollutants are HAPs. Again, it's an awfully roundabout way of excluding HAPs, but only excluding 112-regulated source emissions of HAPs from 111(d) is a perfectly plausible thing for Congress to say. I would even venture to say it's quite a bit more plausible than excluding any source from regulation for its emission of non-criteria, non-HAP pollutants if that source is already regulated for emitting HAPs. While petitioners vaguely gesture in the direction of the 1990 House's supposed intent to root out duplicative regulation of regulated sources, it doesn't make much sense to say that, even though carbon dioxide is a regulable air pollutant, the EPA can't touch coal-fired plants' emissions of it under 111(d) because it's already regulating coal-fired plants' emission of mercury and the like (well, trying to) under 112. It's tough to see how both regulating a coal-fired plant's emission of mercury acid and its emission of greenhouse gases is duplicative. In short, the more one looks at 111(d)'s language and thinks about what that language is doing, it starts to read a lot more like "don't buy chicken at a supermarket that doesn't have a sale [on chicken]" than a provision so unambiguously limited to the broad prohibition on its face that EPA had to doubt its enactment into law.